ON THE FRACTIONATION OF BUTTERMILK BY MICROFILTRATION MEMBRANES

Transcription

1 PIERRE MORIN ON THE FRACTIONATION OF BUTTERMILK BY MICROFILTRATION MEMBRANES Thèse présentée à la Faculté des études supérieures de l'université Laval dans le cadre du programme de doctorat en Sciences et Technologie des aliments pour l'obtention du grade de Philosophiae Doctor (Ph.D.) DEPARTEMENT DES SCIENCES DES ALIMENTS ET DE NUTRITION FACULTÉ DES SCIENCES DE L'AGRICULTURE ET DE L'ALIMENTATION UNIVERSITÉ LAVAL QUÉBEC AOÛT 2006 Pierre Morin, 2006

2 11 RESUME Le babeurre, sous-produit de fabrication du beurre, est riche en composés mineurs associés à la membrane du globule de gras (MFGM). Ces composés possèdent d'excellentes propriétés fonctionnelles et certaines propriétés biologiques uniques. Le but de ce projet était de développer une approche de séparation des composés de MFGM par microfiltration. Plusieurs conditions de filtration (température, diamètre de pores, type de babeurre) on été étudiées. Les résultats on démontré que la présence de micelles de caséines dans le babeurre nuisait à la séparation de la MFGM en raison des tailles comparables des composés. Des différences de comportement en MF du babeurre ayant subi des traitements de séchage par rapport au babeurre frais on été notées mais une étude de la microstructure du babeurre en fonction des étapes du procédé industriel n'a permis de dévoiler que des changements compositionnels subtils et non structurels de la MFGM. L'utilisation d'un babeurre dérivé d'une crème de lactosérum dépourvue de caséine a permis d'améliorer la séparation de la MFGM. Enfin une approche de lavage de la crème avant le barattage a permis une concentration efficace des composés de la MFGM par microfiltration et ce en plus de générer des flux de perméation 2 fois plus élevés que dans les procédés étudiés auparavant. Les procédés étudiés dans ce projet ont permis de mieux comprendre la séparation des composés de la MFGM à partir de babeurre et de produire des fractions pouvant potentiellement être exploitées pour leur propriétés fonctionnelles et nutraceutiques.

3 III ABSTRACT Buttermilk is the by product from churning of cream to make butter. This by product is rich in minor components associated with the milkfat globule membrane (MFGM). MFGM components demonstrate excellent functional properties as well as some significant biological properties. The goal of this project was to develop an approach for the séparation of MFGM material from buttermilk by microfiltration. Various fîltration parameters (température, pore size, buttermilk type) were studied. Results showed that the présence of casein micelles in buttermilk was detrimental to the séparation of MFGM because of the similitude of both components sizes. However, différences in the behavior of buttermilk during MF were noted with the use of reconstituted buttermilk as opposed to fresh buttermilk. A study of the composition and microstructure of buttermilk after each step of buttermilk powder processing revealed subtle changes in buttermilk composition but it was impossible to correlate those changes with a modification of buttermilk microstructure. The use of a buttermilk derived from churning of whey cream deprived in caseins, allowed for an improvement of the séparation of MFGM. Finally an approach of cream washing prior to churning was shown to be efficient to remove caseins from the buttermilk and therefore improve séparation of MFGM. Permeation fluxes were also increased by 2 folds while using washed cream buttermilk as opposed to normal cream buttermilk. The processes studied in this project allowed for a better understanding of MFGM séparation in buttermilk and the création of fractions with potential functional and nutraceutical properties.

4 IV FOREWORD This work has been supported in part by Natural Science and Engineering Research Council of Canada, California Dairy Research Foundation, California State University Agriculture Research Initiative and Dairy Management Inc. The work carried out in this thesis was aimed at characterization and fractionation of buttermilk by MF. This thesis contains ail the results obtained during the course of realization of the project. Results are presented in the form of 4 published or submitted articles. This work represents an achievement of 4 years of research that could not even hâve been carried out without the support and understanding of many important people that I shall therefore acknowledge. First of ail, I would like to express my gratitude to my principal advisor Dr. Yves Pouliot. Dr. Pouliot, with his great advices and sound guidance gave me ail the tools I needed to successfully go through a difficult research area. He gave me my first chance in research and I will always be grateful for that. The confidence he placed in me always felt spécial. I can't count the so many reassuring discussions I had with him and I always knew I was very fortunate to be under his guidance. I sincerely hope that our path will cross in a near future. I would also like to thank my co-advisor Dr. Rafaël Jiménez-Florès. Dr. Jimenez through his contagious enthusiasm always helped me to diversify my views. He generously welcomed me in his laboratory and home for times during the course of my thesis. He offered me an expérience abroad that forever will be carved in my memories. Dr. Jiménez-Flores has been an extraordinary advisor and the friendship we developed will remain for years to corne.

5 Many thanks are also directed to Dr Michel Britten for accepting to do pre-reading of this thesis. Advices and help of Dr. Britten were also greatly appreciated throughout my thesis. Other members of my thesis comity, Dr Paul Paquin and Dr Jacques Rolland are also acknowledged. Groundwork could not hâve been realized without the help of those extraordinary people at STELA Research group (Anne-Françoise Allain, Edith Tousigant, Alain Gaudreault, Pascal Cliché, Mélanie Martineau, Jocelyne Giasson and Bernard Béliveau), Institute for Nutraceutical and Functional Foods (INAF) (Karine Coenen, Dr. Dominique Michaud) and Dairy Products Technology Center (Salvador Uson, Jerry Mattas, Isabelle Sodini). Ail thèse fine research facilities and wonderful people made for fantastic learning and working venues. It's hard to accomplish anything in life without the care and support only your loved ones can give you. I am very fortunate to hâve parents that gave me the best éducation possible and made me become who I am today and I hope I made them proud. They gave me the gift of éducation which is the greatest gift of ail and for that I want to express them ail my gratitude. Their unconditional love and support gave me the best conditions possible to be successful in what I do. I cannot express how important the support of my loving wife, Marie-Hélène, was for me throughout thèse years. Her incredible comprehensiveness and understanding gave me the strength and confidence that I was doing the right thing by taking this path of career. The tough times during thèse 4 years didn't seem so tough with her by my side. If it wasn't for Marie-Hélène, I honestly don't think I'il be where I am today.

6 VI I also must acknowledge the incredible support I had from friends and colleague at STELA, INAF and DPTC. I sincerely had the time of my life at those three superior institutions where I met people that were not only colleague but also great friends.

7 Vtl A Louise et Guy, Parce que vous m'avez donné le plus beau cadeau de tous...

8 vin TABLE OF CONTENTS RÉSUMÉ ii ABSTRACT iii FOREWORD iv EQUATION LIST xii TABLE LIST xiii FIGURE LIST xiv ABREVIATIONS xvii CHAPTER1 Introduction 1 CHAPTER 2 Literature review The milk fat globule membrane Origin MF GM composition MFGM structure MFGM isolation Properties of MFGM components Functional properties of MFGM components Biological activities of MFGM components Cream, butter making and buttermilk Cream for butter manufacture : The churning process Processing buttermilk into powder Evaporation Spray-drying Buttermilk composition Protein content Casein micelles Whey proteins MFGM proteins Lipid content Glycerophospholipids Sphingolipids and gangliosides 30

9 IX Other components, Lactose Minerais and salts Membrane séparations Principles of membrane séparations Driving forces on pressure-driven processes Filtration ranges (classification) Filtration membranes, equipment & process parameters Membrane materials & configuration of modules Modes of opération Performance évaluation of membrane process Fouling in membrane filtration Stratégies to minimize fouling Applications of membrane filtration in dairy processing Fractionation of buttermilk using membrane processes Ultrafïltration of buttermilk Microfiltration of buttermilk 52 CHAPTER 3 Hypothesis, goal and objectives 54 CHAPTER 4 Effect of température and pore size on the fractionation of fresh and reconstituted buttermilk by microfiltration Résumé Abstract Introduction Materials and methods Microfiltration Procédure Cleaning Procédure Composition Analysis Results and discussion Température Effect Pore Size Effect Effect of Buttermilk Source Conclusion Acknowledgements 71

10 CHAPTER 5 Effect of processing on the composition and structure of buttermilk and its milk fat globule membranes Résumé Abstract Introduction Materials and methods Processing conditions Isolation ofmfgms Analyticalprocédures Compositional analyses Transmission électron microscopy (TEM) Statistical analysis Results and discussion Effect ofpasteurizing cream on the composition of buttermilk Effect of processing steps on the composition of buttermilk Effect of cream pasteur ization on the composition of MF GM isolâtes Effect of buttermilk processing steps on the composition of MF GM isolâtes Effect of processing on the microstructure of buttermilk Conclusion Acknowledgements 93 CHAPTER 6 A Comparative study of the fractionation of regular buttermilk and whey buttermilk by microfiltration Résumé Abstract Introduction Materials and methods Buttermilk production Microfiltration Chemical analysis Calculations Statistical analysis 101

11 XI 6.5 Results and discussion Analytical data Microfûtration permeation flux data Transmission of components during MF Concentration factors Conclusions Acknowledgements 114 CHAPTER 7 Effect of washing the cream on buttermilk fractionation using ceramic microfiltration membranes Résumé Abstract Introduction Materials and methods Préparation of buttermilks Washed cream buttermilk Normal cream buttermilk Microfiltration of buttermilks Microfiltration equipment Microfiltration experiments Analytical procédures Statistical analysis Results and discussion Effect of cream washing on cream and buttermilk composition Microfiltration flux and membrane fouling Composition of MF fractions Conclusions Acknowledgments 135 CHAPTER 8 General Conclusions Achievements and original contributions Signifïcance of the results, Perspectives 140 CHAPTER 9 Références 142

12 XII EQUATION LIST Equation 2.1 Volumetric concentration factor calculation 39 Equation 2.2 Diafiltration factor (in diavolumes) calculation 39 Equation 2.3 Rejection coefficient calculation 41 Equation 2.4 Transmission coefficient calculation 42 Equation 2.5 Calculation of flux 42 Equation 2.6 Practical calculation of J. 43 Equation 2.7 Calculation of J accordingto the "résistance in séries" model 43 Equation 4.1 Transmission calculation 60 Equation 6.1 Transmission calculation 100 Equation 6.2 Concentration factor calculation 101 Equation 6.3 Permeation flux calculation in the mass transfer mathematical model Equation 6.4 Mass transfer coefficient calculation 106 Equation 7.1 Permeation flux and résistances calculation according to the résistance in séries model 122 Equation 7.2 Fouling coeficient (f c ) calculation 122

13 Xlll TABLE LIST Table 2.1 Composition of the MFGM 5 Table 2.2 Information obtained by several authors on the MFGM structure 7 Table 2.3 Effect of cream pasteurization 12 Table 2.4 Chemical composition and physicochemical properties of bovine buttermilk and skim milk 19 Table 2.5 Lipid composition of whole raw milk, buttermilk and whey 27 Table 2.6 Minerais and salts composition and distribution in milk 33 Table 4.1 Composition of fresh buttermilk, permeates and retentates from MF 0.8 um at différent températures. Means are given in % of dry matter (DM) 65 Table 4.2 Compositions of reconstituted buttermilk, permeates and retentates from MF at différent pore sizes. Values are given in % of dry matter (DM) 68 Table 5.1 Composition of buttermilk after each processing step 81 Table 5.2 Phospholipid composition of buttermilk after each processing step 82 Table 5.3 Composition of MFGM isolâtes from buttermilk 86 Table 5.4 Protein composition of MFGM isolâtes from buttermilk 86 Table 5.5 Phospholipid composition of MFGM isolâtes from buttermilk 87 Table 6.1 Composition of initial products 103 Table 6.2 Average transmission of components through the membrane 111 Table 6.3 Average concentration factors in the final retentate powders 113 Table 7.1 Average composition of pasteurized creams and buttermilks from normal and washed cream process 125 Table 7.2 Résistances calculation after MF 4X of normal and washed cream buttermilk 129 Table 7.3 Average composition of permeates and retentates (% dry basis) 132 Table 7.4 Rejection % during MF and DF of washed and normal cream buttermilk 133

14 XIV FIGURE LIST Figure 2.1. Electron micrograph of fat globule extrusion from goat mammary gland 4 Figure 2.2 Schematic représentation of the MFGM 8 Figure 2.3 Typical température - time maturation curves in butter making 13 Figure 2.4 Stages in formation of butter 15 Figure 2.5 Continuous butter churn 16 Figure 2.6 Schematic représentation of bovine MFGM 23 Figure 2.7 Phosphatidylethanolamine structure 28 Figure 2.8 Phosphatidylcholine structure 29 Figure 2.9 Phosphatidylserine structure 29 Figure 2.10 Phosphatidylinositol structure 30 Figure 2.11 Sphingomyelin structure 31 Figure 2.12 Salts equilibrium in milk 34 Figure 2.13 Schematic représentation of phenomena occurring during membrane filtration 35 Figure 2.14 Rejection of milk components depending on membrane filtration class 36 Figure 2.15 Configurations of membrane filtration modules 38 Figure 2.16 Modes of opération of filtration processes 41 Figure 2.17 Dead-end filtration and tangential flow filtration 45 Figure 2.18 Conventional tangential filtration versus UTP - co-current permeate flow tangential filtration 47 Figure 4.1 Binary gradient used in phospholipid analysis by HPLC-ELSD 62 Figure 4.2 Phospholipid profile of fresh (A) and reconstituted (B) buttermilk obtained by HPLC-ELSD analysis. (NL = neutral lipids, PPL = phospholipids, PE = phosphatidylethanolamine, PI = Phosphatidylinositol, PC = phosphatidylcholine, SM = sphingomyelin) 63 Figure 4.3 Permeation fluxes during MF (0.8 um) of fresh buttermilk at 50 C (A), 25 C ( ) and 7 C ( ) 64 Figure 4.4 Transmission of proteins ( ), lipids ( ) and phospholipids ( A ) during MF (0.8 jim) of fresh buttermilk at différent températures 66

15 XV Figure 4.5 Permeation fluxes during MF of reconstituted buttermilk (50 C) using différent pore sizes: ( ) 0.1 um at 0.7 bar TMP, ( ) 0.8 um at 0.7 bar TMP and (*) 1.4 ^mat 0.5 bar TMP 67 Figure 4.6 Transmission of proteins ( ), lipids ( ) and phospholipids ( A ) during MF of reconstituted buttermilk at différent pore sizes 69 Figure 5.1 Diagram of buttermilk processing 78 Figure 5.2 Schematic représentation of MFGM 84 Figure 5.3 SDS-PAGE of MFGM isolâtes from raw cream buttermilk after séparation (1), pasteurization (2), evaporation (3) and spray-drying (4) and pasteurized cream buttermilk after séparation (5), pasteurization (6), evaporation (7) and spray-drying (8). MW: molecular weight standard 88 Figure 5.4 Transmission électron micrographs of initial buttermilk (A, E), pasteurized buttermilk (B, F), evaporated buttermilk (C, G) and spray-dried buttermilk (D, H). Bars = 0.2 um 91 Figure 6.1 Average permeation flux curve during (a) VCMF and (b) DFMF of regular buttermilk ( ) and whey buttermilk ( ). J A VG = average flux ± standard déviation. 104 Figure 6.2 Transmission of lipids ( ), proteins (A) and ash ( ) through the VCMF (a) and DFMF (b) process as function of the VCF or DF reached for regular buttermilk ( ) and whey buttermilk (---). * Indicates significant (p < 0.05) transmission decrease throughout the process 107 Figure 6.3 SDS-PAGE (12%) profile of whey buttermilk (1), Whey buttermilk VCMF permeate (2), Whey buttermilk VCMF retentate (3), Whey buttermilk DFMF permeate (4), Whey buttermilk DFMF retentate (5), Regular buttermilk (6), Regular buttermilk VCMF permeate (7), Regular buttermilk VCMF retentate (8), Regular buttermilk DFMF permeate (9), Regular buttermilk DFMF retentate (10) 112 Figure 7.1 SDS-PAGE (13.5%) of normal cream samples (A) and washed cream samples (B). Lanes 1 are pasteurized creams, lanes 2 buttermilks, lanes 3 4X MF retentates, lanes 4 4X MF permeates, lanes 5 6X DF retentates and lanes 6 6X DF permeates. MW - molecular weight standards, HMW - high molecular weight material, MFGM

16 XVI - milk fat globule proteins, CN - caseins, WP - whey proteins, FABP - fatty acid binding protein (MFGM) 126 Figure 7.2 Permeate flux during MF ( ) and DF ( ) of washed cream and normal cream buttermilk. Bars indicate standard déviation. Dénotes highly significant différence (p<0.01) and * dénotes significant différence ( p< 0.05) 128

17 ABREVIATIONS a-la P-Lg ADPH AMF BTN CA CD36 CLD CF CP ELSD HPLC J MF MFGM MUC1 MWCO NF PA PAS III PAS 6/7 PC PE PES Alpha-lactalbumin Beta-lactoglobulin Adipophilin Anhydrous milk fat Butyrophillin Cellulose acétate Cluster of differentiation 36 Cytoplasmic lipid droplets Concentration factor Concentration polarization Electroevaporative light scattering detector High pressure liquid chromatography Permeation flux Microfiltration Milk fat globule membrane Mucin 1 Molecular weight cut-off Nanofiltration Polyamide Periodic acid Schiff III Periodic acid Schiff 6/7 Phosphatidyl choline Phosphatidylethanolamine Polyethersulfone

18 XV111 PI PPL PS PSU PVDF RO SM TMP UF UTP WPC WPI XDH/XO Phosphatidylinositol Phospholipids Phosphatidylserine Polysulfone Polyvinyldiene fluoride Reverse osmosis Sphingomyelin Transmembrane pressure Ultrafiltration Uniform transmembrane pressure Whey protein eoncentrate Whey protein isolate Xathine dehydrogenase/oxidase

19 CHAPTER1 Introduction In the context of optimal utilization of milk solids in the dairy industry, many byproducts hâve found novel applications either as ingrédients in foods Systems, nutraceuticals or natural health products. Whey is the perfect example of a by-product becoming an important ingrédient in the modem food industry. The multiple applications of whey as a food ingrédient has revolutionized the économies of cheese manufacture. In the same industry however, the by-product from butter manufacture, buttermilk, has still not found nearly as many applications. Still today, buttermilk solids are often used as low value milk solids in formulations. Most of time, buttermilk is spray-dried and stocked for months and eventually, because of its poor stability to storage, becomes oxidized and unusable for food application and therefore buttermilk is used in animal nutrition. In the United-States, buttermilk powder production is estimated to be 30.1 million Kg (USDA, 2005) while production in Canada has been fluctuating between 3.1 and 4.5 million Kg for the last 5 years (Statistics Canada, 2005). As for the cheese industry, butter manufacture économies could be revolutionized if its by product could be exploited to its full potential. Buttermilk composition is often reported to be similar to that of skim milk, which is not entirely accurate. Buttermilk's composition is fairly similar to skim milk but fine analysis of buttermilks components reveals a unique concentration in material associated with the milk fat globule membrane (MFGM) that maybe only matched in butter sérum obtained in anhydrous milk fat production (AMF). MFGM components include mainly membrane proteins and polar lipids (phospholipids and sphingolipids). Data coming from médical and pharmaceutical studies tend to show that thèse components could possess bioactive properties ranging from anti-viral to anti-cancer effects. Furthermore, the various components of buttermilk could be used in many functional foods applications. Polar lipids of buttermilk can be used as "natural" emulsifiers and buttermilk protein content can be used to enrich cheeses and improve yields.

20 Fractionation of buttermilk components could be the key to fully exploit their potential. However, fractionation of buttermilk at industrial scale can only be done by a limited number of technologies. Microfiltration (MF) is one of the most promising fractionation methods. MF membranes are already used in dairy industries Worldwide. Altough the interest for MF fractionation of buttermilk is growing, successful industrial applications hâve been scarce. Among the reasons for the few promising results available is the lack of knowledge of buttermilk microstructure. The rather low permeation flux during MF of buttermilk is constantly limiting the potential of this technology to émerge industrially. Poor selectivity represents a barrier to fractionation of buttermilk by MF, mainly because of interactions occurring between caseins, whey proteins and MFGM fragments. Although interactions between skim milk proteins are being better understood today, interactions involving the MFGM are still subject of spéculations and little information is available today. The goal of this work was to develop a membrane-based fractionation approach based on the knowledge of buttermilk properties and the effect of processing variable to obtain tailored ingrédient for spécifie applications. The l st chapter of this thesis provides information on the basis of the présent research. Chapter 2 présents the state of the knowledge on buttermilk origin, composition and fractionation by membrane technology. Chapter 3 covers the hypothesis, goal and objectives of this thesis. Chapter 4 présents the first attempt at fractionation of buttermilk by MF. Différent membrane pore sizes, températures of filtration and the use of fresh and reconstituted buttermilk are compared. In Chapter 5, the effect of industrial processes on the composition and micro structure of buttermilk and MFGM isolated from thèses buttermilks is presented. Chapter 6 contains the work carried out on a research year spent at Dairy products Technology Center of California Polytechnic State University in San Luis Obispo, California. The work carried out was on the MF of buttermilk derived from a cream deprived of casein micelles (whey cream) as compared to buttermilk derived

21 from regular cream. In Chapter 7, MF of buttermilk from a process of washing the cream prior to churning is comparée to fresh buttermilk fractionation. Chapter 8 consists of a gênerai conclusion on the work présentée in this thesis. Ail références used in this thesis hâve been listed in Chapter 9.

22 CHAPTER 2 Literature review 2.1 The milk fat globule membrane Origin Milk fat (triglycérides) is synthesized in the endoplasmic reticulum of the mammary gland. Triglycérides are then accumulated in cytoplasmic lipid droplets (CLD) (Danthine, Blecker, Paquot, Innocente, & Deroanne, 2000). CLD's are then transported to the walls of the seereting eell where they are enveloped by the eell membrane and exereted outside the eell by exocytosis (Keenan, 2001). Figure 2.1 shows a milk fat globule exereted from a goat mammary gland. Figure 2.1. Electron micrograph of fat globule extrusion from goat mammary gland from Wooding (1971) This model of origin and sécrétion of the milk fat globule is still the most aeeepted today (Danthine, Blecker, Paquot, Innocente, & Deroanne, 2000). MFGM is therefore composed of an extern double layer of phospholipid and protein that originates from the mammary gland seereting cells and an interior monolayer of phospholipid and protein derivred from he CLD's

23 2.1.2 MFGM composition Proteins and lipids account for more than 90 % of the dry matter of the MFGM (Danthine, Blecker, Paquot, Innocente, & Deroanne, 2000). Discrepancies can be observed in the exact proportion of proteins and lipids reported in many studies but thèses are most likely due to variations in MFGM isolation procédures used by workers. However, it has to be noted that factors like breed and âge of the animal, feeding, stage of lactation and seasonal variation can induce variability in the MFGM composition (Danthine, Blecker, Paquot, Innocente, & Deroanne, 2000). Data on composition of MFGM is shown in Table 2.1 Table 2.1 Composition of the MFGM mg/100 g of fat globules %inithe MFGM Component Walstra et (2006) al Goff&Hill (1993) Walstra et (2006) al Goff&Hill (1993) Proteins Lipids Phospholipids Cholestérol Water Carotenoids + Vitamin A Fe Cu A high proportion of neutral lipids in the MFGM has béer, reported in many studies (Kanno & Kim, 1990; Keenan, Moon, & Dylewski, 1983; Malin, Foglia, Basch, Thompson, & Vail, 1993). This high proportion of neutral lipids is believed to be the

24 resuit of contamination of the membrane by triglycérides during the churning process (Danthine, Blecker, Paquot, Innocente, & Deroanne, 2000) MFGM structure The MFGM can be viewed as a triple layer membrane with a total thiekness of about 10 to 20 nm (Walstra, Wouters, & Geurts, 2006). There is évidence that MFGM components are distributed asymmetrically in the membrane (Evers, 2004). Considering that the MFGM acts as a natural emulsifier, it is conceivable that the most hydrophobic MFGM material is loeated closer to the lipid core and the highly glyeosylated membrane proteins are loeated at the sérum exposed face of the membrane. Electronic mieroscopy, immunomicroscopy and freeze-etehing electronie mieroscopy has allowed to elucidate the origin of the milk fat globule and its structure (Danthine, Blecker, Paquot, Innocente, & Deroanne, 2000). Information gathered on the structure of the MFGM are summarized in Table 2.2

25 Table 2.2 Information obtained by several authors on the MFGM structure. Observation Technique Références Electron dense layer on the intern face ofthe membrane XDH/XO and BTN are major components ofthe électron dense layer Surface offat globules is irregular and shows "crescents" Asymétrie disposition of phospholipids: PC and SM on the intern face and PE, PI and PS on the extern face Electronic microscopy Immuno-microscopy Détergent release of membrane bound components Freeze-etching electronic microscopy Spécifie phospholipases Wooding(1971) Frankeetal(1981) Freudenstein et al (1979) Buchheim(1986) Deeth(1997)cited by Danthine et al (2000) The data presented in Table 2.2 allowed Danthine et al. (2000) to propose a schematic représentation of the milk fat globule. An adaptation of the model proposed by Danthine et al (2000) has been published by Michalski, Michel, Sainmont, & Briard (2002) and is shown in Figure 2.2. It has to be noted however that this model and ail others that hâve been proposed is highly uncertain as MFGM is still poorly understood (Walstra, Wouters, & Geurts, 2006).

26 Triglycertdes membfsne Doubl«layer (éertved from Milk fat globule membran* g point triglycérides, Low HLB prsospholipids Xrit ( << uylil. ' 5'Nuclen i-,> High HLB phosphofipids Figure 2.2 Schematic représentation of the MFGM from Michalski, Michel, Sainmont, & Briard (2002) MFGM isolation A number of methods hâve been developed at laboratory scale for isolation of the MFGM from milk but no industrial or pilot scale process has been reported. Most laboratory methods comprise of 4 steps: 1) cream séparation from milk, 2) cream washing, 3) washed cream destabilization, and finally 4) MFGM fragments récupération. The first step to prépare native MFGM is to separate the cream from milk. This can easily be done by centrifugation of unhomogenized cream. The cream obtained still contains skim milk proteins, lactose and minerais so it needs to be further washed with a saline buffer or skim milk ultrafiltrate. The use of 0.15 mol.l" 1 NaCl buffer for cream washing instead of pure water, allows for better élimination of skim milk material without destabilizing the fat globules (Basch, Greenberg, & Farrell, 1985). Three repeated washing of the cream are sufficient to remove skim milk material and further washing appears to cause losses of MFGM material (Anderson & Brooker, 1975). Destabilizing of the washed fat globules can be carried out either by mechanical churning or freeze-

27 thawing procédure (Danthine, Blecker, Paquot, Innocente, & Deroanne, 2000). The last step is the récupération of the MFGM fragments found in the aqueous phase (buttermilk) after churning of fat globules. The most utilized and efficient method to recover MFGM fragment is ultracentrifugation at x g for 60 to 90 minutes. This method allows for complète récupération of MFGM fragments as a dense pellet (Patton & Keenan, 1975). Some modifications to the steps of MFGM isolation hâve also been proposed. Patton & Huston (1986) reported a one-step séparation of purified milk fat globules without the need of répétitive washing. The methods consists in delicately placing un-cooled raw milk supplemented with 5 g. 100 ml-1 of sucrose under a saline buffer into a tube and then centrifuge the tube to recover a cream layer free of skim milk material. The récupération step can also be modified using ammonium sulfate to precipitate the membrane and recover it by low speed centrifugation ( x gl 30 min) (Kanno & Kim, 1990). However, this technique requires ammonium sulfate removal by dialysis. Ail of thèse techniques are suitable for whole milk but are problematic in the case of buttermilk which already contains MFGM fragments along with skim milk material. Ultracentrifugation of buttermilk necessarily yields an isolate with important amount of casein micelles (Corredig & Dalgleish, 1997a). An approach has been developed by Corredig and Dalgleish (1997a) to circumvent this issue. The authors proposed the use of sodium citrate to disrupt casein micelles prior to ultracentrifugation of buttermilk. Citrate is a calcium chelating agent that can remove calcium from the micelles creating a breakdown of its structure allowing for caseins to remain in the supernatant after ultracentrifugation. Isolâtes obtained using this method, hâve a composition close to those obtained from raw milk. The différences observed are likely caused by the fact that industrial buttermilk has been used as a starting material, so processing treatments (evaporative concentration, spray-drying) could hâve caused many changes in the MFGM.

28 Properties of MFGM components Functional properties of MFGM components Functional properties of pure MFGM isolated from untreated raw milk hâve neyer been described. Very little work has been made on the characterization of the functional properties of MFGM derived from other material including buttermilk. Reasons for this low scientifïc interest résides most likely in the difficulties of obtaining large amount of MFGM material and that MFGM accounts for a really small part of dairy products solids. Because of their relatively high hydrophobicity, and because of their important rôle in milk fat globule stability, MFGM proteins hâve been hypothesized to be natural emulsifïers. Kanno (1989) has shown that buttermilk supplemented with MFGM proteins greatly enhanced its emulsification properties. However, because of the intensive heat treatments applied to commercially available cream and buttermilk, emulsifying characteristics of MFGM isolâtes were found to be relatively low (Corredig & Dalgleish, 1997a; Corredig & Dalgleish, 1998a). On the other hand, isolâtes of MFGM produced by MF of commercial buttermilk appear to show a better emulsion capacity than whole buttermilk protein concentrate (Roesch, Rincon, & Corredig, 2004). No data is available on other functional properties like solubility, gelation properties and foaming properties Biological activities of MFGM components Data coming from the médical fïeld seems to demonstrate that buttermilk components might possess important biological properties. MFGM material appears to be an important source of nutraceuticals components including polar lipids and membrane proteins (Spitsberg, 2005).

29 Il The physiological effects of sphingomyelin (SM), phosphatidylcholine (PC) and phosphatidylserine (PS) in human health hâve been the most extensively studied. SM has been found to exert anticarcinogenic activity in animal models. Dillehay, Webb, Schmelz, & Merril Jr. (1994) hâve used dietary SM in the diet of rats where colon cancer was induced by 1,2-Dimethylhydrazine. In their study, the authors found that the number of tumors found in the animais fed with SM was 70% lower than rats fed with the control diet. In long term feeding of the animais, the number of adenocarcinomas found was lowered by 27% when SM was added to the diet. However, there has been no study yet on the effect of SM consumption on human subjects. PC via its dégradation products, mainly choline, has been found to possess many biological activities. Positive effects on hepatic function, cognitive function, cardiovascular diseases, reproduction function, physical performance and memory hâve been found for PC (Kidd, 2002). For example, Niederau, Strohmeyer, Heintges, Peter, & Gopfert (1998), hâve observed in their doubleblind clinical study that patients with hepatitis C to which were given a PC supplément, responded better to interferon-a treatments than patient receiving a placebo. Phosphatidylserine has been linked with cognitive function enhancement for the elderly (Crook et al., 1991) and as a potential stress reducer (Ruenberg, 2002). MFGM concentrate as a whole as also been found to possess biological activities. An MFGM concentrate obtained by a combined process of MF and supercritical fluid extraction has shown the ability to inhibit the attachment of rotaviruses to cells in vitro (Ochonicky, Donovan, Kuhlenschmidt, Jiménez-Flores, & Kuhlenschmidt, 2005)

30 Cream, butter making and buttermilk Cream for butter manufacture Fresh raw milk has an average fat content around 4 % (Walstra, Wouters, & Geurts, 2006). The fat in raw milk can be concentrated via centrifugal séparation of the cream. Depending on various séparation parameters including centrifugal speed and température, lipid content of the cream for buttermaking can range between 30 to 60% and even up to 75% lipids in the anhydrous milk fat manufacture (Bylund, 1995). In traditional buttermaking process, cream is standardized at 35 to 40% milk fat while in continuous butter process cream can reach 40 to 45 % milk fat (Mahaut, Jeantet, Brûlé, & Schuck, 2000). The cream is then pasteurized at températures ranging from 90 to 110 C for 15 to 20 seconds. Pasteurization of the cream is necessary to destroy pathogenic microorganisms and thermorésistant enzymes (lipases, oxydoreductases). Pasteurization of the cream has many effects on its components (Table 2.3) Table 2.3 Effect of cream pasteurization Effect of cream pasteurization dumping of fat globules Rupture of the MFGM (excessive heat treatments) Unfolding of P-Lg and exposure of active thiol groups Formation of sulfuric odors (H2S) Destruction of Vitamin E & K (20-30%) Copper (from plasma) fixation on the MFGM from Mahaut, Jeantet, Brûlé, & Schuck (2000)

31 13 Sometimes cream is vacuum de-aerated prior to churning. This step helps to minimize subséquent fouling of heating equipment by removing dissolved gazes (Mahaut, Jeantet, Brûlé, & Schuck, 2000). This procédure also allows for the removal of some unwanted aromatic compounds responsible for sulfuric and cooked flavors. Physical maturation of the cream after pasteurization is used to allow proper crystallization of the milk fat (Frede & Buchheim, 1994). This procédure is done by cooling the cream at a spécifie température for a spécifie amount of time. During physical maturation, the objective is to obtain an adéquate ratio of solid and liquid milk fat in order to obtain butter that is not too soft but not to hard. Furthermore, physical maturation allows for maximal recovery of the milk fat in the butter instead of in the buttermilk (Pointurier & Adda, 1969). Typical temperature-time curves for butter maturation are shown in Figure 2.3. Winter and summer curves are différent to circumvent the seasonal variations in milkfat composition and melting properties. 25, Summer Winter -A Time (h) Figure 2.3 Typical température - time maturation curves in butter making adaptedfrom Mahaut, Jeantet, Brûlé, & Schuck (2000) In the case of cultured butter, a biological maturation step is added to the process. This biological maturation is done by adding a lactic starter to the cream. The lactic starter

32 14 induces spécifie aromas and flavors and allows for better conservation of the product (ph < 4.7) (Mahaut, Jeantet, Brûlé, & Schuck, 2000). Biological maturing of the cream is usually done at the same time as physical maturation. It has to be noted that other sources of cream can be used for buttermaking. For example, cream derived from whey (whey cream) can be used to produce whey butter. Residual lipids removal from whey has been the subject of many studies (Baumy, Gestin, Fauquant, Boyaval, & Maubois, 1990; Fauquant, Pierre, & Brûlé, 1985; Fauquant, Vieco, Brûlé, & Maubois, 1985; Théodet & Gandemer, 1994). Whey is often concentrated by UF membranes to produce whey protein concentrate and residual lipids restrict the degree of protein purity of the concentrate as well as causing prématuré fouling of UF membranes. Very few studies hâve investigated the properties of whey butter. It has been shown that whey butter sensory characteristics are not significantly différent from normal and cultured butter apart from the fact that whey butter (Jinjarak, Olabi, Jimenez-Flores, & Walker, 2006). However, the texture of whey butter was found to be softer than that of normal and cultured butter possibly because of the higher unsaturated lipid content of whey cream (Jinjarak, Olabi, Jimenez-Flores, & Walker, 2006) The churning process Destabilization of the cream is obtained by mechanical churning. During the churning process, a séries of phenomena occurs. As air is incorporated in the cream, proteins (Plactoglobulin, a-lactalbumin and P-casein) unfold and form an interfacial layer around air bubbles (Anderson & Brooker, 1975) creating an unstable foam. Because of the lower surface tension of the MFGM as opposed to skim milk proteins, MFGM displaces skim milk proteins at the air/serum interface (Frede & Buchheim, 1994).

33 Working Milk Cream Small grains Large grains Butter Figure 2.4 Stages in formation of butter White represents milk fat and black represents the sérum (adapted from Mulder & Walstra (1974)). The mechanical stress caused by churning also induces MFGM to rupture and liberate the crystallized lipid content of the globules. Upon further churning the foam is destabilized, fat dumping and phase inversion occurs leaving a mass of butter and the resulting aqueous phase, buttermilk. Most of the MFGM material, skim milk proteins, minerais, lactose and water are recovered in the buttermilk. However, a certain proportion of the water and MFGM remains trapped in the butter in the form of small droplets. As butter is mechanically worked, the water droplets are dispersed evenly in the butter mass (Figure 2.4). Ail the phenomena discussed above apply to batch churning of cream (Frede & Buchheim, 1994) but nowadays butter is mainly manufactured by the Fritz or continuous process (Figure 2.5). In the Fritz process, cream (40-50% milk fat) is fed into a churning cylinder fitted with beaters driven by a variable-speed motor (1). Phase inversion takes place very rapidly, with the butter grains and buttermilk passing on to the séparation section (2). Butter working begins in this section while the butter is being conveyed by the screw conveyor to the squeeze-drying section (3). In the squeeze-drying section, the butter passes through a conical channel and perforated plate to remove any remaining buttermilk. The butter grains continue to the second working section (4), which opérâtes

34 16 at the speed of the screw in the first working section to achieve optimum working of the butter. Upon reaching the injection section, a high-pressure injector may add sait (5). Cream Figure 2.5 Continuous butter churn From Bylund (1995) Typically, churning of 1 Kg of 40% milk fat cream will yield 0.5 Kg of both butter and liquid buttermilk. Given the large quantities of butter that is manufactured Worldwide, large volumes of buttermilk has to be handle industrially. Liquid buttermilk is prone to oxidation so this by product is concentrated by evaporation and spray-dried Processing buttermilk into powder Buttermilk represents a voluminous by product for the dairy industry, and because it contains low total solids and is very sensitive to dégradation and oxidation, it is usually processed into a powder. Generally buttermilk concentration and drying is handled about the same way as skim milk. Buttermilk is usually concentrated to 40-50% total solids before being fed to a spay-drier. However, because of its low value, buttermilk can be stored for a long time before spray-drying and therefore requires an additional pasteurization step.

35 Evaporation The main reason for evaporating buttermilk before drying is because evaporation requires far less energy that drying (Walstra, Wouters, & Geurts, 2006) which results in a substantial economy as compared to direct spray-drying of buttermilk. Nowadays, most evaporators found in the dairy industry are falling flow multi-effect evaporators (Mahaut, Jeantet, Brûlé, & Schuck, 2000). Thèses evaporators are used under vacuum (30 kpa) to allow boiling of the fluid at températures below 100 C. In the first effect of the evaporator, an external source of heat is applied and the subséquent effeets use the vapor generated by the evaporation of water from the preceding effect. Thèse evaporators are designed to remove as much water possible with the lowest possible energy cost and that, along with minimal heat damage to the product (Ye, Singh, Taylor, & Anema, 2004). The evaporation step can induce several changes in milk components. Because of the combined effect of the heat and the concentration of proteins and lactose, Maillard reactions can occur (Walstra, Wouters, & Geurts, 2006). The heat treatment and the concentrative effect can also induce interactions between [3-Lactoglobulin (P-Lg) and MFGM and induce interaction of caseins with the fat globule (Ye, Singh, Taylor, & Anema, 2004). However, no data is available on the effect of evaporation on buttermilk. Reverse osmosis can also be used to remove water from buttermilk. Despite a lower demand of energy (-30 kj/kg of evaporated water for reverse osmosis versus kj for vacumm evaporator) reverse osmosis is not able to concentrate dairy products to high total solids (~45-50%) because of the dramatic increase of viscosity as solids get concentrât éd Spray-drying Once buttermilk has been concentrated to % total solids, it is usually spay-dried. Other types of drier are still used in the dairy industry like roller driers but nowadays most of the dairy powders are spray-dried. Spray driers consists of a conic chamber

36 18 where air is circulated with an inlet and outlet température of about 200 C and 100 C respectively. Inside the chamber, the product is sprayed into a very fine mist. As the fine particles of products are in contact with the air, they quickly dry and fall into cyclone where air and powder particles are separated. This treatment produces powder particles that are generally round and smooth as opposed to highly irregular and compact for roller dried powder (Carie, 1994). Powders produced by spray drying are often categorized as low, médium or high heat. This classification is based on the degree of denaturation of the whey proteins observed in the powder (Walstra, Wouters, & Geurts, 2006). It is not known if spray-drying of buttermilk has an effect on its composition or structure. However, Astaire, Ward, German, & Jiménez-Flores (2003) hâve reported that the behavior of buttermilk reconstituted from a spray-dried powder during MF was significantly différent from that of fresh buttermilk. 2.3 Buttermilk composition Buttermilk and skim milk share a similar content of total proteins, lactose and minerais (Table 2.4). Apart from the higher fat content observed in buttermilk, the gênerai composition of both products appears quite similar. However, the composition of buttermilk is characterized by the présence of fat globules membrane fragments. Thèse MFGM particles are composed mainly of phospholipids, membrane proteins and cholestérol. Ail thèse components account for an important part of buttermilk, explaining its unique characteristics.

37 19 Table 2.4 Chemical composition and physicoehemical properties of bovine buttermilk and skim milk Component Buttermilk Skim milk Total solids Total proteins Total lipids Lactose Minerais Density (kg.m 3 ) Viscosity (20 C, mpa.s) Acidity ( D) From Ramachandra Rao, Lewis, & Grandison (1995) and Walstra, Wouters, & Geurts (2006) Protein content The protein content of buttermilk is relatively similar to that of skim milk except that buttermilk contains important proportion of proteins derived from the MFGM. Skim milk proteins (caseins and whey proteins) make up for around 90% of ail proteins (Surel, 1993; Turcot, Turgeon, & St-Gelais, 2001) and the remaining is composed of minor proteins including MFGM proteins Casein micelles Caseins are the major proteins found in buttermilk, making up to ~ 75% of total protein content (Walstra, Wouters, & Geurts, 2006). Caseins can be separated in 4 major proteins: a S, a S 2, P and K-casein. Dégradation of P-casein by the plasmin leads to the

38 20 formation of a fïfth class of casein, y-casein. The ratio of ct s i, a S 2, P and K-casein in milk is roughly 11:3:10:4 (Walstra, Wouters, & Geurts, 2006). As well as in milk, caseins are found mostly in micellar form in buttermilk. There are still spéculations about the structure of thèse micelles but several information are available today. The surface of the casein micelle is covered by a "hairy layer" which consists of the C-terminal chain of K- casein. This layer provides the micelle with an important négative charge as well as a highly hydrophilic surface. The core of the micelle consists mainly of a s and p-casein. Nanoclusters of calcium phosphate are also found inside the micelle and help maintaining the micelle structure. The forces that keep the micelle components linked together are mainly hydrophobic and ionic bonds. Although casein micelles are remarkably stable in milk, dissociation of the micelles can occur depending on various physicochemical conditions. Cooling is known to induce dissociation of P-casein from the micelle (Rosé, 1968). This dissociation of P-casein is partly associated with the lower yield when cheese is made from cooled milk. The decreasing strength of hydrophobic bonds at low température is causing the release of P- casein into the sérum (Famelart, Hardy, & Brûle, 1989). Acidification also leads to demineralization of the micelles inducing a reduced stability and eventually précipitation when ph reaches the pi of the caseins around 4.6 (Dalgleish & Law, 1988). Heating of milk also induces changes in the micelle. However, depending on the température of heating the effects can be réversible (Holt, 1995). At températures over 80 C, losses of colloidal calcium phosphate are irréversible as milk stone (calcium phosphate) can be formed on the surface of heat exchangers. Changes in colloidal calcium phosphate equilibrium also occurs when température is below 80 C but are réversible since no milk stone is formed.

39 Whey proteins Whey proteins are found in the sérum of milk as well as buttermilk and they are found in large proportion in whey, the by product from cheese making. P-Lg and a-lactalbumin (ot- La) are the two major proteins accounting for more than 75% of ail whey proteins. Other proteins are bovine sérum albumin, various immunoglobulins, lactoperoxydase and lactoferrin. fi-lactoglobulin P-Lg is the single most concentrated whey protein in buttermilk. This protein is member of the lipocalin super family and has affinity for hydrophobic substances. [3-lactoglobulin contains 2 disulphide bonds and one free sulphidryl group (Walstra, Wouters, & Geurts, 2006). The free sulphidryl is buried inside the tertiary structure of the protein in its native state but becomes highly reactive at ph above 7.5. In whole milk, pmactoglobulin interacts strongly with the MFGM through disulphide bond exchange after relatively mild heat treatment (Kim & Jiménez-Flores, 1995; Ye, Singh, Taylor, & Anema, 2002). a-lactalbumin a-lactalbumin (a-la) is a compact whey protein that is stabilized by 4 disulphide bonds (Hong & Creamer, 2002). a-la function in milk appears to be related to lactose synthesis (Walstra, Wouters, & Geurts, 2006). The concentration of a-la in milk is roughly half of that of P-LG. It has also been shown that this protein interacts strongly with the MFGM upon heat treatment of milk or cream (Dalgleish & Banks, 1991; Houlihan, Goddard, Kitchen, & Masters, 1992; Kim & Jiménez-Flores, 1995).

40 MFGM proteins MFGM proteins hâve not generated much interest until recently. Possible reason for this is their relatively low concentration in milk products and the lack of méthodologies for the extraction and analysis of thèses proteins. However, it is known that thèse proteins play a fondamental rôle in the stability and integrity of fat globules in whole and unhomogenized milk (Mulder & Walstra, 1974). MFGM proteins are mainly analyzed using SDS-PAGE gels and with the advances in proteomics, two-dimensional (2-D) gels of MFGM proteins from dairy products should be feasible and could help to better characterize thèses poorly known proteins. MFGM proteins consist of an extremely diverse group of proteins with molecular weight ranging from 15 to 250 kda (Patton & Keenan, 1975). The main proteins of bovine MFGM are: Mucin 1 (MUC1), Xanthine dehydrogenase/oxidase (XDH/XO), Periodic acid Schiff III (PAS III), Cluster of Differentiation (CD36), Butyrophilin (BTN), Adipophilin (ADPH), Periodic acid Schiff 6/7 (PAS 6/7). Ail thèses proteins are dispersed through the MFGM in spécifie positions depending of their physical and physiological properties (Figure 2.6).

41 ml Bllayer membrane Coal I,ipui corc MFOJM MUC1 PP3 CD36 Butyrophïlin Lactadherin MUC15 Xarrthine dehydrogenase/oxi<lss«figure 2.6 Schematic représentation of bovine MFGM front Jiménez-Flores & Morin (2005) Mucin 1 Bovine MUCl is a highly glycosylated protein situated at the sérum exposed side of the MFGM. The carbohydrate portion of MUCl is estimated to account for 50 % of its molecular weight (Mather, 2000). Fucose, galactose, mannose, N-acetylglucosamine, and sialic acid are the glycans found on bovine MUCl (Snow, Colton, & Carraway, 1977). However, the biological function of MUCl is still not clear. It is believed that because of its highly glycosylated filamentous portion situated outside the MFGM, MUCl could act as a protection against physical damage and invasive pathogenie microorganisms (Patton, Gendler, & Spicer, 1995). MUCl can be detected in SDS-PAGE gels of low concentration as a faint Coomassie blue staining band appearing at around 200 kda (Mather, 2000). In bovine milk, MUCl is mainly found in the cream fraction reaching concentration of 40 mg/l of milk (Patton, Gendler, & Spicer, 1995). In native fresh milk, MUCl is tightly bound to the MFGM. However, it as been observed that the sérum

42 24 exposed portion of the protein can be displaced from its membrane anchor by a short treatment at 80 C and recovered as a soluble protein in the supernatant (Buchheim, Welsch, Huston, & Patton, 1988). Xanthine dehydrogenase/oxidase XDH/XO can be detected at around 155 kda in Coomassie blue stained SDS-PAGE gels (Mather, 2000). XDH/XO is most abundant enzyme found in MFGM (Mather, 2000). In milk XDH/XO is présent in the form of an homodimer of about 300 kda (Veerkamp & Maatman, 1995). Each monomer comprises 1332 amino acids with a molecular weight of 146,6 kda and a pi of 7.7. The monomers contains each a molybopterin cofactor, two Fe2/S2 clusters and one FAD molécule (Massey & Harris, 1997). Xanthine oxidase oxidizes a wide variety of purines by incorporation into the substrate of oxygen derived from water. The realeased reducing équivalent are then transferred into the redox centers into the protein (Mather, 2000). The biological fonction of XDH/XO is uncertain but it is believed that it could act as an antibacterial agent providing a source of H2O2 for lactoperoxydase. Periodic acid SchifflII PAS III is a glycoprotein of 95 to 100 kda that has not been characterized extensively. In fact the only information available on PAS III in milk is that it is concentrated on the sérum exposed surface of the MFGM (Mather, 2000). Cluster of Differentiation CD 36 accounts for about 5% of the proteins found in bovine MFGM (Greenwalt, Watt, So, & Jiwani, 1990). This protein is almost fully recovered in the MFGM pellet after centrifugation of buttermilk (Mather, 2000). Biological function of CD36 may be as a

43 25 scavenger receptor binding to apoptotic cells inducing their élimination by phagocytosis (Mather, 2000). However, it is still unclear why CD36 is expressed to the sérum exposed surface of the MFGM. CD36 contains 2 cysteines residues in the N and C-terminal structure of the protein. Thèse cysteines are acylated in the human form of the protein (Tao, Wagner, & Lublin, 1996) and the acyl chains could help stabilize anchoring of the protein in the lipid bilayer (Mather, 2000). However, this hypothesis has not been investigated in bovine MFGM. Butyrophilin The term butyrophilin is derived from the Greek butyros/philos meaning affinity for butterfat which is an important characteristic of this protein. BTN is the most abundant MFGM protein in bovine milk. ranging from 20 to 43 % of ail MFGM proteins depending on the breed of the cow (Mather, 2000). BTN comprise 526 amino acids residues with a molecular weight of 59,3 kda and is member of the Ig superfamily (Gardinier, Ami guet, Linington, & Matthieu, 1992). In milk, BTN is firmly bound to the MFGM and is résistant to extraction with chaotropic agents and détergents (Mather, Weber, & Keenan, 1977). BTN has been found to interact strongly with XDH/XO (Ishii étal., 1995). Adipophilin Adipophilin has been overlooked in past years because of its relative insolubility in SDS buffers and its very similar molecular weight to another major MFGM protein, PAS 6/7 (Heid, Schnolzer, & Keenan, 1996). The fonction of this protein is still unknown. ADPH may hâve a rôle in fat globule sécrétion along with BTN and XDH/XO (Heid, Moll, Schwetlick, Rackwitz, & Keenan, 1998). ADPH contains 5 to 6 molécules of fatty acids per molécule of ADPH (Heid, Schnolzer, & Keenan, 1996) most likely bound by ester bonds. This relative affinity to fatty acids is reflected in its close positioning to the lipid core of fat globule in the membrane (Figure 2.6).

44 26 Periodic acid Schiff6/7 PAS 6/7 is a 47,5 kda protein of 427 amino acids with a pi close to 7.0 (Mather, 2000). PAS 6/7 is believed to be situated on the sérum exposed face of the MFGM because it can be displaced from washed fat globules using aqueous solution containing salts or chaotropic agents (Mather, 2000). PAS 6/7 is also found in soluble form in skim milk (Butler, Pringnitz, Martens, & Crouch, 1980). The fonction of this protein in mouse and human where it is named lactadherin is believed to be phospholipid binding (Fortunato et al., 2003). Moreover, récent data tend to show that lactadherin binds to rotaviruses and may protect the neonate from gastroenteritis (Newburg et al., 1998). This information gives insight that PAS 6/7 physiological rôle might be more related to immunity than to milk fat sécrétion Lipid content Buttermilk lipid content ranges usually between 0.3 and 0.7%. The profile of lipids found in buttermilk is. similar to that of skim milk but contains higher proportion of various polar lipids. Glycerophospholipids (phospholipids), sphingolipids and gangliosides are generally refered to as polar lipids in dairy products. The origin of thèses polar lipids is mainly the MFGM fragments recovered in buttermilk. Buttermilk also contains some small milk fat globules that hâve resisted to the churning process (Pointurier, 1986). Table 2.5 shows the lipid composition of buttermilk as compared to milk and cheese whey.

45 27 Table 2.5 Lipid composition of whole raw milk, buttermilk and whey Whole raw milk a Buttermilk" Whey b Total lipids (%) Phospholipids PE PI Phospholipid class c (% of total phospholipids) PS PC SM CER Cholestérol a from Walstra, Wouters, & Geurts (2006) b from Théodet & Gandemer (1994) c from Rombaut, Camp, & Dewettinck (2005) PE = phosphatydilethanolamine, PI = Phosphatidylinositol, PS = Phoaphatidylserine, PC = Phosphatidylcholine, SM = Sphingomyelin, CER = Ceramides (Lactosylceramide + glucosylceramide). As it can be observed in Table 2.5, buttermilk is a by-product rich in phospholipids as compared to cheese whey. However, the phospholipid class distribution is relatively constant in dairy products. In the MFGM, phospholipid class distribution is very similar to what has been found in whole dairy products (Danthine, Blecker, Paquot, Innocente, & Deroanne, 2000). This supports the idea that phospholipids are mainly found in the MFGM Glycerophospholipids Glycerophospholipids species présent in buttermilk are phosphatidylethanolamine (PE), phospahtidylcholine (PC), phosphatidylserine (PS) and phosphatidylinositol (PI). In fresh milk, phospholipids concentration can reach mg.ml" 1 while in butiermilk

46 concentration can reach mg.ml" 1. However, proportion of the main phospholipids species remains approximately the same as for whole milk or cheese whey (Table 2.6). Phosphatidylethanolamine PE is the most concentrated phospholipid specie in buttermilk and accounts for around 40 % of ail phospholipids (Christie, Noble, & Davies, 1987). As well as for other phospholipids species, PE is constituted with a glycerol basis on which two fatty acids in position 1 (sn-1) and 2 (sn-2) of the glycerol are attached by ester bounds (Figure 2.7). The main fatty acids composing PE in buttermilk are Cl8:0 and C18:l.(Hay & Morrisson, 1971) This specie of phospholipid is mainly found in the MFGM and thus, is a key building block of membrane. o OH Figure 2.7 Phosphatidylethanolamine structure Phosphatidylcholine Commonly called lecithin, PC is often the most concentrated phospholipids in animal tissue. However, in buttermilk, PC is second to PE with a concentration reaching around 30 %. PC is an intégral component of lipoproteins. PC is a non acidic lipid characterized by the présence of a choline residue in position 3 of the glycerol. Lysophosphatidylcholine (LPC) can also be found in buttermilk at very low

47 29 concentration as a resuit of hydrolysis of PC ester bound sn-1 (phospholipase Ai) or sn~ 2 (pohospholipase A2). o np Figure 2.8 Phosphatidylcholine structure Phosphatidylserine Mainly found in brain tissue of animais, PS is weakly concentrated in buttermilk. PS concentration reaches about 3 % of ail phospholipids in whole milk while in buttermilk it represents about 10 % (Rombaut, Camp, & Dewettinck, 2005). PS is a slightly acidic lipid with an L-serine residue on position 3 of the glycerol. 0 0 R2 U o I O II -0-P-O OH NH, OH Figure 2.9 Phosphatidylserine structure

48 30 Phospatidylinositol Characterized by the présence of an inositol residue, PI is a strongly acidic lipid. Its concentration is about 6 % in whole milk as well as in buttermiik. The main fatty acids bound to the glycerol moiety of PI are Cl 8:0 and Cl 8:1 (Fong, Noms, & MacGibbon, 2006). 0 O L 0 II -O-P-0 OH HO OH OH Figure 2.10 Phosphatidylinositol structure Sphingolipids and gangliosides Sphingolipids and gangliosides differ from glycerophospholipids in terms of structure because they contain a ceramide backbone instead of a glycerol. Thèses molécules are présent in small quantities in animal tissue. In buttermiik the main sphingolipid specie is sphingomyelin and various gangliosides species are also présent in minute quantities. Sphingomyelin Sphingomyelin content of buttermiîk is about one third of ail phospholipids. Biological rôle of SM includes cell growth régulation, development, adhésion, aging and aging related disease but also apoptosis (Astaire, Ward, German, & Jiménez-Flores, 2003). SM

49 31 is the analogue of glycerophospholipid PC, thus SM has a phosphorylcholine linked by ester bond in position 1 of the ceramide backbone. The fatty acids components of SM in whole milk are principally C22:0 and C23:0, C24:0 (Surel, 1993). Figure 2.11 Sphingomyelin structure Gangliosides Buttermilk contains small amounts of highly complex ceramide polyhexosides regrouped in the name of gangliosides. They hâve first been found in ganglions cells of central nervous System. Various gangliosides hâve been found in buttermilk (Huang, 1973). Buttermilk is a unique source of 9-O-acetyl-GD3. a ganglioside of important biological interest (Bonafede, Macala, Constantine-Paton, & Yu, 1989). This ganglioside could induce production of antibodies that hâve been proposed as diagnostic tool for melanoma (Bonafede, Macala, Constantine-Paton, & Yu, 1989). Reported concentrations of 9-0- acetyl-gd3 in buttermilk are ranging from 1 to 22 mg.kg" 1 of buttermilk solids.

50 Other components Lactose Among other components found in buttermilk, lactose is the most abundant carbohydrate. Lactose is a disaccharide composed of D-glucose and D-galactose. The synthesis of lactose occurs in the Golgi vesicles of the lactating cell (Walstra, Wouters, & Geurts, 2006). a-lactalbumin is known to promote lactose synthesis via modification of the action of galactosyl-transferase. Lactose content in buttermilk can reach from 37 to 44% of dry matter (Walstra, Wouters, & Geurts, 2006). Typically, lactose can be separated from milk, whey or even buttermilk to obtain crystalline lactose sold to pharmaceutical industries as a filling agent for pills Minerais and salts Minerai composition in buttermilk is similar to that of skim milk but a lower level in the concentration of calcium (O'Connell & Fox, 2000; Surel, 1993) and phosphorus (Surel, 1993) has been observed. O'Connell & Fox (2000) reported that the lower amount of calcium in buttermilk was partly responsible for the longer rennet coagulation time of buttermilk, as compared to skim milk, along with lower amounts of (3-lactoglobulin and higher content of non-micellar K-casein. However, O'Connell & Fox (2000) did not investigate the effect of phospholipids and MFGM components on rennet coagulation time. Minerais in milk, as well as in buttermilk, are distributed between the sérum phase and the casein micelles. Table 2.6 shows the average content of the major minerais and its distribution between sérum and colloidal phase.

51 33 Table 2.6 Minerais and salts composition and distribution in milk Component mg/ 100g of milk % in the sérum phase % micellar Na K Ca Mg Cl COÎ so PO Citrate From Walstra, Wouters, & Geurts (2006) Sait equilibrium in milk and buttermilk are similar. Calcium phosphate, is very sensitive to various processing parameters. Modification of the ph, température treatmerits (cooling or heating) as well as concentration of buttermilk affects the calcium phosphate equilibrium. Figure 2.12 shows the equilibrium of milk salts. As it can be seen, lowering the ph (increasing [H + ]) of milk leads to demineralization of the casein micelle and therefore decreases its stability. Decreasing the température of milk is associated with réversible demineralization of casein micelle. This is caused mainly by an increase in calcium phosphate solubility which drags the equilibrium towards the sérum phase (Pierre & Brûle, 1981). On the other hand, heating milk créâtes the reverse effect on equilibrium and calcium phosphate solubility decreases (Gaucheron, Le Graët, & Schuck, 2004). If heat treatment is over 80 C, then there is formation of insoluble calcium phosphate that may precipitate onto the walls of industrial equipment (Gaucheron, Le Graët, & Schuck, 2004).

52 34 H 2 PO 4 Sérum phase Ca [260] ppm Colloidal phase Ca [800] ppm Figure 2.12 Salts equilibrium in milk 2.4 Membrane séparations Membrane filtration is a fractionation technology that makes use of a semi porous membrane that acts as a barrier to the transportation of species in a fluid (Cheryan, 1998). Membrane séparation processes are important in the food industry and especially in the dairy sector. Thèses processes allow for better utilization of milk solids but need to be well understood to take advantage of their full potential Principles of membrane séparations Driving forces on pressure-driven processes During the course of filtration a number of phenomena occurs at the membrane surface (Figure 2.13). As pressure is applied and fluid circulated across the membrane, a layer usually referred to as concentration polarization layer (CP) is formed at the surface of the

53 35 membrane. This layer is composed of molécules rejeeted by the membrane that are brought to the surface of the membrane by the convective flow. The thickness of the CP layer is influenced by the back-diffusive flow induced by the shear rate created by the fluid high recirculation speed. Présence of molécules at the surface can lead to fouling of the membrane and associated decrease of the permeate flux (J). Fouling will be further characterized at section PERMEATE Convective Fiow S Back-Diffusive Flow Figure 2.13 Schematic représentation of phenomena occurring during membrane filtration from Cheryan (1998) Filtration ranges (classification) Membrane filtration can be separated in 4 classes (Figure 2.14): Microfiltration (MF), Ultrafiltration (UF), Nanofiltration (NF) and Reverse osmosis (RO). In thèse filtration classes, the driving force is the applied pressure (Cheryan, 1998).

54 36 MF membranes are able to separate suspended partieles from dissolved solids. The pore sizes used in MF are roughly between 0.1 and 10 um. UF can effectively separate between dissolved molécules. In the case of UF and NF it is more common to express the nominal molecular weight cut-off (MWCO). The MWCO range of UF is from 1000 to Da (Cheryan, 1998). In NF, séparation is also influenced by the electrostatic charge of the membrane (Lapointe, 2004). MWCO of NF membrane can vary from 200 to more than 1000 Da. In RO, the idéal membrane allows only water to permeate (Cheryan, 1998) although it has been observed that small molécules and even lactose can permeate RO membranes (Walstra, Wouters, & Geurts, 2006). Considering the material to be separated, the pressures required to overcome osmosis in RO are much greater than the pressure needed in MF. In gênerai, pressures below 300 kpa are required for MF whereas pressure between 3000 and kpa are involved in RO (Gésan, 1993) NPN Lactose Froieîos Partîtes MF..... : : : v <*?> \ : Figure 2.14 Rejection of milk components depending on membrane filtration class from Saboya & Maubois (2000)

55 Filtration membranes, equipment & process parameters Membrane materials & configuration of modules Materials The materials used to manufacture membranes are either polymeric or minerai. The main polymeric materials used for membrane manufacture are: cellulose acétate (CA), polyamide (PA), polyvinylidene fluoride (PVDF) and polysulfone (PS). Minerai membranes are mostly made out of carbon or ceramic derivative (zirconium, a-alumina). The major inconvenience of polymeric as compared to minerai membranes is their rather instability to the harsh conditions needed during cleaning procédures. For example, ceramic membranes can withstand a température of 130 C and ph 1-14 while cellulosic membranes can only be used at température below 35 C and ph 3-8 (Cheryan, 1998). Minerai membranes are therefore more résistant and can be used for years without showing any damage. However, the cost of minerai membrane is higher than that of polymeric membranes. For example, the cost for a polymeric based membrane filtration System ranges from 225 to 350 $ per m 2 while the same System equipped with minerai membranes would imply costs of $ per m 2 (Cheryan, 1998). Based on the long period of utilization, minerai membrane can still be viewed as an économie solution despite the initial investment. Configuration Configuration of filtration modules can be plate and frame, tubular, hollow fibers, or spiral wound (Figure 2.15). In the plate and frame configuration, the feed is pumped into narrow channels separated by membranes on each side of the feed. The spiral wound membrane configuration is mostly used in industrial application because of the high ratio of filtration area versus size of the module and because of its low cost (Cheryan, 1998). In the spiral wound configuration, a spacer is added over the membrane to ensure that the

56 38 enrolled layers of membrane do not stick together and let the feed circulate through the membrane in a turbulent flow (Bylund, 1995). This configuration has only been developed with polymeric membranes. Minerai membranes are mostly available in tubular configuration. Plate and frame Psrmeate Spiral wound Pernv:il& ifioiaitxl tivei.- '" '' -'.{> ipporîlng lubee / Membtnra 1 Hollow fîbers Figure 2.15 Configurations of membrane filtration modules adapted from Bylund(1995) Modes of opération Membrane séparation can be conducted in many modes of opération depending on desired degree of séparation, duration of process between cleaning cycles and mostly depending on the costs of opérations. Filtration process can be carried out in

57 39 concentration or diafiltration mode. The concentration mode implies a réduction of the initial feed volume to a desired level often expressed as volumetric concentration factor (VCF) and calculated according to équation 2.1 where Vo is the initial feed volume and VR is the volume of retentate. Equation 2.1 Volumetric concentration factor calculation. In the diafiltration mode, the volume loss to permeate is replaced by water and filtration is continued. Diafiltration can be done in continuous or discontinuous modes. In the continuous diafiltration process, water is added to the retentate at the same rate as the permeation flux, thus keeping a constant volume. Discontinuous diafiltration is done by concentration the fluid to the desired VCF and then replacing the lost volume of permeates by water and re-concentrate the diluted fluid. The diafiltration factor is often expressed as number of diavolumes calculated using équation 2.2 where Vp is the volume of permeated fluid and Vo is the volume of initial feed. Equation 2.2 Diafiltration factor (in diavolumes) calculation. The advantages of diafiîtration versus VCF mode is that the increase in viscosity caused by concentration is prevented by water addition. Diafiltration therefore allows for higher flux, increasing purification of solutés and reduced pumping costs. However, considering the important volumes of water added in a diafiltration process, handling of large volumes of permeate can sometime provc to be problematic.

58 40 Industrial filtration processes can also be carried out in batch, continuous or semi continuous modes. Among the modes of filtration, principally four (4) approaches are used which are: batch opération, single pass processing, feed and bleed and multistage recycle opération (Figure 2.16) (Cheryan, 1998). Batch processing is the simplest of ail modes of opération and can be used with partial recycling of retentate and constant addition of feed. Single pass processes are rarely used because of low VCF reached and high membrane area needed to reach desired VCF (Cheryan, 1998). Feed and bleed process is mainly used in continuous industrial filtration. This process is essentially a compromise between batch and single pass modes. Feed and bleed filtration allows to reach the final retentate concentration as the feed is pumped into the retentate recirculation loop. However, the flux is generally lower than the average flux of batch processing because the concentration of the feed in the recirculation loop is maintained at the same level as the final concentrate of batch process (Cheryan, 1998). Multistage recycle consists in a séries of membranes using the retentate of the previous séries as a feed. Thus the first stage or membrane séries opérâtes at low feed concentration and high flux while the last série operate at final feed concentration and low flux. This mode of opération is usually economical because it can be operated for several hours daily with only short cleaning cycles (Cheryan, 1998).

59 41 Batch Feed : Retentate Single pass Feed and bleed Permeate Retentate Permeate Multistase recycle Permeate Figure 2.16 Modes of opération of filtration processes Performance évaluation of membrane process The performance of a membrane process is usually measured by its selectivity and its productivity. The selectivity of a process is usually calculated using rejection or transmission coefficients. Rejection coefficient is calculated with équation 2.3. The terms C P and CR correspond to the concentration of a component in the permeate and retentate respectively. Rejection coefficient ranges from 0 to 1 where 0 indicates that the membrane does not retain the component while 1 is an indication that the membrane totally retains the component in the retentate. Equation 2.3 Rejection coefficient calculation a

60 42 Transmission coefficient can be calculated according to équation 2.4 and is the opposite of the rétention coefficient. Transmission and rétention coefficient are often reported as percentage. Equation 2.4 Transmission coefficient calculation T = l-r Rejection rate of membranes are affected by many factors including transmembrane pressure, turbulence near the membrane surface, température, concentration of the solutés, ph, ionic strength, shape and conformation of the molécule (Cheryan, 1998). The other important parameter to measure to evaluate the performance of a membrane process is permeation flux. Permeation flux is defined by the amount of liquid that permeates a membrane expressed as unit of volume or mass per unit of time and membrane surface. Permeation flux is calculated according to équation 2.5. Equation 2.5 Calculation of flux J = Q p r T In équation 2.5, J is the permeation flux, s is the membrane porosity, dp is the mean pore diameter of the membrane, Py is the transmembrane pressure, Àx is the length of the channel (i.e the thickness of the membrane) and i is the viscosity of the fluid permeating the membrane (Cheryan, 1998). In practice however, J is measured as équation 2.6

61 43 Equation 2.6 Practical calculation of J T AV 1 At A Where AV is the volume of permeate eollected in a measured period of time (At) and per membrane area (A). Flux is dépendent of many processing factors and modelization of the flux is complex because of the inability to precisely model the phenomena occurring at the vicinity of the membrane (Cheryan, 1998). One approach to modelize the flux during MF is the "résistance in séries" concept. According to this concept, flux is limited by a séries of résistances to permeation of the fluid through the membrane. Résistances includes, résistance of the membrane (RM), résistance caused by fouling (R F ) and résistance caused by the CP layer (R G ) (Equation 2.7). Equation 2.7 Calculation of J according to the "résistance in séries" model P T J = RM +R F+ R G Measurement of RM can be done by measuring the pure water flux of the membrane and is commonly used as a vérification of the effectiveness of a cleaning procédure after filtration of a fluid (Cheryan, 1998).

62 Fouling in membrane filtration Fouling of filtration membrane is still the major faetor that limits the use of membrane filtration processes (James, Jing, & Chen, 2003). As fouling occurs, J decreases and ultimately costly and long term damaging cleaning procédures hâve to be carried out. When filtration conditions are kept constant (recirculation speed, transmembrane pressure), RM and RG should remain constant. The decrease of J in Equation 2.7 is therefore correlated with an increase of RF. Membrane fouling résistance, the term Rp in Equation 2.7, can be decomposed in réversible fouling and irréversible fouling. The différence between the two terms is that irréversible fouling can be removed only by chemical cleaning of the membrane (Cheryan, 1998). Membrane properties affecting fouling reported by Cheryan (1998) are hydrophilicity of the membrane, surface topography of the membrane, charge of the membrane and pore size. The solutés in the fluid to filter can deeply affect fouling of membranes. Proteins are believed to be major foulants of UF membranes (James, Jing, & Chen, 2003). Only 5 minutes of contact between milk and a polysulfonic membrane without any permeation resulted in a three fold decrease of the flux (Tong, Barbano, & Rudan, 1988). Proteins contain both hydrophobic and hydrophilic domains as well as a variable density of charges. Furthermore ail thèse properties are affected by ph, ionic strength, shear and heat treatment (Marshall & Daufin, 1995). Minerais, especially calcium, are also believed to play an important rôle in the fouling of filtration membranes. Précipitation of calcium phosphate on the membrane can induce dramatic fouling but ionic calcium can also induce fouling as it can act as a bridge between the membrane and the proteins (Cheryan, 1998; Vetier, Bennasar, Parodo de la fuente, & Nabias, 1986).

63 Stratégies to minimize fouling There are two ways of operating a filtration: dead-end filtration and tangential flow filtration (Figure 2.17). In dead-end filtration (or frontal filtration) the pressure is applied perpendicularly to the membrane as opposed to parallel in tangential filtration. In dead end filtration, J déclines rapidly because of the accumulation of material at the surface of the membrane (cake) that causes résistance. Industrially, most membrane processes are based on tangential flow of the feed. :.. \\-n- ~~ Utt'f Figure 2.17 Dead-end filtration and tangential flow filtration from Brainerd (2001) Even if the fluid to filter circulâtes tangentially to the membrane surface, phenomena shown at Figure 2.13 still occurs. In order to limit the detrimental effect of the CP and fouling on permeation flux, many technological solutions hâve been proposed including turbulence promoters, backflushing devices and uniform transmembrane pressure filtration (Cheryan, 1998). Turbulence promoters are devices inserted in the feed ehannel of the membrane module to ensure a maximum shear rate at the surface of the membrane thus, reducing the effect

64 46 of CP and fouling. The spacers found in spiral wound membranes can be considered as turbulence promoters although they are originally designed to separate enrolled membrane sheets and allow the feed to circulate in between (Cheryan, 1998). Back flushing can be done in a membrane unit simply by blocking the permeate exit. The permeate outlet being blocked, the pressure build-up in the permeate side of the membrane induces the permeate to go back in the feed channel allowing the removal of material deposited on the surface of the membrane. However, back flushing with periodic pumping of permeate is much more efficient at removing the deposit on the surface of the membrane because of the possibility to use a pressure superior to that of the inlet feed. However, if the flux réduction is caused by pore blocking rather than déposition of material at the surface of the membrane, backflushing may not be efficient to restore original flux (Cheryan, 1998). Uniform transmembrane pressure (UTP) fïltration concept has been first proposed by Sandblom (1974). This hydraulic concept is based on the fact that low transmembrane pressure (TMP) reduces fouling of the membranes. However, during tangential filtration high shear rate is needed resulting therefore in the establishment of high TMP at membrane inlet and low TMP at the membrane outlet leading to heterogenic fouling of the membrane (Figure 2.18, left). To avoid this phenomenon, Sanblom (1974) proposed to use high speed co-current recirculation of permeate (Figure 2.18, right). By adding recirculation of permeate on the permeate side of the membrane, it is possible to obtain a pressure drop equal to that in the feed channel, thus creating a constant and low TMP throughout the length of the membrane. Alpha-Laval introduced successfully this concept for MF of skimmed raw milk (Bacto-catch) for removal of microorganisms and the performance of this process has been dramatic with flux reaching 500 to 900 L.h" 1.m 2 maintained for several hours (Saboya & Maubois, 2000).

65 47 CONVENTIONAL MF CO-CURRENT PERMEATE FLOW 1^1 p» l:ll i Tirne Figure 2.18 Conventional tangential filtration versus UTP - co-current permeate flow tangential filtration adaptedfront Cheryan (1998) Lately, membrane manufacturers hâve developed membranes that allowed for a better control of permeation flux es and fouling without the use of UTP modules with permeate recirculation. Membranes with a porosity gradient allows for uniform permeation flux on the length of the membrane and therefore a better control of the fouling on the surface of the membrane (Saboya & Maubois, 2000). Porosity gradient technology has however only been developed for tubular ceramic membranes. Other devices including vibrating modules, pulsative flow or electrical field hâve been applied to filtration module but most of thèse devices hâve yet to be shown successful at industrial scale Applications of membrane filtration in dairy processing The dairy industry is the largest user of membrane technology among the food industry (Cheryan, 1998). Each filtration class, from RO to MF finds applications in the dairy industry.

66 48 RO RO is mainly used as a concentrating step before drying of whey or low total solids dairy fractions. RO is often preferred to evaporative drying because it consumes less energy up to a total solid levels of 15 to 20 % (Walstra, Wouters, & Geurts, 2006). An important drawback of RO as a water removal technology is that concentration of solutés can not be as important as in evaporative concentration because of the increasing viscosity of concentrated fluids that severely restricts permeation flux (Cheryan, 1998). Typically, 30 to 35 % solids is the maximum level that can be reached with RO membranes before the increase in viscosity becomes highly problematic. NF Nanofiltration membranes are relatively new to the dairy industry. The main application of NF in dairy is the demineralization of acid or sweet whey (Jeantet, Rodriguez, & Garem, 2000). This demineralization technique is still not optimally used because of low permeation of multivalent ions and fouling sensitivity of thèse membranes. Other applications of NF are being developed like fractionation of complex mixture of peptides from whey proteins hydrolyzates for example (Lapointe, Gauthier, Pouliot, & Bouchard, 2005; Pouliot, Wijers, Gauthier, & Nadeau, 1999). UF Among membrane séparation classes, UF is probably the most important membrane process in the dairy industry (Cheryan, 1998). One of the main dairy applications of UF is in the cheese manufacture. UF can effectively concentrate milk prior to cheese making. The major benefïts of this procédure in cheese making is an increase of yield (10-30 %), reduced rennet requirement and a reduced amount of milk to handle (Maubois & Mocquot, 1975). Whey and milk proteins can also be effectively concentrated using UF.

67 49 The use of UF to manufacture whey protein concentrate (WPC) and milk protein concentrate (MPC) has been widely reported (Huffman & Harper, 1999). MF MF became an important process for the dairy industry around the 1980's (Saboya & Maubois, 2000). The développements of résistant and efficient ceramic filtration membrane made this process industrially available. Furthermore, the concept of "uniform transmembrane pressure (UTP) proposed by Sandblom (Sandblom, 1974) allowed for a much better control of the rapid fouling of MF membranes by dairy products. The three main applications of MF in the dairy industry are: bacterial removal from skim milk, préparation of micellar casein from skim milk and fat removal from whey. Bacterial removal In bacterial removal of skim milk, membranes of 0.5 to 1.4 jim are used with low TMP (~ 50 kpa) and can yield a permeate with less than 10 CFU.mL" 1. Permeation fluxes are usually around 500 L.h'.m" 2 (Saboya & Maubois, 2000). Permeate (skim milk) is then blended with the adéquate amount of pasteurized cream and minimal pasteurization is applied. The resulting milk has a claimed shelf life of 35 days and no cooked flavor (Eino, 1997). Micellar casein Micellar casein is used mainly to fortify milk before cheese making (Saboya & Maubois, 2000). MF membranes of um pore sizes can be used to produce a micellar casein using about the same parameters as for bacterial removal of milk. The permate produced by this technology is very similar to cheese whey in composition and is claimed to be stérile and virus free (Gautier, Rouault, Mejean, Fauquant, & Maubois, 1994). Micellar

68 50 casein can be used for isolation of 3-casein (Pouliot, Pouliot, Britten, Maubois, & Fauquant, 1994) and for the préparation of various nutraceuticals products derived from milk proteins (Saboya & Maubois, 2000) Fat removal from whey Whey protein concentrate (WPC) and isolâtes (WPI) are two major ingrédients in today's food industry. Most of thèse WPC and WPI's are produced by UF (Gésan, 1993). The présence of residual lipids in whey causes important problems to the manufacture of WPC and WPI. Residual lipids in whey reduce the performance of UF System as a resuit of fouling and a lower purity of WPC and WPI is obtained because of high lipid rétention in UF membranes (Fauquant, Vieco, Brûlé, & Maubois, 1985; Morr & Ha, 1993). MF membranes of 0.8 um pore size can be used after thermocalcic aggregation of the lipids (Fauquant, Vieco, Brûlé, & Maubois, 1985) to remove lipids and yield a whey with minimal amount of residual fat that can be concentrated to a WPC or WPI. 2.5 Fractionation of buttermilk using membrane processes Considering the low value of buttermilk solids, a fractionation approach is désirable to concentrate or isolate components of interest for either the functional food industry or the newly emerging nutraceuticals market. Only a handful of technologies are industrially available today to isolate or concentrate MFGM material. Among thèse, solvent extraction might be the better suited approach to isolate polar lipids from buttermilk. However, this approach makes use of high volumes of toxic and costly organic solvents which renders the feasibility and the économies of the process unattractive for the dairy industry.

69 51 Membrane filtration is a well established technology in the dairy industry. This fractionation technology is readily available in dairy plants and could be an intégral part of a process aiming at adding value to buttermilk by fractionation Ultrafiltration of buttermilk Ultrafiltration of buttermilk has been investigated for its use in food products rather than for MFGM fractionation. Several studies hâve reported the use of ultrafiltered buttermilk in cheese process (Joshi & Thakar, 1993; Mistry, Metzger, & Maubois, 1996; Poduval & Mistry, 1999; Raval & Mistry, 1999; Turcot, Turgeon, & St-Gelais, 2001) and in low fat yogurt (Trachoo & Mistry, 1998). Buttermilk represents an attractive source of casein to fortify cheese milk because of its relatively low value and the présence of phospholipids in buttermilk can help improve the texture of low fat cheeses (Turcot, Turgeon, & St- Gelais, 2001). Ultrafiltration of buttermilk allows the préparation of an ingrédient rich in milk proteins that can be added to milk before cheese processing. Results of adding ultrafiltered buttermilk to low fat cheeses are generally positive on texture and water rétention when the substitution of milk for UF buttermilk is around 5% (Poduval & Mistry, 1999). However, when more than 5 % UF buttermilk is added, excessive humidity in cheeses and poor curd fusion has been observed (Mistry, Metzger, & Maubois, 1996). Explanations for this resuit hâve not been suggested by the authors but could be related to the présence of high amounts of phospholipids acting as water rétention agents in the cheeses. Fundamental study of the UF of buttermilk has been very scarce. Ramachandra Rao, Lewis & Grandison (1995) reported the flux pattern in UF of buttermilk and found that the initial flux as well as the average flux was dépendent of the ph. At its natural ph (6.6), flux pattem of buttermilk was stable at around 40 L.h~'.m- 2. In this study, the authors only measured the rejection of calcium during filtration and found that calcium

70 rejection decreased with decreasing ph of filtration. Thèse results can be explained by the shift of calcium from casein micelle to the soluble phase created by the ph drop Microfiltration of buttermilk Because of its ability to separate partiel es (MFGM components and caseins) from soluble matter, MF has been used by several authors in order to concentrate MFGM from buttermilk. It became quickly évident that casein micelles may be deleterious to MFGM concentration by MF (Surel, 1993). In their study of defatting of buttermilk, Surel & Famelart (1995) found that casein and MFGM components were both retained or transmitted at the same rate while using ceramic MF membrane of various pore sizes. The System they used in their study was not equipped with fouling control devices and could hâve become quickly fouled as their permeation flux was relatively low (110 to 400 L.h'.m 2 ). In order to overcome that problem, Sachdeva and Buchheim (1997) reported the use rennet coagulation of buttermilk. Briefly, they used rennet to coagulate casein micelles in buttermilk and then applied MF and UF to whey to concentrate phospholipids. The authors reported a recovery of 77.5% of the phospholipids from buttermilk. Coagulation of buttermilk was however difficult to obtain and required the addition of important amount of calcium. Sachdeva and Buchheim (1997) explained the poor rennet coagulation of buttermilk by the decrease in soluble calcium and the interaction of P- lactoglobulin with K-casein following the extensive heat treatment applied to buttermilk components during pasteurization of the cream and spray drying of buttermilk. While this approach appears to be an efficient way to recover phospholipids from buttermilk, the byproduct created (buttermilk weak curd) could be difficult to further process. No data on associated recovery MFGM proteins were presented in that study.

71 53 In another approach, Corredig, Roesch, & Dalgleish (2003) used sodium citrate to disrupt casein micelle structure before membrane fïltration. This approach was based on the laboratory scale method reported by Corredig & Dalgleish (1997a) in which citrate was added to buttermilk before ultracentrifugation. In their study, the authors used PVDF UF membrane of and Da MCWO and a 0.1 um polyethersulfone (PES) MF membrane. While using the MF membrane, Corredig, Roesch, & Dalgleish (2003) reported an effective concentration of MFGM proteins as judged by protein analysis by 4 to 20% acrylamide gradient sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). When the same procédure was used but without the addition of citrate, casein micelles were retained as well as MFGM proteins. The authors did not report any value on the amount of phospholipid found in their isolâtes nor did they quantify the MFGM proteins. Furthermore, this approach necessarily yields a permeate with high concentration of citrate, about 4 times more than milk (Surel & Famelart, 1995). Dealing with important volumes of this by-product could be troublesome for dairy proeessing plants. Hydrolysis of buttermilk proteins by trypsin and chymotrypsin has also been reported by Roesch & Corredig (2002). The principle of this approach was to breakdown casein micelles via enzymatic hydrolysis followed by séparation of MFGM by MF. From the results presented in this study it is hard to conclude on the efficiency of the method as no phospholipid or MFGM protein profiles are presented for the isolâtes. Moreover, MFGM proteins are known to be sensitive to hydrolysis by trypsin and chymotrypsin especially XDH/XO and BTN (Corredig & Dalgleish, 1997b). Unless the goal of the approach is only the concentration of phospholipids from MFGM, the use of enzymatic hydrolysis for isolation of whole MFGM fragments is therefore questionable.

72 54 CHAPTER 3 Hypothesis, goal and objectives As opposed to other dairy by-products, studies on the fractionation of buttermilk hâve been scarce. The instability and variability of this by-product might explain the little scientific interest for buttermilk. Adding to this is the fact that buttermilk colloidal state (MFGM interactions with other solids) is poorly understood. However, with the développements of better fractionation technologies, namely membrane filtration, buttermilk fractionation offers a promising approach for fully exploit its potential and better understand its compositional and structural properties. A fractionation approach could generate ingrédients with better functionality and potential health benefits. The gênerai hypothesis of this doctoral thesis is that MF membranes can be used to achieve séparation of MFGM components from other buttermilk solids. While MF alone may not be able to effïciently separate MFGM, optimization of the parameters of filtration can improve the selectivity of the process. Additionally, modification of the composition of the cream prior to churning can improve the ability of MF to selectively fractionate components of the resulting buttermilk. The goal of the work presented hère was to develop a fractionation approach, based on MF, to generate MFGM concentrâtes from buttermilk. The approach to be developed had to be practical for the dairy industry in terms of feasibility and accessibility. Objectives for this work were: 1. To evaluate the effect of MF séparation variables on the fractionation of buttermilk components

73 55 2. To understand the effect of processing variables on buttermilk and MFGM composition and structure 3. To evaluate the potential use of buttermilk derived from whey cream as a substrate for MFGM concentration by MF 4. To characterize the effect of removing caseins in the cream on buttermilk fractionation by MF.

74 56 CHAPTER 4 Effect of température and pore size on the fractionation of fresh and reconstituted buttermilk by microfîltration The work presented in this chapter was aimed at obtaining a profile of séparation of buttermilk components while using différent membranes pore sizes and filtration températures. We also investigated in this project the effect of using fresh versus reconstituted buttermilk on the results of fractionation by MF. Writing of this manuscript was supervised by Dr Yves Pouliot and Dr Rafaël Jiménez- Flores who also are co authors. Results obtained in this part of the project were published essentially in this form in Journal of Dairy Science (87), This paper was also presented in poster form at 2004 ADSA Joint Annual Meeting in Phoenix, Arizona.

75 Résumé L'objectif de cette recherche était d'évaluer l'effet de la température (7, 25 et 50 C) et du diamètre de pore de la membrane (0.1 ^m, 0.8 [im et 1.4 um) sur la séparation des protéines et lipides (lipides neutres et phospholipides) au cours de la microfiltration (MF) de babeurre frais et reconstitué. Le babeurre a été microfiltré en utilisant un module de filtration pilote équipé de membranes de céramique. Les expériences de MF on été effectuées en mode UTP (pression transmembranaire constante). Les changements de température de filtration n'ont pas eu d'impact significatif sur la transmission de protéines mais l'utilisation de hautes températures de filtration a réduit la transmission des lipides et phospholipides. Un facteur de concentration (CF) maximum a été atteint pour les lipides à 25 C alors que le CF des protéines n'a pas été affecté par la température. L'utilisation de la membrane de 0.1 um a induit une basse transmission des lipides (10%) et de protéines (~ 20%). Les diamètres de pores supérieurs (0.8um et 1.4 um) ont conduit a des transmissions de protéines, lipides et phospholipides plus élevées (> 50%) et induit des flux de perméation plus élevés. Les transmissions de protéines et de lipides étaient significativement différentes pour le babeurre frais et le babeurre reconstitué. Cette étude a démontré que la température, le diamètre de pore ainsi que le type de babeurre affectent la séparation et que la MF seule ne peut parvenir à une séparation efficace des composés du babeurre en vue de créer de nouveaux ingrédients à valeur ajouté.

76 5 H 4.2 Abstract The objective of this research was to evaluate the effect of température (7, 25 and 50 C) and pore size (0.1 )j,m, 0.8 um and 1.4 p,m) on the séparation of proteins and lipids (neutral lipids and phospholipids) during microfiltration (MF) of fresh or reconstituted buttermilk. Buttermilk was subjected to MF using a pilot-scale unit mounted with ceramic membranes. MF runs were carried out in a uniform transmembrane pressure (UTP) mode. Changes in processing température had no significant impact on protein transmission, while increasing température reduced both lipid and phospholipid transmission. A maximum concentration factor (CF) for lipids was reached at 25 C, as protein CF remained essentially unaffected by température. The use of the smaller pore size (0.1 um) resulted in low lipid (10%) and protein (~ 20%) transmission. Larger pore sizes (0.8uxn and 1.4 um) resulted in higher levels of protein, lipid, and phospholipid transmission (> 50%) and gave high permeation fluxes. Transmission of both proteins and lipids were markedly différent when using fresh buttermilk as opposed to reconstituted buttermilk. This study showed that MF température, pore size, and buttermilk type influence fractionation but that MF alone can not achieve optimal séparation of lipids and proteins for the production of novel ingrédients from buttermilk. 4.3 Introduction Buttermilk is characterized by a fat composition rich in milk fat globule membrane (MFGM) components, such as phospholipids, sphingolipids, arid various glycoproteins (Vyas, Astaire, & Jiménez-Flores, 2002). When cream is churned, MFGM components are broken down, most of the hydrophilic components are expulsed into the aqueous phase, and buttermilk is created. The high concentration of phospholipids, such as sphingomyelin, phosphatidylethanolamine, and phosphatidylcholine found in buttermilk is related to the présence of MFGM fragments that migrated to the aqueous phase. This particularity has created an important interest in buttermilk, as evinced by the many

77 59 papers written on that subject (Ramachandra Rao, Lewis, & Grandison, 1995; Raval & Mistry, 1999; Roesch & Corredig, 2002; Surel & Famelart, 1995; Vyas, Astaire, & Jiménez-Flores, 2002). Currently, buttermilk is mainly used for its emulsifying properties in various food products (O'Connell & Fox, 2000). However, récent advances in dairy science hâve shown that some components of buttermilk could also be exploited as health ingrédients. For example, sphingomyelin through its bioactive dérivâtes hâve been found to play important rôles in transmembrane signal transduction, cell régulation, and apoptosis (Parodi, 1997). Moreover, Dillehay et al. (1994) hâve found that dietary sphingomyelin could inhibit 1,2-dimethylhydrazine-induced colon cancer in mice. Important bioactive properties of phosphatidylcholine and phosphatidylserine hâve also been recently reviewed (Kidd, 2002). Since buttermilk is a unique dairy source of phospholipids, it could be a good substrate for fractionation. Sachdeva and Buchheim (1997) reported a method for the recovery of phospholipids using whey from buttermilk acid or rennet coagulation. Purification of phospholipids by thermocalcic aggregation (Fauquant, Vieco, Brûlé, & Maubois, 1985) followed by membrane processing (Baumy, Gestin, Fauquant, Boyaval, & Maubois, 1990) has already been demonstrated from cheese whey, but the présence of casein micelles in buttermilk makes this approach unsuitable. Buttermilk is a substrate that is highiy susceptible to oxidation, which partially explains why it has to be rapidly used or spray-dried (O'Connell & Fox, 2000). The effects of various drying treatments followed by re-hydration are not known and might be of importance if membrane processing is considered. Data from the recovery of MFGM fragments from buttermilk are still lacking (Jensen, 2002) and before going any further in fractionation, effective séparation between lipids and proteins is necessary. Because of its ability to separate particles in suspension (Saboya & Maubois, 2000), MF could be used as the first step of a process for obtaining a lipid concentrate from buttermilk. However, optimum MF processing conditions for better yield between lipid and protein remain to be defined. This paper reports the effect of température and pore size in MF of fresh and reconstituted buttermilk.

78 4.4 Materials and methods Microfiltration Procédure Fresh pasteurized buttermilk was purchased from a local butter factory (Agropur, Plessisville, Canada). Dried buttermilk (Parmalat, Victoriaville, Canada) was reconstituted at the total solid level of the fresh buttermilk (8.4%) using deionized water at 40 C under agitation for 1 h and then cooled overnight at 4 C. MF pilot unit was a TetraPak MSF1 (Lund, Sweden) working in UTP mode equipped with Membralox (Bazet, France) ceramic membranes of 1.4, 0.8, and 0.1 im average pore size. Each MF run was carried out with 50 kg of fresh buttermilk for the température study and 50 kg of reconstituted buttermilk for the pore size study. Température effect and pore size effect runs were carried out with différent buttermilk sources, because of the limited amounts of fresh buttermilk that was available for the experiments. The transmembrane pressure (TMP) during the runs was kept constant at 0.5 to 0.7 bars. For the température study, the fresh buttermilk température was adjusted to 7, 25, and 50 C under agitation in a stainless steel vat for 1 h before MF and the pore size used for this study was 0.8 \xm. Températures of MF were 49.1 ± 0.9, 25.1 ± 1.3 and 7.3 ± 0.5. In the pore size study, reconstituted buttermilk température was adjusted to 50 C under agitation for 1 h in a stainless steel vat. Flux was monitored with a stopwatch and a graduated cylinder. Each experiment was carried out in triplicate. A sample of retentate and permeate from each replicate was collected at the end of each run and frozen (-20 C) until analysis. Transmission rate (%) of proteins, total lipids and phospholipids in dry basis, was calculated on using équation 4.1. Equation 4.1 Transmission calculation Tr(%) = (Cp/Cr)*ioo

79 61 where Cp is the concentration of a component in the permeate side and Cr, the concentration of the same component in the retentate side Cleaning Procédure Cleaning of the MF system was achieved by rinsing the membrane with deionized water followed by alkaline cleaning (Ultrasil 25, EcoLab, Canada; 1.5 % at 75 C / 45 min) combined with 200 ppm of chloride. The System was rinsed with warm deionized water (50-60 C) until reaching normal water ph and then acid détergent (Ultrasil 76, EcoLab, Canada; 0.3 % nitric acid) was circulated for 30 min at 50 C. Finally the system was rinsed with deionized water until reaching normal water ph. The cleaning procédure was repeated until initial water permeation flux was reached Composition Analysis Dry matter, protein and lipids Dry matter content of ail permeates, retentates, and initial buttermilk sample were obtained using a microwave dryer (uwave, Omnimark, Tempe, AZ). In brief, 3 g of sample were placed on a weighing paper and dried in the microwave oven until a constant weight was reached. The protein content was obtained using Kjeldahl (IDF, 1993) Nitrogen Détermination with 6.38 as the protein conversion factor. The lipid content was obtained gravimetrically using the Mojonnier extraction method (IDF, 1987). For each experiment, extracted lipids (triplicates) were combined and diluted to 10 mg/ml in chloroform/mefhanol (2:1) and were stored in a freezer at (-20 C) until analysis.

80 62 Phospholipid analysis Phospholipids from ail samples were analyzed by HPLC (Waters 600, Milford, MA) with an evaporative light-scattering detector (ELSD) (SEDEX 75, Sedere, France). The injector was a Rheodyne Model 7725i (Cotati, CA), the column was a Zorbax Sil 5 um (4.6 i.d x 150 mm, Agilent Technology, Palo Alto, CA). The data were collected and treated using Millennium chromatographic software (Waters). Mobile phases used were A: chloroform/methanol/ammonium hydroxide (80:19.5:0.5) and B: chloroform/methanol/water/ammonium hydroxide (60:34:5.5:0.5) and the binary gradient used is illustrated in Figure K \ 0 t \6 24 il 40 1 imc (ni») Figure 4.1 Binary gradient used in phospholipid analysis by HPLC-ELSD Lipid samples were diluted from 10 mg/ml to 2 mg/ml before injection. Neutral lipids were eluted in the first 6 min of the run and phospholipids were eluted between the 7 l and the 30 th minutes. Runs were 41 min long. Fresh and reconstituted buttermilk phospholipids profiles are shown in Figure 4.2. Components were identified and quantified using calibration curves made with phosphatidylethanolamine, phosphatidylinositol, phosphatidylcholine, and sphingomyelin standards (Sigma, Chicago, IL). Ail solvents and reagents were HPLC grade.

81 63 Figure 4.2 Phospholipid profile of fresh (A) and reconstituted (B) buttermilk obtained by HPLC-ELSD analysis. (NL = neutral lipids, PPL = phospholipids, PE = phosphatidylethanolamine, PI = Phosphatidylinositol, PC = phosphatidylcholine, SM = sphingomyelin) 4.5 Results and discussion Température Effect Processing Température adjustments during buttermilk MF induced important différences in permeation fluxes and the volumetric concentration factors (VCF) achieved were 8X, 6X, and 1.6X for MF at 50 C, 25 C, and 7 C respectively. Flux curves are presented in Figure 4.3. High fluxes obtained at 50 C are comparable to those classically obtained in MF of dairy fluids (Cheryan, 1998; Famelart & Surel, 1994). The increase in the

82 * \ " ' " ^ - -,, r 64 viscosity of buttermilk lowered flux values at 25 C and 7 C (Cheryan, 1998). However, fluxes were lower, but more stable, while proeessing at 7 C as compared to 50 C. Technical difficulties were encountered at the pilot plant during cold (7 C) proeessing, which included température increase caused by fluid circulation in the pumps and fluid friction in the membrane resulting from the high shear rate (>6 m/s). This increase of température was noticeable after 2 h of proeessing so the process had to be stopped at the VFC reached after 2 h. 600 ; A (NI L.h ! J X 3 C ; 0 ~~~'"B +^~ , > _ # i (, -._, r_. _. T T r- - ^ _r_ T-. n,_ Time (min) 75 i - r 90 Figure 4.3 Permeation fluxes during MF (0.8 u.m) of fresh buttermilk at 50 C (±), 25 C ( )and7 C(4). Composition Compositional analysis of the différent retentates and permeates revealed important différences between lipids and phospholipids concentration, while less pronounced différences were observed between protein concentration at the various températures. Permeate and retentate composition is presented in Table 4.1 and comportent transmissions are provided in Figure 4.4. Lipids were concentrated in the retentates 1.07 X, 2.19X, and 1.61X for MF at 7, 25 and 50 C respectively in comparison to original buttermilk. Lipid transmission at 25 C was signifïcantly (P < 0.05) lower than at 7 C but not signifïcantly différent than at 50 C. However it appears that lowering température from 50 C to 7 C increase the lipid transmission. Hydrophobic interactions, which are known to be stronger at high températures, could induce more aggregation between lipid

83 65 particles and increase particle size at high température and therefore reduce the transmission of those particles. Weaker hydrophobic interactions at low température are also known to induce P-casein dissociation (Dalgleish & Law, 1988). Table 4.1 Composition of fresh buttermilk, permeates and retentates from MF 0.8 um at différent températures. Means are given in % of dry matter (DM). Protein (%DM) Fresh buttermilk MF Permeates O.Xum 7 C a 25 C a 50 C a MF Retentates 0.8 um 7 C C b 50 C b Lipids (%DM) : a 6.05 b 9.02 a b ab PPL (% DM) PE %of PI PPL 2 PC SM ab Subcolumn in the same line values with différent superscript differ (P < 0.05). 1 PPL = phospholipids, CV (%) calculated on 4 injections of each standard were 1.63, 2.94, 7.47, and 2.63 for PE, PI, PC, and SM respectively. Percentage of each class into total phospholipids (PE = phosphatidylethanolamine, PI = phosphatidylinositol, PC = phosphatidylcholine, SM = sphingomyelin). The slight decrease in protein concentration in retentates between 50 C and 7 C may be related to that phenomenon. However, the pore size used in this part of the experiments (0.8 um) is probably too large to show important différences in protein rétention by température modifications only as protein concentrations in the permeates at the three températures were not significantly différent. Phospholipid analysis showed that MF at 25 C and 50 C increased the concentration of ail main classes by about 2 fold. MF at 7 C did not provide the concentration of phospholipids in the retentate. This resuit shows that phospholipids transmission is closely related to that of total lipids. This resuit is not unexpected considering the fact that phospholipids in buttermilk are associated with

84 66 triglycérides and various membrane proteins forming MFGM fragments (Sachdeva & Buchheim, 1997). Furthermore, cold température could facilitate phospholipid dissociation of MFGM fragments enabling higher transmission rates of both phospholipids and MFGM fragments. Nevertheless, process température has a considérable impact on MFGM structure (Dufour, Subirade, Loupil, & Riaublanc, 1999) and even if MFGM is présent in form of fragments in buttermilk, it appears that modifying température of processing could induce changes in particles sizes and aggregation, which would lead to variations in transmission levels of both lipids and phospholipids during MF. Température combined with other physicochemical modification (ph, hydrolysis, or chelating agents) could be a way to disrupt furthermore the linear corrélation between lipid and protein rétention that has been reported by Surel and Famelart (1995) Température ( C) Figure 4.4 Transmission of proteins ( ), lipids ( ) and phospholipids (0.8 im) of fresh buttermilk at différent températures. during MF

85 Pore Size Effect Processing VCF were 4X for the 0.1 um, and 8X for 0.8 um and 1.4 um membranes. As it was the case in the study of the température effect, pore size variations during buttermilk MF led to important changes in permeation fluxes (Figure 4.5). MF at 1.4 um was carried at 0.5 bar TMP, while MF at 0.8 um and 0.1 um were carried at 0.7 bar TMP, and at 50 C in ail cases. This explains in part why MF at 0.8 showed higher fluxes than MF at 1.4 um. The flux pattern for MF at 1.4 um suggests that we could hâve obtained higher fluxes with higher TMP. However, an increase in TMP could hâve led to more fouling since large pore membranes are more susceptible to fouling because of the high amounts of particles at the surface of the membrane (Cheryan, 1998). Low fluxes obtained with MF of reconstituted buttermilk with the 0.1 um membrane are higher than those reported by others (Surel & Famelart, 1995) with fresh buttermilk on a 0.2 um membrane mounted on a back-pulse MF unit. Our experiments were done using a MF equipment with cocurrent recirculation of permeate (Bactocatch, TetraPak, Lund, Sweden), which gives higher fluxes than a back-pulse module (Cheryan, 1998) Time (min) Figure 4.5 Permeation fluxes during MF of reconstituted buttermilk (50 C) using différent pore sizes: ( ) 0.1 um at 0.7 bar TMP, ( ) 0.8 um at 0.7 bar TMP and (±) 1.4 um at 0.5 bar TMP.

86 68 Composition Pore sizes had an influence on both protein and total lipid levels. Retentate and permeate composition is presented in Table 4.2. Protein and lipid content was increased in the MF retentate at 0.1 um by 1.77 and 1.93 fold respectively, which indicates that the 0.1 um membrane did not change significantly the corrélation between protein and lipid concentration. However, the low proportion of lipids found in MF permeate using the 0.1 um membrane shows the potential of this membrane. The low transmission of lipids (10%) (Figure 4.6) indicates that this membrane can concentrate effectively lipids in the retentate side. However, low transmission for protein content was also noticed, indicating that this pore size without any other treatments cannot separate lipids from proteins. This resuit is not surprising considering that low pore diameter membranes (= 0.2 um) are often used to produce micellar casein concentrate or native phosphocaseinate (Maubois & Olivier, 1992; Pouliot, Pouliot, Britten, Maubois, & Fauquant, 1994). Table 4.2 Compositions of reeonstituted buttermilk, permeates and retentates from MF at différent pore sizes. Values are given in % of dry matter (DM). Reeonstituted MF Permeates MF Retentates buttermilk 0.1 fini 0.8 fini 1.4 fini 0.1 uni 0.8 fini 1.4 fini Protein (%DM) " a a a b a Lipids (%DM) b 5.2 l a 5.69 a a 8.49 b 9.47 b PPL r ( %DM)' %of PE ri PPL 2 PC SM ab Subcolumn in the same line values with différent superscript differ (P< 0.05). 1 PPL = phospholipids, CV (%) calculated on 4 injections of each standard were 1.63, 2.94, 7.47, and 2.63 for PE, PI, PC, and SM respectively. Percentage of each class into total phospholipids (PE = phosphatidylethanolamine, PI = phosphatidylinositol, PC = phosphatidyleholine, SM = sphingomyelin).

87 69 Phospholipids were concentrated 2.2X in the retentate using the 0.1 im membrane and the low transmission level observed for phospholipids (5.8%) is once again consistent with lipid transmission and probably results from associations between lipids and phospholipids as well in reconstituted buttermilk. Lipid concentration was not that effective and not significantly différent using 0.8 or 1.4 um membranes as CF reached were 1.4 X and 1.56X respectively. Proteins were transmitted more effectively using those pore sizes, indicating that the main mechanism for protein transmission is size exclusion. o i Pore diamcter (fini) 1.5 Figure 4.6 Transmission of proteins ( ), lipids ( ) and phospholipids ( < reconstituted buttermilk at différent pore sizes. during MF of A lowerprotein transmission was noticed at 1.4 fxm compared with that at 0.8 \im, which indicates déviation from Ferry's law. This filtration law indicates that for a particle of a given diameter, an increase in membrane pore size leads to higher transmission. The observed déviation from this law could be attributed to membrane fouling (i.e., gel layer formation at membrane surface). The range of pore sizes tested did not make it possible to détermine a critical pore size in which différences in transmission of both lipid and protein occurred, most likely somewhere between 0.1 and 0.8 \x.m. Only two pore sizes in that range are however provided by the manufacturer (0.2 and 0.5 (xm). Large pore membranes (0.8 and 1.4 um) give high fluxes and a good transmission of proteins, but

88 70 the high transmission of lipids remains a problem. Despite the low fluxes obtained, the 0.1 im membrane is the only membrane that provided an important concentration of lipids and phospholipids. The only way to enhance protein and lipid séparation with this membrane would be to reduce the size of protein particles (casein micelles) through chemical treatments (EDTA or citrate) or enzymatic cleavage. Citrate addition induces demineralization (calcium) of casein micelles and can induce a decrease in micelle sizes (Johnston & Murphy, 1992), but solubilisation of entire colloidal calcium in order to disrupt casein micelles would require concentration that are four times higher than those found in natural milk (Surel & Famelart, 1995) Effect of Buttermilk Source Composition. From our previous results, it appears that, under the same MF conditions (0.8 um and 50 C), results from fresh and reconstituted buttermilk séparation were markedly différent (Table 4.1 and 4.2, Figures 4.4 and 4.6). Proteins were more concentrated in the MF retentate from reconstituted buttermilk than retentates from fresh buttermilk. Moreover, the transmission level of proteins using fresh buttermilk reached around 80% as opposed to 60% with reconstituted buttermilk. Processing history is the only obvious différence between fresh and reconstituted buttermilks. The various treatments for making buttermilk powder (mainly evaporative concentration and spray drying) are likely to induce some aggregation of proteins that would hâve an impact on MF of reconstituted buttermilk. Lipid concentration in retentate was more effective with fresh buttermilk (1.6 IX) than with reconstituted buttermilk (1.4 X). Transmission of lipids through the membrane was significantly (P < 0.05) higher in MF of reconstituted buttermilk (61% vs. 44%). The impact was even greater on phospholipid transmission. In MF of reconstituted buttermilk, 78% of phospholipids were transmitted through the membrane, while only 45% were transmitted using fresh buttermiîk. This resuit shows that the corrélation between lipid and phospholipid transmission is broken in reconstituted buttermilk, but not in the case of fresh buttermilk.

89 71 The reason for thèse différences is not clear, but they could be induced by the additional processing steps in the case of reconstituted buttermilk. Those treatments could induce rearrangement or rupture of a part of the MFGM fragments, which in turn reduce lipid particle size. Microscopic observation of both fresh and reconstituted buttermilk could shed light on the effect of drying buttermilk on protein aggregation and lipid particle changes. 4.6 Conclusion Buttermilk is a complex dairy fluid. Knowledge of its colloidal properties is still scarce. However, it remains that température changes in buttermilk MF induce important changes in both flux behavior and permeate and retentate composition. Higher protein transmission coupled with low fat transmission was observed at 25 C. Pore size also affects buttermilk fractionation with better resuit obtained using 0.1 um membrane despite the high rétention of proteins. The drying of buttermilk seems to affect both lipid and protein and results suggest that processing with fresh buttermilk could lead to a better séparation between lipids and proteins. Microscopic analysis of buttermilk under various conditions is required to further our understanding of colloidal properties of both dried and fresh buttermilk. Combination of processes (hydrolysis, température treatments) could be helpful in controlling séparation and work is in progress to evaluate the potential of this type of process combination. 4.7 Acknowledgements The authors would like to thank A.-F. Alîain for her assistance during the analysis of the phospholipids. This work was funded in part by the Natural Science and Engineering Research Council of Canada (NSERC), the California Dairy Research Foundation (CDRF) and the California State University Agriculture Research Initiative (CSU-ARI).

90 72 CHAPTER 5 Effect of processing on the composition and structure of buttermilk and its milk fat globule membranes Results presented in Chapter 4 highlighted the différence in reconstituted and fresh buttermilk. It was suggested that processing of buttermilk into a powder can induce changes in buttermilk structural characteristics that could explain the différence in transmissions of proteins, lipids and phospholipids when using reconstituted as compared to fresh buttermilk. The work presented in this chapter aimed at simulating processing steps for manufacture of buttermilk powder in order to understand the processing steps effect on composition and microstructure of buttermilk and MFGM isolated from buttermilk. Writing of this manuscript was supervised by Dr Yves Pouliot and Dr Rafaël Jiménez- Flores who also are co authors. This manuscript has been accepted for publication essentially in this form in International Dairy Journal. Results hâve also been presented in poster form at 2006 ADSA Annual Joint meeting in Minneapolis (Minnesota, USA).

91 Résumé L'effet de la pasteurisation de la crème sur la composition et la microstructure du babeurre après pasteurisation, évaporation et séchage par atomisation a été étudié. La composition de la MFGM isolée des échantillons de babeurres a aussi été étudiée. La pasteurisation de la crème a induit une teneur en lipide supérieure dans le babeurre. Le contenu en phospholipides du babeurre a diminué de 38.2 et 40.6 % dans le babeurre de crème crue et de crème pasteurisée, respectivement après séchage par atomisation. Le principal effet observé au niveau de la pasteurisation de la crème était l'incorporation de protéine sérique aux isolats de MFGM. Une diminution similaire à celle du babeurre au niveau des phospholipides a été observée dans les isolats de MFGM suite au séchage par atomisation. Les observations de la microstructure du babeurre par microscopie électronique à transmission ont permis de mettre en évidence une grande hétérogénéité dans la microstructure du babeurre mais n'a pas permis de révéler des effets des étapes de procédés appliqués au babeurre.

92 Abstract The effect of cream pasteurization on the composition and microstructure of buttermilk after pasteurization, evaporation and spray drying was studied. The composition of MFGM isolated from buttermilk samples was also characterized. Pasteurization of cream resulted in higher lipid recovery in the buttermilk. Spray-drying of buttermilk had a signifïcant effect on phospholipid content and composition. The phospholipid content decreased by 38.2% and 40.6% respectively in buttermilk from raw and pasteurized cream. Pasteurization of cream resulted in the highest increase in whey protein recovery in MFGM isolâtes compared to ail other processing steps applied on buttermilk. A réduction in phospholipid content was also observed in MFGM isolâtes following spraydrying. Transmission électron microscopy of the microstructure of buttermilks revealed extremely heterogeneous microstructures but failed to reveal any effect of the treatments. 5.3 Introduction Buttermilk is the liquid phase released during churning of cream in the process of butter making. This liquid phase contains most of the water-soluble components of cream. Milk proteins, lactose, minerai and some lipids are recovered in buttermilk. The milk fat globule membrane (MFGM) fragments released during the disruption of fat globules are also recovered in buttermilk. MFGM is composed mainly of proteins, phospholipids and minerais (Walstra, Wouters, & Geurts, 2006). The MFGM fragments are of particular importance considering the various health-related properties described for its components (Spitsberg, 2005). For example, it was recently reported that MFGM fractions obtained from buttermilk and whey buttermilk hâve an anti-viral effect on rotaviruses strains (Ochonicky, Donovan, Kuhlenschmidt, Jiménez-Flores, & Kuhlenschmidt, 2005). Phospholipids hâve also been reported as having potential physiological effects on brain health (Kidd, 2000), cholestérol binding in vivo (Noh & Koo, 2004), stress management

93 75 (Ruenberg, 2002) and inhibition of tumour growth (Schmelz, Sullards, Dillehay, & Merrill, 2000). Industrial treatments are known to hâve a major impact on MFGMs (van Boekel & Walstra, 1995) Heat is arguably the single most important factor affecting MFGMs. Adsorption of copper from milk plasma, aggregation of MFGM proteins, loss of MFGM proteins and phospholipids, and adsorption of caseins and whey proteins on the surface of MFGMs hâve been reported (Evers, 2004). Interactions between whey proteins and MFGMs hâve also been reported and are believed to be partly caused by sulphydryldisulphide interactions (Dalgleish & Banks, 1991; Kim & Jiménez-Flores, 1995; Lee & Sherbon, 2002; Ye, Singh, Taylor, & Anema, 2002). Houlihan et al. (1992) found both P- lactoglobulin (P-LG) and a-lactalbumin (ct-la) bound to MFGMs after heating whole milk at 80 C. Ye, Singh, James and Anema (2004) proposed that interactions between P- LG and MFGM proteins were température dépendent. Thermal denaturation of both P- LG and MFGM proteins results in disulphide linkage of MFGM aggregates and P-LG complexes. In another study, Lee & Sherbon (2002) reported that MFGMs in milk heated for 3 min at 80 C contain approximately 0.03 g of P-LG and g of a-la per 100 g of fat globules. Thèse amounts increased with increased heating time. Lee & Sherbon (2002) also reported a loss of about 20% of lipids when MFGMs are heated. However, the authors did not report the composition of the lipids lost during the heating process. It is believed that the migration of lipids from MFGMs to the sérum only oecurs in présence of sérum components (Houlihan, Goddard, Kitchen, & Masters, 1992). To date, ail the data collected on the effects of heat treatments on MFGMs hâve been obtained after heat treatment of whole raw milk, which implies that the milk fat globules were structurally intact during heating. However, in the case of buttermilk, MFGMs are not globular but rather sheet-like (Corredig & Dalgleish, 1997a). MFGM components in both the inner and outer layers of the membrane are exposed to buttermilk sérum. The interactions between components might be drastically différent from that of whole milk

94 76 fat globules. At the industrial scale, buttermilk is often subjected to severe conditions (long holding time before evaporation and spray-drying, higher pasteurization températures for sanitary reasons), which are likely to induce changes in buttermilk microstructure. Buttermilk can be considered as a concentrated source of MFGM components. A number of studies aimed at concentrating or isolating MFGMs from buttermilk hâve been reported (Jiménez-Flores & Morin, 2005; Morin, Jiménez-Flores, & Pouliot, 2004; Sachdeva & Buchheim, 1997; Surel & Famelart, 1995). Major variations in fractionation yields hâve been observed with buttermilks that had been obtained from différent sources or had undergone différent industrial treatments (Astaire, Ward, German, & Jiménez- Flores, 2003; Morin, Jiménez-Flores, & Pouliot, 2004). Morin, Jimenez-Flores & Pouliot (2004) used MF membranes to separate phospholipids from MFGMs and showed that the passage of phospholipids through the MF membranes was greatly affected by filtration température (Morin, Jiménez-Flores, & Pouliot, 2004). In the same study, using the same filtration procédure, the authors found that phospholipid transmission was reduced by 50% when fresh pasteurized buttermilk as opposed to reconstituted powdered buttermilk (Morin, Jiménez-Flores, & Pouliot, 2004). The major différence between the products was the processing history, indicating that processing steps may hâve a major impact on buttermilk phospholipids and MFGMs. Little is known about the composition and microstructure of buttermilk and MFGMs, which might explain the limited success of MFGM fractionation by MF. Understanding the changes that occur in buttermilk composition and microstructure as a fonction of processing history could help to improve MF fractionation and enhance our understanding of the properties of buttermilk as a functional ingrédient. The présent study was aimed at investigating the effects of pasteurization of cream and the buttermilk processing steps, namely pasteurization, evaporation and spray-drying, on buttermilk

95 77 composition and microstructure as well as on the composition of MFGMs isolated from buttermilk. 5.4 Materials and methods Processing conditions The expérimental procédure used for this study is summarized in Figure 5.1. Fresh raw cream (110 L) was obtained from a local dairy (Natrel, QC, Canada) and divided into two batches on réception. One batch was pasteurized at 85 C for 20 s using a tubular pasteurizer (Actini, Evian-Les-Bains, France). The pasteurized cream and raw cream were then stored at 10 C overnight for maturation. Both cream samples were subsequently churned in a rotary churn at 26 rpm and 13 C. The raw and pasteurized creams broke down within an average of 23 and 27 min respectively. The buttermilks were recovered in milk cans after separating the butter fines using a stainless steel fïlter. Residual lipids from both buttermilks were removed by centrifugation using a milk separator (Alfa-Laval, Lund, Sweden) running at 6000 rpm and 18 C. Sodium azide (0.02% (w/v)) (Fisher Scientific, Nepean, ON, Canada) was added as a preservative and samples were withdrawn for analysis. Both buttermilks were pasteurized at 72 C for 20 s using a tubular pasteurizer (Actini) and samples were collected. Both pasteurized buttermilks were then evaporated using a falling-flow evaporator at 60 C until 20% solids was reached as measured using a handheld refractometer. Before evaporation, 30 ppm of anti-foaming agent (Dow Corning, Varennes, QC, Canada) was added and samples of concentrated buttermilk were collected for analysis. Lastly, the concentrated buttermilks were spray-dried using a pilot plant spray-dryer (Niro A/S, Hudson, WI, USA). Inlet air température was set at 195 C and outlet température was set at 85 C. Samples of the powders were collected for analysis. The results are the averages of two replicates.

96 7S Raw créa m Churning Séparation Pasteurization Evaporation Spray-dry ing Butter Butter Pasteurization Churning Séparation Pasteurization Evaporation Spray-drying Figure 5.1 Diagram of buttermilk processing Isolation of MFGMs MFGMs were isolated from buttermilk samples (350 ml) using a slightly modifïed version of the procédure of Corredig & Dalgleish (1997a). Sodium citrate was used to dissociate casein micelles and the MFGM fragments were collected in the pellet by centrifugation. Briefly, sodium citrate (2% w/v) was added to buttermilks from raw and pasteurized creams. The buttermilks were stored at 4 C overnight for maximum micelle dissociation. They were then centrifuged at 50,000 x g at 4 C for 2h. The pellets were collected on Whatman #1 filter paper and were rinsed with 25 ml of deionized water. The pellets were then re-suspended in 100 ml of deionized water using a bench-top homogenizer (Polytron PT 3100, Brinkman, NY, USA). The suspensions were

97 subsequently freeze-dried and stored at -20 C. The same procédure was followed for concentrated samples, except that the samples were suspended to 10% total solids using deionized water and were equilibrated overnight at 4 C before isolating the MFGMs Analytical procédures Compositional analyses Buttermilks and MFGM isolâtes were analyzed for total protein using a nitrogen analyzer (Leco FP-528, Leco Corp., St. Joseph, MI, USA) and a protein conversion factor of Total solids were obtained by microwave dying (Smart System 5, CEM Corp., Matthews, NC, USA) and ash was measured by incinération in a muffle furnace at 550 C for 20 h. Lipids were extracted using the Mojonnier ether extraction procédure, and lipid extracts were diluted to 5 mg/ml in 2:1 chlorofornrmethanol and stored at -20 C until further analysis. The lipid profiles of the buttermilks was obtained by HPLC-ELSD as described previously (Morin, Jiménez-Flores, & Pouliot, 2004). MFGM isolate protein profiles were obtained by SDS-PAGE (Laemmli, 1970) and protein class distribution was estimated by densitometry (Gel Doc XR, Bio-Rad, Hercules, CA, USA). AU reagents were HPLC grade Transmission électron microscopy (TEM) The structure of buttermilk samples was observed by TEM using the procédure described by Corredig & Dalgleish. (1997a). Briefly, samples of buttermilk were centrifuged at 15 C at 50,000 x g for 2 h. The pellets were recovered and fixed in glutaraldehyde and post-fixed in osmium tetraoxide. The samples were then dehydrated in a graded ethanol baths and embedded in epoxy resin. The thin sections were stained using lead citrate. The transmission électron microscope (JEOL 1230, JEOL products, Peabody, MA, USA) was

98 80 used at 80 KV. Micrographs shown are représentative of views of two différent observation grids for each sample of each replicate Statistical analysis Statistical analyses were performed on the two replicates using Minitab 14.0 software (State Collège, PA). A gênerai linear model (GLM) procédure was used to analyze the effect of cream pasteurization and the processing steps (séparation, pasteurization, evaporation and spray-drying) on the buttermilk and the MFGMs isolated from the buttermilk. When a significant processing effect (p < 0.05) was found, a one way- ANOVA with Tuckey's comparison test was carried out on the means to evaluate the effect of each processing step. 5.5 Results and discussion Effect of pasteurizing cream on the composition of buttermilk The composition of the différent buttermilks is reported in Table 5.1. The protein, lipid and ash contents of the buttermilks were in the ranges of previously reported data (Astaire, Ward, German, & Jiménez-Flores, 2003; Mistry, Metzger, & Maubois, 1996). However, the total lipid content of buttermilk from pasteurized cream was higher than reported values. The main compositional différence observed between buttermilks from raw and pasteurized cream was the total fat content. Buttermilk from pasteurized cream contained twice as much fat as buttermilk from raw cream. Given the fact that the same skimming conditions were used for both creams, the higher fat content in the buttermilk from pasteurized cream was somewhat unexpected. This observation may be related to the fact that pasteurization of cream induces the déposition of caseins and whey proteins

99 SI at the surface of fat globules and therefore limits the coalescence of the globules (Dalgleish & Banks, 1991; Houlihan, Goddard, Kitchen, & Masters, 1992; Ye, Singh, Taylor, & Anema, 2004). This déposition might be induced either by the heat treatment itself or by partial homogenization in the pasteurization equipment. This explanation is supported by the longer churning time required with pasteurized cream (27 min vs. 23 min). Small, undisrupted fat globules would therefore be recovered in the buttermilk. It is also known that pasteurization of cream at températures higher than 78 C leads to complète inactivation of agglutinins (Walstra, Wouters, & Geurts, 2006). Therefore, the high fat content in the pasteurized cream buttermilk is likely to be much harder to skim considering that inactivated agglutinin cannot induce creaming. Pasteurization resulted in significantly lower amounts of proteins (p < 0.01), ash (p < 0.01) and phospholipids (p < 0.01) (Table 5.2) in buttermilk. Thèse lower values resulted from the higher total fat content of the pasteurized cream buttermilk since ail the data are reported as a proportion of total solids. Indeed, no significant différence was observed between samples when the data were compared on a non fat basis (not shown). Table 5.1 Composition of buttermilk after each processing step % Dry Matter Proteins Lipids Ash Cream type No heat 85 C/20 s No heat 85 C/20 s No heat 85 C/20 s Séparation Pasteurization Evaporation Spray-drying

100 Table 5.2 Phospholipid composition of buttermilk after each processing step Cream type Séparation Pasteurization Evaporation Spray-drying Total No heat 1.49 a 1.42 a 1.54 a 0.92 b Phospholipids dry matter) 85 C/ 20s 1.28 a 1.19 a 1.18 a 0.76 b No heat PE 85 C/ 20 s PC No 85 a heat 20 s A % Total 1Phospholipids No heat SM 85 C/ 20 s No heat PS 85 C/ 20 s PE: phosphatidylethanolamine, PS: phosphatidylserine, PI: phosphatidylinositol, PC: phosphatidylcholine, SM; sphingomyelin. a Means with différent superscripts in the same column are significantly différent (p<0.05) No heat PI 85 C/ 20 s

101 5.5.2 Effect of proeessing steps on the composition of buttermilk The compositional data reported in Table 5.1 provide évidence that the effect of the proeessing steps on the composition of buttermilk was minimal for total proteins, lipids and ash. The phospholipid content and profile of the buttermilks at each proeessing step are shown in Table 5.2. The phospholipid content decreased significantly for both buttermilks after spray-drying. In the case of raw cream buttermilk the relative decrease compared to skimmed buttermilk was 38.2% (from 1.49% to 0.92%) and 40.6 % (from 1.28 % to 0.76 %) for pasteurized cream buttermilk. Most phospholipids in buttermilk are contained in MFGM fragments. A loss of ail major phospholipids would thus be an indication of a loss of whole MFGM fragments. We observed that the proportion of phosphatidylethanolamine (PE) decreased during spray-drying. Phosphatidylserine (PS) and phosphatidylinositol (PI) also decreased somewhat but to a much smaller extent while the relative distribution of phosphatidyleholine (PC) and sphingomyelin (SM) increased. The decrease in phospholipid content seemed to be attributable mainly to PE since PS and PI accounted for only a very small proportion of total buttermilk phospholipids. According to the most récent structural model of MFGMs, the three phospholipids that decreased in our trials are located in the inner leaflet of the membrane and are in contact with the monolayer of phospholipids surrounding the triglycéride core (Figure 5.2) (Danthine, Blecker, Paquot, Innocente, & Deroanne, 2000). In milk, this portion of the membrane is buried inside the fat globule, thus preventing interactions with sérum proteins and caseins. However, in the case of buttermilk, the inner leaflet of the membrane is exposed to the sérum and proeessing can induce interactions and possible dislocation of the membrane. The fate of the phospholipids lost from buttermilk after spray-drying is unclear. The observed decrease suggests that proeessing modified the aggregation of the phospholipids, resulting in decreased solubility in the extraction solvents. It has been reported that MFGM proteins are very reactive at relatively low

102 températures (60 C) (Ye, Singh, Taylor, & Anema, 2002) and that heating at 80 C causes a total loss of PAS 6/7 from MFGM (Lee & Sherbon, 2002). PAS 6/7 is an MFGM protein. Phospholipid monolayer Phospholipid bilayer-inner leaflet (PE, PS and PI) MFGM Phospholipid bilayer-outer leaflet (PC and SM) Figure 5.2 Schematic représentation of MFGM. Lipids: TG: tryglycerides, PE: phosphatidylethanolamine, PS: phosphatidylserine, PI: phosphatidylinositol, PC: phosphatidylcholine, SM: sphingomyelin. Proteins: XO: xanthine oxidase, PAS 6/7: periodic acid shift 6/7, BTN: butyrophilin. Adapted from Danthine et al.(2000)) In human milk fat, PAS 6/7 has been identified as lactadherin and is believed to be involved in phospholipid binding (Fortunato et al., 2003). Heat treatment combined with the rapid water removal and increased ionic strength created by spray-drying could induce the formation of complexes between milk proteins, MFGM proteins and phospholipids, rendering the phospholipids non-extractable by the Mojonnier procédure, thus causing an apparent decrease in phospholipids. Whether the observed decrease in phospholipid content is caused by interactions of spécifie phospholipids with sérum material or with MFGM fragments that render extraction more difficult requires further

103 85 investigation. Phospholipid extraction using ether, as with the Mojonnier method, is a standard method for dairy products and has been used in other buttermilk fractionation studies (Astaire, Ward, German, & Jiménez-Flores, 2003; Jiménez-Flores & Morin, 2005; Morin, Jiménez-Flores, & Pouliot, 2004; Sachdeva & Buchheim, 1997). However, in light of our results, spécial attention should be paid to lipid extraction procédures in order to verify the effectiveness of the method used for extracting buttermilk phospholipids after the différent processing steps Effect of cream pasteurization on the composition of MFGM isolâtes The compositions of the MFGM isolâtes are shown in Table 5.3 while the protein and phospholipid compositions of the isolâtes are shown in Tables 5.4 and 5.5 respectively. Pasteurization of the cream significantly increased (p < 0.01) protein recovery in the MFGM isolâtes from buttermilk. Protein profiling by SDS-PAGE (Figure 5.3 and Table 5.4) showed that whey proteins in the MFGM isolâtes accounted for this increase in protein content. Pasteurization of the cream resulted in an almost 3-fold increase in the amount of whey proteins bound to the MFGMs in the resulting buttermilk compared to buttermilk produced from raw cream (20.4 vs. 7.34%). The heat treatment of the cream resulted in high incorporation of P-LG in the MFGM isolate (Figure 5.3, lane 1 vs. lane 5) and P-LG appears to be the main protein in the MFGM isolâtes from pasteurized cream as judged by the staining intensity in the SDS-PAGE gel. This suggests that any approach used to isolate or concentrate MFGMs by MF or other physical séparation techniques using pasteurized cream buttermilk as a substrate will necessarily yield an MFGM isolate or concentrate with a high concentration of associated P-LG. The interaction of P-LG with MFGM proteins when heated has been previously reported (Kim & Jiménez-Flores, 1995; Ye, Anema, & Singh, 2004; Ye, Singh, Taylor, & Anema, 2002; Ye, Singh, Taylor, & Anema, 2004), and is mainly caused by disulphide bonding. Heat treatment of the cream also results in significantly less ash being recovered from MFGMs. MFGM powder from raw cream is reddish-brown as compared to yellow-white

104 86 from pasteurized cream. This has also been reported by Corredig & Dalgleish (1998b) and is believed to be caused by a decrease in the amount of iron in the MFGMs, which may also be partly responsible for the différence in the amount of total ash. Table 5.3 Composition of MFGM isolâtes from buttermilk Cream type Séparation Pasteur hation Evaporation Spray-drying Proteins No heat 85 C/20s % Dry Matter Lipids No heat 85 C/20s No Ash heat 85 C/20s Table 5.4 Protein composition of MFGM isolâtes from buttermilk Cream type Séparation Pasteurization Evaporation Spray-drying Casein No heat 85 C/20s » Protein in Dry Matter* Whey No heat 85 C/20 s No 7.35" 13.7 bc 12.7 b 15.6 e MFGM heat 85 C/20 s * Protein (casein, whey and MFGM) distribution estimated by Coomassie blue staining intensity in SDS- PAGE gels. a Means with différent superscripts in the same column are significantly différent (/?<0.05)

105 Table 5.5 Phospholipid composition of MFGM isolâtes from buttermilk Total Phospholipids dry matter) 85 C/ 20 s 6.32 a % i Total Phospholipids PE PC SM PS PI No No 85 C/ No 85 C/ No 85 C/ No 85 C/ No 85 C/ Cream type Séparation Pasteurization Evaporation Spray-drying heat a 5.78 ab 4.43 b heat s heat 28.8 a 27.0 a 27.8 a 35.0 b 20 s heat s heat s 0.00' heat s PE: phosphatidylethanolamine, PS: phosphatidylserine, PI: phosphatidylinositol, PC: phosphatidylcholine, SM: sphingomyelin = PS peak not detected by HPLC-ELSD. a Means with différent superscripts in the same column are significantly différent (p<0.05)

106 MW 1 I I i I 1 i! i MFGM proteins Caseins 20 7 m Wliey proteins Figure 5.3 SDS-PAGE of MFGM isolâtes from raw cream buttermilk after séparation (1), pasteurization (2), evaporation (3) and spray-drying (4) and pasteurized cream buttermilk after séparation (5), pasteurization (6), evaporation (7) and spray-drying (8). MW: molecular weight standard Effect of buttermilk processing steps on the composition of MFGM isolâtes The processing steps of buttermilk resulted in no signifïcant changes in the total protein, lipid and ash composition of the MFGM isolâtes (Table 5.3). However, buttermilk pasteurization induced a signifïcant increase (p < 0.01) in whey protein content in the MFGMs isolated from raw cream buttermilk (Table 5.4) but the amount never reached that in the MFGMs isolated from pasteurized cream. The intensity of the heat treatment of the cream was more important (85 C/20 s) than any of the subséquent processing steps. Pasteurization of the cream may therefore resuit in more interactions between MFGMs and whey proteins. This resuit is in accordance with the findings of Corredig & Dalgleish (1998b), who reported that heat treatment of cream is a critical step in

107 determining the composition and functional properties of MFGM proteins recovered from buttermilk. Results also showed a slight decrease in the MFGM protein content of the isolâtes from both buttermilks. This decrease was not statistically significant given that only two replicates were performed and that the quantification was based on densitometric analyses of SDS-PAGE gels. Incorporation of whey proteins could simply hâve decrease the proportion of MFGM proteins in dry basis in the isolâtes. A high proportion of caseins was recovered in the MFGM isolâtes, accounting for 9.5% to 14.2% of the dry matter (Table 5.4). The origin of thèse caseins may hâve been incidental (a resuit of the isolation procédure) or caused by variables inhérent to processing. The method used for MFGM isolation employs centrifugation to recover the MFGMs. Citrate (2%), which acts as a calcium chelating agent, is added to break down the casein micelle structure in order to retain the caseins in the sérum (Corredig & Dalgleish, 1998b). However, MFGMs may capture and drag some of the casein with them into the pellet. The high proportion of casein in the MFGM fraction suggests that the same problem may occur with membrane filtration of buttermilk and thus explain the poor séparation of MFGMs and caseins using this procédure. It is well known that the homogenization of milk results in the incorporation of casein micelles in MFGMs (Walstra, Wouters, & Geurts, 2006). During churning of the cream, partial homogenization may occur, causing caseins to interact with MFGMs and to be recovered in the MFGM fraction. The casein présent in MFGMs isolated from whole milk is mostly the resuit of mechanical treatments rather than heating (Dalgleish & Banks, 1991). This might also be the case for buttermilk. Even after extensive washing, skim milk proteins like P-casein and ot-la can still be detected in rather large proportions (13% of total proteins of the fat globules) in human fat globules using. a proteomic approach (Fortunato et al., 2003). This further supports the fact that skim milk proteins may interact strongly with MFGMs even before milk is collected.

108 90 A decrease in the total phospholipid content of MFGM isolâtes also occurred (Table 6.5) after spray-drying. This was mostly characterized by a decrease in PE, which was also observed in the buttermilks. It is thus likely that the decrease was mainly causée! by changes in the MFGM fragments rather than in free phospholipids in the buttermilk. Conflicting results on the effect of heating on phospholipids from whole milk fat globules hâve been reported (Evers, 2004). No data is available to compare the effect of heating on phospholipids in ruptured milk fat globules Effect of proeessing on the microstructure of buttermilk In an attempt to correlate the observed compositional changes in buttermilk with changes in buttermilk microstructure, transmission électron microscopy (TEM) was used to examine buttermilk samples after the différent proeessing steps. TEM micrographs are shown in Figure 5.4. The size of the MFGM fragments was highly variable, ranging from 0.1 im to 2-3 [am, and should be related to the size of the initial fat globule before churning as well as to the strength of the bonds between the MFGM molécules. Membrane fragments were also found to be flexible, as a large proportion of the fragments appeared folded and deformed. Based solely on the TEM observations, it is not possible to détermine whether the heat treatment of the cream or the proeessing steps to produce the buttermilk had any effect on the microstructure of the MFGM fragments. However, some rather intriguing observations were made. Figure 5.4c (buttermilk) shows a fragment of MFGM that seemed to affect the dispersion of the other components of the buttermilk. There is a void on one side of the membrane but not on the other. Patton & Keenan (1975) reported that isolating MFGMs at low températures induces the adsorption of high-melting triglycérides to the interior face of the MFGMs. Therefore, the void observed in the TEM micrographs could be a layer of lipids that remains unstained during TEM sample préparation using lead citrate.

109 91 ; Raw cream buttermilk Pasteurized cream buttermilk # E.-3BB ï#* Casein micellesfe * MFGM «HP, < : Vf ^^», a. ' >? ï! «"t**:j»... ;i «V-.» - i< ^i **.;* ^JOI Figure 5.4 Transmission électron micrographs of initial buttermilk (A, E), pasteurized buttermilk (B, F), evaporated buttermilk (C, G) and spray-dried buttermilk (D, H). Bars = 0.2 u.m

110 92 The TEM micrographs also show that MFGM fragments can fold and trap casein micelles inside a reconstituted globule (Figures 5.4 d, e and f). While it might appear that the caseins were simply physically trapped by the MFGM fragments, this observation indicates that MFGMs can interact with casein micelles. In the context of buttermilk fractionation, thèse interactions are important for concentrating MFGMs and for buttermilk functional properties. The poor séparation of MFGMs and casein micelles during MF of buttermilk may be partially due to such interactions. Moreover, several studies hâve reported that buttermilk perforais poorly during rennet coagulation for cheese making (Sachdeva & Buchheim, 1997; Turcot, Turgeon, & St-Gelais, 2001). The interactions of MFGM fragments with casein micelles might prevent gel formation and create a weak coagulum. The observed loss of phospholipids after spray-drying of the buttermilks could not be correlated with the TEM observations of the buttermilks. Either the loss of phospholipids did not affect the structure of the MFGM fragments or the changes were too subtle to be clearly seen in the TEM micrographs since the phospholipids may simply hâve become unextractable. 5.6 Conclusion Pasteurization of cream appeared to be the critical step in the modification of MFGM composition, especially in terms of total lipid content. Spray-drying seemed to hâve a strong influence on the amount and distribution of phospholipids recovered in buttermilk. The exact nature of the observed changes in phospholipid content is not known but might reflect modifications in the structure of MFGMs. Although modifications in MFGM structure were not clear based upon direct TEM observations of buttermilk, the influence of MFGM fragments on the dispersion and interactions with other buttermilk solids was highlighted. The présent study provides évidence that the processing of buttermilk can cause major changes in both buttermilk and MFGM fragments. The TEM micrographs support the view that the heterogeneous microstructure of buttermilk explains the unsatisfactory results of MF fractionation of buttermilk.

111 Acknowledgements We would like to thank Dr. Michel Britten (Food Research and Development Centre, St- Hyacinthe, QC) for the phospholipid analysis and critical review of this manuscript and R. Janvier for his assistance with the TEM analyses. This work was funded by the Natural Science and Engineering Research Council of Canada (NSERC).

112 CHAPTER 6 A Comparative study of the fractionation of regular buttermilk and whey buttermilk by microfiltration Based upon previous work indicating that casein micelles might restrict the concentration of MFGM during buttermilk MF, collaboration was established with Hilmar Chesse Co. (Hilmar, CA) in order to study the churning of whey cream free of caseins resulting therefore in whey buttermilk and study the MF of the latter product. This chapter is to our knowledge, the fîrst to address the fractionation of buttermilk obtained from churning of whey cream. The work contained in this chapter was undertaken at Dairy Products Technology Center of California Polytechnic State University in San Luis Obispo, California. Work was supervised by Dr Rafaël Jiménez-Flores during a one year research fellowship in his laboratory. Writing of this manuscript was supervised by Dr Rafaël Jiménez-Flores and Dr Yves Pouliot who also are co authors. Work was also carried out with the help of an undergraduate student, Salvador Uson. Results obtained in this part of the project hâve been published essentially in this form in Journal of Food Engineering. Results were also presented in poster form at the 2005 IDF World Dairy Summit in Vancouver, British- Columbia, Canada.

113 Résumé Le babeurre doux (issue de crème douce) et le babeurre de lactosérum (issu de crème de lactosérum) sont deux importants sous-produits laitiers qui sont que très peu étudiés et donc mal caractérisés. Ces deux produits sont cependant uniques car ils contiennent une teneur élevée en lipides mineurs (phospholipides) d'intérêt nutraceutique. La MF sur membrane de céramique (0.45 um) pour le fractionnement du babeurre doux et du babeurre de lactosérum a été étudiée et ce, en deux modes de filtration soit la concentration volumique et la diafiltration continue. Il a été observé que la composition en phospholipides du babeurre de lactosérum était similaire à celle du babeurre doux (0.1 versus 0.14% respectivement) mais que sa teneur en protéines était significativement plus basse dû à l'absence de caséines. Cependant, une variabilité importante de la teneur en lipides de la crème de lactosérum a été observée donnant lieu à des variations de teneurs en lipides dans le babeurre de lactosérum. En MF, le babeurre de lactosérum a induit des flux de perméation plus élevés que le babeurre doux et ce, autant en concentration volumique qu'en diafiltration continue. Le calcul des taux de transmission des composés au travers de la membrane a démontré que dans les conditions testées, les protéines traversaient la membrane à un taux de 20 à 30% alors que les lipides étaient transmis à des taux avoisinants les 3 à 8%. Une augmentation de la teneur en phospholipides par rapport au babeurre initial a été observée dans le rétentat de MF et ce pour les deux types de babeurre. L'efficacité générale du procédé a été limitée par la présence de micelles de caséines retenues par la membrane dans le cas du babeurre doux. L'analyse des profils protéiques des perméats et rétentats de MF a démontré que la transmission des protéines associées à la membrane du globule de gras est inférieure dans le cas du babeurre de lactosérum suggérant des différences structurales entre les deux produits ou encore l'influence des micelles de caséines dans la séparation en MF.

114 Abstract The use of a ceramic MF membrane for the fractionation of buttermilk and whey buttermilk obtained from pilot scale churning of cream and whey cream from industrial sources has been studied. Whey buttermilk contained comparable amounts of phospholipids compared to regular buttermilk but its protein content was lower due to the absence of caseins. However, it was found that lipid content of whey cream did vary significantly between lots resulting in important variations in the fat content of whey buttermilk. A two-fold MF concentration of regular buttermilk doubled its phospholipids content whereas that of whey buttermilk was increased by 50%. The overall effïciency of the MF processing of regular buttermilk was limited by the amount of caseins retained by the MF membrane. Analysis of the protein profile of permeates and retentates showed that the transmission of milk fat globule membrane proteins by the MF membrane was lower when using whey buttermilk as compared to regular buttermilk possibly indicating the influence of casein micelles in fractionation or some structural différences between both products. 6.3 Introduction Buttermilk is a by-product from butter manufacture that finds applications in various food products. Uses of buttermilk in food Systems are closely related to its particular composition in emulsifying components such as phospholipids which can act as emulsifiers in salad dressings. However, in most applications, buttermilk is used because of its typical flavor such as in baked goods (O'Connell & Fox, 2000). Interest is growing on that particular by product because of its unique composition (Astaire, Ward, German, & Jiménez-Flores, 2003; Corredig & Dalgleish, 1997a; Corredig, Roesch, & Dalgleish, 2003; Morin, Jiménez-Flores, & Pouliot, 2004; Sachdeva & Buchheim, 1997). When milk fat globules are broken during churning of cream, the membrane covering the lipid

115 core is excluded from the lipid matrix and recovered in buttermilk along with most of the proteins, lactose and minerais contained in the aqueous phase of cream. The milk fat globule membrane (MFGM) is rich in various proteins and phospholipids which hâve some potential for both functional and nutraceutical applications. For example, it was shown that sphingomyelin (SM) could help in the prévention of various diseases including colon cancer (Schmelz, Sullards, Dillehay, & Merrill, 2000) and that phosphatidylcholine (PC) could interfère with the development of hepatic diseases (Niederau, Strohmeyer, Heintges, Peter, & Gopfert, 1998). Although mechanisms are not yet fully understood, there is growing need for fractions enriched in thèse components to further study their effect in human health but also their effect in food Systems. Various attempts hâve been made to fractionate buttermilk in order to create an enriched fraction in MFGM components. Surel and Famelart (1995) were the first to report the use of MF to fractionate buttermilk. The major issue they reported for the use of MF to fractionate buttermilk was the similar size of the casein micelles and the MFGM components présent in buttermilk. In order to overcome this problem, Sachdeva and Buchheim (1997) used renetting and acid coagulation of buttermilk to remove caseins prior to fractionation by a combination of MF and ultrafiltration (UF). The authors reported a recovery of 70-77% of the total phospholipids of buttermilk using this process. However, thèse results were highly dépendent of various coagulation factors. Corredig et al. (2003) hâve reported the use of citrate to disrupt the casein micelles followed by either UF or MF on polysulfone membranes. Their results showed that this approach seems to be effective to concentrate MFGM proteins in the retentate. However, their data did not show the efficiency of this process to recover other MFGM components including phospholipids. The high levels of citrate in the permeate limit the potential use of this process for the recovery of MFGM from buttermilk.

116 98 An approach using whey cream as a starting material in order to concentrate MFGM components could also be considered. Whey cream is obtained by séparation of fat from cheese whey by centrifugation. The whey cream obtained is mainly used to standardize milk fat prior to cheese making but can also be used to produce whey butter (Fox, Guinée, Cogan, & McSweeney, 2000) and thus, a by-product, whey buttermilk. Because the objective of the cheese maker is to maximize caseins recovery in the cheese, virtually no caseins can be found in cheese whey, therefore the buttermilk recovered from churning of whey cream is expected to contain very low amount of caseins. This is likely to enhance the membrane séparation of MFGM components from the other milk solids. There are currently no uses reported for buttermilk obtained from whey cream butter, thus MFGM isolation can be a promising way of to add value this by-product. The objective of the présent work was to compare MF of whey buttermilk and regular buttermilk using two différent filtration modes, namely volumetric concentration (VCMF) and diafiltration (DFMF). The effïciency of séparation was compared in terms of selectivity by assessing compositional différences of initial products and fractions (lipids, proteins, phospholipids), and in terms of productivity using MF permeation flux data. 6.4 Materials and methods Buttermilk production For each trial, 200 L of fresh regular manufacturing cream was purchased from Foster Farms (Modesto, CA) and 200 L of whey cream was graciously donated by Hilmar Cheese Co. (Hilmar, CA). Both cream were churned to butter using a continuous pilot scale butter churn (Egli, Switzerland). Buttermilk was recovered in milk cans after butter

117 99 fines were removed by filtration through cheese cloth. The buttermilk production was repeated 3 times with three différent lots of both creams Microfiltration Batehes of 22.7 kg (50 lbs) of fresh regular buttermilk were used for the VCMF and DFMF process. The same batch System was used for fresh whey buttermilk. The remaining buttermilk was spray dried using a Niro Filterlab Spray Dryer (Hudson, WI). The MF System used is described elsewhere (Astaire, Ward, German, & Jiménez-Flores, 2003). Two tubular membranes (0.7 m2 total surface) were fitted in parallel on the module for ail experiments. MF ceramic membranes (Tami Sunflower Design, 0.45jj.m pore size, Tami Industries, France) were fitted in US filters (Warrendale, PA) stainless steel housings. Ail runs were carried out at low température (8-10 C) at a transmembrane pressure of 80 to 95 kpa. The cross-flow rate was 87.1 L.min-1 (3.1 m.sec-1). The System was not equipped with uniform transmembrane pressure control. The VCMF process consisted in buttermilk concentration by MF until a volumetric concentration factor (VCF) of 2X was reached. The DFMF process consisted in continuously adding chilled tap water (4 C) to the retentate to replace the extraeted permeate until reaching a two-fold diafiltration (DF) factor (45.4 kg of water added). Samples of retentates and permeates for were collected for composition analysis when 25, 50, 75 and 100% of the VCF or DF was reached. Permeate flux (L.h-l.m-2) was also measured at fixed intervais during ail experiments. The final permeates and retentates from ail experiments were spray-dried. Ail MF trials runs were performed in triplicate Chemical analysis Cream, butter, liquid buttermilk, samples collected during MF, retentate and permeate powders and buttermilk powders were analyzed for total solids, total protein, total lipids,

118 100 ash content. Total solids were obtained using direct oven drying method in a forced air oven at 102 C and ash content was obtained by incinération at 550 C in a muffle furnace (Marshall, 1992). Protein was obtained using macro-kjeldahl method with 6.38 as a nitrogen to protein conversion factor (Marshall, 1992). Protein profile was established for the buttermilk powders, final retentate and permeate powders by SDS-Page according to Laemmli (1970) using a Mini-Protean III System (Biorad, Hercules, CA). Ail samples were diluted to 3.3 mg.ml-1 of protein with reducing sample buffer and 15 ul were loaded into a 12% acrylamide SDS-PAGE gel. Gels were stained with Coomassie Blue R-250 (Biorad, Hercules, CA) and proteins hâve been identified by their molecular weight according to Mather (2000). Lipids were obtained using the Mojonnier ether extraction method (Marshall, 1992). Extracted lipid were then diluted to 10 mg of lipids per ml with 1:2 chloroform-methanol and kept in a freezer (-20 ) until analysis. Phospholipids were analyzed by HPLC (System Gold, Beckman-Coulter, Mississauga, Ontario, Canada) with an electro evaporative light scattering detector (ELSD) (SEDEX 75, SEDERE, France) as described elsewhere (Morin, Jiménez-Flores, & Pouliot, 2004). Ail reagents were electrophoresis or HPLC grade Calculations Transmission (Tr) of proteins, lipids, phospholipids and ash through the membrane was calculated according to the équation 6.1 : Equation 6.1 Transmission calculation % Tr = (Cp/Cr)ioo where Cp is the concentration of a component in the permeate and Cr the concentration of the same component in the retentate (Cheryan, 1998). Concentration factor (CF) was also calculated for components using équation 6.2 where Cf is the final concentration reached in the retentate and Co the concentration in the initial feed.

119 Statistical analysis Equation 6.2 Concentration factor calculation CF= Cf/Co Ail statistical analysis was done by ANOVA with Tuckey's pairwise comparison using Minitab 14.0 software (Minitab Inc., PA). Results were considered statistically différent at p< Results and discussion Analytical data The composition of cream, butter and initial buttermilk samples are presented at Table 6.1. No significant différences were found between levels of lipids of both regular and whey cream. A significantly lower protein (p<0.01) and ash (p<0.05) content for the whey cream was observed. The ash content was also lower in the whey butter and whey buttermilk. The lower amount of caseins in whey cream explains thèse différences since caseins contains about 8% of minerais, mainly calcium phosphate (Walstra, Geurts, Noomen, Jellema, & van Boekel, 1999). A high variability in the total fat content of whey cream (32.34 ± 14.31%) was observed. Variations between 30 and 70% hâve been previously reported (Pointurier & Adda, 1969). This variability may origin from a number of processing characteristics such as the type of cheese that générâtes whey and by différences of operating conditions of the whey cream separator. This phenomenon induced a higher variability of total lipids content in the whey buttermilk but not in the

120 102 whey butter. Table 6.1 also shows that the phospholipids content of whey buttermilk was not significantly différent from that of regular buttermilk (p=0.627). However, considering the lower total solids in whey buttermilk, phospholipids levels would be representing a more important portion of the solids in whey buttermilk. The main phospholipids that were observed in regular and whey buttermilk were phosphatidylethanolamine (PE), phosphatidylcholine (PC), sphingomyelin (SM), phosphatidylinositol (PI) and phosphatidylserine (PS). The protein content of whey buttermilk was significantly lower (p<0.01) than that of regular buttermilk as explained by the absence of caseins in the whey buttermilk. As opposed to the lipid portion, the protein content of the whey buttermilk did not show any important variability (0.99 ± 0.02%) Microfîltration permeation flux data Permeation flux has been monitored during ail volumetric concentration and diafïltration runs. Figure 6.1 shows the average flux curve observed during the trials. Although average flux was significantly higher with whey buttermilk (p<0.05), the observed flux décline (Jo - Jf) for ail 4 experiments were not significantly différent. However, it was observed that VCMF induced a more important flux décline than DFMF while being not statistically significant (p=0.055). The lower average flux observed for the regular buttermilk could be caused by the accumulation of caseins micelles at the surface of the membrane creating a secondary layer (cake) inducing more résistance to permeation and therefore lower permeation flux (Vetier, Bennasar, Parodo de la fuente, & Nabias, 1986). This phenomenon could be occurring also with whey buttermilk but the évolution of the concentration polarization (CP) layer composition is likely to be différent considering the low amount of caseins. Another possible explanation for the higher flux obtained while using whey buttermilk is the différence in total solids in the initial feed (Table 6.1).

121 Table 6.1 Composition of initial products Cream Butter Buttermilk Whey Regular Whey Regular Whey Regular Total solids (%) ± ± ± ± ±0.92 a ±0.59 b Proteins (%) 0. 89±0.36 a 1.94±0.14 b 0.50 ± ± ± 0.02 a 2.93 ± 0.06 b Lipids (%) ± ± ± ± ± ±0.13 Phospholipids (%) nd* nd nd nd 0.10± ±0.04 PE PI (% of total) **PS nd 3. 8 PC SM Ash (%) ± 0.00 a 0.43 ± 0.03 b 0.08±0.01 a 0.14±0.01 b ± 0.09 a 0.66±0.02 b a Values in same line for each product with différent superscript differs significantly (p <0.05) not detected PE = phosphatidylethanolamine, PI = phosphatidylinositol, PS = Phosphatidylserine, PC = phosphatidylcholine, SM = sphingomyelin ± indicates standard déviation

122 J AVG =23.4 ± 5.2 B J AVG =25.8 ±2.8 J AVG =20.0 ± 4.9 J AVG =15.0 ± Time (rrin) Time (rrin) Figure 6.1 Average permeation flux curve during (a) VCMF and (b) DFMF of regular buttermilk ( ) and whey buttermilk ( ). J A VG = average flux ± standard déviation.

123 105 Thèse différences can be expected on the basis of the mass transfer mathematical model predicting flux in the région near the membrane (Cheryan, 1998): Equation 6.3 Permeation flux calculation in the mass transfer mathematical model J=kln(Cg/Cb) where J is the flux, k the mass transfer coefficient, Cg the gel concentration (concentration at the membrane surface) and Cb the bulk concentration (feed). The more important flux décline observed for VCMF compared to DFMF can also be explained and predicted by the mass transfer model (Cheryan, 1998). In the DFMF mode, the constant water addition in the DFMF process keeps the feed concentration unchanged or slightly reduced. In fact, DFMF induces slightly higher and more stable permeation fluxes. Surel (1993) reported flux values of around 120 L.h-l.m-2 with a 0.5 im pore size ceramic membrane in MF of regular buttermilk at 50 C. The flux values obtained in our experiments are much lower (~ L.rf'.rrf 2 ) and this is mainly due to the processing température. Our previous results (Morin, Jiménez-Flores, & Pouliot, 2004) indicated that in the case of reconstituted buttermilk, the cold MF process (7 C) induced permeation fluxes 5X lower than high température process (50 C). The same effect is also reported by Astaire et al. (2003). The main effect of température on the feed is on the fluid density, viscosity and most importantly on the diffusivity of solutés. For example Cheryan (1998) reported that diffusivity of proteins decreases of 3-3.5% per C decrease. The diffusivity has an important effect on the mass transfer coefficient (k) in équation 6.3.

124 106 The mass transfer coefficient is calculated with équation 6.4 where D is the diffusivity and S is the thickness of the boundary layer where the concentration gradient between the membrane and the bulk is found. The diffusivity (D) of the solutés in buttermilk below 10 C is much lower than at 50 C thus; the mass transfer coefficient (k) is lower which induce lower flux (J). Equation 6.4 Mass transfer coefficient calculation Transmission of components during MF Changes in transmission of proteins, total lipids and ash through the MF membrane were monitored at 4 différent points during each MF run (Figure 6.2). It can be observed that protein transmission decreases significantly throughout the DFMF process with both regular (p<0.01) and whey buttermilk (p<0.01). Protein transmission also decreased significantly in the VCMF process with whey buttermilk. This indieates the establishment of CP layer, possibly composed of proteins which could induce the formation of a secondary layer at the membrane's vicinity. A significant decrease of transmission of ash was observed in the DFMF of regular buttermilk (p<0.01). This is attributed to the low transmission of casein micelles since this resuit is not observed in the case of whey buttermilk where minerais are mostly free in solution. Transmission of lipids through the 0.45 um MF membrane was stable and low (<10%) throughout ail experiments with both whey and regular buttermilk. Phospholipids transmission (not shown) was also followed during the whole process but no clear trend could be observed in the case of whey buttermilk as transmission values were highly variable throughout the process. A possible reason for that is the fact that whey cream contains important portion of aggregates heterogeneous in size. This mixture of globules may rupture into a wide range of

125 107 membrane fragments sizes in the churning process. Furthermore, phospholipids found in whey buttermilk could be originating from the starter used in the cheese process (Umemoto & Sato, 1973) and thèse phospholipids could be on a completely différent aggregated form. The phospholipids transmissions in the case of regular buttermilk (also not shown) did vary to a lesser extent. The membrane fragments in both products could also be drastically influenced by the pumping and the high shear forces at the surface of the membrane : t * * 0.00 : CF DF 1.5 Figure 6.2 Transmission of lipids ( ), proteins (A) and ash ( ) through the VCMF (a) and DFMF (b) process as function of the VCF or DF reached for regular buttermilk ( ) and whey buttermilk ( ). * Indicates signifïcant (p < 0.05) transmission decrease throughout the process.

126 108 The average transmission of components has been calculated and is shown in Table 6.2. Significant effect of the MF mode on lipid transmission was observed (p<0.01) and the effect of the buttermilk type was also significant on lipid transmission (p<0.01). The DFMF process induced a higher transmission of lipids as compared to VCMF process. Since higher and more stable fluxes were observed for DFMF (Figure 6.1), lower fouling of the MF membrane allowed higher transmission of lipid. When using whey buttermilk, lipids were transmitted to a lower level as compared to regular buttermilk. Although différences being smaller, they remained highly significant (p<0.01) and the main reason for this could be a différent type of aggregation of lipids in the case whey buttermilk but also the présence of casein micelles in the CP layer in the case of regular buttermilk. Also it is known that the fat recovered in the whey cream is partly composed of free lipids (Fox, Guinée, Cogan, & McSweeney, 2000) and thèse may hâve a greater tendency to aggregate with time and maybe form vesicles in combination with MFGM fragments. Moreover, from the lipid profile of regular buttermilk, others hâve found that buttermilk contained around 74% of long chain (> 14 carbons) fatty acids (Scott, Duncan, Sumner, & Waterman, 2003) whereas this number was close to 90% for commercial whey (Boyd, Drye, & Hansen, 1999). Surel (1993) has shown that long chain fatty are better retained by MF (0.2 j,m) as compared to short chain fatty acids. The slight différence in transmission of lipids in the regular and whey buttermilk could be related to their relative content of long-chain fatty acids. The présence of casein micelles in the CP layer during MF of regular buttermilk could induce a more porous layer therefore inducing lipids to permeate more easily. Transmission of lipids in the order of 60% has been reported with regular buttermilk using a 0.5 um ceramic membrane at 50 C (Surel & Famelart, 1995). The low température used in this experiment might hâve prevented transmission of lipids through the membrane which is in agreements with others (Astaire, Ward, German, & Jiménez-Flores, 2003). Previous work showed that transmission of lipids was slightly higher at low température (Morin, Jiménez-Flores, & Pouliot, 2004) but this was not observed in the présent work. A possible explanation is the fact that the initial buttermilk used in this trial was produced in the pilot plant and excess of lipids was not removed by centrifugation like it would normally happen when industrial buttermilk is used, as it was with our previous work. Increasing the amount of lipids increase the propensity to

127 109 aggregation (Walstra, Geurts, Noomen, Jellema, & van Boekel, 1999) and creating therefore vesicles or particles that are rejected by the membrane. Despite transmission of phospholipids did vary throughout the process and no clear trend could be observed for both products, différences were noted according to filtration mode in total phospholipids for regular buttermilk (Table 6.2). The DFMF process using regular buttermilk induced more phospholipids transmission as compared to the VCMF process. This resuit is an indication that the occurrence of CP layer and possibly the fouling of the membrane during VCMF could alter the ability of phospholipids to permeate through the MF membrane. No significant différences between phospholipids profiles (PE, PC, and SM) were observed in the four treatments, given that variations between the three replicates were observed. As the transmission of total phospholipids was increased, ail the main classes of phospholipids showed an increase. Therefore, phospholipids classes tend to permeate equally the membrane showing that MF, in the conditions tested, does not show specificity for any class of phospholipids but more on the total amount that permeates. Protein transmission has been found to be significantly affected by the MF mode (p<0.05) and by the buttermilk type (p<0.01). However, the most observable effect was the type of buttermilk. While using whey buttermilk, we achieved a higher transmission of proteins through the MF membrane, compared to regular buttermilk. By observing the SDS-PAGE profile of the permeates from ail experiments (Figure 6.3), it appears that the type of proteins permeating the MF membrane in the case of whey buttermilk are whey proteins and a non-identified protein (most likely IgG light chain) at around 29 Kda to a lesser extent (Farrell Jr et al., 2004). While MFGM proteins were detected in the permeates using regular buttermilk, almost no MFGM proteins can be observed in the permeates. This results shows that MF of whey buttermilk allowed concentration of MFGM components, at least based on the MFGM proteins. Furthermore, this resuit indicates that MFGM proteins might exist in a différent state in whey buttermilk has

128 110 compared with regular buttermilk. The différent processing steps of cheese making and whey cream séparation might induce significant changes in their structure which prevent them to permeate through a MF membrane of 0.45um. Protein profile of permeates and retentates from DFMF and VCMF for both buttermilks do not appear to be any différent, showing that the type of MF mode did not affect the protein composition of the permeates and retentates but rather the total protein content. The protein profile of the regular buttermilk permeates shows important amount of caseins but thèse caseins remains the principal proteins in the retentate suggesting that increasing the VCF or the DF could help to reduce the amount in the retentates. However, some MFGM proteins were observed in the permeates suggesting that important amount of thèse proteins could also be lost in the permeate if a higher VCF or DF is reached. Ash transmission was also significantly affected by both the type of MF (p<0.05) and buttermilk type (p<0.01). Diafiltration is known to induce more transmission of ash in milk UF (Cheryan, 1998) and this is also observed in MF of buttermilk in the présent work. The bound minerais to the casein micelles reduce the level of transmission of ash in the case of regular buttermilk. Furthermore, NaCl added to the curd during cheese making was possibly recovered in the whey cream and therefore in the whey buttermilk. This added amount of NaCl would not hâve been retained by the MF membrane and therefore increased the transmission rate of ash.

129 Table 6.2 Average transmission of components through the membrane Lipids Phospholipids Proteins Ash PE PC SM %Tr Whey buttermilk Regular buttermilk VCMF DFMF VCMF DFMF 3.02±0.96 a ± ab ± ± ± ± 4.79 e ± 7.64 e ± 2.03 ; l ± 5.18' 5.82 ±: 1.70 ab ± 2.70 a 8. 50: tl.56 b : t b ± ± ± ± ± ± ± ± ± ± ± 5.98 b b ± 2.79 a : ± 4.29 ab ± 2.84 a ± 9.66 a a Values in the same line with différent superscript differs significantly (p < 0.05). ± indicates standard déviation

130 112 MW 1 o -tn 12 XO BTN PAS 6/7 I il " CN's IgG light chain P-Lg Figure 63 SDS-PAGE (12%) profile of whey buttermilk (1), Whey buttermilk VCMF permeate (2), Whey buttermilk VCMF retentate (3), Whey buttermilk DFMF permeate (4), Whey buttermilk DFMF retentate (5), Regular buttermilk (6), Regular buttermilk VCMF permeate (7), Regular buttermilk VCMF retentate (8), Regular buttermilk DFMF permeate (9), Regular buttermilk DFMF retentate (10). MW: molecular weight marker, XO: xanthine oxidase, BTN: butyrophilin, PAS 6/7: periodic acid Schiff 6/7, CN: caseins, p-lg: beta-lactoglobulin. Proteins hâve been identified according to Mather (2000) Concentration factors Concentration factors of ail components in retentate were calculated in order to illustrate the overall effïciency of the MF séparations (Table 6.3). Concentration factor were calculated in dry basis by comparison of the concentration of the components in the final retentate powder as compared to the initial liquid buttermilk collected during the churning of the creams. The CF for both lipids and proteins was signifïcantly affected by the type of MF (p<0.01) but also by the type of buttermilk (p<0.05). The continuous dilution involved in DFMF allowed to "wash out" solutés and solids of size smaller than the MF membrane pores more effectively resulting in higher CF for retained components simply because they end up representing a higher proportion of the retentate total solids.

131 113 Table 6.3 Average concentration factors in the final retentate powders Whey buttermilk Regular buttermilk Final CF's Lipids Proteins Phospholipids Ash VCMF 1.49 ± 0.08 a 1.24± ± ±0.60 DFMF 2.31±0.25 b 1.62±0.13 b 1.51 ± ±0.42 VCMF 1.37± 0.08 a ± 0.09 ab 2.03 ± ±0.05 a Values in the same line with différent superscript differs signifïcantly (p < 0.05). ± indicates standard déviation DFMF 1.81±0.26 ab 1.83±0.13 b 2.25 ± ±0.06 Results show that while using regular buttermilk, the calculated final CF for proteins and lipids are very close. This is in agreement with many reports of problems in séparation of lipids and proteins because of the similar size of thèse two components (Morin, Jiménez- Flores, & Pouliot, 2004; Roesch & Corredig, 2002; Sachdeva & Buchheim, 1997; Surel & Famelart, 1995). However, while using whey buttermilk, this problem does not occur since CF of lipids is higher than that of protein. Moreover, the increase in the CF of proteins could be partly due to an increase in MFGM proteins bound to the retained phospholipids (Figure 6.3). The CF reached for phospholipids were definitely higher when using regular buttermilk while not being statistically différent (p=0.064) because of the variability in the. total phospholipids amount in whey buttermilk. This resuit is an indication that the phospholipids in whey buttermilk might be less aggregated which could prevent them to be as well retained by the MF membrane as in the case of regular buttermilk. Also, the most likely différent composition of the CP layer at the surface of the membrane could modify phospholipids transmission if caseins are présent in the case of regular buttermilk, but could also be a facilitator if a higher percentage of lipids are found as it could be the case in whey buttermilk. The affinity of phospholipids for the boundary layer could play an important rôle in their rétention by MF which would show the importance of controlling the parameters of the process in order to use this inévitable boundary layer to our advantage.

132 Conclusions Our results show that using whey buttermilk as a starting material for concentration of MFGM components by MF helps minimizing séparation problems associated by the présence of caseins and therefore helps creating an MFGM concentrate of increasing purity. However, this study highlighted the important variability in the lipid portion of whey cream which results in variation of lipid and phospholipids content in buttermilk. This work is the fïrst, to our knowledge, to attempt fractionation of whey buttermilk and before making any attempt to further develop that kind of process, attention should be given to standardize the whey cream prior to butter making as this problem was not observed in regular buttermilk from industrial manufacturing cream. DFMF helped to increase the séparation efficiency (flux) but not the selectivity of fractionation. The added processing time, but furthermore, the added volumes of water to process may limit the feasibility of this process. Results obtained while using regular buttermilk were similar to those reported in the literature and modification of the product before MF is definitely needed in order to improve séparation. Work is already in progress to develop a novel modification approach. 6.7 Acknowledgements This work has been funded in part by Dairy Management Inc., California Dairy Research Foundation, Hilmar Cheese Co. and Natural Science and Engineering Research Council of Canada. The authors would like to express gratitude to Jerry Mattas and Salvador Uson for their technical assistance. The help for phospholipids analysis of Dr Michel Britten and Hélène Giroux from the Food research and development Center (St- Hyacinthe, QC) was greatly appreciated. The authors would also like to thank Dr. Isabelle Sodini for helpful discussions during the préparation of this manuscript.

133 115 CHAPTER 7 Effect of washing the cream on buttermilk fractionation using ceramic microfiltration membranes As it was previously observed in chapter 4 and 5, casein micelles restricts the concentration of MFGM components. A process developed by Britten & Lamothe (2005) based on double centrifugation of the cream to remove casein from the cream prior to butter manufacture was investigated. The goal of this approach was to remove casein micelles by physical means while using equipment readily available in the dairy industry. This work was supervised by Dr Yves Pouliot and co-supervised by Dr Jimenez-Flores whose are both co-authors of this manuscript. This manuscript will be submitted for publication to Journal of Dairy Science in a short future.

134 Résumé Le babeurre, sous-produit de la fabrication du beurre, a généré beaucoup d'attention dernièrement en raison de sa forte concentration en composés associés à la MFGM. La microfïltration a été fréquemment utilisée en raison de sa capacité à séparer les particules en suspension des solides dissouts. Cependant, la présence de micelles de caséines dans le babeurre nuit à la séparation des particules de MFGM en MF. L'utilisation d'une crème lavée à l'aide de perméat d'ultrafiltration de lait écrémé pour la production d'un babeurre avec un taux inférieur de caséine ainsi que son fractionnement par MF a été étudié. Les résultats obtenus ont démontré que le lavage de la crème avant le barattage diminuait la concentration en protéines du babeurre par 74 %. L'analyse du profil protéique du babeurre a révélé que les caséines ainsi que les protéines sériques étaient les protéines principalement retirées suite au lavage. Les flux de perméation obtenus en MF avec le babeurre de crème lavée étaient 2 fois plus élevés qu'avec le babeurre de crème normale. Le rétentat de MF de babeurre de crème lavée contenait une proportion plus élevé en composés de la MFGM démontrant le potentiel du lavage de la crème avant barattage combiné à la MF pour la concentration des composés de MFGM. Cette étude a permis de préciser l'implication des mieelles de caséines sur la séparation des composés de la MFGM par microfïltration et d'obtenir de nouvelles informations sur l'implication des micelles de caséines sur les mécanismes régissant le flux de perméation en MF.

135 Abstract The by-product from butter manufacture, buttermilk, has gained much attention lately because of its high concentration of MFGM components. Microfiltration has been studied for buttermilk fractionation because of its ability to separate particles from dissolved solutés. However, the présence in this by-product of skim milk solids, especially casein micelles, restricts concentration of MFGM. The use of cream washed with skim milk utltrafiltrate to produce buttermilk with lower casein content was studied as well as fractionation of this buttermilk by MF. Results hâve sowed that washing the cream prior to churning yields a buttermilk with 74% less protein than normal cream buttermilk. Analysis of the protein profile of washed cream buttermilk revealed that caseins and whey proteins were the main classes of proteins removed. MF of washed cream buttermilk resulted in permeation fluxes 2 folds higher than with normal cream buttermilk. The retentates from MF of washed cream buttermilk contained superior proportion of phospholipids and MFGM material showing the efficiency of the MF combined with cream washing for MFGM concentration. The results presented in this study highlighted the implication of casein micelles in the séparation of MFGM components as well as their implication in mechanisms guiding permeation flux in MF. 7.3 Introduction When cream is churn into butter, the aqueous phase of the cream is released from the butter mass and recovered as a by product called buttermilk. The Worldwide production of butter is growing and has reached 8.2 x 10 6 metric tons in 2005 (FAOSTAT, 2006). Churning of 1 Kg of 40% milkfat cream will generate roughly equal parts of butter and buttermilk so Worldwide production of buttermilk can be estimated to be close to butter production. In contrast with cheese whey, buttermilk does find many applications and is mainly processed into a more stable product, buttermilk powder. Buttermilk powder applications are mainly related lo its functional properties. Buttermilk can be added to

136 118 prepared dried mixes, baking goods or to various other dairy products (International Dairy Foods Association, 2003). Despite thèse various applications of buttermilk powder, its added value is still minimal and its full potential as yet to be exploited. Buttermilk composition is similar to skim milk with the exception that buttermilk contains a higher proportion of phospholipids (Christie, Noble, & Davies, 1987). When milk fat globules are broken by the high shear rate of the butter churn, parts of the membrane (MFGM) is recovered in the buttermilk. The MFGM is rich in various membrane proteins and phospholipids (Walstra, Wouters, & Geurts, 2006). Data coming from médical field suggests that MFGM might possess various biological properties of interest. In a récent review, Spitsberg (2005) described the potential nutraceutical proterties of various components of the MFGM. BRCA 1 and BRCA 2, both high molecular weight minor MFGM proteins, hâve been associated with cancer cell growth inhibition while Butyrophilin was tested for suppression of multiple sclerosis. Phospholipids from MFGM could also play an important rôle in cognitive function (Ruenberg, 2002) and Sphingomyelin has been linked with colon cancer cell inhibition (Schmelz, Sullards, Dillehay, & Merrill, 2000). In buttermilk, MFGM components represent less than 5% of the solids, so a fractionation approach is necessary to isolate or concentrate thèse components. Lately many attempts at fractionation of buttermilk hâve been made. Most of thèse attempts make use of membrane filtration to selectively separate MFGM components from other buttermilk solids. Microfiltration membranes hâve been used to fractionate buttermilk using différent températures and membrane pore size (Astaire, Ward, German, & Jiménez- Flores, 2003; Morin, Jiménez-Flores, & Pouliot, 2004; Surel & Famelart, 1995). Ail thèse studies led to the conclusion that casein micelles found in buttermilk are retained at very similar rates as MFGM components. By microfiltration solely, it appears impossible to separate the abundant casein micelles from MFGM. This implies the use of other stratégies to remove caseins prior to filtration. Considering the highly variable sizes of

137 119 the MFGM fragments in buttermilk (0.1 to several im) (Morin, Jiménez-Flores, & Pouliot, 2006) it is almost impossible to retain ail MFGM fragments in the retenate using membranes with pore sizes greater than 0.1 uni. However, the use of low pore sizes does not allow permeation of caseins. Thus, the challenge of buttermilk MF is to find an optimal point to retain maximal amount of MFGM components while removing other buttermilk solids. Corredig et al. (2003) reported a process making use of sodium citrate to disrupt casein micelles prior to membrane filtration. However, the use of high concentration of citrate induces a permeate with 4X the concentration found in milk (Surel & Famelart, 1995). Others hâve used rennet coagulation of buttermilk followed by membrane filtration of the generated whey (Sachdeva & Buchheim, 1997) but buttermilk coagulation is hard to obtain and needs high levels of calcium chloride added. In a previous work (Morin, Pouliot, & Jiménez-Flores, 2006), it has been shown that buttermilk obtained from churning of whey cream represents an important source of MFGM components that can be concentrated by microfiltration. However, the low volumes of whey cream produced around the world and the rapidly oxidized whey butter generated can decrease the value of this approach. In a récent work, Britten and Lamothe (2005) hâve developed an approach based on cream washing to remove casein from the cream prior to churning. In the process, cream is separated from milk and then mixed to a ratio of 1:10 with milk ultrafiltrate and then the cream is separated again yielding a cream with low amount of casein. Buttermilk from washed cream has therefore much less caseins and should be an idéal substrate for MFGM concentration by microfiltration. The goal of the présent study was to evaluate the effect of washing the cream on the buttermilk microfiltration in terms of hydrodynamic filtration parameters and composition of the fractions obtained.

138 Materials and methods Préparation of buttermilks Washed cream buttermilk. Fresh raw milk (550 L) was obtained frora a local dairy farm (Ferme SMA, Beauport, QC, Canada). Raw milk was heated to 38 C in a UHT System (Chalinox, Sorel, QC, Canada) and was then separated into skim milk (~ 485 Kg) and cream (~ 65 Kg) using a milk separator (De Laval Model N 619, Lund, Sweden) running at x g. Skim milk outlet of the separator was restricted to yield a cream of ~ 30 % fat. Skim milk was ultrafiltrated on two Da spiral wound membranes mounted in parallel on a pilot scale System (Koch Filtration System, Koch Membrane Systems Inc., Wilmington, MA). Inlet and outlet pressures were set at 345 and 207 kpa respectively. Température of filtration was 50 C and permeation flux during UF was 44.6 ±2.1 L.h'.m" Kg of skim milk ultrafiltration permeate (SMUF) was collected. SMUF was rapidly cooled to 4 C and stored overnight. Cream (~ 65 Kg) was diluted with 5 volumes of SMUF and gently mixed for 30 minutes before being separated again into cream and skim. Cream exited the separator with a fat content of 67 ± 2 %. Cream was then standardized to 40 % fat with the skim fraction. The cream was then batch-pasteurized in a steam-jacketed tank at 66 C for 30 min and then cooled rapidly to 13 C. Pasteurized washed cream was then churned in a rotary butter churn into butter and buttermilk. Churning was carried at 26 rpm and 13 C. Buttermilk was collected and filtered through a stainless steel mesh. Butter was further churned to expel trapped buttermilk. Buttermilk was then quickly heated to 38 C and separated to remove excess of lipids. This protocol has been performed in triplicate.

139 Normal cream buttermilk Fresh raw milk (1600 L) was obtained from a local dairy farm (Ferme SMA, Beauport, QC, Canada). Raw milk was heated to 38 C in a UHT System (Chalinox, Sorel, QC, Canada) and was then separated into skim milk (~ 1500 Kg) and cream (~ 100 Kg) using a milk separator (Alfa-Laval Model N 619, Lund, Sweden) running at x g. Fat content of the cream was standardized to 40 % by blending with skim milk resulting in a volume of ~ 160 L and was batch-pasteurized as described above. Cream was divided in three lots of 50 Kg each. Each lot of cream was churned; buttermilk was recovered and separated as described above Microfïltration of buttermilks Microfïltration equipment MF module used for buttermilk MF was a Tetra Alcross Pilot Ml (Tetra Pak Filtration Systems, Lund, Sweden) with permeate co-current recirculation allowing for low and uniform transmembrane pressure (TMP). The membrane used in this study was a 0.5 um ceramic Membralox P35-37 membrane (Pall Corporation, Mississauga, ON, Canada) with 0.35 m 2 filtration surface. Clean DI water flux (50 C/ 100 kpa TMP) of the membrane was L.h'.m" 2 which corresponds to a hydraulic résistance of kpa/l.h'.m" 2. Before each run the membrane was conditioned by alkaline - chlorine washing using Ultrasil 25 (0.01%, ph 13) and XY-12 (200 ppm Cl") cleaning solution (Ecolab, Mississauga, ON, Canada).

140 Microfiltration experiments For each run of MF, 20 Kg of buttermilk was used. Buttermilk was concentrated to 5 Kg by MF (MF 4X). Parameters for MF were 50 C, 50 kpa TMP with a pressure drop of 140 kpa through the length of the membrane. Once the MF 4X was reached, the retentate was collected from the System and a sample was withdrawn (100 ml). The System was then rinsed with DI water until reaching a clear permeate and permeate flux was measured to evaluate the irréversible fouling résistance of the membrane. The retentate was then diluted with 6 diavolumes of DI water (30 kg) at 50 C and placed back in the MF System. The diluted MF 4X retentates was then concentrated back to 5 Kg resulting in a diafiltration factor of two (DF 6X). Pressure and température parameters for diafiltration were the same as for MF. Permeation fluxes were measured during MF and DF using a graduated cylinder and a stopwatch. Hydraulic résistance of the membrane (RM), résistance caused by réversible fouling and concentration polarization layer (RRF + RG) and résistance caused by irréversible fouling was calculated using équation 7.1 in which J is the flux and Pj is the transmembrane pressure (Cheryan, 1998) Equation 7.1 Permeation flux and résistances calculation according to the résistance in séries model J- P R M +R G + R RF + R IF Fouling coefficients was calculated as proposed by Ramachandra Rao et al. (1995) according to équation 7.2 where J w is the pure water flux. Equation 7.2 Fouling coeficient (f c ) calculation /e=h after rin sing

141 123 A / c of 0 indicates a membrane with no détectable irréversible fouling and a f c of 1 indicates a completely fouled membrane Analytical procédures Raw milks, creams, buttermilks and MF fractions were analyzed for total protein using a nitrogen analyzer (Leco FP-528, Leco Corp., St. Joseph, MI, USA) and a protein conversion factor of Total solids were obtained by oven drying for 15 hours at 102 C and ash was measured by incinération in a muffle furnace at 550 C for 20 h. Lipids were extracted using the Mojonnier ether extraction procédure, and lipid extracts were diluted to 5 mg/ml in 2:1 chloroform: methanol and stored at -20 C until further analysis. The lipid profiles of the buttermilks was obtained by HPLC-ELSD as described previously (Morin, Jiménez-Flores, & Pouliot, 2004). Standards used for calibration were phosphatidylethanolamine (PE), phosphatidylcholine (PC), phosphatidylserine (PS), phosphatidylinositol (PI) and sphingomyeline (SM) (Sigma-Aldrich, Oakville, ON, Canada). Protein profiles were obtained by SDS-PAGE (Laemmli, 1970) on a 13.5% acrylamide gel. Samples from the 3 replicates were pooled and ail samples were diluted to 3 mg.ml" of protein with reducing sample buffer. Samples where then heated into boiling water for 10 min. Cream samples were centrifuged at x g for 5 min and the loading volume was collected below the lipid layer in order to avoid loading an excess of lipid in the gel. Wells of the gels were loaded with 25 ^L of samples resulting in a 75 \ig of proteins loaded per well. Relative protein classes (MFGM, caseins and whey proteins) distribution

142 124 in the samples was obtained by densitometry using Quantity One software from Bio-Rad (Mississauga, ON, Canada). Protein classes were determined according to their molecular weight by comparison with a molecular weight standard (Précision Plus unstained protein standard, Bio-Rad, Mississauga, ON, Canada) Statistical analysis Results were analyzed statistically using Minitab 14.0 software package. ANOVA's were performed and mean comparisons were carried out using Tuckey's test. Results were considered signifïcantly différent at p < 0.05 and highly significant at p < Results and discussion Effect of cream washing on cream and buttermilk composition Washed and normal creams compositions were différent and this resulted in a signifïcantly différent churning time (p<0.05). The required time for emulsion breakdown (expressed in min per Kg of cream) of washed and normal creams was 1.01 ± 0.09 versus 2.80 ± 0.80 respectively. When cream is churned into butter, a séries of phenomena occurs. As air is incorporated in the cream, skim milk proteins unfold and form an interfacial layer at the surface of the air bubbles and then MFGM displaces the skim milk proteins which induces rupture of the milk fat globules (Frede & Buchheim, 1994). Because of the washing procédure applied to the cream, the amount of skim milk proteins available to stabilize the foam interface was lower which resulted in faster MFGM rupture and therefore faster churning time. The chuming times were longer than what we reported before with the same equipment (Morin, Jiménez-Flores, & Pouliot, 2006). The reason for this différence is mainly that no cold maturation of cream was used

143 125 our experiments. Cold maturation of the cream insures proper crystallization of the triglycérides inside the milk fat globules to obtain the desired butter texture and limits the fat losses to buttermilk (Mahaut, Jeantet, Brûlé, & Schuck, 2000). Cold maturation was not applied in our experiments since the texture of the butter was not the main focus and that buttermilk was skimmed before MF. Composition of pasteurized creams and buttermilks are listed at Table 7.1. The most important effect observed was the decrease in the protein content. Britten & Lamothe (2005) demonstrated a decrease of 80% in the protein content of buttermilk made from cream washed with 10 volumes of milk ultrafiltrate. Our results are in accordance with results of Britten & Lamothe (2005) since the observed decrease in our experiments was 74% with cream washed with 5 volumes of milk ultrafiltrate. Table 7.1 Average composition of pasteurized creams and buttermilks from normal and washed cream process Total solids (%) Total proteins (%) Total lipids (%) Phospholipids (%) Class distribution Ash (%) (/o) PE PC PS PI SM Normal n.d Pasteurized cream Washed ± ± ±3.04 n.d 0.31 ±0.04 Normal 9.12±0.17** 3.46 ±0.05** 0.51 ±0.02** 0.13 ±0.00** ±0.55** ± 1.92** 6.18 ±0.45* ± ± ±0.06** Buttermilk Washed 6.38 ± ± ± ± ± ± ± ± ± ±0.03 One sample of pasteurized cream was taken as the cream was divided in three lots only after pasteurization. ± indicates standard déviation (n=3) n.d. not determined Indicates significant (p < 0.05) différence between normal and washed cream process Indicates highly significant (p < 0.01) différence between normal and washed cream process 1.61

144 126 The protein profile of both creams and buttermilk were markedly différent. As shown in the SDS-PAGE gel (Figure 7.1a & 7.1b lanes 1-2), the proportion of caseins found in washed cream and buttermilk was lower than that of normal cream and buttermilk. Whey proteins, mainly [3-Lactoglobulin, was also reduced by this procédure but MFGM proteins proportion was affected by the washing procédure as they appear slightly more concentrated in buttermilk from washed cream. In washed cream buttermilk, the ratio of MFGM proteins to skimmed milk proteins as estimated by densitometry was 2.31 as opposed to 0.85 for normal cream buttermilk. The washing procédure therefore not only allows for important removal of proteins (74%) but the proteins removed are mainly skimmed milk proteins yielding buttermilk with increasing relative proportion of MFGM proteins. A MW! I I kda MW ! I i ï I I B HMW MFGM CN's 'M&'&JjZÎL WP MFGM (FABP) Figure 7.1 SDS-PAGE (13.5%) of normal cream samples (A) and washed cream samples (B). Lanes 1 are pasteurized creams, lanes 2 buttermilks, lanes 3 4X MF retentates, lanes 4 4X MF permeates, lanes 5 6X DF retentates and lanes 6 6X DF permeates. MW - molecular weight standards, HMW - high molecular weight material, MFGM - milk fat globule proteins, CN - caseins, WP - whey proteins, FABP - fatty acid binding protein (MFGM). Phospholipid analysis of cream samples has been attempted but the very high ratio of neutral lipids to phospholipids in samples restricted phospholipid quantification using our HLPC-ELSD method. However, in similar work, Britten and Lamothe (2005) hâve