Comparative proteomic profiling of 2 muscles from 5 different pure pig. arrays and analyzed using surface-enhanced laser desorption/ionization

Transcription

1 Published December 4, 2014 Comparative proteomic profiling of 2 muscles from 5 different pure pig breeds using surface-enhanced laser desorption/ionization time-of-flight proteomics technology 1 N. Mach,* 2 E. Keuning,* L. Kruijt,* M. Hortós, J. Arnau, and M. F. W. te Pas* *Animal Breeding and Genomics Centre, Wageningen UR Livestock Research, PO Box 65, 8200 AB Lelystad, the Netherlands; and IRTA-Food Technology, Monells, Girona, Spain ABSTRACT: The objectives of this study were to evaluate the influence of different pure pig breeds and muscle types on the expression of muscle proteins, as well as their interactions, and second, to find biomarkers for breed and muscle types. A total of 126 male pigs, including 43 Landrace, 21 Duroc, 43 Large White, 13 Pietrain, and 6 Belgian Landrace, were slaughtered at the age of 174 ± 6 d. Samples from the semimembranosus muscle (SM) and LM were collected 24 h postmortem. Proteomic spectra were generated on an anion exchanger (Q10), a cation exchanger (CM10), and on immobilized metal affinity capture (IMAC30) ProteinChip arrays and analyzed using surface-enhanced laser desorption/ionization time-of-flight mass spectrometry ProteinChip techniques. Breed and muscle type did not affect the number of peaks per spectrum but, interestingly, affected the average intensity of the peaks. Of these peaks, a total of 4 proved to be potential protein biomarkers to differentiate LM or SM muscles, and 2 to classify specific breed types. Additionally, several peaks influenced by the interaction between muscle and breed types could correctly classify pig muscles according to their breed. Further studies need to be carried out to validate and identify these potential protein biomarkers for breed and muscle types in finishing pigs. Key words: biomarker, breed, muscle, pig, proteomic, surface-enhanced laser desorption/ionization time-of-flight mass spectrometry 2010 American Society of Animal Science. All rights reserved. J. Anim. Sci : doi: /jas INTRODUCTION In meat science, proteomics supplies tools that allow the identification of protein patterns and single protein biomarkers (Bendixen, 2005). The aim of proteomics is often to find biomarkers, molecules that describe the biological status of a cell, tissue or organ, given 1 The authors gratefully acknowledge the financial support of the European Community under the Sixth Framework Programme for Research, Technological Development and Demonstration Activities, for the Integrated Project Q-PORKCHAINS FOOD- CT , as well as additional funding received from the Kennisbasis Onderzoek of the Dutch Agricultural Ministry (KB ). The authors are also grateful for the financial support of the Comissionat per a Universitats i Recerca del DIUE of the Government of Catalonia, and to David Lloret from the Department of Computer Science and Applied Mathematics of Universitat de Girona, and Narcís Sais from IRTA-Food Technology for their assistance. The views expressed in this publication are the responsibility of the authors and do not necessarily reflect the views of the European Commission. 2 Corresponding author: nuria.mach@wur.nl Received June 30, Accepted December 14, a specific condition. Comparisons of protein expression patterns or single protein biomarkers in muscle tissue have currently gained much attention in meat science. They can be used to increase understanding of factors associated with muscle development, composition, functioning, and metabolic processes (Pospiech et al., 2007; Hollung et al., 2009), such as genetic and rearing background of the pigs (Kwasiborski et al., 2008a; Hollung et al., 2009), sex (Kwasiborski et al., 2008a,b), management production practices (Hollung et al., 2009), feeding strategies (Lametsch et al., 2006), and processing conditions (Edwards et al., 2003, 2006; Ryu et al., 2006). Aside from molecular knowledge, the identification of protein expression patterns or single protein biomarkers in muscle tissue can allow correct classification of pigs according to a specific condition, such as breed, rearing environment, and sex (Kwasiborski et al., 2009), and subsequently be useful as markers for traceability proposes or genetic research (Kwasiborski et al., 2009). However, detailed information on protein patterns and single protein biomarkers in different breeds and muscles types, with particular focus on their interaction, is still lacking. Furthermore, although 1522

2 Muscle protein expression in pigs dimensional electrophoresis and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry is a powerful approach to protein separation, it has limitations when dealing with small molecular weight proteins, very acidic or basic proteins, very hydrophobic proteins, or proteins in low abundance (Schipper et al., 2007). In addition, the technique requires a relatively large amount of sample and is labor-intensive. Furthermore, good gel-to-gel reproducibility is difficult to achieve. An alternative approach to identify protein expression patterns or single protein biomarkers in muscle tissue is surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF-MS), which uses special chromatographic-like probe surfaces (ProteinChip arrays) to bind proteins with complementary physicochemical properties (O Gorman et al., 2006). The value of SELDI-TOF-MS lies in its ability to obtain and compare spectra from a large number of samples in a relatively short time with little sample preparation or sophisticated chromatography. It produces spectra of complex protein mixtures based on the mass-to-charge-ratio of the proteins and on their binding affinity to the ProteinChip arrays. Differentially expressed proteins are determined from these protein patterns by comparing peak intensities (O Gorman et al., 2006). Because it is not necessary to know the identities of the proteins for the purpose of differential classification, this technology provides an alternative platform for the differential display of multiple potential biomarkers. The technology has been used successfully to detect several human disease-associated proteins and protein expression patterns in a variety of biological tissues and body fluids (Wadsworth et al., 2004; Yang et al., 2005; Driemel et al., 2007; Schipper et al., 2007). Arising from these reasons, the objectives of this study were to evaluate the influence of different breed and muscle types on the expression of muscle proteins, as well as their interactions, and secondly, to find biomarkers for breed and muscle types. MATERIALS AND METHODS The animals used in this study were managed following the principles and specific guidelines of the IRTA Animal Care Committee. Animals and Sample Collection One hundred twenty-six male pigs of the Landrace (LS; n = 43), Duroc (DU; n = 21), Large White (LW; n = 43), Pietrain (PI; n = 13), and Belgian Landrace (BL; n = 6) breeds were fattened under identical conditions in the Pig Testing Station (IRTA-CAP) in Monells (Girona, Spain). Pure breeds were chosen to be able to analyze the contribution of each breed separately. The day before slaughter, pigs (174 ± 6 d) were weighed. The average final BW was ± 1.57 for LM, ± 2.25 for DU, ± 1.55 for LW, ± 2.98 for PI, and 97.1 ± 4.22 for BL. The pigs were then fasted on-farm during 9 h and transported for 1.5 h from the IRTA-CAP to the ESFOSA commercial slaughterhouse (Vic, Spain). Once there, animals from different pens were not mixed. The waiting time at the slaughterhouse for the slaughter of the first pig was approximately 30 min and 2 h for the last pig slaughtered because they were slaughtered individually by exsanguination after stunning with 90% CO 2 for 2 min. The animals from different breeds were slaughtered alternatively. After 24 h of carcass chilling, a sample of LM and semimembranosus muscle (SM) was removed from each animal, frozen in liquid nitrogen, and stored at 80 C until protein extraction. The LM was chosen as the indicator muscle in meat quality studies and as reference muscle to detect changes in protein expression between different breeds. In contrast, the SM was chosen because it is an important muscle of the ham and is easily sampled without damaging the ham before further dry curing and aging processes. Preparation of Protein Extracts for SELDI-TOF Analyses Muscle samples were taken from the freezer, and the water soluble fraction of proteins was isolated. Samples were weighed (30 to 50 mg), placed in 1.5 ml of lysis buffer [10 mm Tris-HCl, ph 7.25, 10 mm KCl, 2% (vol/vol) Triton X-100, 1 mm PMSF], and homogenized (Ultra Turrax T25, IKA Labortechnik, Staufen, Germany) under ice to avoid mechanical heating of the samples. The resulting sample homogenates were briefly centrifuged (20 min, 4 C, 12,000 g) to remove insoluble debris. The supernatant was then assayed for total protein content using a commercial protein assay kit with BSA as standard (Bio-Rad, Veenendaal, the Netherlands). SELDI-TOF-MS Analyses For the SELDI-TOF analyses, all samples were analyzed in duplicate. The strong anion exchanger (Q10), weak cation exchanger (CM10), and immobilized metal affinity capture (IMAC30) ProteinChip arrays and binding buffer combinations were prepared according to the manufacturer s instructions (Bio-Rad Laboratories Inc., Hercules, CA). The Q10, CM10, and IMAC30 ProteinChip arrays were especially selected because they produce good quality proteome patterns with an optimal numbers of peaks (te Pas et al., 2009). The different ProteinChip arrays were equilibrated with the respective binding buffers containing 0.1% Triton. The binding/washing buffer for the Q10 contained 0.1 M sodium acetate (ph 6.0), and that for the CM10 contained 0.1 M sodium acetate (ph 5.0). Before applying the samples to the IMAC30 ProteinChip array, the active spots of the array were preactivated with 100 μl of 0.1 M copper sulfate solution according to the manufacturer s instructions (Bio-Rad Laboratories Inc.). Twenty micrograms of protein was suspended in

3 1524 a 200-μL volume of binding buffer. Then 100 μl of sample was loaded to each well of the array and allowed to bind (60 min, room temperature, and on a platform shaker) to the array. After the binding step, the entire array was washed 3 times with the respective binding buffers (5 min, room temperature, with agitation) and then twice with deionized water. After briefly drying the arrays, 0.8 μl of a saturated solution of 4-hydroxy-3, 5-dimethoxy-cinnamic acid (sinapinic acid, Sigma, St. Louis, MO) dissolved in 50% (vol/vol) acetonitrile and 0.5% (vol/vol) trifluoracetic acid, was applied twice to each of the active spots of the array, and was allowed to thoroughly dry. The different ProteinChip arrays were then placed in the SELDI ProteinChip Biology System Reader 4,000 (Bio-Rad Laboratories Inc.). The laser intensity was 3,000 nj. The SELDI ProteinChip spectra were analyzed using the ProteinChip 3.0 software (Ciphergen Biosystems, Fremont, CA). The instrument was calibrated by using the All-in-One ProteinStandard II (Bio-Rad). To summarize, calibrated spectra were initially normalized to minimize the variability in spectra obtained from different times. The normalization process involved taking the total ion current used for all the spots, averaging the intensity (μa), and adjusting the intensity scales for all the spots so that all spectra could be displayed on the same scale (Hong et al., 2005). Baseline substraction was conducted before normalization, as recommended by Ciphergen Biosystems (Fremont, CA). Baseline substraction offsets in the spectra set the result of electrical noise and the noise from energy absorbing molecules (Hong et al., 2005). The peaks ranging from 2,000 to 50,000 mass-tocharge-ratio (m/z) were detected and extracted by the ProteinChip 3.0 software. The spectral region from 0 to 2,000 m/z was unreliable for normalization and peak detection due to matrix interference and was therefore excluded from the analyses (Hong et al., 2005). The spectral region above 50,000 m/z was also not included in the analyses because the intensity of the laser was too small for this region. The first step for the initial peak detection was the creation of clusters by the ProteinChip 3.0 software (Bio-Rad Laboratories Inc.). Within a designated mass-to-charge ratio, the program grouped peaks that were present across multiple spectra at the same narrow window of 0.3% m/z value. Each cluster was then treated as a single protein. Furthermore, the initial cluster detection scan was carried out using a signal-to-noise (S/N) ratio of 5. A second cluster detection scan was then performed using a S/N ratio 2, valley depth ratio = 2. After spectral preprocessing using ProteinChip 3.0 software, mass spectral data were exported as a profile matrix and used to evaluate the influence of different breed and muscle types on the expression of muscle proteins using SAS (SAS Inst. Inc., Cary, NC). Differentially expressed proteins were determined by comparing the peak intensities. Therefore, the analyses quantify which peaks were upregulated or downregulated between breed and muscle types and which of these could be considered as Mach et al. potential protein biomarkers. To calculate the accuracy of the potential protein biomarkers identified by SEL- DI-TOF, proteins that were detected in 100% of the animals, presented a P-value of the main effect <0.001, and a P-value of the interaction between muscle and breed types >0.10, were subjected to CiphergenExpress 3.0 software for the cluster blots and receiver operation characteristic (ROC) curve analysis. The ROC curve analysis was chosen as a measure of the accuracy of biomarkers because it includes all possible cut points and shows the relationship between the sensitivity of a biomarker and its specificity (Obuchowski, 2003). The area under the ROC curve (AUC) was also calculated. The maximum area under a ROC curve is Peaks with an AUC greater than 0.75 were confirmed protein biomarkers (Clarke et al., 2003). Statistical Analyses Mass spectral data were analyzed using mixed-effects ANOVA (SAS Inst. Inc.). The statistical model included breed type, muscle type, and their interactions, as fixed effects, and animal as random effect. A separate model was fitted for each mass-to-charge-ratio within the ProteinChip arrays type. Differences among breeds were adjusted with Tukey s method. Results are presented as least squares means ± SEM. Significant differences were declared at P-value <0.05. RESULTS AND DISCUSSION Number of Peaks per Spectrum The average and SD of peak intensity for each breed and muscle type on CM10, Q10, and IMAC30 ProteinChip arrays are presented in Tables 1, 2, and 3, respectively. The number of peaks per spectrum (36.6 ± 4.50) was not affected by breed and muscle type. Concerning the number of peaks per spectrum and different ProteinChip array, on CM10 ProteinChip array, 41 different peaks per spectra were seen overall, whereas the Q10 ProteinChip array displayed 31 peaks per spectrum, of which most peaks were observed in the range below 20,893 m/z. Finally, using the IMAC30 ProteinChip array, 37 peaks per spectrum were identified. Occasionally peaks were located at slightly different positions in the spectra. This may reflect different isoforms of the proteins caused by modifications that influence the charge of the protein, or it may also reflect that the protein was fragmented. However, in most cases the peak had the same m/z ratio, suggesting that this was a full-length protein that had not been affected by biochemical processes (Sayd et al., 2006). Protein Expression of Different Muscle Types Although the number of peaks remained invariant to muscle and breed types, these factors strongly in-

4 Muscle protein expression in pigs 1525 Table 1. Peak intensities (μa) of 2 different muscles 1 from 5 different pig pure breeds 2 on the CM10 ProteinChip array Peak, 3 m/z LM SM DU BL LS LW PI DU BL LS LW PI SEM Breed Muscle Breed muscle P-value 4 2, < , <0.001 < , b 8.66 a 4.13 b 3.73 b 5.21 ab b a b b ab <0.001 < , < , bc a 6.46 b 5.18 c 9.92 a bc a b c a <0.001 < , , < , b 8.68 a 6.07 ab 5.55 b 7.43 ab 5.88 b a 7.45 ab 5.87 b 8.88 b < , , b 3.75 b 9.84 a 6.04 b 3.08 b b 6.45 b a b 7.88 b <0.001 < , b 6.27 a 4.88 b 5.77 a 5.70 ab 4.34 b 5.65 a 4.35 b 5.42 a 5.07 ab , < , b 4.02 a 4.53 a 3.94 a 4.35 a 2.75 b 5.14 a 4.00 a 4.13 a 4.11 a < , b 1.65 c 4.48 b 5.61 a 2.26 b 3.70 b 1.34 c 3.00 b 5.07 a 3.30 b < , < , < , < , <0.001 < , , b 5.42 a 2.73 b 3.50 b 4.37 ab 3.98 b 6.22 a 3.56 b 3.49 b 4.56 ab < , a a b a a a a 8.61 b a a , < , < , b 2.57 a 1.97 b 2.16 b 2.41 ab 1.20 b 2.09 a 1.10 b 1.21 b 1.64 ab <0.001 < , b a 8.75 b 8.81 b a 5.44 b 8.92 a 5.13 b 5.89 b 8.37 a <0.001 < , b 2.85 a 1.99 b 2.14 b 2.69 a 1.75 b 2.19 a 1.70 b 1.84 b 2.07 a <0.001 < , a 2.31 a 1.80 b 1.93 b 2.09 ab 1.81 a 2.14 a 1.54 b 1.55 b 1.58 ab 0.07 <0.001 < , <0.001 < , , , , , <0.001 < , < , , , , , , , a c Within rows, mean values not bearing a common superscript differ (P < 0.05). 1 Muscle type: SM = semimembranosus muscle. 2 Breed type: DU = Duroc; BL = Belgian Landrace; LS = Landrace; LW = Large White; PI = Pietrain. 3 Included n = 126 male pigs. m/z = mass-to-charge-ratio. 4 Effect of breed type, muscle type, and the interaction between breed and muscle.

5 1526 Mach et al. Table 2. Peak intensities (μa) of 2 different muscles 1 from 5 different pig pure breeds 2 on Q10 ProteinChip array LM SM P-value 4 Peak, 3 m/z DU BL LS LW PI DU BL LS LW PI SEM Breed Muscle Breed muscle 3, a 1.73 b 4.61 a 3.98 a 2.54 b 6.99 a 3.20 b 6.04 a 6.27 a 3.12 b , b 2.25 b 4.75 b 7.52 a 4.23 b 5.63 b 5.30 b 2.99 b 7.22 a 3.12 b < , , < , b 2.52 b 5.12 b 7.90 a 5.17 b 4.51 b 6.20 b 2.94 b 7.55 a 4.19 b , , b 2.61 ab 2.11 b 4.33 a 3.05 ab 1.62 b 3.64 ab 1.21 b 3.82 a 2.44 ab < , < , < , , , <0.001 < , , <0.001 < , , <0.001 < , , ab 2.86 a 2.64 b 2.87 a 2.83 a 1.91 ab 2.66 a 1.71 b 2.00 a 2.33 a < , < , , , , , , , < , , , , , a,b Within rows, mean values not bearing a common superscript differ (P < 0.05). 1 Muscle type: SM = semimembranosus muscle. 2 Breed type: DU = Duroc; BL = Belgian Landrace; LS = Landrace; LW = Large White; PI = Pietrain. 3 Included n = 126 male pigs. m/z = mass-to-charge-ratio. 4 Effect of breed type, muscle type, and the interaction between breed and muscle.

6 Muscle protein expression in pigs 1527 Table 3. Peak intensities (μa) of 2 different muscles 1 from 5 different pig pure breeds 2 on IMAC30 ProteinChip array LM SM P-value 4 Peak, 3 m/z DU BL LS LW PI DU BL LS LW PI SEM Breed Muscle Breed muscle 3, <0.001 <0.001 < , <0.001 <0.001 < , , < , < , b 3.24 c 7.97 a 5.35 b 2.86 c 6.36 b 3.03 c 7.46 a 5.99 b 3.66 c < , b 5.80 c b a 5.56 c b 9.14 c b a 8.99 c < , < , , , < , < , , , , , , <0.001 <0.001 < , < , < , , < , , , , , , < , < , < , , , , , , , a c Within rows, mean values not bearing a common superscript differ (P < 0.05). 1 Muscle type: SM = semimembranosus muscle. 2 Breed type: DU = Duroc; BL = Belgian Landrace; LS = Landrace; LW = Large White; PI = Pietrain. 3 Included n = 126 male pigs. m/z = mass-to-charge-ratio. 4 Effect of breed type, muscle type, and the interaction between breed and muscle.

7 1528 fluenced the average intensity of peaks. Muscle type affected a total of 23 out of 41 identified peaks on the CM10 ProteinChip array (Table 1), 9 of 31 identified peaks on the Q10 ProteinChip (Table 2), and lastly, 14 peaks out of 37 on the IMAC30 ProteinChip (Table 3). Although little is known about protein expression patterns in response to different muscle tissue, the effect of muscle type on protein expression patterns is most probably associated with large variation in the amount of muscle composition. The type and size of fibers, which are quantitatively the most significant compartment of skeletal muscle (Lefaucheur et al., 2002), differ in morphological, contractile, and metabolic characteristics. In pig muscles, it has been shown that slowtwitch fibers (type I) have a greater aerobic capacity, have less glycolytic capacity, and contain a greater level of intracellular lipid and myoglobin than fast-twitch fibers (type II; Essen-Gustavsson et al., 1994). The type II fibers are further subdivided in fast IIA, IIX, and IIB fibers. The proportion of fiber types varies between muscles. The LM of pigs consists of approximately 13% of slow-twitch oxidative red fibers, 17% of fast-twitch oxidative (type IIA and IIX) fibers, and about 70% of fast-twitch glycolytic (type IIB) and white fibers. In contrast, SM muscle contained a greater percentage of type I fibers and has a greater oxidative capacity than LM (Karlsson et al., 1993). Furthermore, our current results show the potential of SELDI-TOF-MS to not only identify protein expression patterns, but also potential protein biomarkers that may be used to classify muscle type in a relatively short time with little sample preparation or sophisticated chromatography. Therefore, after the identification of potential protein biomarkers, their accuracy was calculated by the ROC analysis. Interestingly, the 8,126 m/z peak on the CM10 ProteinChip array proved to be a potential protein biomarker for muscle type (Table 1 and Figure 1), its average intensity was 2.52 times greater in SM compared with LM. The sensitivity and specificity for this peak was 99 and 68%, respectively, and the ROC curve gave an AUC of 0.93 (Figure 2). The AUC indicated a 93% of probability that a randomly selected SM sample had greater average 8,126 m/z peak expression than a randomly selected LM sample. The ROC curve came close to the upper left corner of the graph and the AUC was 0.93, so the potential protein biomarker had very good predictive ability. Figure 3 further shows the capacity of this biomarker to correctly classify the different muscle types. Likewise, the 4,151 m/z peak on the Q10 ProteinChip was demonstrated to be a potential protein biomarker able to discriminate LM from SM muscles. The average intensity of the 4,151 m/z peak was 5.08 times greater in SM compared with LM muscle (Table 2), with a sensitivity and specificity of 91 and 69%, respectively, and the AUC of ROC curve of at least 0.80 (Figure 4). The bar plot represented in the Figure 5 further confirms the accuracy of this biomarker. Additionally, on the IMAC30 ProteinChip, the peaks 16,898 and 17,105 m/z were confirmed to Mach et al. be potential protein biomarkers for muscle type. The average intensity of the 16,898 was 4.21 times greater (P < 0.001) in LM than in SM, whereas the average intensity of the 17,105 m/z peak was 4.33 times greater (P < 0.001) in LM than in SM (Table 3). The discrimination with ROC between muscle types for 16,898 and 17,105 m/z peaks was sensitive (99%) and specific (62%), with an AUC of 0.97 (Figures 6 and 7). Moreover, Figures 8 and 9 confirmed that the ability of these potential protein biomarkers to discriminate between muscles was very good. In the present study, the ROC curve for all the identified potential protein biomarkers extended well into the upper left corner of the graph, and the area under the graph was >0.90. Additionally, the interaction factor between muscle type and breed type was not significant (P > 0.10) for all of these 4 potential protein biomarkers. These findings are of essential interest to confirm the good predictive ability of the potential biomarker identified. In many human proteomic studies, as a result of the complexity of the tissues analyzed, no single protein biomarkers can be distinguished among sample groups (Wadsworth et al., 2004; Yang et al., 2005), making it necessary for a more complex pattern of multiple protein biomarkers. The suitability of the 4 detected potential protein biomarkers to correctly classify muscle types needs to be further evaluated by testing the model with mass spectral data from more carcasses and different breeds. However, because the analyses were performed with 126 animals, and proteins were differentially expressed between muscles (P < 0.001) with increased sensitivity (>90%) and a notable specificity (>60%), it can be assumed that the proposed biomarkers will help to discriminate muscle type and, thus, that proteins may be used for traceability muscle purposes as an alternative to DNA-based traceability techniques (Kwasiborski et al., 2009) or as markers to further understand metabolic and functional characteristics between muscle types. Furthermore, such promising results clearly suggest that it is important to identify these potential protein biomarkers because they may provide further information for understanding the metabolic and functional processes when comparing LM and SM. Protein Expression of Different Breeds Breed type affected a total of 24 out of 41 identified peaks on the CM10 ProteinChip arrays (Table 1), 12 of 31 peaks on the Q10 ProteinChip array (Table 2), and 9 of 37 peaks on the IMAC30 ProteinChip array (Table 3). It is well established that genetics determines variation of protein expression patterns of finishing pigs (Kwasiborski et al., 2008a,b; Hollung et al., 2009; Xu et al., 2009), which may be attributed to differences in transcriptome expression of metabolic and structural proteins (Hollung et al., 2009), but also differences in hormonal, intracellular, or functional characteristics (Kwasiborski et al., 2009). Differences in protein expression have been reported by Hollung et al. (2009)

8 Muscle protein expression in pigs 1529 Figure 1. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry spectra of 2 proteins significantly affected by muscle type: discrimination by CM10 ProteinChip arrays. SM = semimembranosus muscle. comparing Norwegian LS and DU breeds, by Kwasiborski et al. (2009) comparing LW and DU breeds, and by Xu et al. (2009) comparing LW and Meishan pigs. However, te Pas et al. (2009) showed that the peaks detected with CM10, Q10, and IMAC30 ProteinChip SELDI-TOF arrays were largely similar for Yorkshire and DU breeds, suggesting that the muscle composition in these 2 pig breeds show large similarities. Remarkably, differences in protein expressions may also be related to additional genetic influences or environmental factors, due to, for instance, different stress reactivity of these pigs (Kwasiborski et al., 2008a,b). After the ROC analysis to calculate the accuracy of the identified potential protein biomarkers, the 5,773 m/z on CM10 ProteinChip arrays (Table 1) helped to distinguish the LS breed from LB breeds, with a sensitivity of 84%, specificity of 50%, and an AUC of 0.79 (Figure 10). Furthermore, the 3,743 m/z peak on the Q10 Pro- Figure 2. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry cluster blot of 8,126 mass-to-charge-ratio (m/z) potential biomarker for muscle type: discrimination by CM10 ProteinChip arrays. Receiver operation characteristic (ROC) curve for the discrimination of muscle types shows an area under the curve (AUC) of 0.93, and a sensitivity and specificity of 99 and 68%, respectively. SM = semimembranosus muscle; n = 252 samples.

9 1530 Mach et al. Figure 3. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry bar plot of 8,126 mass-to-charge-ratio (m/z) potential biomarker for muscle type: discrimination by CM10 ProteinChip arrays. SM = semimembranosus muscle; n = 240 samples. Figure 4. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry cluster blot of 4,151 mass-to-charge-ratio (m/z) potential biomarker for muscle type: discrimination by Q10 ProteinChip arrays. Receiver operation characteristic (ROC) curve for the discrimination of muscle types shows an area under the curve (AUC) of 0.80, and a sensitivity and specificity of 91 and 69%, respectively. SM = semimembranosus muscle; n = 252 samples. Figure 5. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry bar plot of 4,151 mass-to-charge-ratio (m/z) potential biomarker for muscle type: discrimination by Q10 ProteinChip arrays. SM = semimembranosus muscle; n = 240 samples.

10 Muscle protein expression in pigs 1531 Figure 6. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry cluster blot of 16,898 mass-to-charge-ratio (m/z) potential biomarker for muscle type: discrimination by IMAC30 ProteinChip arrays. Receiver operation characteristic (ROC) curve for the discrimination of muscle types shows an area under the curve (AUC) of 0.97, and a sensitivity and specificity of 99 and 62%, respectively. SM = semimembranosus muscle; n = 252 samples. teinchip array confirmed to be potentially good protein biomarker to differentiate LW from LB (Table 2) with a sensitivity of 84%, specificity of 54%, and an AUC of 0.75 (Figure 11). Although these 2 biomarkers presented moderate predictive ability, they should provide a good basis for the study of muscle growth or metabolic and functional processes between breeds. Further studies are needed to increase our understanding of the biochemical causes and consequences of changes in protein expression between breeds and, thus, find biomarkers that may be useful for traceability purposes or as markers for genetic research (Kwasiborski et al., 2009). Protein Expression of the Interaction Between Breed and Muscle Types Results from the interactions between breed and muscle types within the 3 different ProteinChip arrays are reported in the Tables 1, 2, and 3. Twelve peaks on the CM10, 8 peaks on the Q10, and 12 peaks on the IMAC30 ProteinChip arrays were influenced by the interactions between breed and muscle types, suggesting that from these peaks, the differences in mean peak intensity for muscle type are not constant across breed types. From the 12 peaks on the CM10 ProteinChip array, 3,098 and 8,043 m/z peaks discriminate pig muscles according to BL breed (Table 1). The 3,098 m/z peak intensity was 3.7-fold less (P < 0.001) in SM than in LM for BL breed, whereas, in contrast, in the other breeds its intensity was less (P < 0.001) in LM than in SM. As for the 8,043 m/z, its peak intensity was 1.6 times greater (P < 0.001) in SM than in LM for the BL breed. However, in the other breeds the intensity was greater (P < 0.001) in LM than in SM muscle. Additionally, the 3,086 and 13,868 m/z peaks on the IMAC30 ProteinChip array were also established to classify muscles in the BL Figure 7. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry cluster blot of 17,105 mass-to-charge-ratio (m/z) potential biomarker for muscle type: discrimination by IMAC30 ProteinChip arrays. Receiver operation characteristic (ROC) curve for the discrimination of muscle types shows an area under the curve (AUC) of 0.97, and a sensitivity and specificity of 99 and 62%, respectively. SM = semimembranosus muscle; n = 252 samples.

11 1532 Mach et al. Figure 8. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry bar plot of 16,898 mass-to-charge-ratio (m/z) potential biomarker for muscle type: discrimination by IMAC30 ProteinChip arrays. SM = semimembranosus muscle; n = 252 samples. Figure 9. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry bar plot of 17,105 mass-to-charge-ratio (m/z) potential biomarker for muscle type: discrimination by IMAC30 ProteinChip arrays. SM = semimembranosus muscle; n = 252 samples. Figure 10. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry cluster blot of 5,773 mass-to-charge-ratio (m/z) potential biomarker that helped to distinguish Landrace from Landrace Belgian breed: discrimination by CM10 ProteinChip arrays. Receiver operation characteristic (ROC) curve for the discrimination of muscle types shows an area under the curve (AUC) of 0.79, and a sensitivity and specificity of 84 and 50%, respectively. LB = Belgian Landrace breed; LS = Landrace breed; n = 98 samples.

12 Muscle protein expression in pigs 1533 Figure 11. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry cluster blot of 3,743 mass-to-charge-ratio (m/z) potential biomarker that helped to distinguish Large White from Belgian Landrace breed: discrimination by Q10 ProteinChip arrays. Receiver operation characteristic (ROC) curve for the discrimination of muscle types shows an area under the curve (AUC) of 0.75, and a sensitivity and specificity of 84 and 54%, respectively. LB = Belgian Landrace breed; LW = Large White breed; n = 98 samples. breed. The differences in mean intensities of 3,086 and 13,868 m/z peaks were much greater in BL (P < 0.001) than in the other breeds (Table 3). On average, it can be concluded from these results that the BL breed presented more highly differentially expressed proteins for muscle classification than the other pure breeds. To our knowledge, there are no studies that report protein expression patterns of the BL breed between muscles and compare them with other different breeds. However, a likely explanation is that the BL breed can be associated with a greater variation in muscle metabolism and structural proteins, within carcasses (McDonagh et al., 2006). Finally, on the IMAC30 ProteinChip array the 4,046 m/z peak was considered to be a helpful protein for muscle classification within the PI breed. The difference in mean intensity response of the 4,046 m/z peak was much greater (P < 0.001) in PI compared with the other breeds (Table 3). Results from the present study demonstrate that SELDI-TOF-MS is a very efficient method to identify potential protein biomarkers and changes in muscle protein expression of different pure pig breeds and muscle types, as well as their interactions. However, further studies are necessary to identify these potential protein biomarkers. In addition, knowledge of their identities will be essential for increasing the understanding of molecular processes and developing a test tool (such as an ELISA procedure based upon the specific biomarker) in industrial processes, traceability, and genetic research. Despite these weaknesses, the potential of SELDI-TOF- MS to yield comprehensive expression of protein patterns in muscle tissue without the need to first carry out protein separation has proved to be a very promising alternative to other currently applied methods. It is also important to note that the results presented in this study were obtained from a laser intensity of 3,000 nj, 3 types of chip surface and wash conditions, and from the protein water-soluble fraction. The use of greater laser intensity, additional types of chip surfaces, and wash conditions that utilize the inherent binding characteristics of protein samples could be useful to generate more protein expression patterns and identify single potential protein biomarkers. In conclusion, our study demonstrates that breed and muscle type did not affect the number of peaks per spectrum, although it did affect the average intensity of several peaks. Interestingly, a total of 4 proteins were confirmed to be potential biomarkers to differentiate muscle types, whereas 2 were found to be potential protein biomarkers to classify specific breed types. Additionally, several peaks influenced by the interaction between muscle and breed types could correctly classify pig muscles according to their breed. Such promising results clearly suggest it is important that further studies are carried out to validate and identify these protein biomarkers because they may provide promising information for understanding biological questions in the muscle tissue, for traceability purposes, or for genetic research. LITERATURE CITED Bendixen, E The use of proteomics in meat science. Meat Sci. 71: Clarke, W., B. C. Silverman, Z. Zhang, D. Chan, A. Klein, and E. Molmenti Characterization of renal allograft rejection by urinary proteomic analysis. Anal. Surg. 237: Driemel, O., U. Murzik, N. Escher, C. Melle, A. Bleul, R. Dahse, T. Reichert, G. Ernst, and F. von Eggeling Protein profiling of oral brush biopsies: S100a8 and s100a9 can differentiate between normal, premalignant, and tumor cells. Proteomics Clin. Appl. 1: Edwards, D. B., R. O. Bates, and W. N. Osburn Evaluation of Duroc- vs. Pietrain-sired pigs for carcass and meat quality measures. J. Anim. Sci. 81: Edwards, D. B., R. J. Tempelman, and R. O. Bates Evaluation of Duroc- vs. Pietrain-sired pigs for growth and composition. J. Anim. Sci. 84: Essen-Gustavsson, B., K. Karlström, K. Lundström, and A. C. Enfält Intramuscular fat and muscle fiber lipid content in

13 1534 Mach et al. halothane-gene-free pigs fed high or low protein diets and its relation to meat quality. Meat Sci. 38: Hollung, K., H. Grove, E. Mosleth Fargestard, M. Singh Sidhy, and P. Berg Comparison of muscle proteome profiles in pure breeds of Norwegian Landrace and Duroc at three different ages. Meat Sci. 81: Hong, H., Y. Dragan, J. Epstein, C. Teitel, B. Chen, Q. Xie, H. Fang, L. Shi, R. Perkins, and W. Tong Quality control and quality assessment of data from surface-enhanced laser desorption/ionization (SELDI) time-of flight (TOF) mass spectrometry (MS). BMC Bioinformatics 6(Suppl. 2):S5 15. Karlsson, A., A. C. Enfalt, B. Essen-Gustavsson, K. Lundstrom, L. Rydhmer, and S. Stern Muscle histochemical and biochemical properties in relation to meat quality during selection for increased lean tissue growth rate in pigs. J. Anim. Sci. 71: Kwasiborski, A., T. Sayd, C. Chambon, V. Santé-Lhoutellier, D. Rocha, and C. Terlouw. 2008a. Pig longissimus lumborum proteome: Part I: Effects of genetic background, raring environment and gender. Meat Sci. 80: Kwasiborski, A., T. Sayd, C. Chambon, V. Santé-Lhoutellier, D. Rocha, and C. Terlouw. 2008b. Pig longissumus lumborum proteome: Part II: Relationships between protein content and meat quality. Meat Sci. 80: Kwasiborski, A., T. Sayd, C. Chambon, V. Santé-Lhoutellier, D. Rocha, and C. Terlouw Specific proteins allow classification of pigs according to sire breed, rearing environment and gender. Livest. Prod. Sci. 122: Lametsch, R., L. Kristensen, M. R. Larsen, M. Therkildsen, N. Oksbjerg, and O. Ertbjerg Changes in the muscle proteome after compensatory growth in pigs. J. Anim. Sci. 84: Lefaucheur, L., P. Ecolan, L. Plantard, and N. Gueguen New insights into muscle fiber types in the pig. J. Histochem. Cytochem. 50: McDonagh, C., A. M. Mullen, J. P. Kerry, and D. J. Troy Evaluation of inherent variation in porcine M. Thoracis et lumborum and M. Semimembranosus. J. Sci. Food Agric. 86: Obuchowski, N Receiver operating characteristic curves and their use in radiology. Radiology 229:3 8. O Gorman, D., J. C. Howard, V. Varallo, P. Cadieux, E. Bowley, K. MacLean, B. Pak, and B. S. Gan Identification of protein biomarkers in Dupuytren s contracture using surface enhanced laser desorption ionization time-of-flight mass spectrometry (SELDI-TOF-MS). Clin. Invest. Med. 29: Pospiech, E., B. Crzés, B. Mikolajczak, E. Iwanska, and A. Lyczynski Proteins of meat as a potential indicator of its quality A review. Polish J. Food. Nutr. Sci. 57: Ryu, Y., M. H. Lee, S. K. Lee, and B. C. Kim Effects of muscle mass and fiber type composition of longissimus dorsi muscle on postmortem metabolic rate and meat quality in pigs. J. Muscle Foods 17: Sayd, T., M. Morzel, C. Chambon, M. Franck, P. Figwer, C. Larzul, P. Le Roy, G. Monin, P. Cherel, and E. Laville Proteome analysis of the sarcoplasmic fraction of pig semimembranosus muscle: Implications on meat color development. J. Agric. Food Chem. 54: Schipper, R., A. Loof, J. de Goot, L. Harthoorn, W. van Heerde, and E. Dransfield Salivary protein/peptide profiling with SELDI-TOF-MS. Ann. N. Y. Acad. Sci. 1098: te Pas, M. F., J. Jansen, K. Broekman, H. Reimert, and H. Heuven Postmortem proteome degradation profiles of longissimus muscle in Yorkshire and Duroc pigs and their relationship with pork quality traits. Meat Sci. doi: /j.meatsci Wadsworth, J. T., K. Somers, L. Cazares, G. Malik, B. Adam, B. Stack, G. Wright, and O. J. Semmes Serum protein profiles to identify head and neck cancer. Clin. Cancer Res. 10: Xu, Y. J., M. Jin, L. Wang, A. Zhang, B. Zuo, D. Xu, Z. Ren, M. Lei, X. Mo, F. Li, R. Zheng, C. Deng, and Y. Xiong Differential proteome analysis of porcine skeletal muscles between Meishan and Large White. J. Anim. Sci. 87: Yang, S. Y., X. Xiao, W. Zhang, L. Zhang, W. Zhang, B. Zhou, B. Chen, and D. C. Hen Application of serum SELDI proteomic patterns in diagnosis of lung cancer. BMC Cancer 5:83 89.