EXTERNAL SCIENTIFIC REPORT

Transcription

1 EXTERNAL SCIENTIFIC REPORT APPROVED: 12 September 2016 doi: /sp.efsa.2017.en-1090 Bioaccumulation and toxicity of mineral oil hydrocarbons in rats - specificity of different subclasses of a broad mixture relevant for human dietary exposures Jean-Pierre Cravedi a, Koni Grob b, Unni Cecilie Nygaard c, Jan Alexander c a French National Institute for Agricultural Research (INRA), Toxalim Unit, Toulouse, France b Official Food Control Authority of the Canton of Zurich, Zurich, Switzerland c Norwegian Institute of Public Health (NIPH), Oslo, Norway Abstract The purpose of this study was to investigate the fate and effects of mineral oil saturated hydrocarbons (MOSH) in female Fischer 344 rats. Animals were fed control diet or diet containing various MOSH mixtures at concentrations ranging from 40 to 4000 mg/kg feed for up to 120 days. MOSH were analysed in liver, spleen, adipose tissue and remaining carcass at different sampling times. In addition to clinical effects, liver microgranulomas, hepatic inflammation, and disruption of the immune function were the main toxicological endpoints investigated. Arthritis symptoms were specifically studied in dark agouti rats. The results indicate that accumulation of MOSH depends on the mixture tested but always occurred predominantly in the liver and to a lesser extent in adipose tissue and spleen. Strong differences exist between liver and adipose tissue in terms of accumulated hydrocarbons: whereas in adipose tissue the accumulated fraction corresponds to the most volatile part of the administered mixture, in the liver, the most volatile as well as the highest boiling part of the mixture are almost absent. Also the types of hydrocarbons differ. When exposure ceases, a significant decrease of MOSH concentration was observed in the liver, but not in adipose tissue. MOSH exposure results in a significant increase in absolute and relative liver weights, the effect being dose-related, but also dependent on the mixture tested. There were large differences in the ability of the different mixtures to induce liver granulomas. The highest incidence was observed with the mixture containing the highest proportion of n-alkanes, suggesting that this fraction could play a significant role in the development of hepatic granulomas. No effects were found on the immune function, irrespective of the mixture or the dose tested. French National Institute for Agricultural Research (INRA), Official Food Control Authority of the Canton of Zurich (KLZ), Norwegian Institute of Public Health (NIPH), 2017 Key words: mineral oil, MOSH, alkanes, accumulation, granuloma, liver, mixture Question number: EFSA-Q Correspondence: biocontam@efsa.europa.eu EFSA Supporting publication 2017:EN-1090

2 The present document has been produced and adopted by the bodies identified above as author(s). In accordance with Article 36 of Regulation (EC) No 178/2002, this task has been carried out exclusively by the author(s) in the context of a grant agreement between the European Food Safety Authority and the author(s). The present document is published complying with the transparency principle to which the Authority is subject. It cannot be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. Acknowledgements: Florence Blas-y-Estrada, Perrine Aubry, Christel Cartier, Xavier Blanc, Raymonde Dou, Suzette Dumoulin and Aurélie Labrador (from INRA), Monica Andreassen, Hege Hjertholm, Prof. Åshild Vege and Prof. Torleiv Rognum (from NIPH), Laura Barp and Maurus Biedermann (from KLZ) are gratefully acknowledged for their valuable contribution to the study. Suggested citation: Cravedi JP, Grob K, Nygaard UC and Alexander J, Bioaccumulation and toxicity of mineral oil hydrocarbons in rats - specificity of different subclasses of a broad mixture relevant for human dietary exposures. EFSA supporting publication 2017:EN pp. doi: /sp.efsa.2017.en-1090 ISSN: French National Institute for Agricultural Research (INRA), Official Food Control Authority of the Canton of Zurich (KLZ), Norwegian Institute of Public Health (NIPH), EFSA Supporting publication 2017:EN-1090

3 Summary The aim of the study was to assess the accumulation and toxicity of mineral oil saturated hydrocarbon (MOSH) mixtures in rats. Two series of experiments were performed. The first series was based on a broad MOSH mixture representative of the whole MOSH range to which humans are exposed via the diet. The second set of experiments was based on three mixtures that differed in their molecular mass range as well as on n-alkanes. The broad MOSH mixture, ranging from about C 14 to C 50, was tested in female Fischer 344 rats dietary exposed to these hydrocarbons at concentrations of 40, 400 and 4000 mg/kg feed. MOSH were analysed in liver, spleen, adipose tissue and the carcass after exposure during 30, 60, 90 and 120 days as well as after 90 d exposure followed by 30 days depuration. Part of the liver was used for histopathological analyses. For testing immune system, 5 days before the end of the experiment all rats exposed for 120 days were injected with an antigen and the response was measured. Dietary exposure to MOSH at the highest level tested resulted in a significant increase in both absolute and relative liver weight. At the highest dose tested, no steady state was reached at the end of the experiment. After 120 days of exposure to 40 mg/kg feed, ca 10% of the ingested MOSH were recovered from the whole body; after 90 days followed by a depuration period of 30 days, approximately 40% of the MOSH body burden was eliminated. The depuration period resulted in a significant decrease of the MOSH concentration in the liver, but not in the adipose tissue. Accumulation of MOSH occurred mainly in the liver and to a lesser extent in adipose tissue and spleen. Concentrations in the tissues increased far less than proportionally with the dose, rendering linear extrapolation to low doses questionable. In liver and spleen, the maximum retention in terms of molecular mass (simulated distillation) was at n C 29 ; in adipose tissue and carcass it was at n-c 15/16. Comprehensive two-dimensional gas chromatography (GCxGC) was used to characterize the MOSH residues in the tissues with the aim of identifying the most strongly accumulated types. In the liver and spleen, the highly branched and cyclic hydrocarbons predominated, whereas in the adipose tissue it was the n-alkanes and species with main n-alkyl moieties. The histological analyses indicated that dietary exposure to the MOSH mixture affects the granuloma formation in the liver of rats, but this was only evident at the highest dose (4000 mg/kg feed) tested. This effect was not observed after 30 or 60 days of treatment, but appeared after 90 or 120 days of treatment. Hepatic granulomas formed in the group exposed to the highest dose for 90 days were not reversible within the 30 days recovery period. In parallel, male and female dark agouti rats were given a single intradermal injection of 200 µl pristane (positive control) or feed containing 4000 mg/kg pristane or the broad MOSH mixture in various concentrations (0 mg/kg, 40 mg/kg, 400 mg/kg or 4000 mg/kg) for 90 days. Markers previously reported to be associated with arthritis development were analysed. Whereas rats intradermally injected with pristane developed arthritis, none of the animals given pristane or MOSH in the feed developed arthritis symptoms. In the second series of experiment, three MOSH mixtures were tested. The first mixture (S-C25) primarily consisted of branched and cyclic MOSH; approximately 27% of the hydrocarbons exceeded n-c 25. L-C25, the second mixture, virtually exclusively consisted of branched and cyclic MOSH ranging from n-c25 to n-c45. The third mixture (L-C25W) consisted of L-C25 mixed in a 1:1 ratio with a wax of similar mass range (n-alkanes ranging from C23 to C45). Female Fischer 344 rats were fed control feed or feed containing the MOSH fractions S-C25, L-C25 and L-C25W in concentrations of 400, 1000 and 4000 mg/kg feed for 120 days. 3 EFSA Supporting publication 2017:EN-1090

4 MOSH were analysed in liver, spleen, adipose tissue and the carcass. Part of the liver was used for histopathological analyses. For testing the immune system, 5 days before the end of the experiment all rats were injected with an antigen. MOSH exposure resulted in a significant increase in absolute and relative weights of both liver and spleen in the L-C25 and L-C25W groups, but not for rats dietary exposed to the S-C25 mixture. As for the broad MOSH mixture, accumulation occurred predominantly in the liver. In liver and spleen, accumulation of the C26-30 MOSH was higher than that of the C20-25 fraction. Wax constituents, such as n-alkanes and n-alkyl monocyclic naphthenes, were generally well retained in adipose tissue. Although their chemical structure was not elucidated, a series of dominant iso-alkanes accumulated in all tissues. No increase in granuloma formation was observed in any of the three dose groups given the L-C25 mixture. However, the groups fed 4000 mg/kg of the S-C25 mixture demonstrated significantly increased levels versus the control feed group. For the groups fed the L-C25W fraction, the granuloma density was significantly higher than in the control group for all doses tested. The numbers of lymphoid cell clusters in the liver parenchyma and in the liver portal tract were not affected by any dose of L-C25, but were significantly increased for the highest dose of the S-C25 mixture. For the L-C25W mixture, a significant increase of the numbers of lymphoid cell clusters in the parenchyma was observed at low and medium doses, whereas the effect was significant only at the highest dose in the portal tract No significant differences were observed for the KLH-specific IgM concentrations in serum due to MOSH exposure, neither after exposure to 40, 400 and 4000 mg/kg feed of the broad mixture of MOSH, nor to 400, 1000 and 4000 mg/kg feed of the three narrow MOSH mixtures. 4 EFSA Supporting publication 2017:EN-1090

5 Table of contents Abstract...1 Summary Introduction Background and Terms of Reference as provided by the requestor Methodological approaches Accumulation and toxic effects in rats of a broad MOSH mixture Methodologies Preparation of the broad MOSH mixture Feed preparation Animal experiments Sample preparation and extraction for chemical analysis Results and discussions Animal data MOSH concentrations in tissues Molecular mass distribution of accumulated MOSH n-alkanes in adipose tissue Characterisation of accumulated MOSH by GCxGC Main conclusions from chemical analyses Histopathological analyses of the livers Immune function analyses Autoimmune arthritis in dark agouti rats Accumulation and toxic effects of narrow MOSH mixtures Methodologies Preparation of the MOSH mixtures Feed preparation Experimental design Analytical methods Results and discussion Clinical data MOSH concentrations Molecular mass distribution Dose dependence of the mass distribution Accumulation relative to exposure Accumulation of n-alkanes Characterisation of the accumulated hydrocarbons by GCxGC Histopathological analyses of liver samples Immune function analyses Conclusions...84 References...86 Abbreviations...89 Appendix A Validation of the methods for chemical analysis...91 Appendix B Average daily feed consumption of rats (broad MOSH mixture experiment)...96 Appendix C Mass spectra of the dominant multibranched hydrocarbons in the broad MOSH mixture...97 Appendix D Histopathological analysis of the livers EFSA Supporting publication 2017:EN-1090

6 1. Introduction 1.1. Background and Terms of Reference as provided by the requestor In 2012, the EFSA Panel on Contaminants in the Food Chain (CONTAM Panel) adopted a Scientific Opinion on the risks for public health related to the presence of mineral oil hydrocarbons in food. 1 Mineral oil hydrocarbons (MOH) are hydrocarbons mainly derived from crude mineral oils, but also synthesised from coal, natural gas or biomass. They are complex mixtures of mineral oil aromatic hydrocarbons (MOAH) and mineral oil saturated hydrocarbons (MOSH). The latter group includes paraffins (consisting of linear and branched alkanes) and naphthenes (mainly consisting of alkylated cycloalkanes). Due to historical reasons and their chemical complexity, MOH mixtures are not fully resolved analytically and different grades are defined, mainly based on their physico-chemical properties, such as viscosity or boiling point. Due to their wide use for various applications, MOH enter food from many sources, following intended use or as contaminants. Depending on the source, the MOH composition varies both in terms of MOSH and MOAH distribution and concentration. Moreover, for many foods different sources can contribute to the presence of MOH. As a result, humans are exposed to a broad range of MOH of complex and mostly unknown composition. In their scientific opinion related to the presence of MOSH in food, the CONTAM Panel noted that MOSH with a carbon number in the range can accumulate both in humans and experimental animals. Moreover, a series of toxicity studies carried out with several food-grade MOH (essentially MOSH) indicated that repeated exposure to most of those grades caused the formation of liver microgranulomas associated with inflammatory response in rats. The various food-grade MOH tested differed both in their bioaccumulation potential and their ability to cause microgranulomas. Since the MOH grades tested were characterised mainly by their physico-chemical properties and little data on chemical compositions were available, it was not possible for the CONTAM Panel to correlate the observed effects with the types of hydrocarbons or to identify substances (or classes of substances) of greater concern within those mixtures. Moreover, none of the mixtures tested could be considered as representative of the broad range of hydrocarbons to which humans are exposed via the food. These limitations prevented the application of the acceptable daily intake (ADI) values previously established for several food-grade MOH for the risk assessment of MOSH present in food. The CONTAM Panel recommended that efforts should be made to base the toxicological evaluation of MOH on molecular mass ranges and structural subclasses, rather than on physico-chemical properties. The generation of additional data to investigate the relevance of the liver microgranulomas observed in rats to humans in relation to the different structural sub-classes of MOSH was recommended. In addition, the CONTAM Panel recommended further investigation of the possible relationship between oral exposure to MOSH and immunological effects in humans. The present call aimed to address the aforementioned recommendations, and in particular to provide more information on toxicity and bioaccumulation of a MOSH mixture representative of the whole MOSH range to which humans are exposed via the diet, and to identify subfractions of this mixture with higher potential for toxicity and bioaccumulation. In addition, the data generated should clarify the possible relationship between immunological effects and the oral exposure to the representative MOSH mixture. This call for proposals aims at a study to assess the bioaccumulation and toxicological potential of mixtures of mineral oil saturated hydrocarbons relevant to human exposure via food. This study will 1 EFSA Panel on Contaminants in the Food Chain (CONTAM), Scientific Opinion on mineral oil hydrocarbons in food. EFSA Journal 2012;10(6):2704, 185 pp. 6 EFSA Supporting publication 2017:EN-1090

7 provide information on the toxicological profile of mineral oil saturated hydrocarbons and will serve a supporting background document for the possible refinement of the scientific opinion on mineral oil hydrocarbons, following recommendations by the CONTAM Panel in the aforementioned scientific opinion. 1 The objectives of the project resulting from this call for proposals were as follows: study of accumulation and toxicity of a broad MOSH mixture representative of the whole MOSH range to which humans are exposed via the diet; identification of the fraction(s) based on carbon number (and/or molecular weight) and chemical structure with higher bioaccumulation potentials; analysis of the correlation between accumulation and formation of hepatic microgranulomas; analysis of the correlation between the changes in immune functions and the formation of microgranulomas; study of the autoimmune response to oral exposure to a broad MOSH mixture representative of the whole MOSH range to which humans are exposed via the diet. This grant was awarded by EFSA to: The French National Institute for Agricultural Research (INRA), Paris, France Beneficiary: The French National Institute for Agricultural Research (INRA), Toulouse, France Grant title: Combined bioaccumulation/toxicity study on a broad mixture of mineral oil saturated hydrocarbons Grant number: GP/EFSA/CONTAM/2013/01-GA1 This study was conducted by a consortium that consisted of: - co-ordinator: the French National Institute for Agricultural Research (INRA), Toulouse, France - co-beneficiary: the Norvegian Institute of Public Health (NIPH), Oslo, Norway - subcontractor: the Official Food Control Authority of the Canton of Zurich (Kantonales Labor Zürich, KLZH), Zürich, Switzerland 1.2. Methodological approaches In order to address the accumulation and toxic effects of MOSH mixtures to which humans are exposed via the diet, several mixtures were investigated in different animal experiments. A first set of experiment, based on a broad MOSH mixture ranging from about C 14 to C 50, was carried out in female Fischer 344 rats and male and female dark agouti rats. Female Fischer 344 rats were selected as they are known to be the most sensitive model to MOSH (Firriolo et al., 1995), whereas the dark agouti strain is reported to develop autoimmune arthritis after injection of MOSH (Sverdrup et al., 1998; Holmdahl et al., 2001). The preparation and the characterisation of the mixture, as well as the experimental designs are presented in Section 2. A second set of experiments was performed on subfractions derived from the broad mixture. The composition of these fractions (referred to as narrow mixtures in this report) was based on the results from the first series of experiments, in order to determine the characteristics of MOSH having the highest potential to bioaccumulate and the most effective toxicological impact. Details on the preparation and the characterisation of the mixture, as well as the experimental designs are reported in Section 3. In addition to generating more information on the retention and depuration of different classes of saturated hydrocarbons in mammals, this project was designed to provide original data on the effect of MOSH on immune functions and on the potential for different categories of MOSH to form granulomas. 7 EFSA Supporting publication 2017:EN-1090

8 2. Accumulation and toxic effects in rats of a broad MOSH mixture 2.1. Methodologies Preparation of the broad MOSH mixture A MOSH mixture ranging from about C 14 to C 50 was prepared with similar concentrations over a wide mass range. The range of molecular masses was intended to be broader than that previously observed in human tissues (approximately C 16 -C 45 ; Barp et al., 2014). The lower mass limit was also determined by the risk of evaporation during preparation of the feed and subsequent storage. The following components were used: 1. Merck, Paraffin highly liquid Ph Eur, BP NF, JP K Fluka Paraffin wax Ph Eur, low viscosity, Fluka Paraffin wax Ph Eur, high viscosity, Shell, Catenex Ph 941 FU As no white mineral oil of sufficient volatility was available, the volatile fraction (around C 16 ) was isolated by vacuum distillation from oil 1: from 700 ml of this oil, the first 25 ml were discarded. Fractions 2 and 3 amounted to 25 ml each. Three such distillations were performed and the fractions combined. The composition of the mixture (in g/kg) used for the experiment is shown in Table 1. Table 1: Combination of white mineral oils to obtain the mixture used for the animal experiments Component g/kg Fraction 2, oil1 88 Fraction 3, oil1 176 Oil Oil Oil The composition as determined by on-line High Performance Liquid Chromatography Gas Chromatography Flame Ionization Detection, HPLC-GC-FID (MOSH fraction) is shown below as an overlay with a mixture of n-alkanes. The molecular masses ranged from n-c 13 to approximately n-c C 20 /min 350 C Figure 1: On-line HPLC-GC-FID chromatogram of the mineral oil mixture for the animal tests overlaid with that of some n-alkanes to show the molecular mass distribution 8 EFSA Supporting publication 2017:EN-1090

9 Feed preparation A standard pelleted diet (AIN-93M, Table 2) as described by Reeves et al. (1993) was selected. Table 2: Composition of AIN-93M rat diet Ingredient % Cornstarch Maltodextrine 15.5 Casein 14.0 Sucrose 10.0 Fiber (cellulose Durieux) 5.0 Soybean oil 4.0 Mineral mix (AIN-93M-MX) 3.5 Vitamin mix (AIN-93-VX) 1.0 L-Cystine 0.18 Choline bitartrate 0.25 Tert-butylhydroquinone Prior to the preparation of the feed, the major ingredients were analysed by on-line HPLC-GC-FID for mineral oil saturated hydrocarbons (MOSH) or polyolefin oligomeric saturated hydrocarbons (POSH, partly from the plastic pouch in which the samples were sent for analysis) as background contamination to rule out a disturbing interference with the experiments. The following results were obtained (Table 3): Table 3: MOSH/POSH concentrations in the feed components Ingredient Cellulose Maltodextrine Sucrose Mineral mix AIN-93M Casein Corn starch Vitamins AIN93VX MOSH/POSH (mg/kg) The contamination of the ingredients was considered acceptable. Before being incorporated into the diet, the broad MOSH mixture was dissolved in soybean oil and stirred for 4 h at 40 C. MOSH were incorporated into the diet at 3 nominal concentrations: 40, 400 and 4000 mg/kg, and in each case an equivalent mass of soybean oil was replaced by the MOSH solution. The same procedure was used to prepare a diet containing pristane (4000 mg/kg) and utilized in the experiment on dark agouti rats (see Section 2.1.3). Diet concentrations were verified at the beginning (March 2014) and the end of the experiment (September 2014). The measured doses were within 88-95% of the nominal concentrations (Table 4). Table 4: MOSH/POSH concentrations in the diets (values are means ± SD, n=3) Nominal Measured concentrations concentration Beginning End (mg/kg feed) ± ± ± ± ± ± ± EFSA Supporting publication 2017:EN-1090

10 Animal experiments Female Fischer 344 rats Experimental design The study was performed at INRA facilities, according to the EU Directive 2010/63 on the protection of animals used for scientific purposes; the protocol was approved by the French ministry of research and education (APAFIS ). Female Fischer 344 rats, 4 weeks of age (average weight = 54.4 ± 7.3 g) at the beginning of the experiment were used in this experiment. Five days before starting the dietary exposure to MOSH, animals were randomly allocated to one of the 19 cages (five individuals per cage), and were maintained at temperature and light control (21 C ± 2 C, 12 h light/dark cycle). Four cages were randomly assigned to the control group and the remaining cages randomly assigned to one of the other groups (five cages per group). Exposure of the animals started on 23/03/2014. Pelleted feed and tap water were available ad libitum. Rat euthanasia (CO 2 inhalation) were scheduled after 1, 2, 3 and 4 months. At each sampling time, 5 animals per group (from five different cages) were sacrificed, resulting in an equivalent reduction of the number of animals per cages. Feed consumption was recorded twice a week and body weights were recorded weekly. Data were statistically analysed with the GraphPad Prism 6 software using a one way ANOVA analysis and a Dunnett s multiple comparisons test. Table 5: Experimental design MOSH concentration in Sampling time (days) diets (mg/kg feed) * 0 (control) 5, Tp, G 5, Tp, G, IF 5, Tp, G, IF 5, Tp, G, IF , Tp, G 5, Tp, G, IF 5, Tp, G, IF 5, Tp, G, IF 5, Tp, G 400 5, T, G 5, T, G, IF 5, T, G, IF 5, T, G, IF 5, T, G , T, G 5, T, G, IF 5, T, G, IF 5, T, G, IF 5, T, G T: Tissue MOH analyses (liver, fat, spleen, remaining carcass) are carried out on 4 animals; Tp: tissues of the 4 animals are pooled before analysis; G: Liver granulomas are analysed in 5 animals; IF: determination of the immune functions using KLH assay (5 animals). *: 90 days of dietary exposure to MOH followed by one month of control diet. Analyses carried out on this group are similar to those performed during the bioaccumulation phase. In order to investigate the possible depuration of accumulated hydrocarbons, 50% of the rats exposed to 40, 400 and 4000 mg/kg MOSH during 3 months were fed the control diet during the last month, resulting in three additional groups (named 90+30, see Table 5), whereas other rats continued to be exposed to the same contaminated diet. Based on feed consumption values (Annex 2), 40, 400 and 4000 mg MOSH /kg feed concentrations correspond to daily doses of about 2, 22 and 222 mg/kg body weight (bw). For immune function analyses, the experiment was conducted according to the OECD test guideline 443. Briefly, 5 days before the end of the experiment, all rats exposed for 120 d were injected (intravenous, iv) with an antigen (300 µg/kg bw Native HMW keyhole limpet hemocyanin, KLH, from Stellar biotechnologies, CA, USA) for determination of the immune function. Immediately after euthanasia, terminal blood sample from vena cava was collected in glass tubes without anticoagulant and serum was prepared and frozen in Eppendorf tubes at 80 C. Samples were shipped in dry ice to NIPH/Oslo. At each sampling time, the liver, spleen and abdominal adipose tissue were excised, weighed and prepared for further analyses. The liver left lobe was fixed in formalin and processed for conventional histopathological examination by mounting in paraffin. Paraffin blocks were shipped to NIPH/Oslo, where the samples were sectioned and stained EFSA Supporting publication 2017:EN-1090

11 Remaining liver, adipose tissue and spleen were placed in polyethylene bags (tested for POSH/MOSH residues) and frozen in liquid nitrogen for chemical analysis. Blood from animals sacrificed after 1, 2 and 3 months was collected in heparinized glass tubes and stored at -20 C. The gastrointestinal tract (including digestive content) was removed from the carcass before homogenization and freezing for potential chemical analysis. Liver, adipose tissue, spleen and remaining carcasses were sent to KLZH for MOSH analysis. The gloves used for the dissection of the animals were checked for the absence of disturbing MOSH and POSH (Appendix-A). Histopathological analyses of the livers - Tissue sectioning Tissue blocks from left liver lobes were sectioned with the microtome adjusted to 4 µm. From each tissue block four sections were stained with hematoxylin-azophloxine-saffron (HAS) for morphometric quantification of granulomas, clusters of inflammatory cells in portal tracts, clusters of inflammatory cells (separate from the granulomas) in liver parenchyma as well as for semiquantitative evaluation of degree of liver cell vacuolization. To avoid counting the same granulomas and clusters of inflammatory cells in more than one tissue section, all together 60 serial sections were cut from each tissue block and the first, the 20th, the 40th and the 60th sections were selected for staining. Based on previously reported tests, it was decided that scoring four sections from each tissue block would give a representative granuloma density from each tissue block. The histological examination was performed using Leica DMLB Microscopes. - Morphometric analyses All sections were evaluated by the same examiner. The inter observer reproducibility and the intra observer reproducibility were performed blindly and evaluated by applying a κ-test. The inter- and intra-observer reproducibility of histopathological evaluation of granuloma were found satisfactory; κ=0.6, p<0.05 (moderate) (Figure 1A, Annex 3), and κ=0.87, p<0.05, (very good) (Figure 1B, Annex 3), respectively. To decide the number of sections necessary to obtain a representative granuloma density from each tissue block, seven sections from two cases were evaluated (Figure 2, Annex 3). Due to the relatively modest variation in granuloma density, we decided to examine four sections from each tissue block. All scoring was performed blinded to the histopathologist. All granulomas and clusters of inflammatory cells in portal tracts and liver parenchyma were counted in each section. Vacuolization of liver cells were semi-quantitatively scored as none, slight (less than 5% of the liver cells), moderate (5-15% of the liver cells) and abundant (more than 15% of the liver cells). These scores were denoted as 0, 1, 2 and 3, respectively, for graphical presentation. The density of granulomas and clusters of inflammatory cells were given per cm2. For measuring the areas of the tissue sections, a Wild Heerbrueg stereo microscope was applied equipped with a 6 times magnification lens and a photo tube with a reduction ocular (0.5). The picture was transmitted to the planimetry unit by a Leica DFC450 camera. Calculation of area was performed by a LEICA application system (LAS). Due to the small group size (n=5) and the non-normal distribution, non-parametric Mann-Whitney U test were performed to investigate between group differences. Dark Agouti male and female rats The aim of this study was to perform a subchronic oral study to determine whether peroral exposure to a broad MOSH mixture can induce clinical signs and biological markers for autoimmune arthritis in a rat model. The arthritis-prone dark agouti (DA) rats was used, since this rat strain has been reported to develop autoimmune arthritis after injection (intraperitoneally or intradermally) of various mineral 11 EFSA Supporting publication 2017:EN-1090

12 oils (Sverdrup et al., 1998, Holmdahl et al., 2001). Percutaneous application of a white mineral oil also demonstrated mild and transient arthritis in this rat model (Sverdrup et al., 1998). In the main experiment, groups of 10 rats (5 males and 5 females) were exposed via the diet to three different doses of the broad MOSH mixture as described in Section (40, 400 and 4000 mg/kg feed) or pristane (4000 mg/kg feed). The negative control group (n=10) received a diet free of MOSH. A positive control group (n=10), also fed the control feed, was exposed to a single intradermal injection of pristane (200 µl). This report includes results both from a pilot study performed to establish the model with the positive control, as well as the main study with dietary exposures to MOSH. Experimental design A pilot study, as previously reported, was performed to establish the model using a positive control. DA female rats, ordered from Harlan, of about 7 weeks of age were acclimatized for two weeks before entering the experiment. The rats were housed two and two together in Makrolon cages placed in Scantainer filter cabinets. The animal facilities were maintained at 21 ± 2 C and 35-75% relative humidity, with a 12:12 h light/dark cycle. MOSH-free feed (AIN 93M pellets, MP Biomedicals), and tap water, was given ad libitum. On day 0, a single intradermal injection of 200 µl pristane (Sigma Aldrich) was applied to six female rats, while six female rats received a single intradermal injection of 200 µl physiological buffer (negative controls). The injections were given at the base of the tail, under Isoflurane gas anaesthetics. The animals were kept for 40 days, unless sacrificed earlier for animal welfare reasons (see below). In the main experiment, groups of 10 DA rats (5 males and 5 females) were exposed via the diet to three different doses of the broad MOSH mixture (40, 400 and 4000 mg/kg feed) or pristane (4000 mg/kg feed). The negative control group (n = 10) received a control feed free of MOSH (AIN 93M pellets, MP Biomedicals). A positive control group for arthritis development (n = 10), also fed the control feed, was exposed to a single intradermal injection of pristane (200 µl). The animals were maintained under the conditions described above, housed 2 or 3 animals per cage, and terminated after 90 days of exposure. The animals in the positive control group were terminated after 40 days unless sacrificed due to strong symptoms (see below). Animal weight was monitored weekly. The feed intake was measured at 1 and 10 weeks after start of treatment by weighing the feed per cage at a week s interval, and calculating the consumed feed per group of animals per week. Clinical arthritis score The clinical manifestation of arthritis in the animals were determined by the use of a macroscopic scoring system ranging from 0 to 4 for each of the four limbs (according to Holm et al., 2002), yielding a score of 0-16 per animal. Monitoring of arthritis scores were performed by the same investigator twice a week during the whole experimental period, blinded with regard to group affiliation. This score was expressed as arthritis incidence, severity (maximum score over the whole period) and day of onset. Rats showing signs of discomfort were given analgesic treatment (Temgesic) subcutaneously (s.c.) two to three times a day. In cases of very strong arthritis symptoms (score > 10 and/or weight loss > 10%) the rats were sacrificed for animal welfare reasons. Serum analyses Due to the subjective character of the arthritis score system, we wanted also to measure some objective, biological markers in the animals, which also may allow for detection of pre-disease states. Sera collected at termination in the pilot study were diluted 1:2 and levels of cytokines interleukin (IL) IL-10, IL-17, IL-1β and IL-6, were analysed using ELISA kits from Invitrogen by Life Technologies and ebiosience respectively, according to the manufacturer s instructions. Serum levels of cytokines IL-22, TNFα (ebioscience, San Diego, CA, USA) and IgG and immunoglobulin M (IgM) rheumatoid factor (RF; MyBioSource, San Diego, CA USA) were analysed by ELISA according to the manufacturer s 12 EFSA Supporting publication 2017:EN-1090

13 instructions. A selection of these cytokines and RF, based on positive findings in the pilot study, was determined in all rat sera from the main experiment. Splenocytes analyses Additional biological markers of arthritis, i.e. expression of TLR2 and TLR3 on splenocytes, were determined in the main experiment. Spleen single cell suspensions were prepared and counted as previously described (Hansen et al., 2011), and re-suspended in a wash buffer containing PBS, 0.1% natriumazide and 1% FCS, for a cell concentration of 2x107 cells/ml. 1 x106 cells from each cell suspension were first incubated with BD Fc blocker (0.5 µl/well; BD Biosciences, CA) at 4 C for five minutes. Thereafter, cells from each animal was stained with a mixture of antibodies against TLR2 (H 175) and TLR3 (N-14) for 30 minutes at 4 C, followed by a wash step and secondary staining with goat anti-rabbit IgG-FITC and donkey anti-goat IgG-APC (all form Santa Cruz Biotechnology Inc., Dallas, USA). One negative control sample per spleen was stained either only with secondary antibodies or with no antibodies. Cells were washed and re-suspended in wash buffer before analyses on a LSRII flow cytometer (BD Bioscience). Statistical analysis In the pilot study, Mann Whitney tests were used to compare the results from the injection and control group. In the main experiment, two-way ANOVAs (with treatment group and sex as the two factors) were performed, followed by pairwise comparisons using Tukey post hoc test. The group injected with 200 µl pristane was not included in the statistical tests since the research question was whether there are any effect of dietary exposures to MOSH or pristane and also since most endpoints are not directly comparable due to different termination day (thus age) for the injection group Sample preparation and extraction for chemical analysis Extraction and sample preparation The analytical method was derived from that used for human tissues (Barp et al., 2014), but downscaled to meet the smaller amounts of sample available. The thawed tissues were immersed in ethanol at a 1:1 (w/w) sample/ethanol ratio and homogenized using a Polytron (PT GT). To 2 g tissue/ethanol homogenate weighed into a 10 ml centrifuge tube, 4 ml ethanol and 20 µl internal standard solution (Biedermann and Grob, 2012a) were added. Single spleens were of approximately 0.5 g and were directly homogenized in 5 ml ethanol. The ground carcasses were extracted without further homogenization, weighing into the centrifuge tube 1 g of sample and adding 5 ml of ethanol. The ethanolic homogenates were intensively shaken and left standing for at least 1 h to allow ethanol entering the particles and exchange with the water. After centrifugation, ethanol was decanted into a 30 ml centrifuge tube and the tissue residue homogenized with 5 ml hexane. After extraction overnight at ambient temperature, the tube was centrifuged and the supernatant hexane added to the previously collected ethanol. To separate the hexane phase from ethanol, approximately 10 ml water was admixed. HPLC-GC-FID analysis MOSH were determined by on-line coupled normal phase HPLC-GC-FID, as described by Biedermann and Grob (2012a). A 90 µl extract in hexane was injected into a 25 cm 2 mm i.d. HPLC silica gel column (LiChrospher Si 60, 5 µm) using hexane as eluent at 300 µl/min. The MOSH fraction was eluted between 1.8 and 3.3 min. The column was back-flushed with dichloromethane at 500 µl/min for 9 min. HPLC-GC transfer occurred by the retention gap technique with partially concurrent eluent evaporation (Grob, 1991). A 7 m 0.53 mm i.d. uncoated, deactivated pre-column was connected to the solvent vapour exit and a 15 m 0.25 mm i.d. separation column coated in the laboratory with a 0.13 µm film of PS-255 (a dimethyl polysiloxane; Fluka, Buchs, Switzerland). The GC oven 13 EFSA Supporting publication 2017:EN-1090

14 temperature was programmed from 55 C (6.5 min) to 350 C at 20 C/min. The area corresponding to the MOSH was determined by integration of the whole hump of largely unresolved material above the baseline transferred from a blank run (obtained by injecting hexane). Internal standards and endogenous hydrocarbons were subtracted from the hump as described in Biedermann and Grob (2012b). Concentrations were calculated referring to the internal standard cyclohexyl cyclohexane (Cycy). The method was validated as described in Annex 1. The blank covering the whole chemical analysis was repeatedly checked simulating the extraction procedure and the chromatographic analysis without tissue. The yield of the extraction was determined for the four rat tissues analysed, liver, spleen, adipose tissue and carcass, by second extractions under more severe conditions. Repeatability and homogeneity of the sample was determined from 4 replicate extractions from the same rat liver homogenate. The relative standard deviation (RSD) was 3%. Four injections of the same extract resulted in a RSD of 0.4%. Results of 3 samples from different parts of the ground carcass had a RSD of 5.2%. Linearity of the data was checked by the analysis of the extract from liver tissue containing 91 mg/kg MOSH. This extract was mixed at various ratios with an extract from cow liver without detectable MOSH to yield concentrations ranging between 1.3 and 91 mg/kg. R2 was The detection limit largely depended on the stability of the baseline, as humps were often broad and flat. Three analyses at 1.3 mg/kg MOSH in liver resulted in a coefficient of variation of 18%. From this, a limit of quantification (LOQ) of approximately 1 mg/kg and a limit of detection (LOD) of about 0.5 mg/kg were estimated. Comprehensive two dimensional gas chromatography Comprehensive two-dimensional GC (GCxGC), as described by Biedermann et al. (2015), was used for the characterization of the MOSH composition. A TRACE GCxGC gas chromatograph from Thermo Scientific (Milan, Italy) was used, first with a cryogenic dual jet modulator (carbon dioxide), later with a two-stage thermal loop-modulator ZX-2 from Zoex Corporation (Houston, US). It was equipped either with a flame ionization detector or a Bench TOF-dx mass spectrometer (Almsco International, Llantrisant, UK). GC involved a 3 m x 0.53 mm i.d. uncoated precolumn deactivated in the laboratory by phenyl dimethyl silylation to enable injection of up to 20 µl by the retention gap technique, connected to a 15 m x 0.25 mm i.d. DB-17 first dimension separation column (50% phenyl methyl polysiloxane of 0.25 µm film thickness; Agilent, Santa Clara, USA). When used with the cryogenic dual jet modulator, the second dimension separation was on a 2.5 m x 0.15 mm i.d. column coated in the laboratory with a µm film of PS-255 (dimethyl polysiloxane from Fluka, Buchs, Switzerland); when used with the two-stage thermal loop-modulator it was a 3.2 m x 0.15 mm i.d. column with the same coating. Cryomodulation was performed on the outlet of the first dimension column at intervals of 5 s using CO 2. The Zoex modulator involved a constant stream of 7.5 L/min air cooled to -84 C and hot air at 400 C pulsed every 6 s for 350 ms on the inlet of the second dimension column. The loop consisted of a 1 m section of the second dimension separation column, leaving 2.2 m of this column for the separation process. MOSH were isolated from the tissue extracts using the on-line HPLC-GC method described above, but transfer from HPLC to GCxGC occurred off-line by collection of the HPLC MOSH fraction from the transfer valve in small glass vials and manual injection into GCxGC. Ninety µl of 200 mg/ml solutions in hexane were injected into HPLC. The MOSH fractions of 450 µl were diluted or concentrated to adjust the amount injected into GCxGC to the amounts of MOSH in the samples. A programmed temperature vaporizing (PTV) injector equipped with an on-column needle guide liner was used for on-column injection. The temperature of the injector was increased at 0.1 C/s from 60 to 340 C (1 min) EFSA Supporting publication 2017:EN-1090

15 For MS, the column outlet was attached to a 30 cm x 0.20 mm i.d. deactivated fused silica transfer line connecting to the ion source (kept at 300 C), with the interface heated to 330 C. Using FID, the carrier gas, hydrogen, was applied at 80 kpa in constant pressure. Using MS, helium was supplied at a constant flow rate of 1.0 ml/min. The oven temperature was increased at 5 C/min from 50 C (3 min) to 320 C (0.5 min). Data were collected for a scan range comprising amu at 40 Hz. GC Image software (Zoex Corporation, Houston, US) was used for processing data. MOSH need to be isolated from other constituents of the tissue extracts. This was achieved by HPLC using the same technique as for the HPLC-GC method described above. However, transfer from HPLC to GCxGC occurred off-line, by collection of the MOSH fraction in small vials and manual injection into GCxGC Results and discussions Animal data Table 6: Effect of treatment on the weight of animals and organs Exposure period (d) Dose (mg/kg feed) Body weight ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 5.98 Liver 5.69 ± ± ± ± 0.25** 6.57 ± ± ± ± ± ± ± 0.47* 6.86 ± 0.11** 6.39 ± ± ± ± 0.46* 6.92 ± ± ± ± 0.32 Spleen 0.44 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± HSI ± ± ± ± ** ± ± ± ± ± ± ± * ± ** ± ± ± ± ** ± ± ± ± Values are in grams and are means from 5 animals ± SD. HSI: hepatosomatic index (liver/body weight ratio) : animals dietary exposed during 90 days to MOSH mixture, then fed for 30 days the control diet. *: significantly different from control (p<0.05); **: significantly different from control (p<0.01). No deaths, abnormal behaviour or altered physical appearance were observed during the study conducted on Fischer 344 rats. Body weight gain (Table 6) and feed intake (see Annex 2) were not affected by dietary exposure to the broad MOSH mixture, included at the highest dose tested. Whereas spleen weight was not affected by MOSH intake, the weight of the liver was significantly higher in animals exposed to the MOSH mixture at 4000 mg/kg for 30, 90 and 120 days than in controls (Table 6). A significant difference was also observed for the hepatosomatic index (HSI, 15 EFSA Supporting publication 2017:EN-1090

16 corresponding to the ratio of liver weight to body weight) in the same groups, as well as in the group exposed during 90 d to 400 mg/kg MOSH, compared to controls. It is concluded that at high levels the broad MOSH mixture may affect liver metabolism, resulting in an increase in absolute and relative liver weights. Similar effects were observed in several studies dealing with the pathological response of rats to MOSH mixtures. In a 90-day study in Fischer 344 rats, Baldwin et al. (1992) found a dose-related increase in absolute liver weight in male or female rats fed diets containing 5000 ppm or more of food grade white oil. Absolute and relative liver weights were significantly increased in 90-day study in Fischer 344 rats fed a dietary dose of 2000 ppm and higher of different food grade white oils and waxes (Smith et al., 1996). The effect more pronounced in females than in males. No effect was observed at 200 ppm. In the same study, absolute and relative spleen weights were also increased, but only at a dietary level of ppm. Griffis et al. (2010) reported that a diet containing 2000 ppm low melting point paraffin wax resulted in a significant increase in absolute and relative hepatic and splenic weights in Fischer 344 rats, but not in Sprague Dawley rats. In our case, the fact that compared to control, no difference was observed in animals exposed to MOSH during 90 d, then fed with the control diet for 30 d, suggests that this effect is reversible MOSH concentrations in tissues The MOSH concentrations (mg/kg) measured in liver, spleen, adipose tissue and carcass are listed in Table 7. Values for rats exposed to 400 and 4000 mg/kg MOSH in the feed are reported for four individual animals (Rep1-Rep4), those from the 0 and 40 mg/kg groups to a single animal (Rep1) as well as the pooled tissues from the three others (Rep2). Mean concentrations were clearly highest for the liver. MOSH concentrations in spleen were slightly higher than in adipose tissue. Those in the carcasses (free of gastrointestinal tract) were low, indicating that there was no major further deposit of MOSH in the rat body. The preferential partition to liver observed at all concentrations and sampling times tested is in agreement with the data reported by Smith et al. (1996). In a 90-d study carried out in Fischer 344 rats fed a diet containing mg/kg of different food grade white oils and waxes, the concentration of MOSH in liver was 4 to 22 fold higher than in perirenal adipose tissue, depending of the properties of the oil. There was one exception, a high melting point wax (average molecular weight of 630 Da; average carbon number of 22-80), for which higher residues levels were noted in adipose tissue than in liver (Smith et al., 1996). At 90 days and for comparable doses tested, the concentrations now measured in the liver were in the same order as found by other authors (Firriolo et al., 1995; Miller et al., 1996; Smith et al., 1996; Trimmer et al., 2004; Griffis et al., 2010). The preferential retention of MOSH in the liver was also observed by Tulliez et al. (1975) in Large White pigs (35 kg body weight at the beginning of the experiment) fed during 3 months a diet containing 1.13 or 3.23% of a mineral oil ranging from about C15 to C28 and containing 10, 73 and 17% of n-, branched- and cyclo-alkanes, respectively. At the end of the experiment with the higher dose, the MOSH concentration in the liver was 1353 mg/kg, but only 293 mg/kg in abdominal fat (lard) and 65.2 mg/kg in spleen. The MOSH concentrations in the human adipose tissues, however, were in the same order as in the liver and spleen. This might be related to the far longer period for accumulation in this tissue, while MOSH entering the liver were partly eliminated. In fact, with the duration of exposure the MOSH concentration in the adipose tissue of the rat tended to increase relative to the liver (Table 7). MOSH were also detected in control animals. The concentrations were relatively high taking into account that the control feed only contained 1.6 mg/kg MOSH EFSA Supporting publication 2017:EN-1090

17 Table 7: MOSH concentrations (mg/kg) in liver, adipose tissue, spleen and carcass for the 30, 60, 90, 120 and day (d) experiments Time (d) Dose Liver Spleen Adipose tissue Carcass (mg/kg feed) Rep1 Rep2 Rep3 Rep4 Average Rep1 Rep2 Rep3 Rep4 Average Rep1 Rep2 Rep3 Rep4 Average Rep1 Rep2 Rep3 Rep4 Average <LOQ <LOQ - - < LOQ < LOQ < LOQ - - < LOQ < LOQ < LOQ - - < LOQ EFSA Supporting publication 2017:EN /2002, this task has been carried out exclusively by theauthor(s) in the context of a grant agreement between the European Food Safety Authority and the output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards theissues addressed and the conclusions reached in the present document, without prejudice to the rights ofthe author(s).

18 Variability between animals Variability between the animals, calculated for the high doses for which the tissues of four individuals had been analysed, was mostly between 10 and 20% (Table 8), which is significantly higher than the measurement uncertainty. The highest variability (up to 43%) was observed in the adipose tissues of the 120 d experiment. This (modest) variability was of interest in the context of the far higher variability of the granuloma formation. Table 8: Variability of the tissue concentrations of the four individual animals per group, expressed as relative standard deviations (RSD, %) Time (d) Dose Liver Average RSD (%) Spleen Average RSD (%) Adipose tissue Average RSD (%) Carcass Average RSD (%) MOSH concentrations versus time The curves of MOSH concentrations over time (Figure 2) show a rapid increase to day 30, followed by a slower increase, which then seems to accelerate again from 90 to 120 d in spleen and adipose tissue. No steady state was reached (uncertain for the liver at the highest dose). Feeding the control diet during 30 d after exposure to MOSH during 90 d (90+30 d) resulted in a reduction of the MOSH concentration in the liver by about 53% for the groups 40 and 400 mg/kg and by 44% for the 4000 mg/kg dose. The reduction in the spleen was weaker. In the adipose tissue, the MOSH concentrations even increased during the 30 d without exposure to MOSH: by 55, 66 and 69% for the 40, 400 and 4000 mg/kg doses, respectively, and were similar to those after continuous exposure up to day 120. These data not only indicate that discontinuation of the exposure does not result in the release of MOSH from the adipose tissue, but also suggests a transfer from other tissues EFSA Supporting publication 2017:EN-1090

19 Fat = abdominal adipose tissue Figure 2: MOSH concentrations (mg/kg) in the tissues versus time of exposure, shown as group means with bars denoting the standard deviations Uptake versus exposure Figure 3 plots the concentrations in the tissue against the dose and shows a strong deviation from linearity: concentrations in the tissue increased far less than proportionally with intake. When the dose was increased from 40 to 4000 mg/kg, the concentrations in the tissues increased by factors of rather than by a factor of 100. This means that linear extrapolation from high dose to a lower exposure results in serious underestimation of the tissue concentration. The control feed contained approximately 1.6 mg/kg MOSH (of a somewhat different composition), but the concentrations in the liver and spleen of the animals fed with it were again far less than 25 times lower as would be expected by constant relative uptake (Table 7). Concentrations in the liver and spleen of the control animals were times higher than expected by linear extrapolation from the 40 mg/kg dose (Table 9). Referring to the 4000 mg/kg dose, the tissue concentrations were times higher. The deviation from linearity was higher for the spleen than the liver EFSA Supporting publication 2017:EN-1090

20 Fat: abdominal adipose tissue. Figure 3: MOSH concentrations (mg/kg) in the tissues versus dose (mg/kg feed) with standard deviations shown as bars Non-linearity was reported in the literature. Firriolo et al. (1995) measured MOSH concentrations of 5.6 and 1.7 mg/g in the livers of Fischer 344 and Sprague Dawley rats, respectively, after feeding 2 g/kg MOSH (different mixture) in the diet for 90 days. For a ten times higher dose, these values only increased to 8.2 and 4.1 mg/g. Table 9: Deviation of the MOSH concentrations in the liver and spleen of the animals fed with the control feed when dividing the measured concentration by that linearly extrapolated from the animals fed with the 40 and the 4000 mg/kg dose Exposure (days) Dose (mg/kg Deviation feed) Liver Spleen This deviation from linearity means that the extrapolation of internal exposure from animal experiments carried out at high concentrations to the far lower human exposures is questionable: the relative uptake into human tissue is likely to be considerably higher, and if a safety factor of 100 were applied for the derivation of health-based guidance values, most of it would be consumed by the inadequate extrapolation from the high doses in animals EFSA Supporting publication 2017:EN-1090

21 The higher relative tissue concentrations at low dose may partly explain the high concentrations in human tissues compared to the estimated exposure, as discussed in Barp et al. (2014) accumulation during far longer times being another factor. Estimated MOSH accumulation in the whole animal Bioconcentration factors (BCFs) for MOSH in liver and abdominal adipose tissue were calculated as the ratio of the MOSH concentration in the tissue to that measured in the feed (Table 10). As for all experiments described below, the pooled tissues of the animals having received the control feed or the 40 mg/kg dose were used, but the tissue of an individual animal was used for the higher doses. For both liver and adipose tissue, statistical differences (p<0.05) were observed between groups exposed to 400 and 4000 mg/kg feed, irrespective of the duration of exposure, suggesting reduced uptake at the highest concentration tested. For the liver, values are substantially higher than those calculated from the data published by Smith et al. (1996) for female Fischer-344 rats exposed during 90 days to various mineral oils and waxes, but similar to those derived from the data reported more recently by Trimmer et al. (2004) and Griffis et al. (2010). For adipose tissue, values are in the same order as those estimated from the data reported by Smith et al. (1996). Table 10: MOSH bioconcentration factors (ratio of the concentration in tissues to the concentration in feed) in liver and adipose tissue Time (d) Dose (mg/kg) Values are based on measured concentrations in feed. SD: standard deviation. Bioconcentration factor Liver Adipose tissue Mean SD Mean SD Table 11 shows the percentages of the MOSH found in the body of the rats compared to the exposure. Exposure was calculated by the feed consumption (summed data averaged for 3-4 days; Table in Appendix-B) multiplied by the MOSH concentrations (data from pooled animals). Amounts in the body were summed up by multiplying the mass of the tissue with the concentrations measured. For the adipose tissue, the mass of the aliquot analysed was used, the remaining fat being part of the carcass. The gastro-intestinal tract was not taken into consideration, as it was removed from the carcass. They included MOSH remaining in the ingested feed as well as those absorbed into the tissue, the latter being a potential deposit from which the MOSH were transferred to the adipose tissue with the delay shown above. For the lowest dose and after 30 d of exposure, the retained MOSH amounted to 10.9% of those ingested. This proportion decreased with extended exposure to 6.2% after 120 d, and from 7.1 to EFSA Supporting publication 2017:EN-1090

22 when control feed was used for 30 d after 90 d of exposure, indicating a substantial depuration during 30 d. The corresponding values for the higher doses were substantially lower (see above). Table 11: Percentage of MOSH recovered from the animals (except gastrointestinal tract) compared to the amount ingested Exposure (d) Dose (mg/kg) 40* (17.2) (10.3) (11.9) (7.0) (9.9) *: In parentheses: percentage for the most retained molecular mass. These data refer to the total of the administered broad MOSH mixture, including the constituents of lowest and highest molecular mass which did not show up in the tissues. The retention of the most efficiently retained MOSH was correspondingly higher. The data in parentheses refer to the fraction of a width in GC retention time corresponding to a single carbon atom at the maximum concentration in the tissues (C 25 for liver, C 24 for spleen and C 18 for fat and carcass). These concentrations were divided by the exposure to the equivalent fraction in the feeds. Exposure at 40 mg/kg was used, since it is closest to the human exposure. Distribution in the rat Roughly half of the MOSH retained in the rats were located in the liver (40 mg/kg dose, being closest to realistic exposure; Table 12). The spleen contained % of the total. The values in the adipose tissue analysed and the carcass were added up since the MOSH in the carcass most probably were primarily in the adipose tissue remaining in the carcass. This was the deposit of the other half of the MOSH. The MOSH in the gastrointestinal tract were not considered. Table 12: Percentage of the MOSH of the 40 mg/kg dose in the tissues analysed at the time points of the analysis Exposure (d) % MOSH in tissue Liver Spleen Adipose tissue + carcass* * without gastrointestinal tract. It is plausible that as the liver is the most efficient organ for MOSH biotransformation, the proportion of the MOSH in the liver was reduced after 30 days of depuration with control feed, whereas the limited metabolic activity, if any, resulted in a relative increase in the percentage of MOSH in adipose tissue and carcass Molecular mass distribution of accumulated MOSH Preliminary remark: GC characterizes the volatility distribution, which is linked to, but not identical with the molecular mass distribution: branched hydrocarbons are eluted before the n-alkanes of the same carbon number (mass), cyclic hydrocarbons (naphthenes) generally afterwards (unless carrying strongly branched alkyl groups). iso-paraffins and alkylated naphthenes are eluted in a range of 22 EFSA Supporting publication 2017:EN-1090

23 retention times corresponding to approximately ±3 carbon atoms from the n-alkane of the same carbon number. This selectivity depends on the stationary phase of the column. It is, however, common practice to use GC retention times relative to n-alkanes on a dimethyl polysiloxane stationary phase to approximate molecular mass distributions of hydrocarbons ( simulated distillation ; ASTM D2887). Chromatographic overlays Figure 4 shows HPLC-GC-FID chromatograms of the four tissue types analysed (120 d exposure; 40 mg/kg dose for liver and spleen, 400 mg/kg dose for adipose tissue and carcass). The chromatogram of the broad MOSH mixture added to the feed is overlaid to point out the part of the MOSH retained by the tissue. Attenuation is adjusted to result in humps of similar height. For the liver (top left), the chromatogram of the lowest dose is shown, since the composition of the MOSH residues varied with the exposure level (see below). The hydrocarbons ranged from about n-c 16 to about n-c 40, with a maximum abundance between n-c 23 and n-c 31. The most volatile as well as the highest boiling part of the mixture were virtually absent. The n-alkanes C 16 -C 22 were also absent, whereas in the later eluted part distinct signals showed up which were almost absent in the chromatogram of the MOSH mixture applied. Fat tissue: abdominal adipose tissue. Internal standards (11, Cycy, 13 and Cho) are marked in chromatograms. Figure 4: Overlay of the MOSH added to feed (black) and found in the different tissues (blue) Compared to the human liver (Barp et al., 2014), the distribution in the rat liver was broadened towards the smaller molecular masses: the human liver contained virtually no constituents below n-c 20. This may be the result of a different MOSH composition humans are exposed to or (more likely) to more time for exhalation. However, the finding of McKee et al. (2012) that no MOSH up to n-c 20 are accumulated in rat livers, and onto which the BfR based the migration limit of 4 mg/kg for the MOSH C 17 -C 20 (Bfr, 2012), is not confirmed (Bfr, 2012). The MOSH detected in spleen had a distribution similar to that in the liver, but the proportion of the MOSH below C 20 and above C 30 was somewhat higher and the maximum more distinctly at n-c 23. The MOSH detected in the adipose tissue corresponded to the most volatile part of the MOSH fed to the animals. They ranged up to about n-c 31 and were centred on n-c 19. The MOSH profile of the carcass was similar to that of the adipose tissue, suggesting that adipose tissue was the main MOSH deposit in the carcass EFSA Supporting publication 2017:EN-1090

24 Concentrations by carbon number For a more quantitative comparison of the molecular mass distributions, the concentrations of the MOSH in the tissues were determined for segments of a width corresponding to a single carbon atom in n-alkanes, cutting at the retention times corresponding to the end of the n-alkane peaks by which the segment was named. In Figure 5, the data is shown for the 90 d exposure period followed by 30 d depuration, as this best singles out the retained hydrocarbons. Concentrations were normalized on the corresponding ones in the feed, with the effect that the highest values were obtained for the lowest dose (absorption being highest). The volatility/mass distributions of the MOSH in liver and (to a reduced extent) in spleen depended on the dose: at the lowest dose (40 mg/kg feed), the concentration was almost constant from n-c 24 to n-c 30, whereas at the highest dose the concentration of the n-c 24 segment was approximately 2.5 times higher than that at n-c 30. As the elimination rate of the high mass constituents hardly increased with the dose, the relative uptake must have selectively decreased with increasing dose. The distributions were again similar in adipose tissue and the carcass. Fat: abdominal adipose tissue. Concentrations refer to the total. Figure 5: MOSH concentrations determined for segments corresponding to a single carbon atom in GC retention time, normalized on the concentrations in the feed for the exposure to MOSH during 90 d followed by 30 d with control feed To determine the most strongly retained hydrocarbons in terms of molecular mass, the areas of the MOSH segments corresponding to a single carbon atom in rat tissues were divided by those of the corresponding sections in the feed and normalized on the maximum value, i.e. a value of 100% means maximum retention in a given tissue. In Figure 6 the data is shown for the d experiment. Since the distribution depended on the dose, the lowest dose (40 mg/kg) was selected, being the nearest to the human exposure EFSA Supporting publication 2017:EN-1090

25 The maximum retention in liver and spleen was at n-c 29, which is within the mass range of the Class I 2 mineral oils so far considered to be of low concern. It is close to the maximum concentrations in the corresponding human tissues that varied between n-c 25 and n-c 28 (Barp et al., 2014). It should be noted that the maxima in humans were not normalized on exposure (since the molecular mass distributions of the MOSH humans are exposed to is unknown), i.e. the maxima in human tissues could have been influenced by the composition of the ingested MOSH. 100 Normalized area ratio (%) Liver Spleen Fat Segment (n-alkane) Fat: abdominal adipose tissue. Figure 6: Maximum retention with regard to molecular mass: areas for single carbon segments in tissues divided by those in the feeds and normalized on the maximum (40 mg/kg exposure during days) In the adipose tissue, the maximum retention was observed for the most volatile segments of the MOSH added (n-c 15/16 ). This is far below the maximum at n-c 23/24 observed in human subcutaneous abdominal adipose tissue (Concin et al., 2008; Barp et al., 2014) n-alkanes in adipose tissue Since the MOSH added to the feed only contained significant amounts of n-alkanes up to about n-c 21, their retention could be investigated merely for this range, and as no n-alkanes of this mass range were detectable in liver and spleen, the focus was on adipose tissue. In the mass range of n-c 14-21, the n-alkanes amounted to 11% of the MOSH mixture, but in the MOSH extracted from the adipose tissue, they reached 30%, indicating preferential accumulation. Data for the individual n-alkanes are shown in Figure 7 (400 mg/kg dose for analytical reasons), pointing out a particularly strong enrichment for the lowest molecular masses. Figure 8 refers to the accumulation of the C n-alkanes after different exposure periods (400 mg/kg dose). The proportion of the n-alkanes decreased from 32% after the exposure during 30 d to 23 and 19% after 90 d and 90 d plus 30 d control diet, respectively. 2 For the definition of Class I MOSH see, e.g. documents/3073.pdf 25 EFSA Supporting publication 2017:EN-1090

26 Fat: abdominal adipose tissue. Figure 7: n-alkanes in the feed (blue) and in the adipose tissue (red); peak areas of individual n-alkanes divided by the total area of MOSH eluted up to n-c 21 expressed as % Figure 8: n-alkanes in the adipose tissue after exposure to MOSH during 30, 90 and d (400 mg/kg dose); percentages of the total of the C 14 -C 21 MOSH Characterisation of accumulated MOSH by GCxGC After the 3-4 months of exposure, only a minor proportion of the administered MOSH was recovered in the rat body (Table 11). An unknown proportion has not been absorbed from the gastrointestinal tract. However, since in the d experiment the retained proportion was decreased by a factor of about 3 compared to the 30 d exposure, there must have been substantial elimination during the first 30 d. Considering the low MOSH solubility in aqueous media, elimination must have gone through a metabolic conversion. Such processes and their rates are expected to depend on the structure of the hydrocarbons, which raises the question, which types of hydrocarbons are reduced or enriched relative to the others. Residues might provide information about the structural elements that are responsible for strong accumulation EFSA Supporting publication 2017:EN-1090

27 Presently comprehensive two-dimensional GC (GCxGC) is the most powerful method for separating complex mixtures of hydrocarbons. As shown for human tissues (Biedermann et al., 2015), it enables characterizing the MOSH by separating cyclic from open chain MOSH and grouping open chain MOSH by the degree of branching and the cyclic ones by the degree of alkylation. In GCxGC the solutes eluted from a first dimension column of conventional dimension are cryotrapped for a short period of time (e.g. 6 s). The trapped material is then released as a sharp band and separated another time by fast chromatography on a short and narrow bore second dimension column coated with a stationary phase of different polarity. Results are mostly recorded in plots with the first dimension separation shown on the x-axis and the second dimension on the y-axis. Signals stand vertical to the plain and their height is visualized by the intensity and/or colour of the spot. Mass spectrometry (MS) enables to pick out certain groups of hydrocarbons, either by their molecular mass or characteristic fragments. For instance, compared to open chain hydrocarbons, monocyclic and polycyclic naphthenes have a mass reduced by 2 units per ring, which is conclusive for rings provided the presence of olefins can be ruled out. The sharp peaks obtained by GCxGC presuppose fast scanning, such as by time of flight (TOF) MS. MS does not enable quantitative determinations, since the standards required for calibration of the response are mostly not available, but GCxGC-MS is a useful tool for comparing MOSH compositions. Quantitative measurements are performed by flame ionization detection (FID), since FID provides approximately equal response per mass for all hydrocarbons. Selectivity of the GCxGC system used Often GCxGC is configured as a separation by volatility on a non-polar first dimension column, followed by separation by polarity in the second dimension. However, this concept is not properly applicable to saturated hydrocarbons, as steric effects dominate polarity. In fact, as shown by Vendeuvre et al. (2004), van der Westhuizen et al. (2010, 2011) and Biedermann et al. (2014), the reverse configuration is preferable for the analysis of the MOSH, i.e. a first separation on a polar stationary phase and a second on a non-polar one. A description of this selectivity was given by Biedermann et al. (2015). The GCxGC-FID plot of the MOSH mixture administered to the rats is shown in Fig. 9. The centres of the signals representing the n-alkanes are interconnected by a more or less horizontal line (similar retention in the second dimension). The series of n-alkanes started after n-c 13 (internal standard) and ended at about n-c 22. Beyond n-c 22, the dominant spots on the line represent little branched hydrocarbons. There are numerous, mostly irregularly positioned multibranched hydrocarbons above the row of n- and iso-alkanes, including pristane and phytane (labelled). Increased branching reduces the retention time in the first dimension separation and increases it in the second, resulting in signals drifting upwards to the left. Below the n- and little branched paraffins are the cyclic hydrocarbons (naphthenes). Some of them, such as the n-alkyl cyclopentanes and hexanes, form distinct signals (pointed out by a line interconnecting their centers). In the background are clouds of unresolved, highly isomerized components. The lower the position in the plot, the more rings a hydrocarbon includes. In fact, the steranes and hopanes (4 and 5 rings, respectively) formed signals near the bottom. Highly branched alkylation or multialkylation of naphthenes results in higher positions in the plot, which is the main reason for the slanted rows from the lower right to the upper left in the left part of the plot. The plot visualizes the complex, but still partly regular composition of the MOSH added to the feed EFSA Supporting publication 2017:EN-1090

28 C13 (n-tridecane), C20 (n-eicosane), Cycy (cyclohexyl cyclohexane) and Cho (cholestane) are internal standards. Figure 9: Comprehensive two-dimensional GC plot (GCxGC-FID) of the MOSH mixture added to the rat feed Hydrocarbons from sampled tissues Figure 10 shows GCxGC-FID plots obtained for the hydrocarbons extracted from liver, spleen and adipose tissue (top, centre and bottom row, respectively) for the 400 mg/kg dose and exposure during 120 d or d. In order to improve comparability of the composition, the injected aliquots and attenuations were adjusted to the total concentrations determined by on-line HPLC-GC-FID. As also observed by on-line HPLC-GC (Figure 4), the hydrocarbons in liver included neither the most volatile nor the highest molecular mass constituents of the mixture added to the feed. Compared to the applied MOSH (Figure 9), most distinct signals are strongly reduced, leaving behind an enhanced cloud of unresolved material, as it was also observed in human livers (Biedermann et al., 2015). By the second dimension retention time, most compounds forming the cloud are likely to belong to the naphthenes. High isomerization of naphthenes implies branching of the alkyl groups, multialkylation of the ring or differing ring conformation if the compound is polycyclic. The multibranched open chain components, however, remained similar in relative intensity. Also their composition remained similar, which means that either all multibranched hydrocarbons persisted or were eliminated to a similar extent. The signals at the height of the n-alkanes almost totally disappeared up to about n-c 25, but were enhanced beyond this point. The main ones do not represent n-alkanes. This phenomenon will be further investigated in later sections. Comparing the plot from the liver after the d exposure with that obtained after the 120 d exposure, a slight further reduction of the volatiles might be noted, particularly for the multibranched paraffins, but no selective elimination among those remaining. In the 120 d exposure experiment, also hydrocarbons ingested shortly before sacrification could have been expected. Their almost complete absence suggests that the difference between the MOSH in the feed and in the liver is due either to selective uptake or rapid elimination of some hydrocarbons and a far slower one of the others. No differences in the MOSH composition were noted between the 400 and the 4000 mg/kg dose (not shown) EFSA Supporting publication 2017:EN-1090

29 Fat tissue: abdominal adipose tissue. Figure 10: GCxGC-FID plots of MOSH from rat tissues exposed to the 400 mg/kg dose during 120 d (left) and 90 d followed by 30 d depuration (right) The hydrocarbons extracted from the spleen differed little from those in the liver. In human tissues (Biedermann et al., 2015), the degradation of the components forming distinct signals was more advanced in the spleen than in the liver, but this was less pronounced in the rat tissues analysed. The plots obtained from the adipose tissue clearly differ from those of the liver and spleen. They are characterized by an increased dominance of the distinct signals compared to the cloud of highly isomerized components. Among the n-alkanes, the natural n-c 29 and n-c 31 are prominent, but the other strong signals starting from about C 23 are iso-alkanes. Apart from the changed composition in terms of mass distribution, most signals are the same as in the mineral oil mixture administered. Composition in broad classes To investigate whether the proportion of the broad classes of (i) multibranched open chain alkanes, (ii) n- plus little branched alkanes and (iii) naphthenes would be changed in the tissues, cuts were placed above and below the band of the n-alkanes (Figure 11). The internal standards Cycy and Cho were cut out. Total areas were integrated over the baseline of the second dimension chromatograms. Integrated areas of the classes were expressed as percentages of the sum of the three classes EFSA Supporting publication 2017:EN-1090

30 Figure 11: Separation of the MOSH in the oil added to the feed into three broad classes Table 13: Summed areas of the multibranched paraffins, the n-alkanes plus little branched paraffins and the naphthenes, separated as shown in Figure 11 and expressed as percentages of the total area Tissue Exposure Multi branched n-alkanes + little paraffins branched paraffins Naphthenes Broad MOSH mixture Liver 4000 mg/kg, 120 d mg/kg, 120 d mg/kg, d mg/kg, d Mean Spleen 4000 mg/kg, 120 d mg/kg, 120 d mg/kg, d mg/kg, d Mean Adipose tissue 4000 mg/kg, 120 d 8, mg/kg, 120 d mg/kg, d mg/kg, d Mean The data in Table 13 refer to 120 d and d exposure and the higher doses (because of the precision of the data). The naphthenes clearly dominated the MOSH in the mixture used. In the liver and spleen, their proportion in the total was similar to that in the MOSH admixed to the feed (top line; roughly 60%). Also for the other two classes, the differences between the administered MOSH mixture and the MOSH residues in liver and spleen were apparently minor. The 30 d depuration period had no significant effect. In the adipose tissue, the percentages of the n- and iso-alkanes were clearly higher than in the administered oil. This was influenced by the enrichment of the n-alkanes. As there were no major differences in the proportions of the 3 classes of MOSH, these three classes (each representing complex mixtures) do not reflect selectivity in the uptake into the organism or in the metabolism and excretion. For instance it means that naphthenes are neither preferentially accumulated nor preferentially eliminated. Since selective enrichment or depletion was noted, they 30 EFSA Supporting publication 2017:EN-1090

31 must have occurred in all three classes to a similar extent, which means that the structural elements determining uptake, degradation and elimination seem to be similarly distributed in the three classes. Data from selected masses Since no preferential accumulation of one of the broad classes of hydrocarbons was observed for liver and spleen, the compositions were studied in more detail by extraction of characteristic ions in GCxGC-MS. Abundance of the selected ions can vary strongly among the components, i.e. signal intensity does not represent concentration, but such analyses enable the detection of changes between the oil added to the feed and the residues in the tissues. Multibranched hydrocarbons The fragment m/z 183 (C13H27, unknown structure) seemed fairly selective for the multibranched paraffins. Figure 12 compares the GCxGC-MS fragmentogram with the FID plot to point out the spots extracted by this ion. There are minor differences in retention times owing to the different carrier gases (helium in GCxGC-MS; hydrogen in GCxGC-FID) and column outlet pressures. The row of n-alkanes helps orientation. Four major peaks of the multibranched hydrocarbons are encircled for comparison with the plots obtained from the tissues shown below. Figure 12: GCxGC plots of the paraffin mixture selectively recorded by the fragment m/z 183 and by FID for comparison Figure 13 shows the residues detected by m/z 183 in the MOSH from liver (top), spleen and adipose tissue for the 120 d exposure (left) and the 90 d exposure with 30 d depuration (right). The plots are 31 EFSA Supporting publication 2017:EN-1090

32 from the 4000 mg/kg dose, but the results from the 400 mg/kg dose were identical. To facilitate the comparison of the residues in the tissues with the broad MOSH mixture added to the feed, the line interconnecting the n-alkanes was introduced. The position of pristane, phytane and n-c 20 are marked in the liver, but these components were not detected. Furthermore, the four dominant multibranched hydrocarbons encircled in Figure 12 are pointed out to facilitate orientation. Fat tissue: abdominal adipose tissue. Figure 13: GCxGC-MS plots of m/z 183 (multibranched hydrocarbons) in rat tissues after exposure during 120 d (left) and d (right) In the liver and the spleen the early eluted peaks recorded by m/z 183 largely disappeared, including pristane and phytane, as seen in the HPLC-GC analysis. Furthermore, the remaining main signals (including the encircled ones) were strongly reduced relative to the cloud of unresolved highly isomerized hydrocarbons (compare with Figure 12). The earlier eluted ones are more strongly reduced than the later ones, presumably because of higher volatility rather than because of structural properties (their structure remained unknown, but they are unlikely to be homologues as they do not show up in regular intervals). Hence no selectivity in degradation is visible in this way. No significant differences between the 120 d and the d experiments were noted, such as signals disappearing during depuration. The plots from the adipose tissues show dominant distinct signals, as in the mixture added to the feed, and include n-alkanes. Pristane and phytane formed strong signals. Although being strongly reduced in intensity, also the encircled multibranched components have the same composition as in the administered oil. Again, the different exposure regimes (time and dose) did not show obvious differences EFSA Supporting publication 2017:EN-1090

33 A closer analysis of the ten dominating components, the mass spectra of which are shown in Appendix-C does not really change the conclusion (Figure 14). The residues after 30 d and the 120 d exposure seems almost identical, with slightly more of the volatiles in the 30 d experiment, in which the MOSH mixture administered during the last days plays a more prominent role. During the 30 d depuration, the loss of volatiles progressed slightly further for the liver and in more pronounced manner for the spleen. Peak 8 was reduced in the spleen, but there is no strong change like a component having largely disappeared. Fat: abdominal adipose tissue. Figure 14: Relevant sections of the GCxGC-MS plots of m/z 183 focusing on the multibranched hydrocarbons in the tissues of rats exposed to the 4000 mg/kg dose for 4 exposure scenarios Monoalkylated monocyclic naphthenes The fragments m/z 68 and m/z 82 are fairly specific for monoalkylated cyclopentanes and cyclohexanes, respectively (Biedermann et al., 2015). In Figure 15 the sum of these two fragments is shown for the administered oil and compared with the FID plot to position these in the whole MOSH mixture. The dominant signals form rows with regular intervals corresponding to a single carbon atom (pointed out by a line). The cyclopentanes and the cyclohexanes of the same carbon numbers are hardly separated in neither of the two dimensions. The row of signals above the n-alkyl cyclopentanes and cyclohexanes represents components the structure of which has not been identified (Biedermann et al., 2015) EFSA Supporting publication 2017:EN-1090

34 Figure 15: GCxGC plots of the broad MOSH mixture added to the feed selectively recorded by the fragments m/z 68 plus 82 and by FID for comparison As shown by Figure 16, in the extracts of liver and spleen the two rows of signals of the n-alkylated monocyclic naphthenes and the unidentified compounds eluted above those as well as the other distinct peaks virtually disappeared, leaving behind an unstructured cloud of apparently highly isomerized material. This supports the interpretation from the retention time that at least this part of the cloud consists of cyclic compounds. The high degree of isomerization must be from the alkylation, i.e. ramification (multialkylated compounds would have hardly produced the same ions). Some of these isomers seem to most strongly resist biotransformation, but it was not possible to further describe the relevant structural element. No difference was observed between the 120 d and the d exposure, which means that no significant amounts of hydrocarbons from the last days of exposure are visible. In the adipose tissue, however, components forming distinct signals persisted, while the cloud of unresolved material remained in the background as in the administered MOSH mixture EFSA Supporting publication 2017:EN-1090

35 Fat tissue: abdominal adipose tissue. Positions of pristane, phytane and C20 are indicated as well as the line to support orientation. Figure 16: GCxGC-MS plots of m/z (mainly cyclopentanes and cyclohexanes) in rat tissues exposed to the 4000 mg/kg dose during 120 d (left) and d (right) Alkylated C26 hydrocarbons in liver and spleen To obtain information about the composition of hydrocarbons of the same carbon number, the molecular ions of the C 26 hydrocarbons were selected from the extracts of the livers of rats exposed to 4000 mg/kg doses during 30, 90, and 120 d (Figure 17). The C 26 hydrocarbons were chosen since they are positioned in the center of the accumulated material such that their composition is relatively little dependent on volatility. As no olefins were removed by epoxidation (Biedermann et al., 2015), mass deficiencies of 2 amu were indicative of a saturated ring, i.e. the plots of m/z 364, 362, 360 and 358 corresponded to the molecular ions of C 26 naphthenes with 1-4 rings (26:1, 26:2, 26:3 and 26:4). There were fragments of larger hydrocarbons with the same masses, but the retention of the corresponding substances was shifted to the upper right (higher retention in the first dimension), forming clouds above the line drawn into some of the plots EFSA Supporting publication 2017:EN-1090

36 Comparison of these naphthenes in the administered MOSH mixture (top row) with those in livers of rats exposed to 4000 mg/kg MOSH for 30, and 120 d. Figure 17: Relevant sections of the GCxGC-MS plots of m/z 364, 362, 360 and 358, corresponding to the molecular ions of C 26 naphthenes with 1-4 rings (26:1, 26:2, 26:3 and 26:4) The intensity of the molecular ion strongly depends on the structure, i.e. the apparent composition is distorted. For instance, the n-alkyl substitution for the monocyclics dominating in the admixed oil, the C 21 -cyclopentane (21cyclo5) and C 20 -cyclohexane (20cyclo6) labelled in the top left plot, is underrepresented. However, the residues in the tissue can be compared with the administered MOSH mixture. Injected amounts and attenuation were adjusted to provide plots of similar intensities for the different tissues, but were the same for the four plots from the same tissue. In the MOSH mixture added to the feed (top row), slanted oval clouds are observed, presumably reaching from the little branched constituents at the lower right to the highly branched, perhaps multiple substituted ones at the upper left. The n-alkyl monocyclics, C 21 -cyclopentane (21cyclo5) and C 20 -cyclohexane (20cyclo6), are at the lower right end of the 26:1 cloud and served as centre point for the lines drawn to facilitate orientation in the plots. With additional rings, this cloud was moved downwards. The 5-ring naphthenes were almost undetectable (not shown). The clouds obtained from the liver have a similar relative intensity as in the added MOSH mixture, ruling out substantial selectivity by ring number. Also the changes in the shape of the clouds are minor: apart from the almost complete disappearance of the n-alkyl monocyclics, some further losses are observed at the lower right of the clouds (pointed out by the vertical lines), indicating easier biotransformation and/or elimination of the compounds with a major straight alkyl chain. This was accentuated with longer exposure and still slightly more pronounced for the d than the 120 d experiment, reflecting depuration EFSA Supporting publication 2017:EN-1090

37 In conclusion, the number of rings did not seem to strongly influence the elimination of the naphthenes. The n-alkyl monocyclics disappeared and there is some indication that also other little branched monoalkylated naphthenes were preferably eliminated. Distinct peaks in alkylated monocyclic naphthenes 7 distinct peaks pointed out and characterized by their mass spectrum. Figure 18: GCxGC-MS from the oil admixed to the feed; plot of m/z 336, the molecular ion of the alkylated monocyclic naphthenes Among the monocyclic naphthenes a series of distinct signals of the C 24 hydrocarbons were characterized by MS (Figure 18). The pair of partially separated peaks at the lower right of the band represents the n-c 19 -cyclopentane and the n-c 18 -cyclohexane and is characterised by dominating m/z 69/97 and 83, respectively. On the basis of the intensity of these fragments, all except of P3 seem to be cyclopentane derivatives EFSA Supporting publication 2017:EN-1090

38 Fat: abdominal adipose tissue. Peak assignments: refer to Figure 19. Figure 19: GCxGC-MS plots of extracted molecular ions of C24:1 for the MOSH mixture applied (4000 mg/kg; 120 d) and the tissues analysed Figure 19 shows the plots of m/z 336 from the liver, spleen and adipose tissues after administration of the MOSH mixture during 30, 90, and 120 d. In liver and spleen, n-c 19 -cyclopentane, n-c 18 -cyclohexane and P1 were virtually absent after the 30 d exposure. The large peaks P2 and P3 in the applied MOSH mixture are still visible, but strongly reduced. P4 seems similarly dominant as it was in the administered oil. After 90 and 120 d, P2 was further reduced, and disappeared after a depuration period of 30 d, indicating elimination within weeks. After the 90 d exposure and 30 d of depuration, only P4 and P5 were left, though at reduced size. These results are consistent with those of Tulliez and Bories (1979) who found an efficient metabolism of n-alkylated monocyclo paraffins in rat through oxidation of the alkyl chain. After the 90 d exposure and 30 d of depuration only P4 and P5 were left, though at reduced size. By their positions in the plot they are likely to be highly branched, multiple-alkylated or both. This suggests decreasing elimination with increasing branching or multisubstitution. In the adipose tissue no such trend is visible (low sensitivity because the low abundance in this range of molecular mass) EFSA Supporting publication 2017:EN-1090

39 Quantitative data on alkylated C24 hydrocarbons Quantitative data for uncorrected MS response was obtained for the clouds of C 24 hydrocarbons represented by the molecular ions for 0-4 rings (experiment analogous to Figure 17). The volumes of the clouds were integrated by drawing a polygon around the relevant cloud (neglecting the fragments of higher mass hydrocarbons), using the baseline in the second dimension chromatograms. The data shown in Table 14 were normalized on the sum of the volumes for a given tissue and exposure. In the MOSH mixture used, the naphthenes with 2 and 4 rings seemed to dominate the mixture (uncorrected response; the apparent absence of saturates is such an artefact). In the liver and spleen, the bicyclic and tricyclic naphthenes were enriched on cost of the mono- as well as tetra- and pentacyclic ones. Table 14: Normalized data (percentages) obtained from molecular masses extracted for C 24 hydrocarbons with 0-5 deficiencies of 2 mass units (uncorrected MS response) Percent of total 4000 mg/kg; 120 d 4000 mg/kg; d MOSH fed Liver Spleen Liver Spleen C24:0 0.1 <0.1 <0.1 <0.1 <0.1 C24: C24: C24: C24: C24: Alkylated C 20 hydrocarbons in adipose tissue Extracts from the adipose tissue were analysed analogously, but owing to the lower molecular masses of the MOSH present, the C 20 hydrocarbons were investigated (Figure 20). Again the top row is from the oil admixed to the feed, the rows below from the adipose tissues of the rats exposed to 4000 mg/kg of the MOSH mixture. No significant differences in the composition of the administered MOSH mixture and the residues in the tissues were found, such as loss at the low right end of the clouds (see vertical lines). However, there seemed to be a relative loss of bicyclic naphthenes EFSA Supporting publication 2017:EN-1090

40 Fat tissue: abdominal adipose tissue. Comparison of the administered MOSH mixture (top row) with the residues in adipose tissue after exposure to 4000 mg/kg MOSH. Figure 20: GCxGC-MS plots of m/z 280, 278, 276 and 274, corresponding to the molecular ions of the C 20 naphthenes with 1-4 rings (20:1, 20:2, 20:3 and 20:4) In conclusion, the compositional changes detected over the period of observation were modest. To some extent, this may be explained by selective uptake into the tissue. However, since there must have been substantial biotransformation, this also suggests strongly different elimination rates: some types of hydrocarbons seem to be eliminated within hours or days, whereas for others no significant elimination was detected over the observation period. Some constituents in the unresolved cloud may be more rapidly eliminated without getting visible even at the high resolution achieved by GCxGC Main conclusions from chemical analyses After 120 d of exposure to 4000 mg/kg of the broad mixture of MOSH, (averaged) concentrations in liver, spleen and adipose tissue reached 5511, 383 and 274 mg/kg, respectively. During a depuration period of 30 d after 90 d of exposure, the MOSH concentrations in liver and spleen decreased by 34 and 36%, respectively, but not in the adipose tissue. Relative standard deviations for the concentrations in liver and spleen were between 5.6 and 22%, averaging 11 and 13%, respectively. Concentrations in liver and spleen still increased from day 90 to 120, indicating that no steady state or plateau had been reached. In the adipose tissue, the increased even seemed to accelerate. Tissue concentrations were far from proportional to the administered doses in view of the following evidence: when increasing the dose from 40 to 4000 mg/kg, they increased between 4.0 and 11.5 times, rather than 100 times. The same was observed when comparing the control feed containing 1.6 mg/kg MOSH with the feed with added 40 mg/kg MOSH: the tissue concentrations increased by factors between 3.1 and 12 instead of by a factor of 25. This means that extrapolation of 40 EFSA Supporting publication 2017:EN-1090

41 the internal exposure from a high experimental dose to a far lower real exposure results in a severe underestimation. The overall retention of the MOSH fed to the animals was 10.9% after 30 d, 6.2% after 120 d and 3.9% after 90 d followed by the 30 d depuration. This refers to the 40 mg/kg dose, whereas the values were lower at the higher dose because of the reduced relative uptake mentioned above. Retention depended on the molecular mass. Referring to the fraction of the highest retention (GC retention window of a width corresponding to a single carbon atom), the retention amounted to 17.2, 9.9 and 7.0%, respectively. Roughly 50% of the retained MOSH were located in the liver. In the spleen it was around 0.6%. The other roughly half was in the adipose tissue and the carcass (as the composition in the adipose tissue and the carcass were largely the same, it is assumed that the MOSH in the carcass was in fatty tissue). The MOSH retained in the adipose tissue were strongly different from those in the liver and spleen. In liver and spleen the maximum relative retention was at n-c 29 (simulated distillation by GC). Hydrocarbons below n-c 19 and above n-c 40 were virtually absent. In the adipose tissue the maximum retention was at the low molecular mass end of the mixture, which was n-c 15. Retention rapidly dropped to n-c 22 and remained at a lower level up to about n-c 34. The MOSH added to the feed only contained significant amounts of n-alkanes up to about n-c 21. They were largely absent in liver and spleen, but after the 30 d exposure, in the MOSH in the adipose tissue their proportion increased from 11% in the MOSH added to the feed to 32%. Their proportion decreased with longer exposure, indicating preferential elimination. GCxGC was used for analysing the composition of the MOSH and the compositional changes from the mixture added to the feed to that in the tissues. A rough classification indicated that the MOSH mixture used in the experiments consisted of 31% n-alkanes and little branched paraffins, 9.9% multibranched paraffins and 59% naphthenes. There was no significant change in the composition by this classification in liver and spleen, but a significant shift to the open chain hydrocarbons in the adipose tissue. As there was substantial elimination in liver and spleen, it means similar elimination in all three classes. A more detailed analysis showed the reduction or elimination of most components forming distinct GCxGC signals in liver and spleen, i.e. of the well-defined, dominant constituents, though with the exception of the multibranched hydrocarbons. In the mixture tested, the n-alkanes, n-alkyl monocyclic naphthenes and constituents with mainly unbranched structure belonged to the most efficiently eliminated species, leaving behind a largely unstructured cloud of unresolved hydrocarbons. This cloud represented highly isomerized hydrocarbons, such as strongly branched alkanes as well as naphthenes with complex alkylation and polycyclics of various conformations. This was in contrast to the adipose tissue, where the distinct signals seemed to gain in intensity compared to the cloud of unresolved constituents in the background, i.e. not only the n-alkanes were enriched. No significant compositional differences were observed for the doses of mg/kg. There were little changes in the composition by varying the duration of the exposure, including the depuration period. Only the alkylated monocyclic naphthenes gave indications that resistance against biotransformation increased with ramification and/or multiple-alkylation. This suggested that for most constituents elimination is either rapid (in the order of hours or days) or slow compared to the observation period (i.e. taking more than a few months) Histopathological analyses of the livers Histopathological scoring of all samples was performed, and is presented in figures 21, 22 and 23. There were no significant differences in granuloma density between the groups after 30 and 60 days of exposure (Figure 21A and 21B). After 90 and 120 days of exposure, the granuloma density was 41 EFSA Supporting publication 2017:EN-1090

42 significantly higher in the group of rats fed MOSH formula 4000 mg/kg feed versus controls (Figure 21 C and D). This difference was still significant for rats fed the highest dose of MOSH formula (4000 mg/kg) for 90 days followed by a 30 days recovery period on control feed (Figure 21 E). The dots represent the value for the individual animals while the bars represent the group median value. Significant group differences are denoted by p-values. Figure 21: Granulomas/cm 2 in livers from rats fed for 30 (A), 60 (B), 90 (C), 120 (D) days with control feed or feed containing MOSH mixture 40 mg/kg, 400 mg/kg and 4000 mg/kg, or for 90 days with the test feed followed by 30 days of control feed (E) The numbers of lymphoid cell clusters in the liver parenchyma/cm 2 were significantly increased at the highest MOSH dose (4000 mg/kg feed) only, and only after 90 days of exposure (Figure 22 C). A nonsignificant tendency of increased numbers at the highest dose(s) were observed after 120 days of exposure, as well as after 90 days exposure followed by a 30 days period on control feed (Figure 22 D and E), since no animals in these groups had zero lymphoid clusters and the group median tend to be higher than for the other groups EFSA Supporting publication 2017:EN-1090

43 The dots represent the value for the individual animals while the bars represent the group median value. Significant group differences are denoted by p-values. Figure 22: The numbers of lymphoid cell clusters in the parenchyma/cm 2 in livers from rats fed for 30 (A), 60 (B), 90 (C), 120 (D) days with control feed or feed containing MOSH mixture 40 mg/kg, 400 mg/kg and 4000 mg/kg, or for 90 days with the test feed followed by 30 days of control feed (E) The numbers of lymphoid cell clusters in the liver portal tracts/cm 2 were significantly increased at the highest MOSH dose (4000 mg/kg feed) after 30 days of exposure, and at the 400 mg/kg dose after 120 days of exposure (Figure 23 A and D). The statistically significant decrease at the highest dose (4000 mg/kg feed) after 60 days of exposure is not regarded as biologically significant due to the very low levels at this time point EFSA Supporting publication 2017:EN-1090

44 The dots represent the value for the individual animals while the bars represent the group median value. Significant group differences are denoted by p-values. Figure 23: The numbers of lymphoid cell clusters in the portal tracts/cm 2 in livers from rats fed for 30 (A), 60 (B), 90 (C), 120 (D) days with control feed or feed containing MOSH mixture 40 mg/kg, 400 mg/kg and 4000 mg/kg, or for 90 days with the test feed followed by 30 days of control feed (E) The degree of vacuolization of the liver cells was semi-quantitatively assessed as described in Section ; examples of different vacuolization degrees (slight, moderate and abundant) are presented in Figure 24. The vacuolization did not appear to differ between the exposure groups (Figure 25). Figure 24: Examples of liver sections showing slight (<5%) (a), medium (10%) (b) and abundant (>15%) (c) vacuolization (2.5 x objective) 44 EFSA Supporting publication 2017:EN-1090

45 The dots represent the value for the individual animals while the lines represent the group median value. Figure 25: The degree of vacuolization of liver cells from rats fed for 30 (A), 60 (B), 90 (C), 120 (D) days with control feed or feed containing MOSH mixture 40, 400 and 4000 mg/kg, or for 90 days with the test feed followed by 30 days of control feed (E) In conclusion, the histological analyses indicate that dietary exposure to MOSH mixture affect the granuloma formation in the liver of rats, but this was only evident at the highest dose (4000 mg/kg feed) tested. This effect is not observed after 30 or 60 days of treatment, but appears after 90 or 120 days of treatment. Due to the high interindividual variation and the group size, the data did not reveal any differences in the number of granulomas per cm 2 after either 90 or 120 days of exposure. Also the number of granulomas in the group treated for 90 days with the highest MOSH dose and a subsequent 30 days on control feed did not differ from the group exposed to the highest dose for 90 days, indicating that granuloma formed are not reversible within the 30 day recovery period. The increase in granuloma formation at the highest dose appeared to be accompanied by increased number of lymphoid clusters in the liver parenchyma, reaching statistically significance after 90 days of exposure, but with a similar trend after treatment for 120 days or treatment for 90 days with subsequent 30 days on control feed. It must be kept in mind that parenchymal and portal lymphoid or inflammatory cell aggregates are commonly observed in the majority of rat livers (McInnes, 2012). These lesions are thought to be caused by bacterial showering from the intestine. This is reflected in the findings in our study, where such inflammatory infiltrates are found in both control animals and those given MOSH mixture in different concentrations. Inflammatory cells in the portal tracts are also a common finding in human livers. Unless they are extensive and part of a hepatitis, they do not seem to be of any clinical relevance EFSA Supporting publication 2017:EN-1090

46 Immune function analyses After diluting the serum 1:2000, KLH-specific IgM were detected in all animals, demonstrating that the KLH immunization protocol was successful in eliciting production of KLH-specific IgM. The IgM concentrations were not significantly different in the four groups of rats (Figure 26A), as determined by ANOVA on Ranks. In the mice fed the highest MOSH concentration (4000 mg/kg feed), however, four of five animals had low IgM levels in serum, while the 5th animal had a remarkably high response compared to all other animals. The trend of reduced antibody concentrations in the highest exposure group (the median value was about halved) was confirmed by the results from the 1:100 serum dilution, where the IgM concentrations were above the upper detection limit for four of five animals in all groups, except for the 4000 mg/kg group, where four of five animals again were low (i.e. within detection limits; Fig. 26B). Sera were diluted 1:2000 (A) and 1:100 (B). The dots represent the value for the individual animals while the bars represent the group median value. The dotted line indicates the upper detection limit for the ELISA assay. Animals were exposed for 120 days to control feed (TEM) or MOSH at concentrations of 40, 400 and 4000 mg/kg feed. Figure 26: KLH-specific IgM antibodies in serum of rats (n=5) 5 days after intravenous injection of KLH In summary, no significant changes due to MOSH exposure were observed for the KLH-specific IgM concentrations in serum. The group exposed to the highest dose, however, tended to show somewhat reduced levels of serum KLH-specific IgM antibodies. Whether this tendency was real effect that did not reach significance due to the small group size (n = 5) or just a chance finding is unknown. Immune function studies often include higher group sizes and high variation and/or presence of outliers are not unusual. Only two previous experiments measuring antigen-specific antibodies after MOSH exposure are known, indicating no or decreased levels of antigen-specific antibody production in Fisher 433 rats (but not in Sprague Dawley) (ImmunoTox, 2001). To investigate effects of environmental exposures on the immune system, functional endpoints are recommended, both by the ICH S8 Immunotoxicity studies for Human Pharmaceuticals and the EPA Health Effects Test Guidelines OPPTS Immunotoxicity, the latter stating that hematology, lymphoid organ weights and histopathology alone are not sufficient to predict immunotoxicity. Therefore, we measured KLH-specific IgM in serum after KLH immunization, which has been recommended as a useful marker of immunosuppressive or -stimulating effects in the OECD guideline 46 EFSA Supporting publication 2017:EN-1090

47 407 Repeated dose 28-day oral study in rodents. A recent study does, however, suggest that KLHspecific IgG is a more sensitive endpoint than KLH-specific IgM (Vandebriel et al., 2014). Switch to IgG production in B-cells (measured after two KLH injections) requires also involvement from T cells and antigen-presenting cells, thus IgG provides a read-out that evaluates the combined functionality of these three cell types. One advantage of the IgM assay over the IgG assay, and the reason why IgM was measured in this project, is it is assumed minimal impact on the other endpoints in the study, allowing us to assess immune function in the same animals. Due to the complexity of the immune system, we can obviously only conclude on a small part of immune function, i.e. IgM production by B cells, from the present experiment. In the next section we describe an experiment investigating another arm of immune function, by investigating the effect of broad MOSH mixture in feed on the induction of autoimmune arthritis in a rat model Autoimmune arthritis in dark agouti rats Results from both the pilot study performed to establish the model with the positive control (see Section 2.1.3), and the main study with dietary exposures to the broad MOSH mixture are reported hereafter. Pilot study - Arthritis incidence, severity and day of onset A single intradermal injection of 200 µl pristane day 0 induced arthritic symptoms in all six female DA-rats (Table 15). Four of the pristane-injected rats were sacrificed during the experiment (on days 19 and 22) due to severe symptoms (score > 10) while the two remaining pristane-injected rats displayed clear but milder symptoms (maximum score 7 and 9) and showed remission of symptoms from day 19 and 22. These rats were monitored for the full 40 day period. None of the control rats intradermally injected with the physiological buffer did show arthritic symptoms (Figure 27). Table 15: Arthritis incidence, maximum score and day of onset in female DA-rats intradermally (i.d) injected on day 0 with 200 µl pristane (n = 6) or physiological buffer (control, n= 6) and monitored twice a week for maximum 40 days Treatment Arthritis incidence Maximum score (a) Day of onset Pristane i.d. 6/6 7, 11, 11, 10, 10, 9 12, 12, 8, 8, 12, 8 Control i.d. 0/6 - - (a): Rats with arthritic score > 10 were sacrificed EFSA Supporting publication 2017:EN-1090

48 16 Arthritic scores Days Figure 27: Arthritic scores (0-16) in female DA-rats intradermally injected day 0 with 200 µl pristane (circles, n = 6) or physiological buffer (squares, n = 6) and monitored twice a week for maximum 40 days - Serum analyses The pristane-injected rats had significantly higher serum levels of IL-17 (Figure 28A) compared to control rats (Mann Whitney test, p < 0.05), while serum levels of IL-1β did not differ between the groups (Figure 28C). IL-10 and IL-6 were under the assay limit of detection for all rats irrespective of treatment (Figure 28B, D). Asterisks (*) denote significantly higher cytokine levels than in the control group, p < 0.05 (Mann Whitney test). Figure 28: Serum levels of IL-17 (a) IL-10 (b), IL-1β (c) and IL-6 (d) in female DA-rats intradermally injected on day 0 with 200 µl pristane (n = 6) or physiological buffer (controls, n = 6) and monitored twice a week for maximum 40 days 48 EFSA Supporting publication 2017:EN-1090

49 While IL-22 levels in serum in general were below the lower detection limit for both groups, TNFα levels were detectable but did not significantly differ between the control and treatment groups of rats (Figure 29 A and B). The presence of IgM-rheumatoid factor (RF) in serum at termination was low, and did not differ significantly between the positive and negative control group (Figure 29 C). IgG-RF, however, were significantly higher in the pristane-injected group (Figure 29 D). In the pilot study, also wet weights of the mesenteric lymph nodes were measured, but no differences were observed between the two groups (data not shown). Based on these results, it was decided to use serum levels of IL-17, TNFα, IL-1β and IgG-RF as serum markers of arthritis induction in the main experiment. The dots represent the value for the individual animals while the lines represent the group median value. The dotted line indicates the lower detection limit for the ELISA assay. Asterisks (*) denote significantly higher levels than in the control group, p < 0.05 (Mann Whitney test). Figure 29: Serum levels at termination of IL-22 (A) TNFa (B), IgM-RF (C) and IgG-RF (D) in female DA-rats intradermally injected on day 0 with 200 µl pristane (n = 6) or physiological buffer (controls, n = 6) Main arthritis study - Body weight and feed intake The main experiment was performed at FHI during the spring 2015, and laboratory analyses were performed during autumn Regarding body weights, there were no group differences between the weight curves for the animals exposed via the feed, demonstrated by slowly increasing, parallel curves (Figure 30). The positive control group, receiving the pristane injection, however, demonstrated a weight loss/levelling off around day 11 after injection. The observed increase for male weights in this group after day 28 is caused by the two individuals showing disease remission (see below) EFSA Supporting publication 2017:EN-1090

50 The injection group received a single injection of 200 µl pristane and control feed. The symbols represent the group median value. Figure 30: Body weights (group medians) for the 5 male and 5 female DA-rats during the 90 days exposure to 0 (control), 40, 400 and 4000 mg/kg feed broad MOSH mixture or 4000 mg/kg feed pristane Feed consumption per group at week 1 and 10 during the experiment is reported in Figure 31. The MOSH or pristane content in the feed did not significantly affect the feed consumption. There seems to be some overall reduction in feed intake from week 1 to week 10, but it is assumed this is caused by a natural pattern among young animals in growth (about 9 weeks of age) compared to adult (about 18 weeks of age) EFSA Supporting publication 2017:EN-1090

51 The symbols represent the group median value. Figure 31: Feed consumption (gram consumed per group per week) for the five male and five female DA-rats in week 1 and week 10 during the 90 days exposure to 0 (control), 40, 400 and 4000 mg/kg feed broad MOSH mixture or 4000 mg/kg feed pristane - Arthritis score Only the positive control group (pristane injection) demonstrated clear arthritis symptoms, with five of five female and five of five male animals demonstrating a significantly increased arthritis score during the first days after injection. Five females and three males had scores above 10 and were terminated. The adversity of the arthritis score, expressed as maximum score per animal during the 90 days of exposure, was in general low for all other groups, and did not significantly differ between groups with the various dietary treatments (Figure 32). The injection group received a single intradermal injection of 200 µl pristane and control feed. The dots represent the value for the individual animals while the lines represent the group median value. # denotes significantly higher level in the positive control (injection) group compared to all other groups, p < 0.05 (ANOVA and Tukeys post hoc test). Figure 32: Maximum arthritic scores (0-16) during the 90 days, in five female (A) and five male (B) DA-rats exposed to 0 (control), 40, 400 and 4000 mg/kg feed containing broad MOSH mixture or 4000 mg/kg feed pristane 51 EFSA Supporting publication 2017:EN-1090

52 - Serum analyses IgG rheumatoid factor (IgG-RF) were measured in serum from blood collected at day 90, or at the day of termination. Without including the injection group in the statistical analyses, there was an overall significant treatment related difference (ANOVA analyses, p=0.028), while there were no significant pairwise differences between the treatment groups (Figure 33). The serum levels of TNFα and IL-1β at termination were low or below the detection limit for all animals (data not shown). IL-17 in serum was significantly elevated in the pristane injection group in the pilot study, and was therefore measured at day 0, 30, 60 and at termination (day 90) in the main study. IL-17 was, however, not detectable in serum before day 90, and there was no difference between the treatments groups (data not shown). The dotted line indicates that the levels in the pristane injected animals (terminated at day 40 or before due to severe arthritis score) should not be directly compared to the rest of the animals (terminated at day 90). Figure 33: IgG rheumatoid factor (IgG-RF) measured in serum at termination for five female (closed symbols) and five male (open symbols) DA rats - Splenocyte analyses Neither % TLR2+ cells nor TLR2 expression per cell showed any significant differences between treatment groups or sexes (data not shown). The same was the case for TLR3 expression, which was low for all samples (data not shown). The following overall conclusions can be drawn from this series of experiment. After a single intradermal injection of pristane all DA rats displayed clinical signs of arthritis. The symptoms, monitored as arthritic scores, developed from day 8, and those rats who were allowed to continue showed remission of symptoms from day 19 and 22, and until day 40. These findings correspond well with Sverdrup et al. (1998), demonstrating similar arthritic development after intradermal injection of rats to different cosmetic products containing mineral oils. This pilot study confirmed that the injection of pristane can serve as a positive control for induction of autoimmune arthritis in the main experiment of Task 5, investigating the effects of MOSH-containing feed on autoimmune arthritis EFSA Supporting publication 2017:EN-1090

53 In the main experiment, both female and male positive control rats demonstrated clear arthritis development after injection of pristane. However, none of the groups exposed to MOSH or pristane via the feed did significantly increase the maximum arthritis score. Arthritis score, however, is a subjective measurement, based on a visual inspection and grading of the four limbs of each animal. Precautions were taken to reduce the subjective impact on the results, by letting the same technician performing the scoring throughout the experiment and blinding this technician with regard to exposure groups (impossible for the injection group due to shaved area in back). In spite of such precautions, some subjectivity will influence the background level of scores. The slightly higher maximum scores in males (group medians at 2 and 3) compared to females (group medians at 1) may be a result of larger animals/paws influencing the observations. In general, weak responses may be hard to detect using this scoring system. Therefore, as more objective markers, several serum and cellular markers associated with arthritis in previous studies were also assessed. Based on the pilot study results, it was decided to use serum levels of IL-17, TNFα, IL-1β and IgG-RF as serum markers of arthritis induction in the main experiment. Also expressions of TLR2 and TLR3 on splenocytes were determined. None of these markers demonstrated convincing effects as a result of MOSH or pristane exposure in feed. 3. Accumulation and toxic effects of narrow MOSH mixtures The objective of this second series of experiments was to address two hypotheses: (i) maximum relative accumulation is around the hydrocarbons C 29 and (ii) n-alkanes have an impact on granuloma formation. Thus, the comparison is between the broad MOSH mixture tested previously and mixtures reduced to that accumulated or forming granulomas. Hydrocarbons at the lowest and highest end of molecular masses covered by the broad MOSH mixture have been shown not to accumulate in human tissues (Barp et al., 2014) and, therefore, were no longer included in the experiments. For the predominantly branched and cyclic MOSH, two products were used which were part of the broad MOSH mixture: they were suitable to distinguish MOSH below and above C 25, and it seemed appropriate to use them since in this way the composition in terms of branching and cyclic constituents was the same as in the previous experiment. In the middle range of molecular masses, two fractions were tested, separated at about n-c 25, since the previous evaluations by the Joint FAO/WHO Expert Committee on Food Additives (JECFA) and the assessments previous to 2012 by EFSA drew the limit between the Class I oils and the oils of the Classes II and III here: JECFA specified a 1000 times lower ADI for oils of Classes II and III than for those of Class I. Class I was distinguished from the others mainly by a 5% distillation point of at least n-c 25, which means that it contains at most 5% hydrocarbons eluted before n-c 25 in simulated distillation. The experiments were intended to clarify to what extent this distinction is justified in terms of accumulation and granuloma formation. The product previously determining the toxicological evaluation was a low melting wax (e.g. EFSA, 2012). Waxes mainly consist of n-alkanes, which were widely thought to be easily metabolized. In fact, in human tissues n-alkanes were only present at low concentrations. The question was which type of hydrocarbons would form the granuloma and impact the immune system. The experiment should clarify whether the elimination of the n-alkanes is really as rapid and what is their role in the formation of granulomas. To this end, a third mixture was tested consisting of the high molecular mass oil mixed in a 1:1 ratio with a wax of similar mass range Methodologies Preparation of the MOSH mixtures The 3 MOSH mixtures were: - S-C25. Fluka Paraffin wax Ph Eur, low viscosity, EFSA Supporting publication 2017:EN-1090

54 - L-C25: Fluka Paraffin wax Ph Eur, high viscosity, L-C25W: 1:1 (w:w) mixture of L-C25 and wax, (Sigma-Aldrich) (mp 65 C) S-C25 as well as L-C25 were labelled as wax, but at ambient temperature they consisted of liquids and are referred to as oil in this report. The wax, W, as added to L-C25. Internal standards in italics: n-alkanes C11, C13 and C17, cyclohexyl cyclohexane (Cycy) and cholestane (Cho). Figure 34: HPLC-GC-FID chromatograms of the three MOSH mixtures tested, namely S-C25 (paraffin wax low viscosity, predominantly eluted below n-c25, upper left), L-C25 (paraffin wax high viscosity, >C25), and L-C25W (1:1 mixture of L-C25 and W, paraffin wax, mp 65 C) The GC-FID chromatograms of the 3 MOSH mixtures tested and the wax added to L-C25 are shown in Figure 34. S-C25 consisted of hydrocarbons eluted from GC between n-c 16 and n-c 34 (GC with a dimethylpolysiloxane stationary phase, simulated distillation). It included a small proportion of n- alkanes ranging from n-c 15 to n-c 24, but primarily consisted of unresolved branched and cyclic MOSH forming a hump. Approximately 27% of the hydrocarbons exceeded n-c 25. L-C25 virtually exclusively consisted of branched and cyclic MOSH ranging from n-c 25 to n-c 45. Approximately 1.5% of its hydrocarbons were eluted below n-c 25. Roughly 80% of the wax (W) consisted of n-alkanes, ranging from C 23 to C 45. Approximately 1.5% of the wax constituents were eluted below n-c Feed preparation As for previous experiments with the broad MOSH mixture described in Chapter 2, a standard pelleted diet (AIN-93M) was selected. MOSH mixtures were dissolved in soybean oil and stirred for 4 h at 40 C. The three MOSH mixtures were added to the feed at nominal concentrations of 400, 1000 and 4000 mg/kg, and in each case an equivalent mass of soybean oil was replaced by the MOSH solution EFSA Supporting publication 2017:EN-1090

55 As granuloma formation was in the focus, the lowest concentration (40 mg/kg) previously selected for the broad MOSH mixture was no longer used. Since the more narrow fractions increased the concentrations in their range, also a dose of 1000 mg/kg was introduced, similar to the 4000 mg/kg dose of the broad mixture in the mass range covered. Feed samples were analysed for their MOSH concentrations. The data are shown in Table 16. Table 16: MOSH concentrations determined in the feeds (mg/kg) MOSH concentration (mg/kg) S-C25 L-C25 L-C25W Experimental design The study was conducted at INRA facilities, according to the EU Directive 2010/63 on the protection of animals used for scientific purposes; the protocol was approved by the French ministry of research and education (APAFIS ). The study was performed on 80 female Fischer 344 rats, 4 weeks of age at the beginning of the experiment. Animals were divided in 20 cages (four animals per cage). After one week acclimation period, animals (n = 8 for each group) were dietary exposed during 120 days to the MOSH fractions listed above. Concentrations tested were 0, 400, 1000 and 4000 mg MOSH /kg feed. Experimental conditions were as described for the broad MOSH mixture study. Feed consumption and body weight were recorded weekly. Data were statistically analysed with the GraphPad Prism 6 software using a one way ANOVA analysis and a Dunnett s multiple comparisons test. After 120 days of exposure, animals were euthanized and the liver, adipose tissue (abdominal), and spleen were sampled, weighed and immediately frozen in liquid nitrogen before storage at -20 C, except for the liver left lobe which was fixed in formalin, and processed for conventional histopathological examination by mounting in paraffin. Paraffin blocks were shipped to NIPH/Oslo, where the samples were sectioned and stained. After removal of the digestive tract, the remaining carcasses were homogenized at low temperature and stored at -20 C. Liver, adipose tissue, spleen and remaining carcasses were sent in polyethylene bags to KLZH for MOSH analysis. For immune function analyses, 5 days before the end of the experiment, each rat was injected (s.c.) with an antigen (KLH: PI from Thermo Scientific, Pierce, 200 µl, 5 mg/ml in PBS buffer) for determination of the immune function. Immediately after euthanasia, blood sample from vena cava was collected in glass tubes without anticoagulant and serum was prepared and frozen in Eppendorf tubes at 80 C. Samples were shipped in dry ice to NIPH/Oslo. At NIPH, the serum samples were thawed, and the concentrations of KLH-specific IgM antibody were determined by enzyme-linked immunosorbent assay (ELISA; ELI-02M, Stellar biotechnologies) according to the manufacturer s recommendations. Histopathological analyses of the livers were performed as described in Section Due to the small group size (n = 8) and the non-normal distribution, non-parametric ANOVA (Kruskal-Wallis) with Dunn s post hoc test correcting for multiple comparisons were used to investigate statistical group differences EFSA Supporting publication 2017:EN-1090

56 Analytical methods For all three doses, MOSH contents were determined by on-line HPLC-GC-FID in spleen, liver, fat and carcass of the pooled tissues from four animals from one cage as well as the same tissues of two individual animals of the other cage (RA1 and RA4). The analytical methods, extraction, on-line HPLC-GC and GCxGC, were basically the same as in the previous experiment on broad MOSH mixture. However, particularly for L-C25W, on-line HPLC-GC required an optimization of the amount injected in order to, on the one hand, enable reliable integration of the hump above the baseline and, on the other, avoid overloading: overloading distorted the peak of the internal standard Cycy and affected the separation of the n-alkanes from the branched and cyclic hydrocarbons on the hump (upper left chromatogram in Figure 35). Injection of different amounts of L-C25W spiked with a series of the n-alkanes indicated an optimum of µg reaching the column (bottom chromatograms). To render the analysis more robust, an additional internal standard (n-c 17 or n-c 20 ) was added for the analysis of L-C25W residues in liver and spleen. To avoid overloading, but also considering the low weight of the individual spleens, the amount of tissue analysed was reduced from 1.0 to 0.1 g tissue. This method was revalidated as reported in Appendix-A. GCxGC used for characterizing compositional changes of the hydrocarbons between the administered MOSH and those recovered from the tissues involved the same selectivity (column combination) as reported by Barp et al. (2016), i.e. a first separation on a polar stationary phase (OV-17) and the second on a non-polar one (dimethylpolysiloxane). Using FID, the analysis can be considered quantitative in terms of composition, since the response is approximately equal for all hydrocarbons. The amounts injected and the attenuations were adjusted to provide plots of similar intensities, i.e. the plots are not indicative of absolute concentrations in tissues (for the quantitative data it is referred to HPLC-GC-FID). Figure 35: Optimization of the amount injected for L-C25W spiked with n-alkanes; expansions from n-c 26 to n-c 32 in the boxes 56 EFSA Supporting publication 2017:EN-1090

57 3.2. Results and discussion Clinical data No clinical symptoms or signs were observed at any dose level during the course of the study. There were no statistically significant differences in body weight between controls and the animals exposed to MOSH fractions (Table 17). No differences in feed intake were observed between groups (data not shown). In the groups dietary exposed to L-C25 and L-C25W MOSH mixtures, the weights of the liver and the spleen were significantly higher than in control, irrespective of the dose tested (Table 17). In contrast, no effect was observed with S-C25 mixture. These data complement those demonstrating that the exposure of female Fischer 344 rats to the MOSH broad mixture results in an increase in absolute and relative liver weights (see Section 2.2.1) and confirm the findings of Griffis et al. (2010) demonstrating that a diet containing 2000 mg/kg low melting point paraffin wax resulted in a significant increase in absolute and relative hepatic and splenic weights in Fischer 344 rats. They suggest that the MOSH fraction responsible for the effect mainly correspond to MOSH ranging from n-c 25 to n-c 45. The presence of n-alkanes in the mixture seems to increase the effect observed on liver weights, since concentrations of 1000 or 400 mg/kg L-C25W mixture, corresponding to a dietary exposure of approximately 55 and 22 mg/kg bw per day, respectively, result in a significant increase in absolute and relative liver weights whereas for the L-C25 mixture the increase in absolute liver weight was only observed at the highest dose tested (Table 17). Table 17: Effect of treatment on the weight of animals and organs Groups Dose Body weight Liver Spleen HSI (mg/kg) Control ± ± ± ± S-C ± ± ± ± ± ± ± ± ± ± ± ± L-C ± ± ± 0.05** ± ** ± ± ± 0.15*** ± *** ± ± 0.70** 1.17 ± 0.11*** ± *** L-C25W ± ± 1.14* 1.07 ± 0.35** ± ** ± ± 1.66*** 1.22 ± 0.20*** ± ** ± ± 1.29*** 1.25 ± 0.23*** ± *** Values are in grams and are means from eight animals ± SD. HIS: hepatosomatic index (liver/body weight ratio). *: significantly different from control (p<0.05); **: significantly different from control (p<0.01); ***: significantly different from control (p<0.001) MOSH concentrations Data in the tissues The MOSH concentrations (mg/kg) in liver, spleen, adipose tissue and carcass for the doses of 0 (control), 400, 1000 and 4000 mg/kg of the three MOSH mixtures in the feed are listed in Table 18 for two single animals of one cage as well as the pooled tissues of four rats of the other. Natural n- alkanes in the tissues of the animals receiving the control feed were subtracted, i.e. the average, e.g mg/kg in the liver, corresponds to the background contamination with MOSH. Averages were calculated weighing the pooled data 4 times. The relative standard deviation (RSD, %) was calculated from the three measurements EFSA Supporting publication 2017:EN-1090

58 Table 18: MOSH concentrations (mg/kg) in liver, spleen, adipose tissue and carcass Fraction Dose RA1 RA4 POOL Average RSD% Liver S-C L-C L-C25W S-C <LOQ 305 Spleen <LOQ L-C L-C25W Adipose tissue S-C L-C L-C25W Carcass S-C <LOQ 45.0 <LOQ 39.6 <LOQ 51.3 <LOQ L-C L-C25W The main results are: 1. For all three MOSH mixtures, the mean concentrations were clearly highest for the liver, followed by spleen and adipose tissue. Although the liver only represents 3-4% of the total body weight, 73-85% of the MOSH were located in this organ. The low concentrations in the carcasses (free of gastro-intestinal tract) confirm that there was no further major MOSH 58 EFSA Supporting publication 2017:EN-1090

59 deposit in the rat body. These data are in line with the trend observed for the broad MOSH mixture. 2. In all tissues, the highest MOSH contents were observed for the S-C25 fraction, the lowest ones for the L-C25 fraction, except for the spleen, in which the residues of L-C25W were even lower. 3. The RSD was mostly in the 5-25% range. As this clearly exceeded the uncertainty of the measurement, this is interpreted as mainly being a variability of the animals. The highest value (31.5%) was observed in the adipose tissue and L-C25 at the 4000 mg/kg dose in the feed. Comparison with broad MOSH mixture The concentrations were generally higher than those measured for the broad MOSH mixture used in the previous experiment at the same dose. A first explanation is that the now applied mixtures fell into the strongly absorbed and accumulated mass range, i.e. were less diluted by eliminated or not absorbed hydrocarbons. For S-C25, however, the increase exceeded that: for the 400 and 4000 mg/kg doses, the concentrations in liver increased by factors of 2.9 and 2.6, respectively. This was on cost of L-C25, the concentration of which decreased from 1,604 to 1,393 mg/kg for the 400 mg/kg dose and from 5,511 to 3,805 mg/kg for the 4000 mg/kg dose. In the adipose tissue, S-C25 concentrations were similar to those of the broad mixture, whereas those of L-C25 decreased by a factor of about 7. This is in line with the much stronger absorption of the low molecular mass hydrocarbons noted in chapter 2. Comparison of the mixtures The differences in concentrations between the tissues depended on the MOSH fraction. On average, the L-C25 concentration was 8.7 times higher in the liver than in the spleen, but it was 21 and 38 times higher for S-C25 and L-C25W, respectively, reflecting different accumulation of the different hydrocarbons. The data in the adipose tissue went into the opposite direction: the L-C25 concentration in liver was 121 times higher, whereas it was 52 times higher for S-C25 and L-C25W. At the 400 mg/kg dose, replacement of half of the L-C25 by the wax (L-C25W) caused an increase of the residues from 1,272 to 4,499 mg/kg in the liver (factor 4.3) and from 14.4 to mg/kg (factor 6.9) in the adipose tissue, indicating high absorption and/or lower metabolism of the n-alkanes. At the 4000 mg/kg dose, the increase only amounted to factors of 1.8 and 5.9, probably due to lower uptake. The opposite was noted for the spleen: a decrease from 190 to 134 and from 419 to 243 mg/kg for the 400 and the 4000 mg/kg dose, respectively. As mentioned above, for a given dose, concentrations in the tissues were higher for S-C25 than L-C25. Table 19 shows the ratios of the concentrations in the four tissues analysed. In liver and spleen, on average over the three doses, 3.6 and 1.6 times higher concentrations were found for S-C25. The strong accumulation of n-alkanes from the L-C25W mixture decreased the S-C25/L-C25W ratio. This accumulation of n-alkanes was also a reason for the higher concentrations of S-C25 in adipose tissue and carcass. The stronger accumulation of the low molecular hydrocarbons had been observed with the broad MOSH mixture, but it was noted that for liver and spleen this was more pronounced at the higher doses. In fact there was no clear trend for the 40 mg/kg dose in the diet (not investigated here) EFSA Supporting publication 2017:EN-1090

60 Table 19: Concentration ratios in the tissues for S-C25/L-C25 and S-C25/L-C25W Dose Liver Spleen Adipose tissue Carcass Concentration ratio S-C25/L-C Concentration ratio S-C25/L-C25W Dependence on the dose Figure 36 plots the MOSH concentrations (mg/kg) in the tissues against the dose (mg/kg in the feed) and confirms the strong deviation from linearity observed with the broad MOSH mixture (see Figure 3): the strongest uptake and/or accumulation is observed for the lowest dose. The curves show different trends for the MOSH mixtures applied and tissue. Fat: abdominal adipose tissue. Vertical bars are standard deviations. Figure 36: MOSH concentrations (mg/kg) in the tissues versus dose (mg/kg) in the feed The increases in the concentration from the 400 to the 4000 mg/kg dose are listed in Table 20. Ranging from 1.6 to 3.9, they are clearly below the factor of 10 to be expected for a linear increase EFSA Supporting publication 2017:EN-1090

61 Table 20: Ratios of the MOSH concentrations in the tissues for the doses of 400 and 4000 mg/kg in the feed Ratio 4000/400 mg/kg dose S-C25 L-C25 L-C25W Liver Spleen Fat Carcass Molecular mass distribution Comparison of chromatograms Figure 37 compares chromatograms from the tissues (upper, red chromatograms) with those of the MOSH added to the feed (lower, black chromatograms). The retention times of some n-alkanes are marked. These and all the following data are from the pooled four animals. S-C25 ranged from n-c 16 to n-c 34, but the MOSH recovered from the liver and spleen (400 mg/kg dose) only started at about n-c 20 (which is in agreement with McKee et al., 2012). The n-alkanes were not detectable in liver and spleen. In the adipose tissue, a shift in the opposite direction was noted, namely towards the more volatile hydrocarbons and notably an enrichment of the n-alkanes (dominating up to C 21 ) and a series of iso-alkanes (dominating as from C 22 ). The same is observed for the carcass, supporting that the residual fat in the carcass is the main deposit of MOSH. L-C25 ranged from about n-c 23 to about n-c 45 and was centred on n-c 35. The distribution of the MOSH residues in the tissues was strongly shifted towards lower molecular masses. In the liver and spleen, the hydrocarbons ranged from n-c 23 to about n-c 40, with the maximum at the retention time of around n-c 30. In the liver, the hump was topped by peaks which were hardly detectable in L-C25. As shown in Figure 38 and further characterized by GCxGC, the main ones do not represent n-alkanes, but open-chain iso-alkanes ranging from C 27 to C 35. Also small peaks of n-alkanes were detected, even though largely absent in L-C25. In the spleen these same peaks are smaller. In the adipose tissue, the shift was even more accentuated: the hump ended at about n-c34 and the maximum was at n-c 28. The signals on top of the hump are even more predominant than in the liver. There is the series of iso-alkanes as in the liver, but also of the n-alkanes, mainly of C 29 and C 31 that are of natural origin in the feed. These natural n-alkanes are also visible in the chromatogram of the liver, but the peaks are small due to the far larger MOSH content. The MOSH composition in the carcass was again similar to that in the adipose tissue EFSA Supporting publication 2017:EN-1090

62 Figure 37: Comparison of HPLC-GC-FID chromatograms of extracts from the tissues (upper chromatograms, in red) with the corresponding MOSH mixtures applied (lower chromatograms, black) In terms of molecular mass distribution, the changes for the L-C25W resemble those of L-C25. In the liver, the n-alkanes were missing up to C 25, then enriched compared to the other MOSH and missing again beyond C 36. The mass range in the spleen was quite the same as in the liver, but the proportion of the n-alkanes on the residual MOSH was reduced. In the adipose tissue the n-alkanes were strongly enriched, combined with an even stronger shift to the low molecular constituents. The same was observed for the carcass EFSA Supporting publication 2017:EN-1090

63 Cho, internal standard. Figure 38: Detail of the HPLC-GC-FID chromatograms of L-C25 and its residues in the tissues, with a dotted vertical line showing the position of n-c 29 and a dotted line of iso-c 30 Data from segments corresponding to single carbon atoms To show the shifts in molecular mass distribution more quantitatively, segments corresponding to a single carbon atom of the MOSH in the feed and the tissues were integrated and normalized on that with the maximum area (Figure 39). Segments were cut immediately after elution of the n-alkanes naming the fraction EFSA Supporting publication 2017:EN-1090

64 Figure 39: Concentrations of segments corresponding to a single carbon atom normalized to the segment of maximum area: tissues compared to the three narrow MOSH mixtures in the feed (4000 mg/kg dose) In the liver and spleen, no significant shift of the maximum was observed for S-C25, but a narrowing of the distribution mainly at the low masses. For L-C25, however, a clear preference for lower mass is shown by a shift of the maximum from n-c 35 to n-c 30 or n-c 31. The same applies to L-C25W. These shifts in mass distribution are in line with the stronger accumulation of S-C25 compared to L-C25 (Table 18). For the adipose tissue and carcass, a strong shift towards the lower molecular mass was observed for all three MOSH mixtures, as had also been noted for the broad MOSH mixture applied in the previous experiments Dose dependence of the mass distribution Figure 40 shows the dose-dependence of the molecular mass distributions per segment corresponding to a single C-atom: the concentrations per segments were divided by the total concentration in the feed. For comparison, in the top row the composition of the MOSH added to the feeds is shown as analysed at the two concentrations in the feed EFSA Supporting publication 2017:EN-1090

65 Fat: abdominal adipose tissue. Concentrations, total MOSH concentrations as in Table 18. Figure 40: MOSH concentrations in segments corresponding to a single carbon atom normalized on the total MOSH concentrations in the feed It is again noted that the relative retention in the tissues is lower at the high dose. Furthermore, as highlighted in the previous tasks for the broad MOSH mixture, at high dose the composition was more strongly shifted towards the lower molecular mass MOSH than at the lower one. In liver and spleen, 65 EFSA Supporting publication 2017:EN-1090

66 this was evident for all three MOSH mixtures; in adipose tissue and carcass the shift was stronger for the L-C25 and L-C25W Accumulation relative to exposure Relative retention of total MOSH The retention of the hydrocarbons in the animals was calculated from the amount of feed consumed during the 120 d exposure multiplied by the concentration of the hydrocarbons in the feed and the weight of these tissues when the animals were sacrificed. The digestive tract had been removed from the carcass and was not included in the calculations. Amounts in the body were summed up by multiplying the mass of the tissue of a given animal with the MOSH concentration measured. For the adipose tissue, the mass of the aliquot sampled and analysed was used, the remaining fat being part of the carcass. Table 21 shows the percentages of the MOSH found in the (pooled) rat bodies compared to the exposure to the three MOSH fractions. For comparison, the related values for the broad MOSH mixture is added as last column. For the 400 mg/kg dose, the most strongly retained mixture was L-C25W (10.1%), followed by the S-C25, while the L-C25 was about 4 times less retained. The difference between L-C25 and L-C25W reflects the strong retention of the n-alkanes from the wax. At increased dose, the retention was decreased (as shown above). Retention of the broad mixture was generally lower due to little retained or absorbed MOSH at the low and high molecular mass ends. Table 21: MOSH recovered in the body of the rats compared to the amount ingested (%) after 120 d exposure Dose (mg/kg) Retention (%) SC25 LC25 LC25W Broad mixture The data in Table 22 refers to the bioconcentration factors (BCFs) for MOSH in liver and abdominal adipose tissue and were calculated as the ratio of the MOSH concentration in the tissue to that measured in the feed. The highest values were observed for the S-C25 and the L-C25W groups, and for the liver these values were 2-3 times higher than those observed for the broad MOSH mixture (see Table 10 for comparison). Table 22: MOSH bioconcentration factors (ratio of the concentration in tissues to the concentration in feed) in liver and adipose tissue Mixture Bioconcentration factors Concentration Liver Adipose tissue feed (mg/kg) Mean SD Mean SD S-C L-C L-C25W Values are based on measured concentrations in feed 66 EFSA Supporting publication 2017:EN-1090

67 Data with regard to the distinction at n-c 25 The previous toxicological evaluation of MOSH by JECFA and EFSA (with the exception of the EFSA opinion from 2012) distinguished those of a molecular mass above and below n-c 25 : in 2002, JECFA specified a 1000 times lower (temporary) ADI for Class II and III oils than for Class I, mainly using the 5% distillation point at C 25 as a criterion (JECFA, 2002). In 2012 JECFA withdrew this t-adi owing to insufficient data (JECFA, 2012). The data from the broad MOSH mixture experiment and confirmed above do not support this classification: the maximum accumulation in liver and spleen was in the range of n-c 28 to n-c 30 (which is close to the maximum MOSH concentration in human tissues). Figure 41 confirms this by the concentration ratio of MOSH in the C 20-C25 and the C 26-C30 range for the three types of tissues and the three MOSH mixtures. The ratios in the tissues were normalized to those of the same range in the feed, i.e. a ratio higher than 1 means an enrichment of the lower molecular constituents, whereas a lower one indicates a preferential accumulation of the higher molecular hydrocarbons. In liver and spleen the ratio was below 1, i.e. the relative retention of the C hydrocarbons was stronger than that in the C range. In adipose tissue and carcass the opposite trend was noted. A similar conclusion is drawn from the relative retentions in the rat body (Table 23): for S-C25, the retention of the hydrocarbons C was higher than that in the range of C 20-C25, with a more pronounced difference at the 400 mg/kg than the 4000 mg/kg dose. It was slightly lower, however, for L-C25. For L-C25W the retention of the C 26-C30 constituents were primarily determined by the performance of the n-alkanes, for which a strong accumulation was determined at the 400 mg/kg dose, but a weaker one for the 4000 mg/kg dose. Figure 41: Ratio of MOSH C 20 -C 25 /MOSH C 26 -C 30 normalized on the ratio in the feed (400 mg/kg concentration) Table 23: Percentages of the C 20 -C 25 and C 26 -C 30 MOSH of the 400 and 4000 mg/kg doses recovered in the rat bodies compared to the amount ingested MOSH mixture Dose (mg/kg) C 20 -C 25 (%) C 26 -C 30 (%) S-C L-C L-C25W EFSA Supporting publication 2017:EN-1090

68 If the relative retention only depended on the molecular mass range, the data should be approximately the same for S-C25 and L-C25. This was, however, not the case. The comparison of L-C25 with L-C25W, having approximately the same molecular mass distribution, indicates a strong effect by the structural composition, here particularly by the n-alkanes Accumulation of n-alkanes Retention relative to other hydrocarbons The retention of n-alkanes was compared with that of the other MOSH in L-C25W. The wax added to L-C25 consisted of approximately 80% n-alkanes next to other hydrocarbons forming a small hump and minor peaks of branched or cyclic constituents (left GC-FID chromatogram in Figure 42). Therefore, the n-alkanes represented less than half of L-C25W. When summing up the n-alkanes above the contour line shown in the right chromatogram, the ratio between the n-alkanes and the remaining hump was 0.8. Figure 42: GC-FID chromatograms of the wax added to L-C25 (left) and of the mixture L-C25W with the contour line above which the n-alkanes were integrated and summed up (right) The ratio of the n-alkanes to the other MOSH was determined in the same way for the tissues of the rats exposed to L-C25W at 400 and 4000 mg/kg (Figure 43). In liver, the ratio was 2.8 for the lower dose, which corresponds to an enrichment of the n-alkanes by a factor of 3.5 (2.8 divided by 0.8). For the higher dose, the enrichment amounted to a factor of 2.3. In the spleen, the n-alkanes were reduced, for the higher dose more clearly than for the lower one. In the adipose tissue the n-alkanes were enriched by a factor of 4.4 at the lower dose and a factor of 6.8 at the higher one. The black line positioned at 0.8 corresponds to the ratio in L-C25W added to the feed. Figure 43: Ratio of the summed n-alkanes to the other MOSH ( hump ) in the tissues of the animals exposed to 400 and 4000 mg/kg L-C25W 68 EFSA Supporting publication 2017:EN-1090

69 The concentrations of the n-alkanes did not increase to the same extent as the other MOSH when increasing the dose in the feed from 400 to 4000 mg/kg. As shown in Table 24, the concentration of the iso- and cyclic-alkanes (other MOSH) in liver is 1.9 times higher at 4000 mg/kg dose, while that of the n-alkanes only increased 1.2 times (both far less than the factor of 10 in the feed). In the spleen, the concentration of the n-alkanes doubled, while the other MOSH did not increase significantly. The n-alkane concentration in fat is 2.5 higher at 4000 mg/kg dose, while that of iso- and cyclo-alkanes only increased 1.6 times. The contrary was observed in the liver, where the concentration of the isoand cyclo-alkanes increased more than that of the n-alkanes. This complex picture indicates once more that the various types of hydrocarbons behave differently with regard to saturation of absorption and metabolism/elimination. Table 24: Area ratios of n-alkanes and hump in tissues for doses of 400 and 4000 mg/kg in the feed Ratio 4000/400 mg/kg dose n-alkanes Other MOSH Liver Splee Adipose tissue Accumulation of the natural n-alkanes from the feed The strong accumulation of the n-alkanes from the wax in the rat tissues, particularly the liver, was against the expected efficient metabolism and their low concentrations found in human tissues, which had been interpreted by assuming efficient elimination. n-alkanes naturally occur in many foods and feed (e.g. about 25 mg in the surface wax of an apple; Belding et al. (1998). Hence, it was hypothesized that the observed strong accumulation was an artifact due to saturation of the metabolism at the high experimental doses. Figure 44: HPLC-GC-FID chromatograms of the control feed and the tissue extracts of the animals fed with the control feed As no dose lower than 400 mg/kg was applied, this was checked for the natural n-alkanes C 29 to C 33 in the control feed. The HPLC-GC-FID chromatograms of the extracts of the control feed and the four tissues of the control animals are shown in Figure 44. The pattern is characterized by the strong 69 EFSA Supporting publication 2017:EN-1090

70 predominance of the odd-numbered species. Some interference by n-alkanes of mineral origin (liver and spleen) and possibly other compounds cannot be ruled out for the minor components. In the feed, the concentrations of the individual n-alkanes C 29 -C 33 were in the range of mg/kg (Table 25). The clearly highest residues were in the liver: the total concentration of the n-alkanes was 37 times above that in the feed. In the liver and spleen, n-c 31 remained the dominant hydrocarbon, as it was in the feed (Figure 44). In the adipose tissue and the carcass, however, n-c 29 dominated, in line with the strong shifts to lower molecular mass constituents noted previously. Table 25: Concentrations (mg/kg) of the natural n-alkanes in the control feed and the tissues of the rats fed with the control feed Concentrations (mg/kg) Feed Liver Spleen Adipose tissue Carcass n-c n-c n-c n-c n-c Total Relative retention of n-alkanes To check the hypothesized saturation of the metabolic system, Table 26 compares the relative retentions in liver and spleen of the n-alkanes C 29 and C 31 from the control feed with those from the 400 and 4000 mg/kg L-C25W diet. The concentrations of n-c 29 and n-c 31 in the control feed were 0.05 and 0.12 mg/kg, respectively, but 5.5 and 11 mg/kg, respectively, in the feed containing 400 mg/kg L-C25W. For a potential saturation of the elimination system, the total amount of n-alkanes would have been more relevant, rendering the difference even larger, since the wax contained several more prominent n-alkanes. The relative retention of the naturally occurring n-c 29 and n-c 31 from the control feed (dose 0) amounted to 10 and 18%, respectively of the amount ingested over 120 d was recovered from the liver. Interestingly, the retention was even higher at the 400 mg/kg dose (23 and 26%, respectively) and strongly decreased again for the highest dose. The retention from the control feed and the 400 mg/kg is that high that not even an almost complete absence of elimination can be ruled out, taking into account incomplete uptake from the feed and the substantial deposition in the other tissues, principally the adipose tissue. The retention of the ingested n-c 31 and n-c 29 in the spleen was far lower than in the liver because of lower concentrations combined with the low mass of this organ. The same strong drop for the retention of n-c 31 at the highest dose is noted as in liver. Table 26: Relative retention (%) of the n-alkanes C 29 and C 31 in liver and spleen in dependence of the exposure; n-alkanes naturally present in the feed (dose 0) or in L-C25W Retention (%) Dose (mg/kg Liver Spleen feed) n-c 29 n-c 31 n-c 29 n-c EFSA Supporting publication 2017:EN-1090

71 In Figures 37 and 38 it was shown that the retention of the n-alkanes from L-C25W in the liver had a rather sharp maximum at n-c 30. It was negligible below n-c 25, as shown for the n-alkanes in L-C25W, but also for those in S-C25. Retention was again low above n-c35. Table 27 shows this more quantitatively in terms of relative retention of the wax components in L-C25W at the 400 mg/kg dose: in the liver, the retention of n-c 31 was more than 10 times higher than that of n-c 25, and at n-c 33 it was reduced again from 10.8 to 7.5%. In the spleen the maximum was at n-c 33, and the level approximately 20 times higher than at n-c 25. In the adipose tissue the maximum was at n-c 27 and the retention of n-c 33 was very low (at a low general retention in that mass range). It is noted that the range of retained n-alkanes is narrower than that of the other MOSH. It seems remarkable that the retention of n-alkanes is highest just in the region of the natural n-alkanes. Table 27: Relative retention (%) of the n-alkanes C 25, C 27, C 30 and C 33 in liver, spleen and adipose tissue from L-C25W (400 mg/kg concentration) Retention (%) Liver Spleen Adipose tissue n-c n-c n-c n-c For the interpretation of these results, the strong accumulation of certain n-alkanes and the selectivity with the steep increase of the retention at C 25 might be the key. n-alkanes below C 25 were absent, as expected for easily biotransformed straight chain hydrocarbons. The melting point of n-c 25 is 54 C, that of n-c C, which might support the hypothesis that the wax components crystallized and precipitated, protecting them from metabolism. Granulomas were observed in the livers of animals exposed to 400 mg/kg L-C25W (see Section 3.2.9), presumably from precipitated wax, but none in those of the control animals, in which n-alkanes (those of natural origin) were also accumulated. This may explain the higher retention at the 400 mg/kg exposure than for the control (Table 26): encapsulated n-alkanes were better protected against metabolism and part of the n-alkanes were removed from equilibration within the body, particularly with adipose tissue. At the highest dose, retention was substantially lower (Table 26), which could be explained by reduced absorption from the digestive tract. It is noted that the retention dropped more strongly for n-c 31 than for n-c 29 when increasing the dose from 400 to 4000 mg/kg, which could reflect a stronger precipitation of n-c 31. This might be of importance for the extrapolation of animal data obtained at high dose to human exposure Characterisation of the accumulated hydrocarbons by GCxGC Composition of the MOSH mixtures Figure 45 shows the relevant sections of the plots of the three MOSH mixtures used for testing. The multibranched open chain hydrocarbons on top of the plots are encircled, the circle being at the same position in all three plots to facilitate comparison. The centres of the signals of the n-alkanes and little branched open chain paraffins as well as of the n-alkyl monocyclic naphthenes are interconnected by lines. The tails at the upper right corner of the plot are from column bleed EFSA Supporting publication 2017:EN-1090

72 Figure 45: GCxGC-FID plots of the MOSH mixtures incorporated into the rat diet. C13, C20, Cycy and Cho were internal standards In monodimensional GC (Figure 37), S-C25 and L-C25 formed similar humps of non-resolved constituents, but GCxGC reveals strong differences in their composition. S-C25 was dominated by distinct signals for the multibranched open chain hydrocarbons in the top region of the plot, the series of n- and little branched iso-alkanes and two series of signals further below, the lower one representing the (scarcely separated) n-alkyl cyclopentanes and cyclohexanes. There are two clouds of unresolved, highly isomerized hydrocarbons in the background, the upper one at the height of the n-alkanes, the other in the region of the naphthenes. L-C25 is not only of higher molecular mass, but also shows less distinct signals, presumably because of de-paraffination, i.e. removal of the waxes (comprising the n-alkanes and other hydrocarbons with major linear parts in their structure, such as the n-alkyl monocyclic naphthenes) in the production process. The plot of L-C25 is dominated by the cloud of unresolved hydrocarbons, most of which are naphthenes with 1-4 rings. A substantial proportion of these resulted from hydrogenation of the MOAH (Biedermann et al., 2015). In L-C25W, the n-alkanes strongly dominated, and owing to overloading of the second dimension GC column, many signals were broadened upwards in the plot (low capacity due to small amount of stationary phase in the second dimension column). Figure 46 highlights a detail in the row of the n- and little branched iso-alkanes already referred to in the HPLC-GC chromatograms of Figures 37 and 38: in S-C25, the n-alkanes (positions marked by vertical dotted lines below the n-alkanes) are the dominating signals up to C 21, but then a series of stronger signals at slightly higher position (marked by dotted lines above these signals) takes over. In the L-C25W, however, the n-alkanes strongly predominate throughout the range of the accumulated MOSH, whereas the iso-alkanes from the L-C25 are hardly visible EFSA Supporting publication 2017:EN-1090

73 Figure 46: Details of the GCxGC-FID plots of S-C25 and L-C25W shown in Figure 45 focusing on the zone of the n-alkanes, with dotted lines for first dimension retention time of the n-alkanes below and certain iso-alkanes above the signals MOSH composition in the rat tissues Figure 47 compares the composition of L-C25W with the related residues in the tissues of the animals exposed to the 400 mg/kg dose. In the liver, the wax components predominated the mixture even more strongly than in the MOSH mixture applied. The n-alkanes are severely overloaded in the second dimension, extending the signals upwards. Between them, little branched iso-alkanes are visible (mainly the 2- and 3-methyl n-alkanes; Biedermann et al., 2015). As shown above, the accumulated n-alkanes reached from about n-c 25 to n-c 36. The distinct peaks below the row of n-alkanes, including the n-alkyl monocyclics, are also enhanced. The molecular size above which this occurred remains uncertain, since in L-C25W these components were detectable only above about C 31. The anyway low proportion of the multibranched hydrocarbons and the cloud in the background virtually disappeared. In the spleen, the range of accumulated n-alkanes was slightly broadened towards higher molecular masses, confirming the data in Table 27. In the adipose tissue and the carcass, the cloud of unresolved hydrocarbons largely disappeared. Fat: abdominal adipose tissue. Figure 47: GC GC-FID plots of the L-C25W mixture and its residues in the tissues (400 mg/kg dose) 73 EFSA Supporting publication 2017:EN-1090

74 Figure 48 compares the composition of the administered S-C25 with the residues in the tissues of the animals exposed to the 400 mg/kg dose. In the liver, the multibranched open chain hydrocarbons gained intensity relative to the total residue. The signals of the n-alkanes, in S-C25 clearly visible up to C 24, largely disappeared up to C 23, but can be observed above C 24, in agreement with the performance of the n-alkanes in L-C25W. The series of iso-alkanes shown in Figure 46 starts predominating the plot as from about n-c24 (details shown in Figure 49). The n-alkyl monocyclics and the row of signals above these virtually disappeared (the weak signals visible on the line specifying the 2nd dimension retention time of the n-alkyl monocyclics represent other, non-identified hydrocarbons). It should be noted that n-alkyl monocyclics present in S-C25 were of lower mass than those accumulated from L-C25W. Two incompletely separated clouds of non-resolved hydrocarbons gained in intensity. The first one is located at the height around and slightly below the n-alkanes. In the separation system used, the n-alkyl monocyclics were eluted below the n-alkanes, but multibranched alkylation and probably also multialkylation brought the second dimension retention into the region of the n-alkanes. The other cloud is located below the n-alkyl monocyclics, which is in the area of the alkylated polycyclic MOSH. The plot obtained from the spleen is similar, though with less n-alkanes (as noted earlier). The plots obtained from the adipose tissue and carcass indicate opposite changes. Apart from the gain in the low molecular constituents, the proportion of the distinct signals belonging to the n-and isoalkanes as well as the n-alkyl monocyclics was enhanced, whereas the cloud in the background was reduced. The MOSH residues in the carcass have virtually the same composition as in the adipose tissue. Fat = abdominal adipose tissue. Figure 48: GC GC-FID plots of the S-C25 MOSH mixture (top) and its residues in the tissues (400 mg/kg dose) 74 EFSA Supporting publication 2017:EN-1090

75 Figure 49: Section of the GC GC-FID plots of the S-C25 and L-C25 residues in the liver from Figure 48, focusing on the n- and iso-alkanes Figure 50 shows the analogous plots for L-C25. In the liver, the residues were dominated by the series of iso-alkanes between C26 and C35 as already shown in Figures 2 and 3. The distinct signals of the multibranched open chain compounds are reduced in relative intensity. The hopanes largely disappeared. Although amounts in L-C25 were small, n-alkanes in the range of the crystallizing species are present (from C 26 to C 35 ). Also the natural n-alkanes C 29 and C 31 from the feed are visible. Several series of distinct signals (unidentified hydrocarbons) below the n-alkanes are enhanced, including the n-alkyl monocyclics from C 28 to C 35. In the spleen, the same trends are seen as in liver. In adipose tissue and carcass, the strong gain in the distinct signals is on cost of the cloud in the background. The multibranched compounds and the cloud are almost absent. This change is probably due to selective absorption, since it is unlikely that just those hydrocarbons are metabolized in the adipose tissue. Fat = abdominal adipose tissue. Figure 50: GC GC-FID plots of the L-C25 mixture and its residues in the tissues (400 mg/kg dose) 75 EFSA Supporting publication 2017:EN-1090

76 Characterisation of the accumulated iso-alkanes The identification of the structure of the iso-alkanes shown in Figures 38 and 49 would be of interest to explain their strong accumulation in a metabolically active tissue such as the liver. According to the second dimension GC retention time, the iso-alkanes are not strongly branched: they are not eluted in the field of the multibranched hydrocarbons, but approximately the height of the 2- and 3-methyl n-alkanes (Biedermann et al., 2015). In the first dimension they are eluted earlier than their n-alkane homologues by an equivalent of approximately 0.8 carbon atoms, which is more than by mono-methyl branching in positions 2 or 3. Figure 51 shows the mass spectrum (GCxGC-MS/EI) of the corresponding C 27 iso-alkane in liver. It is characterized by low abundance of the molecular ion (m/z 380; far lower than for the corresponding n-alkane). The loss of methyl and ethyl is pronounced. Loss of ethyl rules out methyl branching in position 2. Methyl branching in position 3 could result in loss of methyl and ethyl, but according to library spectra, the abundance of the ions from loss of methyl is far lower and there are more pronounced signals by loss of ethylene after loss of ethyl (m/z 351 for C 27 ). In case of a 3-methyl branching, loss of 43 amu (m/z 337) would have to be explained by fragmentation at the other end of the chain. The distribution of the fragments by further degradation is homogeneous, which renders branching further in the chain unlikely. Even numbered fragments in the intermediate mass range are pronounced (even more so in the spectrum of the analogous C 23 iso-alkane (not shown). Figure 51: Mass spectrum (GCxGC-MS) of the C 27 iso-alkane belonging to the series of isoalkanes pointed out in Figure 49 (liver; L-C25, 4000 mg/kg dose) Two reasons were considered for the enrichment of these iso-alkanes in liver and spleen: resistance to enzymatic attack due to structural hindrance and precipitation inhibiting the attack by enzymes. The accumulation of the n-alkanes above n-c 25 and n-alkyl monocyclics from L-C25W could be explained by crystallization that protects these basically rapidly degraded hydrocarbons (Tulliez and Bories, 1979; Tulliez et al., 1981) from elimination. That of the strongly branched paraffins or highly isomerized alkylated naphthenes could be explained by slow or blocked metabolism EFSA Supporting publication 2017:EN-1090

77 The iso-alkanes in question had a similar restricted mass range as the accumulated n-alkanes, in particular starting at about C 27. This selectivity suggests crystallization. However, as shown in Figure 52, deparaffination of mineral oils in the refining transfers them virtually exclusively into the liquid phase; they are practically absent in the wax, indicating a low tendency of crystallization. This points to resistance against metabolism because of a specific branching. It would probably require branching at both ends. GCxGC retention times indicate a largely linear structure. Branching at the two ends might be different, explaining the loss of 43 amu (m/z 337 for the C 27 molecule). Figure 52 documents the trend to crystallize for a raw mineral oil (section of the GCxGC-FID plot between n-c 26 and n-c 33 ). On the left, it is shown that the n-alkanes (pointed out by dotted lines for n-c 28 ) are largely removed from the paraffin oil (L-C25), whereas the weak signal of iso-c 29 in the raw oil becomes dominant in paraffin oil. The n-alkyl monocyclics (pairs of signals at the bottom of the plots) are still present at C 26, but removed at higher mass (or melting point). In the wax (right), the n- alkanes and n-alkyl monocyclics are enriched whereas the iso-alkanes are virtually absent. Dashed and solid lines interconnecting the n- and iso-alkanes, as well as the n-alkyl monocyclics, respectively. Figure 52: Sections of GCxGC-FID plots focusing on the hydrocarbons between n-c 26 and n-c 33 ; MOSH isolated from a crude mineral oil (top) as well as from a paraffin oil (L-C25) and a wax (that in L-C25W; bottom) Accumulation of these iso-alkanes was strong, as pointed out in Figure 38 and confirmed here by the ratio of iso-alkanes to the surrounding total hydrocarbons in GCxGC-FID plots of L-C25. In L-C25, the area ratio of the total of the hydrocarbons between n-c 31 and n-c 32 (vertical section in the plot) to iso C 32 was 109. In the liver it was merely 11, indicating a 10 times higher relative retention than for the total of hydrocarbons in the same mass range (400 mg/kg dose). As the total retention after the 120 d exposure to the 400 mg/kg dose of L-C25 in the rat body was 2.5% (Table 21), the retention of these n-alkanes seems to be high (absorption from the gastrointestinal track being unknown). In the spleen the ratio changed from 109 to 50, i.e. the enrichment was twice as efficient compared to the total. The same type of analysis was performed for adipose tissue, but due to the prevalence of lower masses, iso-c 29 was selected. The ratio changed from 55 to 4.8, which means that the accumulation was 11.5 times stronger than for the sum of all hydrocarbons in the same mass range. This strong change of the composition of the residues in the adipose tissue (enrichment of n-alkanes from S-C25 as well as iso- alkanes from S-C25 and L-C25) is remarkable EFSA Supporting publication 2017:EN-1090

78 If the accumulation of this series of iso-alkanes is really caused by resistance to degradation, this would be a promising molecule to elucidate the relevant structural elements for resistance. Precipitation and granuloma formation In the livers of the rats exposed to 4000 mg/kg S-C25, granulomas were found, but not in those exposed to 400 mg/kg (Section 3.2.9). Wax components, primarily n-alkanes above C 25, are able to precipitate and could be involved in granuloma formation. In this context it was of interest to investigate whether granuloma formation had an influence on the composition of the residues, in particular whether the wax constituents are more enriched in tissues with granulomas. Figure 53 compares the GCxGC-FID plots from the livers and spleens (for which granuloma formation was not investigated) of rats exposed to 400 and 4000 mg/kg S-C25. No differences were evident, in particular the abundance of the n-alkanes and the n-alkyl monocyclics was not increased at the 4000 mg/kg dose. Signals of the n-alkanes highlighted by arrows. Figure 53: GCxGC-FID plots comparing the residues in the liver and the spleen of rats exposed to 400 mg/kg S-C25 (no granuloma detectable) and 4000 mg/kg S-C25 (with granulomas) Histopathological analyses of liver samples Histopathological scoring of all samples was performed. Examples of granuloma, portal tract lymphoid infiltration and vacuolization are shown in Figure 54, and the results are presented in Figures 55, 56, 57 and EFSA Supporting publication 2017:EN-1090

79 Figure 54: A) Example of liver granuloma B) Example of portal tract lymphoid infiltration C) Example of liver section showing slight (10%) vacuolization After 120 days of exposure via feed, no increase in granuloma formation was observed in any of the three dose groups given the L-C25 mixture (Figure 55). The groups fed 4000 mg/kg of the S-C25 mixture, however, demonstrated significantly increased levels versus the control feed group. For the groups fed the L-C25W mixture, the granuloma density was significantly higher than in the control group for all three doses EFSA Supporting publication 2017:EN-1090

80 The dots represent the value for the individual animals while the lines represent the group median value. Significant differences between the control and treatment groups are denoted by * (p<0.05). Figure 55: Granulomas/cm 2 in livers from rats fed control feed or feed containing the MOSH mixtures S-C25, L-C25 and L-C25W in doses of 400, 1000 and 4000 mg/kg feed for 120 days The numbers of lymphoid cell clusters in the liver parenchyma/cm 2 were not affected by any dose of L-C25, but were significantly increased for the two highest doses of the S-C25 and the lowest dose for the L-C25W fraction (Figure 56). While there was an apparent dose-response-relationship observed in the livers of the rats receiving the S-C25 fraction, no clear differences between the groups fed the three doses of the L-C25W fraction were observed. Like the lymphoid cell clusters in the parenchyma, the numbers of lymphoid cell clusters in the liver portal tracts/cm 2 were not affected by any dose of L-C25. However, all groups receiving the L-C25W fraction and the highest dose of the S-C25 fraction had significantly increased lymphoid cell clusters in the portal tracts compared to the control group (Figure 49). The lymphoid cell clusters were also significantly higher in the group fed the highest versus the medium dose of the S-C25 fractions. The degree of vacuolization of the liver cells was semi-quantitatively assessed as previously described. The results are presented in Figure 58, by use of the category variables 0-4 for the vacuolization grades no, slight, moderate and abundant, respectively. While there was only a slightly increased vacuolization in the groups fed the high dose (4000 mg/kg) S-C25 and the low and medium dose of L-C25W, the degree of vacuolization was statistically significantly increased, scored to abundant, for the high dose of L-C25W EFSA Supporting publication 2017:EN-1090

81 The dots represent the value for the individual animals while the lines represent the group median value. Significant differences between the control and treatment groups are denoted by * (p<0.05). Figure 56: The numbers of lymphoid cell clusters in the parenchyma/cm 2 in livers from rats fed for 120 days with control feed or feed containing the MOSH mixtures S-C25, L-C25 and L-C25W in doses of 400, 1000 and 4000 mg/kg feed The dots represent the value for the individual animals while the lines represent the group median value. Significant differences between treatment and the control groups are denoted by * (p<0.05), while denotes significant group differences (p<0.05) versus the medium dose of the respective MOSH fraction. Figure 57: The numbers of lymphoid cell clusters in the portal tracts/cm 2 in livers from rats fed for 120 days with control feed or feed containing the MOSH mixtures S-C25, L-C25 and L-C25W in doses of 400, 1000 and 4000 mg/kg feed 81 EFSA Supporting publication 2017:EN-1090

82 The dots represent the value for the individual animals while the lines represent the group median value. Significant differences between the control and treatment groups are denoted by * (p<0.05). Figure 58: The degree of vacuolization of liver cells from rats fed for 120 days with control feed or feed containing the MOSH mixtures S-C25, L-C25 and L-C25W in doses of 400, 1000 and 4000 mg/kg feed Taken together, there were large differences in the ability of the different MOSH fractions to induce liver granulomas, with no granuloma formation after ingestion of feed containing the L-C25 fraction, increased amounts of granuloma formation induced by S-C25 ingestion, while very strong granuloma formation was observed after ingestion of the wax-containing L-C25W mixture. This pattern appeared to be similar for the lymphoid cell clusters in the portal tract. In fact, the correlation (Pearson) between the granuloma formation and the lymphoid cell clusters in the portal tract had a correlation coefficient of (p<0.001). While 120 days of exposure to 4000 mg/kg of the broad MOSH mixture reported above induced between 2 and 19 granulomas/cm 2, most of the animals exposed to 4000 mg/kg S-C25 mixture or to all doses of the L-C25W displayed between 50 and several hundred granulomas/cm 2. This illustrates the apparent impact of the qualitative composition of the MOSH fractions, in particular the fractions containing > 25C n-alkanes, on the granuloma formation. In the groups receiving the L-C25W mixture, the granuloma formation in the liver sections were often so dense and partly confluent that it was hard to distinguish between single granulomas, which may introduce some uncertainty in the actual values. Further analyses of the 4 sections per animal will be scored to improve the precision of the dataset, but so far, the data do not suggest that the numbers are very different from slide to slide from the same rat, and the present conclusions are most probably valid. Two of the animals in the 4000 mg/kg L-C25W group displayed crystal rods in the liver, not observed in the other animals. As noted already in Section 2.2.7, parenchymal and portal lymphoid or inflammatory cell aggregates are commonly observed in the majority of rat livers (McInnes, 2012). This is reflected in the findings concerning broad MOSH mixture, where such inflammatory infiltrates were found in both control animals as well as those given MOSH mixture in different concentrations. The picture in this experiment, however, is somewhat different; the formation of inflammatory cell aggregates seems to 82 EFSA Supporting publication 2017:EN-1090

83 be increased along with strong granuloma formation. In general, the granulomas observed do have lymphocytes associated with them, while the lymphoid cells counted in this study are groups of cells not associated with the granulomas. The inflammatory cells in the portal tracts are also a common finding in human livers. Unless they are extensive and part of a hepatitis, they do not seem to be of any clinical relevance in humans Immune function analyses The serum levels of KLH-specific antibodies for many of the animals were above the detection limit when diluting the serum samples 1:2000 (data not shown), a dilution optimal for the sera in broad MOSH mixture experiment. The higher concentration of KLH-specific IgM antibodies in the present experiment may be due to a change in KLH provider as well as injection route (s.c. versus i.v.). The samples were therefore re-analysed at a dilution of 1:5000. No significant differences between the groups were observed for KLH-specific IgM antibodies, neither by use of one way ANOVA nor ANOVA on Ranks (Figure 59). Animals were exposed for 120 days to control feed or MOSH fractions at concentrations of 400, 1000 and 4000 mg/kg feed. Sera were diluted 1:5000. The dots represent the value for the individual animals while the lines represent the group median value. The results are expressed as as μg/ml serum. The dotted line indicates the upper detection limit for the ELISA assay. Figure 59: KLH-specific IgM antibodies in serum of rats (n = 8) 5 days after subcutaneous injection of KLH (Pierce) Overall, no significant differences were observed for the KLH-specific IgM concentrations in serum due to MOSH exposure, neither after exposure to 40, 400 and 4000 mg/kg feed of the broad mixture of MOSH nor to 400, 1000 and 4000 mg/kg feed of the three narrow MOSH mixtures. Only two previous experiments measuring antigen-specific antibodies after MOSH exposure are known, indicating no or decreased levels of antigen-specific antibody production in Fisher 433 rats (but not in Sprague Dawley) (ImmunoTox, 2001) EFSA Supporting publication 2017:EN-1090

84 The freeze-thaw of the serum samples due to need for reanalyzes with higher dilutions may have affected the antibody levels to some degree, but would most likely be independent of treatment group. It is important to acknowledge that due to the complexity of the immune system, we can obviously only conclude on a small part of immune function. Although specific IgM has been recommended as a useful marker of immunosuppressive or -stimulating effects in the OECD guideline 443 Extended One-Generation Reproductive Toxicity Study, this endpoint needs to be combined with other endpoints reflecting immune function to conclude on an immunotoxic potential. One advantage of the IgM assay, and the reason why IgM was measured in this project, is the single KLH injection s assumed minimal impact on the other endpoints in the study, allowing us to assess immune function in the same animals. 4. Conclusions The aim of this study was (1) to assess the accumulation and toxicity of MOSH mixtures representative of the whole MOSH range to which humans are exposed via the diet, (2) to identify the fractions and sub-categories with higher bioaccumulation and toxicological potentials, (3) to analyze of the link between accumulation and different toxicological endpoints including the formation of hepatic microgranulomas, and the changes in immune function and autoimmune responses. After 120 d of dietary exposure to 4000 mg/kg of a broad mixture of MOSH (ranging from about C 14 to C 50 ), average concentrations in liver, spleen and adipose tissue reached 5511, 383 and 274 mg/kg, respectively. During a depuration period of 30 d after 90 d of exposure, the MOSH concentration in liver and spleen decreased by 34 and 36%, respectively, whereas it even seemed to slightly increase in the adipose tissue. Relative standard deviations for the concentrations in liver and spleen were between 5.6 and 22%, averaging 11 and 13%, respectively. Concentrations in liver and spleen still increased from day 90 to 120, indicating that no plateau had been reached. In the adipose tissue the increase even seemed to accelerate. Hence there is no indication that a steady state was reached after 120 d of exposure. Tissue concentrations were far from proportional to the administered doses: when increasing the dose from 40 to 4000 mg/kg, they increased between 4.0 and 11.5 times, rather than 100 times. The same was observed when comparing the control feed containing 1.6 mg/kg MOSH with the feed with added 40 mg/kg MOSH: the tissue concentrations increased by factors between 3.1 and 12 instead of by a factor of 25. This means that extrapolation of the internal exposure from a high experimental dose to a far lower real exposure results is severe underestimation. The overall retention of the MOSH fed to the animals was 10.9% after 30 days, 6.2% after 120 days and 3.9% after 90 d followed by the 30 d depuration. This refers to the 40 mg/kg dose, whereas the values were lower at the higher dose because of the reduced relative uptake mentioned above. Roughly 50% of the retained MOSH were located in the liver. In the spleen it was around 0.6%. The other roughly half was in the adipose tissue and the carcass The composition of the MOSH retained in the adipose tissue was strongly different from that in the liver and spleen. In liver and spleen, the maximum relative retention was at n-c 29 (simulated distillation by GC). Hydrocarbons below n-c 19 and above n-c 40 were virtually absent. In the adipose tissue, the maximum retention was at the low molecular mass end of the mixture, which was n-c 15. Retention rapidly dropped to n-c 22 and remained at a lower level up to about n-c EFSA Supporting publication 2017:EN-1090

85 The broad MOSH mixture added to the feed only contained significant amounts of n-alkanes up to about n-c 21. They were largely absent in liver and spleen. After the 30 d exposure, their proportion in the adipose tissue increased from 11% in the MOSH added to the feed to 32% (referring to the mass range n-c ). Their proportion decreased with longer exposure, suggesting preferential biotransformation and/or elimination. GCxGC was used for analysing the composition of the MOSH and the compositional changes from the mixture added to the feed to that in the tissues. A rough classification indicated that the broad MOSH mixture consisted of 31% n-alkanes and little branched paraffins, 9.9% multibranched paraffins and 59% naphthenes. There was no significant change in the composition by this classification in liver and spleen, but a significant shift to the open chain hydrocarbons in the adipose tissue. As there was substantial elimination in liver and spleen, it means similar elimination in all three classes. A more detailed analysis showed the reduction or elimination of most components forming distinct GCxGC signals in liver and spleen, i.e. of the well-defined, dominant constituents, though with the exception of the multibranched hydrocarbons. In the mixture tested, the n-alkanes, n-alkyl monocyclic naphthenes and constituents with mainly unbranched structure belonged to the most efficiently eliminated species, leaving behind a largely unstructured cloud of unresolved hydrocarbons. This cloud represented highly isomerized hydrocarbons, such as strongly branched alkanes as well as naphthenes with complex alkylation and polycyclics of various conformations. This was in contrast to the adipose tissue, where the distinct signals seemed to gain in intensity compared to the cloud of unresolved constituents in the background, i.e. not only the n-alkanes were enriched. No significant compositional differences were observed for the doses of mg/kg feed of the broad mixture. There were little changes in the composition by varying the duration of the exposure, including the depuration period. Only the alkylated monocyclic naphthenes gave indications that resistance against biotransformation increased with ramification and/or multiplealkylation. This suggested that for most constituents, elimination is either rapid (in the order of hours or days) or slow compared to the observation period (i.e. taking more than a few month). The histological analyses indicate that dietary exposure to MOSH broad mixture increase the granuloma formation in the liver of rats, but this was only evident at the highest dose (4000 mg/kg feed) tested. This effect was not observed after 30 or 60 days of treatment, but appeared after 90 or 120 days of treatment. Hepatic granulomas formed in the group exposed to the highest dose for 90 days were not reversible within the 30 days recovery period. None of the rats given pristane or MOSH in the feed developed arthritis measured as symptoms or markers for disease. MOSH exposure resulted in a significant increase in absolute and relative weights of both liver and spleen in the group of rats exposed to the mixture L-C25, consisting of branched and cyclic MOSH ranging from n-c 25 to n-c 45, and to the mixture L-C25W, consisting of L-C25 mixed in a 1:1 ratio with a wax of similar mass range (n-alkanes ranging from C 23 to C 45 ). Limited or no effects were observed with the broad MOSH mixture or the mixture having only 27% of the hydrocarbons exceeding n-c 25 (S-C25), respectively. n-alkanes up to C 25 were not detected in liver and spleen. For all mixtures, accumulation occurred predominantly in the liver. In liver and spleen, accumulation of the C MOSH was higher than that of the C fraction. A series of dominant iso-alkanes accumulated in all tissues, but their chemical structure was not elucidated EFSA Supporting publication 2017:EN-1090

86 No increase in granuloma formation was observed in any of the three dose groups given the L-C25 mixture (well deparaffinated oil, virtually free of n-alkanes). However, in the groups fed 4000 mg/kg of the S-C25 mixture, significantly increased granuloma formation was noted compared to the control feed group. For the groups fed the L-C25W fraction, the granuloma density was significantly higher than in the control group for all doses tested. Granuloma formation seemed to presuppose the presence of n-alkanes above a certain threshold concentration in the tissue. The numbers of lymphoid cell clusters in the liver parenchyma and in the liver portal tract were not affected by any dose of L-C25, but were significantly increased for the highest dose of the S-C25 mixture. For the L-C25W mixture a significant increase of the numbers of lymphoid cell clusters in the parenchyma was observed at low and medium doses, whereas the effect was significant only at the highest dose in the portal tract MOSH exposure, irrespective of mixture tested, had no impact on the immune response following challenge with KLH measured as KLH-specific IgM concentration in serum. References Barp L, Kornauth C, Würger T, Rudas M, Biedermann M, Reiner A, Concin N and Grob K, Mineral oil in human tissues, Part I: concentrations and molecular mass distributions. Food and Chemical Toxicology, 72, Belding RD, Blankenship SM, Young E and Leidy RB, Compositional and variability of epicuticular waxes in apple cultivars. Journal of the American Society for Horticultural Sciences, 123, BfR (Bundesinstitut für Risikobewertung), German Federal Institute for Risk Assessment. 10. Sitzung der BfR-Kommission für Bedarfsgegenstände, Protokoll vom 29. November 2012, p. 6. Available online: Biedermann M and Grob K, 2012a. On-line coupled high performance liquid chromatography gas chromatography (HPLC-GC) for the analysis of mineral oil; Part 1: method of analysis in foods, environmental samples and other matrices. A review. Journal of Chromatography A, 1255, Biedermann M and Grob K, 2012b. On-line coupled high performance liquid chromatography gas chromatography (HPLC-GC) for the analysis of mineral oil. Part 2: migrated from paperboard into dry foods: interpretation of chromatograms. A review. Journal of Chromatography A, 1255, Biedermann M, Barp L, Kornauth C, Würger T, Rudas M, Reiner A, Concin N and Grob K, Mineral oil in human tissues, part II: characterization of the accumulated hydrocarbons. Science of the Total Environment, 506, Biedermann M, Castillo R, Riquet A-M and Grob K, Comprehensive two-dimensional gas chromatography for determining the effect of electron beam treatment of polypropylene used for food packaging. Polymer Degradation and Stability, 99, Blewitt RW, Bradbury K, Greenall MJ and Burrow H, Hepatic damage associated with mineral oil deposits. Gut, 18, Boitnott JK and Margolis S, Saturated hydrocarbons in human tissues III. Oil droplets in the liver and spleen. Hopkins Medical Journal, 122, Concin N, Hofstetter G, Plattner B, Tomovski C, Fiselier K, Gerritzen K, Fessler S, Windbichler G, Zeimet A, Ulmer H, Siegl H, Rieger K, Concin H and Grob K, Mineral oil paraffins in human body fat and milk. Food and Chemical Toxicology, 46, EFSA Supporting publication 2017:EN-1090

87 Cruickshank B, Follicular (mineral oil) lipidosis: I. Epidemiologic studies of involvement of the spleen. Human Pathology, 15, Dincsoy HP, Weesner RE and MacGee J, Lipogranulomas in non-fatty human livers. A mineral oil induced environmental disease. American Journal of Clinical Pathology, 78, Dowdy DL, Mckone TE and Hsieh DPH, Prediction of chemical biotransfer of organic chemicals from cattle diet into beef and milk using the molecular connectivity index. Environmental Science and Technology, 30, Firriolo JM, Morris CF, Trimmer GW, Twitty LD, Smith JH and Freeman JJ, Comparative 90-day feeding study with low-viscosity white mineral oil in Fischer-344 and Sprague-Dawley-derived CRL:CD rats. Toxicologic Pathology, 23, Griffis LC, Twerdok LE, Francke-Carroll S, Biles RW, Schroeder RE, Bolte H, Faust H, Hall WC and Rojko J, Comparative 90-day dietary study of paraffin wax in Fischer-344 and Sprague Dawley rats. Food and Chemical Toxicology, 48, Grob K, On-Line Coupled LC-GC. Hüthig, Heidelberg, ISBN Grob K, Vass M, Biedermann M and Neukom HP, Contamination of animal feed and food from animal origin with mineral oil hydrocarbons. Food Additives and Contaminants, 18, Hansen JS, Alberg T, Rasmussen H, Lovik M and Nygaard UC, Determinants of experimental allergic responses: interactions between allergen dose, sex and age. Scandinavian Journal of Immunology, 73, Holm BC, Svelander L, Bucht A and Lorentzen JC, The arthritogenic adjuvant squalene does not accumulate in joints, but gives rise to pathogenic cells in both draining and non-draining lymph nodes. Clinical & Experimental Immunology, 127, Holmdahl R, Lorentzen JC, Lu S, Olofsson P, Wester L, Holmberg J and Pettersson U, Arthritis induced in rats with nonimmunogenic adjuvants as models for rheumatoid arthritis. Immunological Reviews, 184, ImmunoTox, Effects of dietary mineral hydrocarbons on T. cell dependent antibody production in two strains of rats. Unpublished study ITI 1298 prepared for American Petroleum Institute, Washington DC, as described by WHO/IPCS, 2003 JECFA (Joint FAO/WHO Expert Committee on Food Additives), Joint FAO/WHO Expert Committee on Food Additives; 59 th report, p ; WHO Technical Report Series 913. Available online: JECFA (Joint FAO/WHO Expert Committee on Food Additives), Joint FAO/WHO Expert Committee on Food Additives. 66 th meeting, Summary report issued on 29 June Available online: McInnes EF, Background lesions in laboratory animals. A Color Atlas. Elsevier, ISBN: , p 26. McKee WW, Drummond JG, Freeman JJ, Letinski DJ and Miller MJ, Light white oils exhibit low tissue accumulation potential and minimal toxicity in F344 rats. International Journal of Toxicology, 31, Nochomovitz LE, Uys CJ and Epstein S, Massive deposition of mineral oil after prolonged ingestion. South African Medical Journal, 49, Reeves PG, Nielsen GC and Fahey Jr., AIN-93 purified diets for laboratory rodents: final report of the American Institute of Nutrition ad hoc writing committee on the reformulation of the AIN 76A rodent diet. Journal of Nutrition, 123, EFSA Supporting publication 2017:EN-1090

88 Salvayre R, Nègre A, Rocchiccioli F, Duboucher C, Maret A, Vieu C, Lageron A, Polonovski J and Douste-Blazy L, A new human pathology with visceral accumulation of long-chain n-alkanes; tissue distribution of the stored compounds and pathophysiological hypotheses. Biochimica et Biophysica Acta, 958, Sverdrup B, Klareskog L and Kleinau S, Common commercial cosmetic products induce arthritis in the DA rat. Environmental Health Perspectives, 106, Tulliez J, Bories G and Peleran JC, Effet de l ingestion prolongée d huile de paraffine chez le porc: rétention sélective, interférence avec le métabolisme du cholestérol. Comptes rendus de l'académie des Sciences Paris, 280, Tulliez JE and Bories GF, Metabolism of naphthenic hydrocarbons. Utilization of a monocyclic paraffin, dodecylcyclohexane, by rat. Lipids, 14, Tulliez JE, Durand E and Peleran JC, Mitochondrial hydroxylation of the cyclohexane ring as a result of β oxidation blockade of a cyclohexyl substituted fatty acid. Lipids, 16, Trimmer GW, Freeman JJ, Priston RA and Urbanus J, Results of chronic dietary toxicity studies of high viscosity (P70H and P100H) white mineral oils in Fischer 344 rats. Toxicologic Pathology, 32, Vandebriel RJ, Tonk EC, de la Fonteyne-Blankestijn LJ, Gremmer ER, Verharen HW, van der Ven LT, van Loveren H and de Jong WH, Immunotoxicity of silver nanoparticles in an intravenous 28 day repeated-dose toxicity study in rats. Particle and Fibre Toxicology, 11:21. DOI: / van der Westhuizen R, Crous R, de Villiers A and Sandra P, Comprehensive two-dimensional gas chromatography for the analysis of Fischer Tropsch oil products. Journal of Chromatography A, 1217, van der Westhuizen R, Ajam M, De Coning P, Beens J, de Villiers A and Sandra P, Comprehensive two-dimensional gas chromatography for the analysis of synthetic and crudederived jet fuels. Journal of Chromatography A, 1218, Vendeuvre C, Bertoncini F, Duval L, Duplan JL, Thiébaut D and Hennion MC, Comparison of conventional GC and GCxGC for the detailed analysis of petrochemical samples. Journal of Chromatography A, 1056, Wanless IR and Geddie WR, Mineral oil lipogranulomata in liver and spleen. Archives of Pathology & Laboratory Medicine, 109, EFSA Supporting publication 2017:EN-1090

89 Abbreviations ADI BCF bw CONTAM Panel Cycy DA FID GC GCxGC HAS HPLC-GC-FID HSI i.d. Ig IL INRA i.v. JECFA KLZH LAS LOD LOQ MOAH MOH MOSH MS NIPH POSH PTV RF acceptable daily intake Bioconcentration factor Body weight EFSA Panel on Contaminants in the Food Chain cyclohexyl cyclohexane Dark agoutibina flame ionization detection Gas chromatography Comprehensive two-dimensional GC hematoxylin-azophloxine-saffron High Performance Liquid Chromatography Gas Chromatography Flame Ionization Detection hepatosomatic index intradermal immunoglobulin Interleukin The French National Institute for Agricultural Research intravenous Joint FAO/WHO Expert Committee on Food Additives the Official Food Control Authority of the Canton of Zurich (Kantonales Labor Zürich) LEICA application system limit of deviation limit of quantification of mineral oil aromatic hydrocarbons Mineral oil hydrocarbons mineral oil saturated hydrocarbons Mass spectrometry Norvegian Institute of Public Health Polyolefin Oligomeric Saturated Hydrocarbons programmed temperature vaporizing rheumatoid factor 89 EFSA Supporting publication 2017:EN-1090

90 RSD s.c. SD TOF W relative standard deviation subcutaneous Standard deviation time of flight wax 90 EFSA Supporting publication 2017:EN-1090

91 Appendix A Validation of the methods for chemical analysis The gloves used for the dissection of the animals were checked for the absence of disturbing MOSH and POSH. In particular, one glove s finger was extracted in hexane, of which 0.5% was analysed (Figure 1). Considering that an amount corresponding to 0.35% of the tissue samples was injected and assuming that three fingers contacted the tissue, it was concluded that the contamination of the tissues by the gloves would have been negligible even if a complete transfer of the MOSH had occurred. Figure 1: MOSH fraction from an extract of gloves used for dissection The blank over the chemical analysis was repeatedly checked simulating the extraction procedure and the chromatographic analysis without tissue. As shown below in two chromatograms (Figure 2), the extraction procedure did not produce any extra peak. Occasionally there was a small hump due to the HPLC-GC system, as shown in blank 2, the effect of which on quantitative determinations was considered negligible. Figure 2: Chromatograms obtained from two different blank extractions; labelled signals correspond to the internal standards The yield of the extraction was determined for the four rat tissues analysed (liver, spleen, adipose tissue and carcass from 4000 mg/kg dose, 60 d exposure) and their inherent MOSH content. After the 1 h equilibration with ethanol, samples were extracted overnight with hexane. Then the centrifuged residues were extracted another time with 5 ml hexane for three days without adding new internal standard. These second extracts contained less than 4% of the MOSH of the first extract as calculated by external standard, but 24% for the carcass (Figure 3). Percentages for the MOSH and the internal standards were equal, indicating that the residue consisted of first extract which was incompletely removed by decanting, i.e. that the extraction was virtually complete EFSA Supporting publication 2017:EN-1090

92 Figure 3: Yield of the extraction: first overnight extraction of the tissues (black) overlaid with those obtained from a second extraction on the residual tissue (in blue) The yield of the transfer from the ethanol/water phase into the hexane after splitting the combined ethanol/hexane extract was tested by adding another 5 ml of hexane % of MOSH of the first partitioning were measured, but also this percentage was equal for the internal standard and the MOSH, indicating that this was from incomplete removal of the hexane phase, the internal standard compensating for this. The Figure 4 shows an overlay of the first and second partitioning for a liver sample from the 400 mg/kg/60 d group. Figure 4: Overlay of the hexane phases from the first and second partitioning with the ethanol/water mixture of a liver sample (400 mg/kg/60 d) Repeatability and homogeneity of the sample was determined from 4 replicate extractions from the same rat liver homogenate (40 mg/kg/60 d pool) containing about 185 mg/kg MOSH. The relative standard deviation of the complete analyses was 3%. Four injections of the same extract resulted in a relative standard deviation of 0.4%. Figure 5 overlies the chromatograms EFSA Supporting publication 2017:EN-1090

93 Figure 5: Repeatability of the analysis starting from a pooled liver homogenate (top) and of HPLC-GC-FID analysis (bottom): overlays of four chromatograms Homogeneity of a milled carcass (4000 mg/kg/30 d) containing about 50 mg/kg MOSH was determined taking three samples from different parts of the whole mass. The relative standard deviation of the complete analyses was 5.2% (Figure 6). Furthermore, the extraction of two aliquots of the same carcass with and without further homogenation by a Polytron gave exactly the same results (65.4 mg/kg). Figure 6: Homogeneity of a carcass sample: overlays of 3 chromatograms Linearity of the data was checked by the analysis of the extract from liver tissue 40 mg/kg/30 d containing 91 mg/kg MOSH. This extract was mixed at various ratios with an extract from cow liver without detectable MOSH to yield concentrations ranging between 1.3 and 91 mg/kg, as reported in Figure 7. R 2 was EFSA Supporting publication 2017:EN-1090

94 Figure 7: Linearity of the data in the range of mg/kg for an extract from rat liver mixed with extract from cow liver virtually free of MOSH The chromatogram in Figure 8 shows 1.3 mg/kg MOSH in liver (extract from rat liver 40 mg/kg/30 d diluted with extract from cow liver). Three analyses at this level resulted in a coefficient of variation of 18%. From this, a limit of quantification (LOQ) of approximately 1 mg/kg and a limit of detection (LOD) of about 0.5 mg/kg were estimated. Figure 8: LOD and LOQ: extract of liver (mixed rat/cow) containing 1.3 mg/kg MOSH used as reference point As part of the studies on the narrow MOSH mixtures, the HPLC-GC method was validated again using the mixtures used in this context. To account for the high concentrations in liver and spleen, the severe overloading of the n-alkanes in the instance of L-C25W and the low weight of the individual spleens, the amount of tissue analysed was reduced from 1.0 to 0.1 g. For three complete analyses, a coefficient of variation (CV%) of 2.1% was obtained (Figure 9) EFSA Supporting publication 2017:EN-1090

95 Figure 9: Overlay of three HPLC-GC-FID chromatograms obtained from the analysis of 0.1 g of liver tissue (4000 mg/kg dose, L-C25) The accuracy of the data obtained was checked by spiking each tissue from the control animals with the three MOSH mixtures approximately at the lowest level of contamination as determined by first measurements, namely 1000 mg/kg for liver, 100 mg/kg for spleen and 10 mg/kg for adipose (Table 1) tissue and carcass. This procedure also checked whether quantitation by the internal standard and using a response factor of 1 would yield the correct results. Table 1: Recoveries for the tissues of interest and the MOSH mixtures spiked near the lowest concentration found Spiking level (mg/kg) Liver 1000 Spleen 100 Fat 10 Carcass 10 Recovery (%) L-C S-C25 L-C25W EFSA Supporting publication 2017:EN-1090

96 Appendix B Average daily feed consumption of rats (broad MOSH mixture experiment) Concentrations (mg/kg feed) Dates Mean SD Mean SD Mean SD Mean SD 25/03/ /03/ /04/ /04/ /04/ /04/ /04/ /04/ /04/ /04/ /04/ /05/ /05/ /05/ /05/ /05/ /05/ /05/ /05/ /06/ /06/ /06/ /06/ /06/ Mean & SD SD: standard deviation EFSA Supporting publication 2017:EN-1090

97 Appendix C Mass spectra of the dominant multibranched hydrocarbons in the broad MOSH mixture GCxGC-MS (EI, m/z 183) plot of the administered oil (top left) and the mass spectra of the dominant components in the region of the multibranched hydrocarbons. The identification was possible only for a few of them EFSA Supporting publication 2017:EN-1090

98 Appendix D Histopathological analysis of the livers Figure 1: Inter-observer (A) and intra-observer (B) reproducibility of histopathological evaluation of granuloma Estimate of the number of sections necessary to obtain a representative granuloma density Seven sections from two cases were evaluated (Figure 2). For the tissue block with the lowest granuloma density, the median for all seven sections was 2/cm 2. When selecting the four sections with the highest granuloma density the median was 2/cm 2. Choosing the four sections with the lowest density, the median was 0/cm 2. Concerning the tissue block with the highest granuloma density, the median granuloma density for all seven sections was 13/cm 2. Selecting the four sections with the highest granuloma density, the median was 20/cm 2, while the median for the four sections with lowest density was 13/cm 2. Based on this relatively modest variation in granuloma density, we decided to examine four sections from each tissue block. The black line denotes the median for all the seven sections in the cases; the red line the median for the four sections with the highest granuloma density and the green line the median for the four sections with the lowest granuloma count. Figure 2: Two tissue blocks from which seven sections were examined 98 EFSA Supporting publication 2017:EN-1090