The use of non-targeted proteomics and in vitro bioassays as a non-destructive approach towards the development of biomarkers of contaminant exposure in threatened marine wildlife

Author Chaousis, Steph

Published 2019-10-31

Thesis Type Thesis (PhD Doctorate)

School School of Environment and Sc

DOI https://doi.org/10.25904/1912/1259

Copyright Statement The author owns the copyright in this thesis, unless stated otherwise.

Downloaded from http://hdl.handle.net/10072/389085

Griffith Research Online https://research-repository.griffith.edu.au

The use of non-targeted proteomics and in vitro bioassays as a non-destructive approach towards the development of biomarkers of contaminant exposure in threatened marine wildlife

Stephanie Chaousis, BSc, Honours

Supervision by Dr. Jason P van de Merwe & Prof. Frederic DL Leusch

School of Environment and Science

Submitted in fulfilment of the requirements of the degree of Doctor of Philosophy on the 24th of April 2019

Abstract Biomarkers of chemical exposure and effect are an important tool for monitoring the health of threatened species that are vulnerable to the adverse effects of prolonged contaminant exposure. However, there are many challenges that have limited the discovery of new biomarkers of chemical exposure in protected species, particularly the constraint on the use of destructive methods such as in vivo experimentation. This thesis first examines the current methods of biomarker discovery as reported in the literature to identify what methods future research in this field should focus on. A systematic quantitative review of methods of non-destructive biomarker discovery in wildlife highlighted the hinderance of these limitations on current research as well as the paucity of studies harnessing in vitro techniques (Chapter 2). The use of in vitro bioassays is not uncommon in ecotoxicology, however, these assays are largely targeted at one or several known markers. Non-targeted proteomics analyses allow for the discovery of new biomarkers, which is important considering wildlife may continually be exposed to new compounds and mixtures. Therefore the application of in vitro and in vivo non- targeted proteomic analysis as a tool for biomarker discovery was examined. To do so, sea turtles were used as a model species due to their priority conservation status and their susceptibility to the adverse effects of contaminant exposure. Cell lines derived from sea turtles exposed in vitro to environmentally relevant contaminants were used as a model for the in vitro analyses. Firstly, experimental sources of variation on protein expression (time, concentration and contaminant type) were explored (Chapter 3) revealing a strong effect of exposure time on observed proteomic changes and little effect from concentration. Furthermore, potential candidate biomarkers were discovered and the relevance of this method to in vivo molecular responses was demonstrated with the observation of known in vivo markers of exposure dysregulated in response to chemical exposure. Then, the influence of biological sources of variation (tissue type)

i were examined with cells derived from several tissue types exposed in vitro to contaminants (Chapter 4). Different tissue types showed a different response to contaminant exposure and known in vivo biomarkers of exposure were again observed.

Furthermore, potential new biomarker candidates were identified that would be beneficial for future research in this field to explore. Finally, non-targeted proteome profiling was applied to biological samples of wild-caught animals to examine its potential in biomarker discovery (Chapter 5). The proteome of the blood plasma of three

Southeast Queensland sea turtle populations exposed to different chemical profiles were compared, and distinct differences in population protein expression were observed.

These differences indicated altered immune states between populations, which could be caused by contaminant and/or pathogen exposure. In conclusion, this novel method of non-targeted proteomic analysis of both in vitro and in vivo samples provides a wealth of information about how contaminants affect sea turtles at the cellular and whole organism level and is a promising avenue to enhance wildlife toxicology. However, it was discovered that it is challenging to draw correlations between in vivo and in vitro global protein expression, and furthermore that sources of experimental variation can greatly influence the outcomes. Therefore, it is vital that future studies on this area focus on reproducibility of these sensitive methods before fully utilising them for directing wildlife toxicology research.

ii Statement of originality This work has not previously been submitted for a degree or diploma in any university.

To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made in the thesis itself.

Signature ______Date: 24/04/2019 Stephanie Chaousis

iii Statement of ethical clearance This work was conducted in accordance with animal ethics requirements approved by the Griffith University Animal Ethics Committee (ENV/13/15/AEC).

Signature ______Date: 24/04/2019 Stephanie Chaousis

iv Table of contents Abstract………………………………………………………………...………………...i Statement of originality………………………………………………………….……...iii Statement of ethical clearance…………………………………………………...……...iv List of tables…………………………………………………………………...…...…..vii List of figures…………………………………………………………………….....…..ix Acknowledgements……………….………………………………………...………….xii Acknowledgement of funding………………………………………………...…….…xiv Contribution of others to this thesis…………………………………....…………….…xv Publications during candidature………………………………….………..…….…...... xv Conference presentations based on thesis………………………………..……...…....xvii Chapter 1: Introduction…………………………………………………………...……..1 1.1 Thesis objectives………………………………………….……….……..….……....7 1.2 Thesis structure…………………………………………………..…………….…....8 1.3 References……………………………………………………...……………..……10 Chapter 2: Charting a path towards non-destructive biomarkers in threatened wildlife: A systematic quantitative literature review…………………………………...……..……19 2.1 Abstract………………………………………………………………………….….20 2.2 Keywords……………………………………………………………….……….….20 2.3 Introduction……………………………………………………….…...………..….20 2.4 Methods……………………………………………………………….……………25 2.4.1 Search tools and parameters………………………………………………….…..25 2.4.2 Categorisations and exclusions…………………………………………...………26 2.4.2.1 Broad experimental categories………………………………………...……….26 2.4.2.2 Detailed experimental categories…………………………………………….…27 2.4.2.3 Classification of methods and quality of experimental design…………….…...28 2.5 Results………………………………………………………………..………...…..29 2.5.1 Search results and experimental categories…………………………………...….29 2.5.2 Classification of methods and quality of experimental design……………...... ….30 2.6 Discussion…………………………………………………………..…………...….32 2.6.1 Experimental categories……………………………………………..………...... 32 2.6.2 Methodological categories…………………………………..…………..…….....36 2.6.2.1 Method 1: Correlation of biomarkers in destructive and non-destructive samples……………………………………………………………………………..…..36 2.6.2.2 Method 2: Correlation of multiple biomarkers in non-destructive samples only……………………………………………………………………………….….....39 2.6.2.3 Method 3: Correlation between site/treatment and detection of biomarkers in non-destructive samples……………………………………………………………...... 44 2.6.2.4 Method 4: Validation of methods for detection of potential non-destructive biomarkers with no contaminant correlations made…………………………..………..45 2.6.2.5 Method 5: In vitro methods using non-destructive samples……………………46 2.6.3 Future directions for non-destructive biomarker discovery in threatened wildlife………………………………………………………………………………….50 2.7 References……………………………………………………………………….…54

v Supporting information for Chapter 2...………………………..………………………66 Chapter 3: Influence of concentration and exposure duration on global protein expression of green sea turtle primary skin cells exposed to polychlorinated biphenyl (PCB) 153 and perfluorononanoic acid (PFNA)…………………………………..…...77 3.1 Abstract……………………………………………………………………………..79 3.2 Key words…………………………………………………………………………..80 3.3 Introduction……………………………………………………………………..….80 3.4 Methods………………………………………………………………………...…..84 3.4.1 Chemicals…………………………………………………………………….…..84 3.4.2 Primary cell culture………………………………………………………………85 3.4.3 Validation that exposure concentrations are not cytotoxic……………….……...85 3.4.4 24 and 48 h exposure bioassays for proteomic analysis………………………….86 3.4.5 Protein extraction…………………………………………………………………86 3.4.6 Protein quantification and digestion…………………………………………..….87 3.4.7 LC-MS/MS…………………………………………………………………….…88 3.4.8 Data analysis………………………………………………………………..….…88 3.5 Results and discussion……………………………………………..………….……91 3.5.1 Cytotoxicity results and cell morphology…………………….…………….….…91 3.5.2 Effect of concentration, exposure time and contaminant on total number of dysregulated proteins……………………………………………………………..….....93 3.5.3 Practical aspects of the methodology used for cellular bioassay proteomic analysis..………………………………………………………………………………..97 3.5.4 Effect of treatments on up and down regulation and compound-specific protein dysregulation………………………………………………………………….……..…98 3.5.5 In vitro expression of known in vivo biomarkers expressed by cells exposed to PCB153 for 24 h………………………………………………………………………101 3.6 Conclusion……………………………………………………………….....……..104 3.7 References……………………………………………………………...……...….105 Supporting information for Chapter 3 ………………………………………..…...... 113 Chapter 4: Changes in global protein expression of five sea turtle (Chelonia mydas) tissue types exposed to PCB 153, PFNA and phenanthrene indicate potential new biomarkers of exposure ………………………………………………………..……..115 4.1 Abstract ……..………………………………………………….…………….…..117 4.2 Keywords….………….………………………………………….……………...... 117 4.3 Introduction…………...…………………………………………………….…….118 4.4 Methods……………..…………………………………………………..……..….120 4.4.1 Chemicals………...……………………………………………………………..120 4.4.2 Primary cell culture ...…………………………………………..……………....121 4.4.3 In vitro exposure to chemical contaminants…………………..………………...122 4.4.4 Extraction of total intracellular and membrane-bound protein content…………122 4.4.5 Protein quantification and digestion by filter aided sample preparation (FASP).123 4.4.6 LC-MS/MS……………………………………………………………………...124 4.4.7 Search of the literature…………………………………………………………..125 4.4.8 Data analysis……………………………………………………………….……125

vi 4.5 Results and discussion…………………………………………………………….127 4.5.1 Overall protein profiles of unexposed cells derived from different tissues……..127 4.5.2 Overall number of differentially expressed proteins between cell types………..130 4.5.3 Identifying biomarkers previously reported in sea turtles…………………...….133 4.5.4 Potential new candidate biomarkers………………………………….………....140 4.6 References………………………………………………………………….….….143 Chapter 5: Non-targeted proteomic analysis of blood from three distinct populations of green sea turtle (Chelonia mydas) reveals significantly altered immune states between populations…………………………………………………………………………….149 5.1 Abstract………………………………………………………………….….……..151 5.2 Keywords………………………………………………………………….…..…..152 5.3 Introduction……………………………………………………………...…...…...152 5.4 Methods………………………………………………………………...…………155 5.4.1 Collection of turtle blood plasma……………………………………………….155 5.4.1.1 Site selection………………………………………………………………….155 5.4.1.2 Sample collection……………………………………………………………..156 5.4.2 Non-targeted proteomic analysis………………..……………………………....157 5.4.2.1 Protein quantification and digestion by Filter Aided Sample Preparation (FASP)………………………………………………………………………………...157 5.4.2.2 Liquid Chromatography-tandem Mass Spectrometry (LC-MS/MS) analysis...158 5.5 Organohalogen analysis…………………………………………………………...158 5.5.1 Samples, chemicals and target compounds……………………………………..158 5.5.2 Sample preparation……..…………………………………………………….…159 5.5.3 Gas Chromatography-Mass Spectrometry (GC-MS) analysis………..………...159 5.5.4 Quality assurance/quality control (QA/QC)…………………………………….159 5.6 Data analysis………………………………………………………………………160 5.6.1 Protein identification and quantification………………………………………..160 5.6.2 Comparison of protein profiles between sites…………………………………..161 5.6.3 Identifying potential markers through analysis of protein function expression...162 5.6.4 Organohalogen concentration in sea turtle plasma……………………………...162 5.7 Results and discussion…………………………………………………………….162 5.7.1 Non-targeted protein analysis of sea turtle plasma……………………………...162 5.7.2 Protein expression profile between sites………………………………………..163 5.7.3 Contaminant exposure and differential protein expression………………….….171 5.8 Conclusions…………………………………………………………………..…...176 5.9 References……………………………………………………………..……….…177 Supporting information for Chapter 5…………………………..…………………….183 Chapter 6: Conclusions, final discussion and future research directions………….….188 6.1 Thesis summary…..……………………….………………………………………188 6.2 Key findings from this thesis……………………………………………………...191 6.3 Future research…………………………………...……………………………….196 6.4 Final remarks………………………………………...……………………………198 6.5 References………………………………………...………………………………198

vii List of tables Table 2.1 Proportion of papers in each of the four categories of studies assigned to the different combinations of destructive methods and/or experimental exposure of animals to contaminants for the development of non-destructive biomarkers. Note that several papers were included multiple times as they contained both method types, therefore the total percent exceeds 100% and number of papers is greater than 82. Numbers are percentage (with the number of papers listed in parenthesis). The total number of papers categorised was 82………………………………………………………………..…….33 Table 2.2 Detailed features (source of animals, success in confirming biomarkers of exposure and effect, and use of in vitro methods) of the studies from each category represented in Table 2.1 defined as “ND & NE” (non-destructive and non- experimental), “ND & E” (non-destructive and experimental), “D & NE” (destructive and non-experimental) and “D & E” (destructive and experimental). See Section 2.4.2.1 for more details on categorisation. The proportion of studies that possessed the features noted in the table from within each of the four categories is represented by a percent with the actual number of studies noted in parentheses. Note that some percentages for the detailed categories can add up to >100% because some papers were assigned to more than one category. The use of cell cultures is automatically non-destructive; therefore, the two destructive broad categories (D&NE and D&E) were deemed ‘non- applicable’ (N/A) under the category of “use of cultured tissue or cells”………...……35 Table 2.3Matrix of eight desirable traits for each of the five methods applied for the discovery of non-destructive biomarkers in threatened wildlife identified in the studies identified in this review……………………………………………………….……..…53 Table S2.1 List of all papers included in the database after a systematic search of the literature was performed. The broad experimental categories, source of test animals used, confirmed or potential biomarkers of exposure and effect, and the method type that each study was categorised into are denoted by a grey circle……………………..67 Table 3.1 Comparison of the coefficient of variation (CoV, calculated as standard deviation / mean) between 24 and 48 h exposure of the number of overall proteins significantly (p<0.0001) differentially expressed at each treatment derived from the data presented in Figure 3.2……...………………………………………………………….96 Table 3.2 The protein identification yields and number of distinct peptides within each false discovery rate (FDR) threshold for all treatments combined at 24 and then 48 h exposure derived from the Protein Pilot search conducted on the independent data acquisition (IDA) ion library…………………………………………………………...98

Table S3.1Calculated EC10 for cell viability for each compound at 24 and 48 hours of exposure expressed in µM and the highest exposure concentration of 100 µg/L expressed in µM for each compound. ‘N/A’ indicates that EC10 could not be calculated due to an insufficient change in cell viability…………………………………………113 Table S3.2. Table indicating the percent of all proteins significantly differentially expressed (p-value<0.0001) at each exposure time that are dysregulated exclusively within that chemical treatment (and not in the other) after exposure at 24 or 48 h…...114

viii Table 4.1Table showing the number of differentially expressed proteins for each tissue type and treatment that had over a fold change of ≥ 2 for up-regulated proteins and £ -2 for down-regulated proteins…………………………………………………………...132 Table 4.2 Table showing results of a literature search for molecular markers of contaminant (PCB, PFNA (or PFC), Phenanthrene (or PAHs)) exposure in marine (a) and freshwater (b) turtles……………………………………………………………...134 Table 4.3. A table indicating the most strongly up- (a) and down- (b) regulated proteins for each tissue type and treatment relative to controls (p<0.01) along with fold change (FC) values and standard error (SE)…………………………………………………..141 Table 5.1 Table indicating the source of chemical contaminants at each study site along with classes of contaminants previously detected at each site…………………..……156 Table 5.2 Number of proteins and peptides identified at 1% and 5% false discovery rate thresholds calculated at both the local and global level……………………………....163 Table 5.3 Summary of the total number of significantly (p<0.0001) differentially expressed proteins between sites and the lowest and highest value of fold change for each comparison (derived from Figure S5.1). FC = fold change……………….…….167 Table 5.4 Functional role of the 10 most strongly up and down regulated from each site comparison as listed in Figure 5.1 as described in previous studies. APP = acute phase protein. *only 15 of the 17 proteins are described here as two of the proteins listed in Figure 5.1 are uncharacterized and therefore no functional information currently exists…………………………………………………………………………………..169 Table 5.5 Concentrations (pg/mL) of organohalogens detected in green sea turtle plasma collected from Hervey Bay (n=25), Port of Gladstone (n=24) and Moreton Bay (n=23). SD = standard deviation, *indicates that this compound was only detected in one replicate at this site hence there is no reported SD. ND = not detected. Note that the following compounds were measured but not detected in any of the plasma samples (with detection limits ranging from 1-4 pg/mL): CB 28, 49, 52, 74, 95, 99, 101, 105, 118, 128, 146, 156, 170, 171, 177, 183, 194, 199, 206 and 209; BDE 28, 99, 100, 153, 154 and 183), p,p’-DDE, p,p’-DDT, chlordanes (oxychlordane, cis-nonachlor, cis- chlordane) and HCH (α-, β-, γ-)……………………………………………………....173 Table S5.1 Results of the principal component analysis (PCA). Dim. = Dimension...186 Table S5.2 a) Results of non-parametric one-way ANOVA (Kruskal-Wallis) comparing concentrations of selected OH between sites and b) results of multiple comparisons indicating which sites were significantly different for sites in which ANOVA indicated statistical significance p<0.01 2-MeO-BDE68, trans-nonachlor (TN) and BDE 47. Significant values are highlighted in bold…………………………………………….187 Table 6.1 Summary of the objectives and main findings from this thesis……………192

ix List of figures Figure 1.1 Visual representation of experimental workflow of thesis by chapter……....8 Figure 2.1 Percent of all studies within each broad category defined as “ND & NE” (non-destructive and non-experimental), “ND & E” (non-destructive and experimental), “D & NE” (destructive and non-experimental) and “D & E” (destructive and experimental) represented by bars (ND&NE, 61%; ND&E, 15%; D&NE, 12%; D&E, 20%). The proportion of papers within each broad category is divided into the five method types, described in section 2.4.2.2, represented as a percent of all papers reviewed (n=82) that used the indicated method type, and fell into the specified broad category, by patterned segments. Method 1: correlation of biomarkers in destructive and non-destructive samples; Method 2: correlation of multiple biomarkers in non- destructive samples only; Method 3: correlation between exposure (according to habitat or treatment) and detection of biomarkers in non-destructive samples (i.e. contaminant levels were not measured in the organism); Method 4: validation of the detection of biomarkers and/or of non-destructive samples, with no correlations made; and Method 5: tissue or cell-based in vitro methods. Note that the sum of all percentages is higher than 100 because several papers fall under multiple categories………………………..40 Figure 3.1 Percent survival of G12s-p (C. mydas skin) cells exposed to A; 0.08 µM – 170 µM PCB 153, and B; 0.5 µM – 800 µM PFNA, over 24 and 48 hours. Error bars represent standard error of the mean (SEM) (n = 6)……………………………………92 Figure 3.2 The total number of proteins differentially expressed (p-value <0.0001) in cells exposed to a) PCB 153 and b) PFNA compared to controls. Error bars represent standard error of the mean (SEM) (n=3)……………………………………………….95 Figure 3.3 Proportion (calculated from the total number of differentially expressed proteins within treatments) of proteins that were up-regulated and down-regulated in cells exposed to a) PCB 153 or b) PFNA. Error bars represent standard error of the mean (SEM) (n = 3)………………………………………………………………….100 Figure 3.4 Fold change of a & c) superoxide dismutase (SOD) and b) glutathione S- transferase compared to the control samples, in cells exposed three concentrations of PCB153 and PFNA for 24 h. Error bars represent SEM (n = 3). Data for GST from PFNA exposure not shown as it was not significantly dysregulated compared to controls at any of the exposure concentrations. *SOD was not dysregulated at these concentrations…………………………………………………………………………103 Figure S3.1 Images showing (a) no morphological changes to control cells after 48 h incubation compared to (b) changes seen in cells after 48 h exposure to PCB153 100 µg/L. Before images were taken of cells between 18-24 hours after seeding in 25cm2 tissue culture flasks before contaminant exposure and after images were taken immediately after 48 h exposure……………………………………………………...113 Figure 4.1 Venn diagram indicating the shared and unique proteins identified by data dependent acquisition (DDA) mass spectrometry of cell lines derived from green sea turtle (Chelonia mydas) liver, kidney, ovary, small intestine and skin tissue. The vertical bar graph below the venn diagram indicates how many proteins were in each list and the horizontal bar scale indicated how many proteins were shared between 1, 2, 3, 4 and all tissue groups. (Made using jvenn online software (Bardou et al., 2014))…………….128

x Figure 4.2 Heatmap displaying fold change of significantly (p<0.01) differentially expressed known protein biomarkers of PCB, PFNA or phenanthrene exposure in marine and freshwater turtles as identified from the literature (see Table 4.2). A fold change of one indicates double the expression of a protein relative to the controls. Standard error is <0.17 for all values. OS=Organism Species GN=Gene Name……..139 Figure 5.1 Principal component analysis (PCA) plotting PC1 (36.5% of the variance) against PC4 (4.2% of the variance). The PC1 axis is indicative of overall protein expression (from low on the left to high on the right), while the PC4 axis is driven mostly by proteins associated with immune responses (from low expression at the bottom to high at the top). Ellipses indicates the region within 1 SD of the group average (68% probability). PG = Port of Gladstone (n=25), HB = Hervey Bay (n=24), MB = Moreton Bay (n=24)……………………………………………………………165 Figure 5.2 Heatmap comparing the protein expression profiles of each site indicating the five mostly strongly up-regulated and five most strongly down-regulated proteins for each site comparison. A positive number (orange) indicates that expression at the second site was higher than at the first site (e.g., adiponectin-like protein was 1.368× more expressed at MB than at HB). A negative number (blue) indicates a lower expression at the second site compared to the first site. A blank cell indicates that expression of this protein was not significantly different between the two sites. Several of the proteins were most strongly up or down regulated for more than one comparison, hence there are only 23 proteins listed. HB = Hervey Bay (n=24), PG = Port of Gladstone (n=24) and MB = Moreton Bay (n=24)……………………………………166 Figure S5.1 Results of the PCA analysis, showing: (a) Scree plot of the top 25 dimensions; (b) PC1 vs PC2; (c) PC1 vs PC3; (d) PC1 vs PC4; and (e) PC1 vs PC5..183 Figure S5.2 Heatmap of protein expression comparisons between the three sites with all proteins significantly (p<0.0001) differentially expressed included. A positive number (blue) indicates that expression at the second site was higher than at the first site (e.g., adiponectin-like protein was 1.368× more expressed at MB than at HB). A negative number (orange) indicates a lower expression at the second site compared to the first site. A blank cell indicates that expression of this protein was not significantly different between the two sites. HB = Hervey Bay (n=24), PG = Port of Gladstone (n=24) and MB = Moreton Bay (n=24)……………………………………………………………184 Figure S5.3Average concentration (pg/mL) of organohalogen grouped by site: Port of Gladstone (n=24), Hervey Bay (n=25) and Moreton Bay (n= 23) with error bars representing SD. The only compounds with significantly different means across sites were 2-MeO-BDE68 (p<0.0001), TN (p=0.0071) and BDE47 (p=0.0079). However, a multiple comparisons test only distinguished significantly different concentrations of 2- MeO-BDE68 between sites (Table S5.2)……………………………………………..185

xi Acknowledgements

This work could only have been achieved with the input and assistance of many mentors, colleagues, peers and family and friends who offered professional and personal support throughout the duration of this project.

A special thank you to my supervisory team Jason van de Merwe and Frederic Leusch for taking me on as a student and having confidence in me and my abilities. I am grateful for the guidance, support and expertise you provided throughout all of the stages of my project and particularly the freedom you allowed me to explore and test my ideas. I have learned an incredible amount from the both of you and couldn’t have asked for a better supervisory team!

I’d like to acknowledge my co-author Amanda Nouwens for her invaluable guidance with non-targeted proteomics and mass spectrometry as well as Peter Josh for his assistance in this aspect of the project too.

Thank you to my co-author Steve Melvin for his input into the interpretation of the data and writing and drafting of manuscripts. A further thank you to co-authors Liesbeth

Weijs, Antonia Weltmeyer and Adrian Covaci for help with PCB analysis and data interpretation.

Thank you to Chris Henderson, Sophie Doell and Nick Goody for volunteering their time to help on field trips to Hervey Bay. Thank you to Don and Lesley for assisting in access to the sampling site in Hervey Bay.

Thank you to Kim Finlayson who assisted greatly with blood sample collection in the field and provided cells for this research, as well as trained me in the techniques of primary cell culture. Also, for being a great companion as a fellow PhD student and friend to chat, vent and laugh with over the last four years!

xii Thank you to the other lab group members Chantal Lanctot, Peta Neale, Erik Prochazka,

Shima Ziajahromi and Ben Matthews for your help in the lab and office and for providing a great, supportive working environment.

This project would not have been possible without the tireless dedication of Dr. Col

Limpus from the Department of Environment and Science (DES). Col and DES provided extensive in-kind help in terms of field sampling, which enabled the collection of blood and tissue samples for this project. A further thank you to all of the other researchers and volunteers who were a part of the DES field trips who assisted in capturing and sampling the turtles with us.

A sincere thank you to all of the organisations that worked with us to provide samples for analysis starting with Marnie Horton, David Blyde and Kelly Wood from Sea

World. Thank you to the veterinary team at Dolphin Marine Magic, Duan March and

Kieran Marshall as well as the staff at Australia Zoo and Australian Seabird Rescue for providing your expertise and assistance in access to samples from rehabilitated wildlife at these facilities. Furthermore, thank you to WWF and the volunteers who assisted in the collection of field samples.

Thank you to my friends and family, and especially to Chris, for being there for me throughout this process, for your belief in my abilities and unending encouragement and support, you made the hard days so much easier.

xiii Acknowledgement of funding

This thesis was funded by Griffith University School of Environment and Science, the

Australian Rivers Institute and the Griffith University Post Graduate Research

Scholarship.

Signature ______Date: 24/04/2019 Stephanie Chaousis

xiv Contribution of others to this Thesis Dr. Jason van de Merwe and Professor Frederic Leusch contributed greatly in guiding project conceptualisation, planning, data interpretation and analysis as well as being co- authors on all prepared work. Dr. Steven Melvin, Dr. Amanda Nouwens, Dr. Liesbeth

Weijs, Antonia Weltmeyer, Professor Adrian Covaci and Dr. Col Limpus, provided assistance with lab work, data analysis and interpretation and drafting of articles in selected chapters.

Publications during candidature Included in this thesis is one paper that has been published (Chapter 2), two papers that have been submitted for publication (Chapter 3 and Chapter 4) and one that will be submitted shortly (Chapter 5), all of which were co-authored with other researchers. My contribution to each co-authored paper is outlined at the beginning of the relevant chapter. The bibliographic details are as follows.

Stephanie Chaousis, Frederic DL Leusch & Jason P van de Merwe (2018) Charting a path towards non-destructive biomarkers in threatened wildlife: A systematic quantitative literature review. Environ Poll. 234: 59-70.

Stephanie Chaousis, Frederic DL Leusch, Amanda Nouwens, Steven D Melvin and

Jason P van de Merwe (under review) Influence of concentration and exposure duration on global protein expression of green sea turtle primary skin cells exposed to polychlorinated biphenyl (PCB) 153 and perfluorononanoic acid (PFNA). Environ Sci

Technol.

xv Stephanie Chaousis, Frederic DL Leusch, Amanda Nouwens, Steven D Melvin and

Jason P van de Merwe (under review) Changes in global protein expression of five sea turtle (Chelonia mydas) tissue types exposed to PCB 153, PFNA and phenanthrene indicate potential new biomarkers of exposure. Ecotox Environ Safe.

Stephanie Chaousis, Frederic DL Leusch, Col Limpus, Amanda Nouwens, Liesbeth

Weijs, Antonia Weltmeyer, Adrian Covaci and Jason P van de Merwe (to be submitted)

Non-targeted proteomic analysis of blood from three distinct populations of green sea turtle (Chelonia mydas) reveals significantly altered immune states between populations. Sci Tot Environ.

Appropriate acknowledgements of those who contributed to the research but did not qualify as authors are included in each paper.

(Signed) ______(Date) 24/04/2019 Stephanie Chaousis

(Countersigned) ______(Date) 23/04/2019 Jason van de Merwe

xvi Conference presentations based on thesis

Stephanie Chaousis, Frederic DL Leusch, Amanda Nouwens, and Jason P van de

Merwe. Changes in protein expression of primary sea turtle cells exposed to contaminants indicate the potential for in vitro proteomics as a high throughput tool to support biomarker discovery. Society for Environmental Toxicology & Chemistry

(SETAC) Europe 28th Annual Meeting. Rome, Italy, May 2018.

Stephanie Chaousis, Frederic DL Leusch, Amanda Nouwens, and Jason P van de

Merwe. Altered protein expression of primary sea turtle cells exposed to contaminants indicates the potential for in vitro proteomics as a high throughput tool to support biomarker discovery in threatened wildlife. The American Society for Biochemistry &

Molecular Biology, San Diego, USA, April 2018.

Stephanie Chaousis, Frederic DL Leusch, Amanda Nouwens, and Jason P van de

Merwe. How can mass spectrometry tools assist the conservation of endangered species? The Queensland 2nd Mass Spectrometry Symposium QIMR Brisbane,

Australia, November 2017.

Stephanie Chaousis, Frederic DL Leusch, Amanda Nouwens, and Jason P van de

Merwe. Combining in vitro techniques and non-targeted omics to discover new biomarkers and enhance non-destructive monitoring of adverse effects of contaminants on wildlife. Society for Environmental Toxicology and Chemistry Australasia (SETAC

AU) Conference 2017 Gold Coast Australia, September 2017.

xvii Stephanie Chaousis, Frederic DL Leusch, and Jason P van de Merwe. A novel approach to establishing biomarkers of chemical contaminant exposure and effect in marine megafauna. The 8th International Conference on Marine Pollution and

Ecotoxicology Hong Kong, July 2016.

Stephanie Chaousis, Frederic DL Leusch, and Jason P van de Merwe. Novel methods for the development of non-destructive biomarkers of chemical contaminant exposure and effect in sea turtles. Australian Rivers Institute (ARI) Higher Degree Research

Student Symposium Brisbane, Australia. June 2016.

xviii Chapter 1 Introduction Anthropogenic chemical pollutants are present in the marine environment from a number of sources, including pesticides in agricultural run-off, industrial compounds in urban and industrial wastewater and consumer chemicals and pharmaceuticals in urban sewage (Daszak et al. 2001, Islam and Tanaka 2004, Schwarzenbach et al. 2006). These anthropogenic chemical contaminants tend to be concentrated around densely populated or industrialised coastal areas (Islam and Tanaka 2004), and many have been shown to accumulate in the tissue of large marine vertebrates such as whales (Nielsen et al. 2000,

Waugh et al. 2014), sea turtles (Gardner et al. 2006, van de Merwe et al. 2010a), dolphins (Cardellicchio et al. 2000, Tornero et al. 2006) and seals (Van Loveren et al.

2000, Ikemoto et al. 2004). These species are particularly vulnerable to chemical accumulation as they are typically long lived, fat storing, positioned at higher trophic levels, and feed in coastal areas that potentially have high concentrations of contamination (Muir et al. 1999, Halpern et al. 2008). Furthermore, species that have a period of fasting, such as whales during migration or sea turtles during breeding season, are potentially exposed to concentrated mixtures of lipophilic chemicals that have accumulated in the lipid stores over time and are remobilised during these periods (van de Merwe et al. 2010b, Bengtson Nash et al. 2013). As a consequence, bioaccumulation and mixture effects can increase the negative impacts of contaminants on these animals.

The chemicals that accumulate in marine wildlife can have a wide range of effects.

Aside from lethal toxicity (Oaks et al. 2004, Relyea and Diecks 2008), contaminants have also been found to have less immediately apparent negative effects in marine fauna, such as immune suppression (Ross et al. 1996, Van Loveren et al. 2000, Keller et al. 2006) and reproductive dysfunction (Reijnders 1986, Addison 1989, Porte et al.

2006). For example, endocrine disrupting chemicals (EDCs) can cause feminisation of males or masculinisation of females of a population through either morphological

1 changes, such as poorly developed gonads (Guillette et al. 1996) or hormonal changes that interfere with sexual development or reproduction (Reijnders 1986, Guillette and

Gunderson 2001). Furthermore, immune suppression can cause individuals to become more vulnerable to other stressors, such as parasites, which can reduce reproductive fitness due to the energetic costs of this parasite load (Booth et al. 1993, Harvell et al.

2004, Van Bressem et al. 2009). Less is known about the effects of contaminants on marine megafauna despite the fact that this group of animals are particularly vulnerable to contaminants (Kelly et al. 2007). These factors, together with additional anthropogenic pressures impacting wildlife, make it challenging to predict long term effects of contaminants on individuals, populations and ultimately, whole ecosystems

(Spurgeon et al. 2010).

Chemical contamination has been cited as a contributing factor to the current decrease in global biodiversity (Bickham et al. 2000). Climate change and other human impacts on marine fauna (e.g., overfishing) are potentially compounded by chemical contaminant exposure that can impact on individual and population fitness through a number of sublethal mechanisms (Gouin et al. 2013, Moe et al. 2013, Hooper et al.

2013, Noyes et al. 2009, Kendall 2010). These subtle effects may cause population decline over time, which might only be noticed when it is too late to reverse the problem. For example, a study of a wild fish population exposed experimentally to a synthetic estrogen used in birth control pills (17α–ethinylestradiol) often found in wastewater, observed the eventual total collapse of the population as male fish were feminised and could no longer reproduce (Kidd et al. 2007). Even further concerning and more difficult to correlate to contaminant exposure are transgenerational effects.

Numerous laboratory-based studies have identified that maternal exposure to contaminants can have flow-on effects to the next few generations of offspring even without direct exposure (Matta et al. 2001, Baccarelli and Bollati 2009, Manikkam et al.

2 2012). This exemplifies the urgency to monitor and mitigate effects of contaminants on vulnerable populations of marine wildlife as a part of conservation and management efforts.

There are a number of studies that have identified correlations between contaminant exposure and adverse effects in threatened wildlife. For example, abnormal hormone levels have been associated with residues of PCBs in seals, dolphins and whales

(Tanabe 2002). A study on a population of Baltic seals showed that reproductive success was significantly lowered in correlation with exposure of this population to organochlorine contaminants (Vos et al. 2000). Furthermore, immune suppression due to contaminant exposure has been observed in experiments with wild seals (de Swart et al. 1996) and was the suspected cause for mass mortalities of seal populations living in northwestern Europe (Van Loveren et al. 2000). More recently, a study on wild Atlantic bottlenose dolphins found a correlation between exposure to perfluoroalkyl compounds and changes in immune parameters (Fair et al. 2013). There is still a need to discover new biomarkers as effects of chemical exposure can vary across different species and a biomarker for one species may not be suitable for another.

Monitoring the health of chemically exposed wildlife, however, requires knowledge of the specific adverse effect(s) of each contaminant. Toxicity testing has historically been conducted using controlled experiments with laboratory animals, although developments in our understanding of mechanistic toxicology is paving the way for a significant change towards in vitro and in silico alternatives (NRC 2007). Studying toxicity in marine wildlife is particularly incompatible with conventional “whole animal” toxicity testing models, as marine megafauna are often threatened species. The method of monitoring when dealing with marine wildlife is thus preferably non- destructive and causes minimal harm. While there are several non-destructive ways to monitor population health such as through morphological (Leusch et al. 2006) or

3 behavioural changes (Melvin and Wilson 2013), these are generally the latest stage at which evidence of adverse effects manifest. It is advantageous if adverse effects of chemical exposure are detected as early as possible. Changes at the molecular level are the first point of physiological effect of exposure on the adverse outcome pathway framework (Ankley et al. 2010), and therefore the best point at which to measure early signs of effect.

Endogenous molecular indicators of negative health, commonly referred to as biomarkers, are a great tool for early detection of health deterioration that has been well established in humans (Etzioni et al. 2003, Vaidya and Bonventre 2010). ‘Biomarker’ is a far reaching and sometimes ambiguous term, and hence has several definitions with varying scopes. However, a widely accepted definition of a biomarker is, “a biological response to a chemical or chemicals that gives a measure of exposure and sometimes, also, of toxic effect” (Peakall 1994). A biomarker of chemical exposure and effect is an endogenous response of an organism exposed to contaminants, such as the up or down regulation of a protein. For example, an increase in circulating enzymes such as glutathione S-transferases is a well-studied marker of exposure to environmental pollutants (Valavanidis et al. 2006, Antunes et al. 2010, Jemec et al. 2010). In addition to the benefit of not causing harm to or death of an animal, non-destructive biomarkers can also be used in temporal monitoring of animal health. Generally, studies that establish biomarkers of effect are laboratory based, however, due to their large size and ethical constraints there are inherent difficulties in studying marine megafauna species such as turtles, dolphins and whales in the lab.

Non-destructive biomarker discovery is often a difficult and long process, especially in the case of biomarkers of chemical exposure in wild animals. Benninghoff (2006) provides some guidelines that are useful to ensure that this process is done in the most efficient and effective way: studies developing biomarkers should aim to produce a

4 biomarker that can be induced and repressed, is specific to chemicals or mixtures of exposure, present a significant enough response that it can be measured routinely, and is highly reproducible and quantifiable. Normally, in vivo lab-based experiments are done to tackle the difficult task of biomarker discovery for contaminant exposure. Such experiments involve the exposure of an animal to controlled amounts of chemicals before biological parameters are measured and any adverse outcomes directly attributed to the exposure. However, this is difficult in the case of marine megafauna as these individuals can be difficult to handle or not permitted to be used for in vivo experiments due to ethical constraints. A potential alternative to in vivo testing is the use of in vitro techniques for biomarker discovery, which look at changes in gene or protein expression of cells following exposure to a contaminant or mixtures of contaminants.

Such methods have been used to develop biomarkers of human diseases with success, particularly protein biomarkers associated with diseases such as cancer and diabetes

(Grønborg et al. 2006, Pavlou and Diamandis 2010, Stastna and Van Eyk 2012).

However, such studies have not generally been used for biomarker discovery in wildlife.

Protein expression is a readily available option to measure changes in cell response and hence in vitro biomarker discovery in wildlife (Skalnikova et al. 2011). Previous studies have observed changes in protein expression of cells derived from animals that have been exposed to contaminants in vivo, indicating that protein expression was altered in the whole animal in response to this exposure (Wei et al. 2008). Alternatively, primary cell lines of these animals could be used for analysing changes in cell secretome

(excreted proteins) in response to contaminant exposure, as an indicator of in vivo response. Critical to using in vitro exposure experiments to develop biomarkers of chemical exposure and effect in marine megafauna is the establishment of whether

5 changes in protein expression in response to contaminant exposure seen in vitro also occur in vivo (whole animals).

Green sea turtles (Chelonia mydas) are an ideal model species for the development of new methods in this field as 1) they are listed globally as threatened, vulnerable or endangered, 2) they are at risk of chemical exposure as they are coastal foragers, fat storing, long lived and have a strong fidelity to their foraging site (Limpus et al. 1992) meaning repeated exposure to the same contaminants, and 3) there are a number of chemical exposure and effect biomarkers already identified for this species. Some preliminary studies have been instigated on biomarker discovery for specific contaminants in sea turtles; however, the number of such studies is quite limited and success rate of biomarker discovery is relatively low (Finlayson et al. 2016). This clearly suggests that there is a significant knowledge gap and still a lot of work to be done on biomarkers of contaminant exposure in sea turtles. Studies aiming to find biomarkers of exposure in sea turtles are limited in number with the first study conducted only in 2003 (Keller 2003). A recent review by Finlayson et al. (2016) found only 13 studies looking at biomarkers in sea turtles. Less than half of these studies were successful in correlating biomarker induction with contaminant exposure indicating there is still a large gap in sea turtle biomarker research.

6 1.1 Thesis objectives

The overarching goal of this thesis was to assess the usefulness of non-targeted proteome profiling both in vitro and in vivo as a tool to assist biomarker discovery in threatened wildlife using sea turtles as a model species.

The main objectives of this thesis were:

1) Assess the current state of non-destructive biomarker discovery methods in wildlife, and consider which methods appear to be more or less successful and which ones are most widely applicable to threatened species.

2a) Optimise the experimental conditions for in vitro exposure followed by non-targeted proteomics analysis. Determine the effects that experimental parameters can have on the proteome.

2b) Assess how well the observed cellular level effects of exposure represent those identified from in vivo exposure.

3a) Determine the impact of biological variation (different tissues) on in vitro protein expression

3b) Compare the whole organism congruence of several different cell lines from the same species.

3c) Identify potential biomarkers of exposure and effect in sea turtles.

4) Assess the usefulness of non-targeted proteomics on wild-caught turtle samples for identifying differences in marker expression between populations with differing exposure profiles.

7

1.2 Thesis structure

This thesis is comprised of six chapters: a general introduction (Chapter 1), a semi- quantitative literature review (Chapter 2), three research chapters investigating various aspects of protein biomarker discovery in sea turtles (Figure 1.1), and a general discussion (Chapter 6).

Figure 1.1 Visual representation of experimental workflow of thesis by chapter.

This thesis includes a series of stand-alone papers (chapters 2-5) that each address one of the objectives listed above and have, or will be, submitted to a peer reviewed journal for publication. There may be some repetition between chapters, particularly in the methods sections, as each chapter needs to stand alone as a scientific manuscript.

8 Chapter 1 (this chapter) introduces the topic in general and provides a brief overview of relevant background information along with the thesis objectives and structure. A more detailed review of the literature is presented in chapter two.

Chapter 2 addresses the first aim of this study, which was to assess the current state of non-destructive biomarker discovery in wildlife. It presents a summary and critical review on the current literature covering the methods that have been applied to non- destructive biomarker discovery in wildlife. The results of a systematic, quantitative literature search are presented, and the methods utilised in the studies categorised based on numerous factors primarily focused on their use of non-destructive techniques. The relative success of each method category in discovering biomarkers of exposure and effect is assessed and gaps in current research on useful methods are identified.

Chapter 3 starts the process of optimising a new method of in vitro biomarker discovery in which alterations in the proteome of sea turtle skin cells after exposure to contaminants is examined. This chapter presents an investigation of sources of technical variability through the testing of several different experimental parameters (exposure duration, concentration and contaminant). Furthermore, in order to assess the relevance of this method to in vivo outcomes, the resulting cellular changes are compared to known markers of exposure from the literature.

Chapter 4 examines how biological sources of variation might impact cellular protein expression. Primary cell lines derived from five different tissue types were exposed to three different environmentally ubiquitous contaminants. The resulting changes in the proteome were compared to each other to determine the similarities and/or differences in the molecular response of each tissue type to contaminant exposure.

The objective of the final research chapter (Chapter 5) was to assess how non-targeted proteomics might be useful in vivo by applying this method to sea turtle blood plasma.

In Chapter 5, the blood plasma proteome from three different sea turtle populations is

9 compared. Distinct differences in protein expression levels between populations are investigated and the potential sources of these differences considered.

Finally, Chapter 6 discusses these results collectively with consideration of the limitations of these studies and suggestions for future research. The references for each chapter are located immediately after the text and all supporting information is located at the end of each relevant chapter.

1.3 References

Addison, R F. 1989. Organochlorines and Marine Mammal Reproduction. Can J Fish Aquat Sci 46:360-368. Anderson, N L., Polanski, M., Pieper, R., Gatlin, T., Tirumalai, R S., Conrads, T P., Veenstra, T D., Adkins, J N., Pounds, J G., Fagan, R. and Lobley, A. 2004. The human plasma proteome: a nonredundant list developed by combination of four separate sources. Mol Cell Proteomics 3:311-326. Ankley, G T., Bennett, R S., Erickson, R J., Hoff, D J., Hornung, M W., Johnson, R D., Mount, D R., Nichols, J W., Russom, C L., Schmieder, P K., Serrrano, J A., Tietge, J E. and Villeneuve, D L. 2010. Adverse outcome pathways: a conceptual framework to support ecotoxicology research and risk assessment. Environ Toxicol Chem. 29:730-741. Arukwe, A. and Goksøyr, A. 2003. Eggshell and egg yolk proteins in fish: hepatic proteins for the next generation: oogenetic, population, and evolutionary implications of endocrine disruption. Comp Hepatol 2(1):4. Attademo, A M., Peltzer, P M., Lajmanovich, R C., Cabagna, M. and Fiorenza, G. 2007. Plasma B-esterase and glutathione S-transferase activity in the toad Chaunus schneideri (Amphibia, Anura) inhabiting rice agroecosystems of Argentina. Ecotoxicology 16:533-539. Baccarelli, A. and Bollati, V. 2009. Epigenetics and environmental chemicals. Curr Opin Pediatr 21:243-251. Bengtson Nash, S M., Waugh, C A. and Schlabach, M. 2013. Metabolic concentration of lipid soluble organochlorine burdens in the blubber of Southern Hemisphere humpback whales through migration and fasting. Environ Sci Technol. 47:9404- 9413.

10 Benninghoff, A D. 2006. Toxicoproteomics--The Next Step in the Evolution of Environmental Biomarkers? Toxicol Sci. 95:1-4. Bickham, J W., Sandhu, S., Hebert, P D., Chikhi, L. and Athwal, R. 2000. Effects of chemical contaminants on genetic diversity in natural populations: implications for biomonitoring and ecotoxicology. Mutat Res 463:33-51. Booth, D T., Clayton, D H. and Block, B A. 1993. Experimental Demonstration of the Energetic Cost of Parasitism in Free-Ranging Hosts. P Roy Soc B-Biol Sci 253:125-129. Broderick, A C., Coyne, M S., Fuller, W J., Glen, F. and Godley, B J. 2007. Fidelity and over-wintering of sea turtles. Proc Biol Sci 274:1533-1538. Casini, S., Fossi, M C., Cavallaro, K., Marsili, L. and Lorenzani, J. 2002. The use of porphyrins as a non-destructive biomarker of exposure to contaminants in two sea lion populations. Mar Environ Res 54:769-773. Champoux, L., Rodrigue, J., Trudeau, S., Boily, M H., Spear, P A. and Hontela, A. 2006. Contamination and biomarkers in the great blue heron, an indicator of the state of the st. lawrence river. Ecotoxicology 15:83-96. Cox, J. and Mann, M. 2007. Is proteomics the new genomics? Cell 130:395-398. Daszak, P., Cunningham, A A., and Hyatt, A D. 2001. Anthropogenic environmental change and the emergence of infectious diseases in wildlife. Acta Trop 78:103- 116. de Swart, R L., Ross, P S., Vos, J G. and Osterhaus, A D M E. 1996. Impaired Immunity in Harbour Seals (Phoca vitulina) Exposed to Bioaccumulated Environmental Contaminants: Review of a Long-term Feeding Study. Environ Health Persp 104(suppl 4):823-828. Denslow, N D. 2000. Vitellogenin as a Biomarker of Exposure for Estrogen or Estrogen Mimics. Ecotoxicology 8:385-398. El-Shahawi, M S., Hamza, A., Bashammakh, A S. and Al-Saggaf, W T. 2010. An overview on the accumulation, distribution, transformations, toxicity and analytical methods for the monitoring of persistent organic pollutants. Talanta 80:1587-1597. Espín, S., Martínez-López, E., Jiménez, P., María-Mojica, P. and García-Fernández, A J. 2015. Delta-aminolevulinic acid dehydratase (δALAD) activity in four free- living bird species exposed to different levels of lead under natural conditions. Environ Res 137:185-198.

11 Etzioni, R., Urban, N., Ramsey, S., McIntosh, M., Schwartz, S., Reid, B., Radich, J., Anderson, G. and Hartwell, L. 2003. The case for early detection. Nat Rev Cancer 3:243-252. Fair, P A., Romano, T., Schaefer, A M., Reif, J S., Bossart, G D., Houde, M., Muir, D., Adams, J., Rice, C. and Hulsey, T C. 2013. Associations between perfluoroalkyl compounds and immune and clinical chemistry parameters in highly exposed bottlenose dolphins (Tursiops truncatus). Environ Toxicol Chem 32:736-746. Finlayson, K A., Leusch, F D. and van de Merwe, J P. 2016. The current state and future directions of marine turtle toxicology research. Environ. Int. 94:113-123. Fossi, C., Depledge, M. and Cordi, B. 1997. Temporal biomarker responses in wild passerine birds exposed to pesticide spray drift. Environ Toxicol Chem 16:2118. Fossi, M C. 1994. Nondestructive biomarkers in ecotoxicology. Environ Health Persp 102 Suppl 12:49-54. Gouin, T., Armitage, J M., Cousins, I T., Muir, D C., Ng, C A., Reid, L. and Tao, S. 2013. Influence of global climate change on chemical fate and bioaccumulation: The role of multimedia models. Environmental toxicology and chemistry, 32(1):20-31. Grønborg, M., Kristiansen, T Z., Iwahori, A., Chang, R., Reddy, R., Sato, N, Molina, H., Jensen, O N., Hruban, R H. and Goggins, M G. 2006. Biomarker discovery from pancreatic cancer secretome using a differential proteomic approach. Mol Cell Prot 5:157-171. Guillette, L J. and Gunderson, M P. 2001. Alterations in development of reproductive and endocrine systems of wildlife populations exposed to endocrine-disrupting contaminants. Reproduction 122:857-864. Guillette, L J., Pickford, D B., Crain, D A., Rooney, A A. and Percival, H F. 1996. Reduction in penis size and plasma testosterone concentrations in juvenile alligators living in a contaminated environment. Gen Comp Endocr 101:32-42. Halpern, B S., Walbridge, S., Selkoe, K A., Kappel, C V., Micheli, F., D'Agrosa, C., Bruno, J F., Casey, K S., Ebert, C. and Fox, H E. 2008. A global map of human impact on marine ecosystems. Science 319:948-952. Hansen, P-D., Dizer, H., Hock, B., Marx, A., Sherry, J., McMaster, M. and Blaise, C. 1998. Vitellogenin–a biomarker for endocrine disruptors. Trac-Trend Anal Chem 17:448-451.

12 Harvell, D., Aronson, R., Baron, N., Connell, J, Dobson, A., Ellner, S., Gerber, L., Kim, K., Kuris, A., McCallum, H., Lafferty, K., McKay, B., Porter, J., Pascual, M., Smith, G., Sutherland, K., and Ward, J. 2004. The rising tide of ocean diseases: unsolved problems and research priorities. Front Ecol Environ 2:375-382. Hooper, M J., Ankley, G T., Cristol, D A., Maryoung, L A., Noyes, P D. and Pinkerton, K.E. 2013 Interactions between chemical and climate stressors: A role for mechanistic toxicology in assessing climate change risks. Environ Toxicol Chem 32(1):32-48. Irwin, L K. 2001. Vitellogenin induction in painted turtle, Chrysemys picta, as a biomarker of exposure to environmental levels of estradiol.Aquat Toxicol 55(1- 2):49-60. Islam, M S. and Tanaka, M. 2004. Impacts of pollution on coastal and marine ecosystems including coastal and marine fisheries and approach for management: a review and synthesis. Mar Pollut Bull 48:624-649. Jiménez, B., Mori, G., Concejero, M A., Merino, R., Casini, S. and Fossi, M C. 2007. Vitellogenin and zona radiata proteins as biomarkers of endocrine disruption in peregrine falcon (Falco peregrinus). Chemosphere 67:S375-S378. Jin, L., Gaus, C. and Escher, B I. 2015. Bioanalytical Approaches to Understanding Toxicological Implications of Mixtures of Persistent Organic Pollutants in Marine Wildlife. Comp Anal Chem 67:57-84. Keller, J M. 2003. Occurence and effects of organochlorine contaminants in sea turtles. Duke University, North Carolina, USA. Keller, J M., McClellan-Green, P D., Kucklick, J R., Keil, D E. and Peden-Adams, M M. 2006. Effects of organochlorine contaminants on loggerhead sea turtle immunity: comparison of a correlative field study and in vitro exposure experiments. Environ Health Persp 114:70-76. Kelly, B C., Ikonomou, M G., Blair, J D., Morin, A E. and Gobas, F A. 2007. Food web–specific biomagnification of persistent organic pollutants. Science 317:236-239. Kendall, R J. 2010. Introduction and Overview.in R. J. Kendall, T. E. Lacher, G. P. Cobb, and S. B. Cox, editors. Wildlife Toxicology: Emerging Contaminant and Biodiversity Issues CRC Press, USA.

13 Kidd, K A., Blanchfield, P J., Mills, K H., Palace, V P., Evans, R E., Lazorchak, J M. and Flick, R W. 2007. Collapse of a fish population after exposure to a synthetic estrogen. Proc Natl Acad Sci USA 104:8897-8901. Kim, B-H. and Takemura, A. 2002. In vitro vitellogenin synthesis in primary cultures of tilapia hepatocytes. Fish Sci 68:123-131. Kim, B H. and Takemura, A. 2003. Culture conditions affect induction of vitellogenin synthesis by estradiol-17 beta in primary cultures of tilapia hepatocytes. Comp Biochem Physiol B Biochem Mol Biol 135:231-239. Leusch, F D., van den Heuvel, M R., Laurie, A D., Chapman, H F., Gooneratne, S R. and Tremblay, L A. 2005. Quantification of vitellogenin mRNA induction in mosquitofish (Gambusia affinis) by reverse transcription real-time polymerase chain reaction (RT-PCR). Biomarkers 10:429-438. Leusch, F D L., Chapman, H F., Kay, G W., Gooneratne, S R. and Tremblay, L A. 2006. Anal fin morphology and gonadal histopathology in mosquitofish (Gambusia holbrooki) exposed to treated municipal sewage effluent. Archives of Environ Cont Toxicol 50:562-574. Limpus, C., Miller, J., Paramenter, C., Reimer, D., McLachlan, N. and Webb, R. 1992. Migration of green (Chelonia mydas) and loggerhead (Caretta caretta) turtles to and from eastern Australian rookeries. Wildlife Res. 19:347-357. Manikkam, M., Guerrero-Bosagna, C., Tracey, R., Haque, M M. and Skinner, M K. 2012. Transgenerational actions of environmental compounds on reproductive disease and identification of epigenetic biomarkers of ancestral exposures. PLoS ONE 7:e31901. Matta, M B., Linse, J., Cairncross, L., Francendese, L. and Kocan, R M. 2001. Reproductive and transgenerational effects of methylmercury or Aroclor 1268 on Fundulus heteroclitus. Environ Toxicol Chem 20:327-335. Melvin, S D. and Wilson, S P. 2013. The utility of behavioral studies for aquatic toxicology testing: a meta-analysis. Chemosphere 93:2217-2223. Miller, A., Hedman, J E., Nyberg, E., Haglund, P., Cousins, I T., Wiberg, K. and Bignert, A. 2013. Temporal trends in dioxins (polychlorinated dibenzo-p-dioxin and dibenzofurans) and dioxin-like polychlorinated biphenyls in Baltic herring (Clupea harengus). Mar Pollut Bull 73:220-230.

14 Moe, S J., De Schamphelaere, K., Clements, W H., Sorensen, M T., Van den Brink, P J. and Liess, M. 2013. Combined and interactive effects of global climate change and toxicants on populations and communities. Environmental toxicology and chemistry, 32(1):49-61. Muir, D., Braune, B., DeMarch, B., Norstrom, R., Wagemann, R., Lockhart, L., Hargrave, B., Bright, D., Addison, R. and Payne, J. 1999. Spatial and temporal trends and effects of contaminants in the Canadian Arctic marine ecosystem: a review. Science of the total environment 230:83-144. National Research Council. 2007. Toxicity Testing in the 21st Century: A Vision and a Strategy. Washington, DC: The National Academies Press. https://doi.org/10.17226/11970. Noyes, P D., McElwee, M K., Miller, H D., Clark, B W., Van Tiem, L A., Walcott, K C., Erwin, K N. and Levin, E D. 2009. The toxicology of climate change: environmental contaminants in a warming world. Environ Int. 35:971-986. Oaks, J L., Gilbert, M., Virani, M Z., Watson, R T., Meteyer, C U., Rideout, B A., Shivaprasad, H L., Ahmed, S., Chaudhry, M J I., Arshad, M., Mahmood, S., Ali, A. and Khan, A A. 2004. Diclofenac residues as the cause of vulture population decline in Pakistan. Nature 427:630-633. Pandey, A. and Mann, M. 2000. Proteomics to study genes and genomes. Nature 405:837-846. Pavlou, M P. and Diamandis, E P. 2010. The cancer cell secretome: a good source for discovering biomarkers? J Proteomics 73:1896-1906. Peakall, D B. 1994. The Role of Biomarkers in Environmental Assessment .1. Introduction. Ecotoxicology 3:157-160. Perkins, E J., Chipman, J K., Edwards, S., Habib, T., Falciani, F., Taylor, R., Van Aggelen, G., Vulpe, C., Antczak, P. and Loguinov, A. 2011. Reverse engineering adverse outcome pathways. Environ Toxicol Chem 30:22-38. Porte, C., Janer, G., Lorusso, L C., Ortiz-Zarragoitia, M., Cajaraville, M P., Fossi, M C. and Canesi, L. 2006. Endocrine disruptors in marine organisms: Approaches and perspectives. Comp Biochem Physiol C 143:303-315. Reijnders, P J. 1986. Reproductive failure in common seals feeding on fish from polluted coastal waters. Nature 324:456-457. Relyea, R A. and Diecks, N. 2008. An unforseen chain of events: lethal effects of pesticides on frogs at sublethal concentrations. Ecol App 18:14.

15 Ross, P S., De Swart, R L., Van Loveren, H., Osterhaus, A D. and Vos, J G. 1996. The immunotoxicity of environmental contaminants to marine wildlife: a review. Ann Rev Fish Dis 6:151-165. Schwarzenbach, R P., Escher, B I., Fenner, K., Hofstetter, T B., Johnson, C A., von Gunten, U. and Wehrli, B. 2006. The challenge of micropollutants in aquatic systems. Science 313:1072-1077. Scott, A P. and Robinson, C D. 2008. Fish vitellogenin as a biological effect marker of oestrogenic endocrine disruption in the open sea. Ad Fish Sc 50:472À490. Scott, P D., Coleman, H M., Colville, A., Lim, R., Matthews, B., McDonald, J A., Miranda, A., Neale, P A., Nugegoda, D., Tremblay, L A and Leusch, F D L. 2017. Assessing the potential for trace organic contaminants commonly found in Australian rivers to induce vitellogenin in the native rainbowfish (Melanotaenia fluviatilis) and the introduced mosquitofish (Gambusia holbrooki ). Aquat Toxicol 185:105-120. Shepard, J L., Olsson, B., Tedengren, M. and Bradley, B P. 2000. Protein expression signatures identified in Mytilus edulis exposed to PCBs, copper and salinity stress. Mar Environ Res 50:337-340. Simms, W. and Ross, P S. 2000. Vitamin A physiology and its application as a biomarker of contaminant-related toxicity in marine mammals: a review. Toxicol Ind Health 16:291-302. Skalnikova, H., Motlik, J., Gadher, S J. and Kovarova, H. 2011. Mapping of the secretome of primary isolates of mammalian cells, stem cells and derived cell lines. Proteomics 11:691-708. Spurgeon, D J., Jones, O A., Dorne, J-L C., Svendsen, C., Swain, S. and Stürzenbaum, S R. 2010. Systems toxicology approaches for understanding the joint effects of environmental chemical mixtures. Sci Total Environ 408:3725-3734. Stastna, M. and Van Eyk, J E. 2012. Secreted proteins as a fundamental source for biomarker discovery. Proteomics 12:722-735. Sumpter, J P. and Jobling, S. 1995. Vitellogenesis as a biomarker for estrogenic contamination of the aquatic environment. Environ Health Persp 103:173. Tanabe, S. 2002. Contamination and toxic effects of persistent endocrine disrupters in marine mammals and birds. Mar Pollut Bull 45:69-77.

16 Taylor, C., Duffy, L K., Terry Bowyer, R. and Blundell, G M. 2000. Profiles of fecal porphyrins in river otters following the Exxon Valdez oil spill. Mar Poll Bull 40:1132-1138. Tornero, V., Borrell, A., Aguilar, A., Forcada, J. and Lockyer, C. 2006. Organochlorine contaminant and retinoid levels in blubber of common dolphins (Delphinus delphis) off northwestern Spain. Environ Poll 140:312-321. Troisi, G., Borjesson, L., Bexton, S. and Robinson, I. 2007. Biomarkers of polycyclic aromatic hydrocarbon (PAH)-associated hemolytic anemia in oiled wildlife. Environ Res 105:324-329. Troisi, G M. 2013. T-cell Responses in Oiled Guillemots and Swans in a Rehabilitation Setting. Arch Environ Con Tox 65:142-148. Vaidya, V S. and Bonventre, J V. 2010. Biomarkers: in medicine, drug discovery, and environmental health. John Wiley & Sons. Van Bressem, M-F., Raga, J A., Di Guardo, G., Jepson, P D., Duignan, P J., Siebert, U., Barrett, T., Santos, M., Moreno, I B. and Siciliano, S. 2009. Emerging infectious diseases in cetaceans worldwide and the possible role of environmental stressors. Dis Aquat Organ 86:143-157. van de Merwe, J P., Hodge, M., Olszowy, H A., Whittier, J M. and Lee, S Y. 2010a. Using blood samples to estimate persistent organic pollutants and metals in green sea turtles (Chelonia mydas). Mar Pollut Bull. 60:579-588. van de Merwe, J P., Hodge, M., Whittier, J M., Ibrahim, K. and Lee, S Y. 2010b. Persistent organic pollutants in the green sea turtle Chelonia mydas: nesting population variation, maternal transfer, and effects on development. Mar Ecol Prog Ser. 403:269-278. Van Loveren, H., Ross, P., Osterhaus, A. and Vos, J. 2000. Contaminant-induced immunosuppression and mass mortalities among harbor seals. Toxicol Lett. 112:319-324. Vos, J G., Dybing, E., Greim, H A., Ladefoged, O., Lambre, C., Tarazona, J V., Brandt, I. and Vethaak, A D. 2000. Health effects of endocrine-disrupting chemicals on wildlife, with special reference to the European situation. Crit Rev Toxicol 30:71-133.

17 Waugh, C A., Nichols, P D., Schlabach, M., Noad, M. and Bengtson Nash, S. 2014. Vertical distribution of lipids, fatty acids and organochlorine contaminants in the blubber of southern hemisphere humpback whales (Megaptera novaeangliae). Mar Environ Res 94:24-31. Wei, Y., Chan, L L., Wang, D., Zhang, H., Wang, J. and Dai, J. 2008. Proteomic analysis of hepatic protein profiles in rare minnow (Gobiocypris rarus) exposed to perfluoroctanoic acid. J Proteome Res 7:1729-1739. Wolska, L., Mechlińska, A., Rogowska, J. and Namieśnik, J. 2012. Sources and Fate of PAHs and PCBs in the Marine Environment. Crit Rev Environ Sci Tec 42:1172- 1189. Wyk, J H. v., Pool, E J. and Leslie, A J., 2003. The effects of anti-androgenic and estrogenic disrupting contaminants on breeding gland (nuptial pad) morphology, plasma testosterone levels, and plasma vitellogenin levels in male xenopus laevis (African Clawed Frog). Arch Environ Con Tox 44:247-256.

18 Chapter 2: Charting a path towards non-destructive biomarkers in threatened wildlife: A systematic quantitative literature review

Statement of Contribution for Chapter 2 This chapter includes a co-authored article that has been published in Environmental

Pollution. The bibliographic details of the co-authored paper are:

Stephanie Chaousis, Frederic DL Leusch & Jason P van de Merwe (2018) Charting a path towards non-destructive biomarkers in threatened wildlife: A systematic quantitative literature review. Environ Poll 234: 59-70.

My contribution to the paper include:

• the design and undertaking of the systematic quantitative literature review,

• data analysis, and

• writing of the manuscript.

My co-authors provided input into the design of the study, interpretation of the data and final polishing of the writing.

This chapter is formatted in the style of the journal of publication.

(Signed) ______(Date) 24/04/2019 Stephanie Chaousis

(Countersigned) ______(Date) 23/04/2019 Jason van de Merwe

19 2.1 Abstract

Threatened species are susceptible to irreversible population decline caused by adverse sub-lethal effects of chemical contaminant exposure. It is therefore vital to develop the necessary tools to predict and detect these effects as early as possible. Biomarkers of contaminant exposure and effect are widely applied to this end, and a significant amount of research has focused on development and validation of sensitive and diagnostic biomarkers. However, progress in the use biomarkers that can be measured using non- destructive techniques has been relatively slow and there are still many difficulties to overcome in the development of sound methods. This paper systematically quantifies and reviews studies that have aimed to develop or validate non-destructive biomarkers in wildlife, and provides an analysis of the successes of these methods based on the invasiveness of the methods, the potential for universal application, cost, and the potential for new biomarker discovery. These data are then used to infer what methods and approaches appear the most effective for successful development of non-destructive biomarkers of contaminant exposure in wildlife. This review highlights that research on non-destructive biomarkers in wildlife is severely lacking, and suggests further exploration of in vitro methods in future studies.

2.2 Keywords

Contaminant; biomarker; non-destructive; wildlife; in vitro

2.3 Introduction

Elucidating current or potential negative effects of chemical pollutants on threatened wildlife is a vital aspect of conservation efforts. Populations of threatened species are already pressured by major anthropogenic impacts such as climate change, habitat destruction and overfishing, and the additional effect of exposure to chemical contaminants has been cited as a contributing factor to the decrease in global biodiversity (Bickham et al., 2000).

20 Aside from the obvious adverse effect of lethal toxicity (Oaks et al., 2004; Relyea and

Diecks, 2008), contaminant exposure can have less immediately apparent negative effects in wildlife, such as immune suppression (Keller et al., 2006; Van Loveren et al.,

2000), reproductive incapacity (Porte et al., 2006), endocrine disruption (Scott et al.,

2014; Soffker and Tyler, 2012) and behavioural toxicity (Lanctot et al., 2016; Melvin and Wilson, 2013). Ultimately, these organism-level effects can lead to consequences at the population and ecosystem levels. For example, Kidd et al. (2007) demonstrated that reduction of reproductive output can lead to the complete collapse of a fish population after exposure to environmentally relevant concentrations of the synthetic hormone ethinylestradiol. The effects of chemical exposure on organisms can be varied, even between contaminants with the same mode of action. For example, endocrine disrupting chemicals can cause either feminisation of the males or masculinisation of the females, through either physical changes, such as abnormal gonad morphology (Kang et al.,

2006; Leusch et al., 2006), or complex hormonal changes that reduce successful reproductive output (Denslow and Sepúlveda, 2007). A decrease in reproductive output can also be indirectly caused by immunotoxicant exposure through various adverse effects. Studies have connected immunotoxins to the deterioration of lymphoid tissue, such as the thymus, causing a weakened innate or adaptive immune response (Desforges et al., 2016; Galloway and Handy, 2003). This reduced immune function can lead to an increased parasite load or pathogen susceptibility, and the resulting accumulated energetic costs of this burden can detract from healthy reproductive function in the organism (Booth et al., 1993; Hillegass et al., 2010; Marcogliese and Pietrock, 2011). In more severe cases, the immune system may be so compromised that life span is dramatically decreased, as was the case with several mass mortality events where tens of thousands of seals became ill and died within a short period of time due to the contraction of a morbillivirus (Mahy et al., 1988; Osterhaus and Vedder, 1988). While

21 the direct cause of death was the virus, it was later established through analysis of the carcasses and in vivo exposure experiments that exposure to mixtures of organochlorine contaminants such as polychlorinated biphenyls (PCBs), polychlorinated dibenzofurans

(PCDFs) and polychlorinated dibenzodioxins (PCDDs) were responsible for decreasing immune system functioning, rendering these individuals more susceptible to the virus

(de Swart, 1994; 1996; Ross et al., 1995; Van Loveren et al., 2000). It is therefore apparent that the effects of contaminant exposure can be subtle, yet cause serious adverse effects in wildlife populations.

Resulting population decline may be irreversible by the time these effects manifest, particularly for threatened species in which population numbers are already low or declining. In light of this, contaminant mitigation efforts need to focus on early detection of chemical exposure and negative effects. One approach to determine the effects of contaminant exposure in organisms is to perform controlled experiments on laboratory animals to correlate exposure and effect. This has been widely done with laboratory animals such as fish (Bandelj et al., 2006; Brockmeier et al., 2016; Melvin,

2016; van der Oost et al., 2003), and while this method is valuable, there are logistical and ethical considerations to this approach when dealing with threatened wildlife. An alternative may be to use closely related (but not endangered) model organisms and assume that effects to a related species would be similar (e.g. de la Casa-Resino et al.,

2013). However, inter-species extrapolation is problematic due to the potential for significant differences in both toxicokinetic and toxicodynamic factors (Bokkers and

Slob, 2007; Nyman et al., 2014).

In addition, laboratory experiments often focus on single compounds or a small number of compounds in mixtures (e.g. Gilman et al., 2003; Solomon et al., 2008; Wang et al.,

2016). However, chemicals are generally present as complex mixtures in the environment and subsequently in wildlife (Jin et al., 2015), which means that adverse

22 effects may manifest at different concentrations and in a different manner in wild populations than in animals exposed in a laboratory setting (de Swart et al., 1996; Kelly et al., 2007). Examining species-specific in situ effects of contaminants in wildlife therefore represents a more meaningful reflection of the effect of real-world exposure, compared to effects observed in laboratory-based controlled environment (Spurgeon et al., 2010). However, moving towards examining in situ exposure presents a new set of challenges, largely due to the fact that many variables cannot be controlled for, and therefore the ability to assign causation of exposure to adverse effects is limited.

Biomarkers of exposure and/or effect are tools often used for early detection of health deterioration, although the term is very general and has been assigned a number of definitions depending on the context. In general, the term biomarker refers to the response of a biological system to a potential threat ranging from an ecosystem, to a population or individual level. Biomarkers of chemical exposure and effect have largely been studied in individuals, predominantly at the molecular level. Separating the terms

‘exposure’ and ‘effect’ can be challenging as is thoroughly discussed by van der Oost et al. (2003). In toxicology, a biomarker of exposure is often considered to be the detection of a contaminant itself or its metabolites, whereas a marker of effect is the alteration of a biochemical or physiological parameter associated with contaminant exposure (Dos

Santos et al. 2016; Espín et al. 2015). However, if this altered response is not proven to be detrimental to the individual, it may not be an indicator of effect and hence will simply be a marker of exposure (e.g. da Silva Corrêa et al. 2016). Some studies do not define the detection of contaminant in an organism as a biomarker but rather more accurately as a ‘bioaccumulation marker’ (van der Oost et al. 2003). However, the reference to biomarkers in this paper will be to both endogenous molecules and chemicals detected in the tissue or blood of organisms with endogenous molecular responses that indicate a proven adverse effect referred to as “biomarkers of effect”.

23 Biomarkers play an important role in early detection of disease in humans (Etzioni et al., 2003), and the methods for clinical biomarker discovery are well established

(Spurgeon et al., 2010; Vaidya and Bonventre, 2010). The application of biomarkers in ecotoxicology provides an alternative to laboratory exposure experiments, and a potentially more effective way to detect negative effects of contaminants on the health of threatened wild populations (Kendall, 2016). Biomarkers of effect have been well established for some contaminants and validated in a number of species, especially aquatic organisms such as fish. These markers cover a number of molecules including biotransformation enzymes (e.g. cytochrome P450), stress proteins (e.g. HSP70), as well as haematological, immunological and reproductive factors (e.g. vitellogenin, a well-established marker of endocrine disruption in fish) (Sumpter and Jobling 1995;

Leusch et al. 2005). However, the validation of these biomarkers for use in wildlife has been limited, especially for those species with high conservation importance (e.g. sea turtles; Finlayson et al., 2016).

One significant hindrance in wildlife biomarker research is that traditional methods for the establishment of biomarkers require manipulation and often sacrifice of live animals

(e.g. Jasinska et al., 2015), which is particularly not desirable for threatened species.

However, studies attempting to establish and validate biomarkers of contaminant exposure using non-destructive samples from wildlife are emerging, and technological developments are inviting new and innovative ways to approach this problem

(Martyniuk and Simmons, 2016). Despite this, there has not been a comprehensive review of the methods used to validate non-destructive measurement of biomarkers in wild vertebrate fauna since Fossi and Leonzio’s (1993) review of this topic over 20 years ago. While there are countless studies that investigate the accumulation and adverse effects of contaminants in wildlife, studies that aim to validate non-destructive measurement of these markers for the detection of adverse effects are not as numerous,

24 or successful. In light of the continuing decrease in global biodiversity, it is crucial that attempts to develop non-destructive tools that could be used for monitoring and early detection of negative impacts, and therefore help prevent further population decline, be as effective as possible. This paper aims to address this issue by systematically reviewing studies that have explicitly attempted to establish and/or validate biomarkers of contaminant exposure and effect in wildlife using non-destructive techniques. Within this, gaps in the knowledge are identified, and guidance on future research in this field are presented.

2.4 Methods

2.4.1 Search tools and parameters

A systematic literature search (Pickering and Byrne 2014) was performed to find studies that have attempted to develop non-destructive biomarkers of contaminant exposure and effect in vertebrate wildlife. Google Scholar, Web of Science and Science Direct databases were used to search for studies that met a set of predetermined criteria. Each study must have included all of the following terms: ‘biomarker’ and ‘contaminant’ in combination with either of the following terms; ‘non-destructive’ or ‘non-invasive’. All papers published prior to October 2016 that met all of these criteria, and that were relevant to vertebrate wildlife, were entered into a database under the categories defined below (Section 2.4.2.1). Only peer-reviewed research papers written in English were considered, and review papers were excluded, although carefully checked to ensure all papers within were included.

During the assessment of relevant papers, studies that were actually destructive in their design but had the end goal of establishing a method to measure a biomarker non- destructively were included, as were studies that measured contaminant concentrations in wildlife, but only if done non-destructively (e.g. skin, fur sample). Studies that detected contaminant concentration in internal tissues that needed to be sampled

25 destructively, or that assessed the effect of contaminants in wildlife but did not develop or discuss biomarkers, were not included.

2.4.2 Categorisation and exclusions

2.4.2.1 Broad experimental categories

Initially, studies were placed in one of four broad categories. The categories were developed based on the criteria that the preferred method for discovery of novel biomarkers in threatened wildlife is one that a) doesn’t involve sacrifice of any animals

(completely non-destructive) and b) doesn’t involve experimental exposure of animals to contaminants. To cover all possible permutations of these two factors, studies were sorted into four categories: 1) non-destructive methods and non-experimental exposure,

2) non-destructive methods and experimental exposure, 3) destructive methods and experimental exposure and, 4) destructive methods and non-experimental exposure.

Distinguishing between experimental and non-experimental exposure was as simple as noting whether or not the test animals were intentionally exposed to contaminants in the search for a biomarker. Distinguishing between non-destructive and destructive studies was more complex as some papers did not explicitly state if animals were euthanized at the end of the experiment or released after sampling. However, for the purposes of this paper, if the method for the biomarker assessment required only non-destructive samples (e.g. skin, fur, blood), the study was assigned as non-destructive, even if the animals had to be sacrificed at the end of the experiment to meet laboratory and/or ethical requirements.

Studies using animals that had died from other causes (e.g. roadkill or recreational hunting), or abandoned/hatched eggs, for the purpose of correlating markers in destructive samples with non-destructive samples, were all considered non-destructive, but were distinguished from the other non-destructive studies in the more in-depth analysis as described below. Similarly, studies involving in vitro (cell or tissue based)

26 methods were all considered to be non-destructive, and were also separated from the other studies as they present a unique approach to biomarker discovery that does not require destructive sampling of the target species at all. Some studies used a combination of techniques simultaneously (e.g. in vitro and destructive sampling), and therefore were assigned to more than one of the categories.

2.4.2.2 Detailed experimental categories

Following assignment to one of the four aforementioned categories, a more in-depth categorisation was performed that considered additional factors: 1) the use of wild, laboratory-reared or already deceased animals, 2) whether a biomarker of exposure was confirmed, 3) whether a biomarker of effect was confirmed, and 4) whether in vitro methods were used. For the purposes of this paper, we defined wild vs. laboratory animals based on the origin of the individuals. Animals were considered “wild” even if they were in rehabilitation centres, semi-wild sanctuaries, or were brought into the laboratory from the wild.

Defining whether a biomarker of chemical exposure was successfully discovered, or if a biomarker of effect was confirmed, proved to be somewhat challenging. While some studies were explicit in confirming or rejecting the proposed biomarker/s, many studies described the “potential” value of biomarker. Often, the results of the study indicated that a biomarker would be a good candidate, but stated that further investigations should take place to confirm the accuracy of the marker. These ‘potential’ biomarker studies were separated from the more definitive studies during the in-depth analyses of the literature so as not to over or under-represent failure or success in these categories.

In order to gain deeper insight into the methods used, the type/s of sample taken for the study was assigned to one of seven categories: blood, excreta, external tissue (e.g. fur, skin, feathers), internal tissue (e.g. liver, brain, gonads), biopsy (e.g. subcutaneous fat,

27 non-destructive liver biopsy), eggs, and other (e.g. morphological traits, behavioural observations).

It is important to note that the type of biomarker validated for each study within these method categories is of interest because some markers will be more easily obtained from non-destructive samples (such as those in excreta), while some will be more contaminant specific than others. However, to analyse each specific marker in relationship to the method category was out of the scope of this review, where the aim was to focus on the non-destructive methods used by the studies.

2.4.2.3 Classification of methods and quality of experimental design

Finally, the methods employed in the studies were evaluated and classified based on the advantages and limitations in the context of biomarker discovery in threatened wildlife.

All papers were categorised into the following five methods used for discovering non- destructive biomarkers of chemical exposure and effect: 1) correlation of biomarkers between destructive and non-destructive samples, 2) correlation between multiple biomarkers in non-destructive samples only, 3) correlation between exposure (according to habitat or treatment) and detection of biomarkers in non-destructive samples (i.e. contaminant levels were not measured in the organism), 4) validation of the detection of biomarkers and/or of non-destructive samples, with no correlations made, and 5) tissue or cell-based in vitro methods.

The advantages and disadvantages of these methodological categories were then critically assessed based on several factors that impact the suitability of methods for non-destructive biomarker discovery in threatened wildlife: 1) involves non-destructive sampling, 2) can be done without experimental exposure, 3) accuracy of the method in indicating environmentally realistic effects, 4) feasibility for universal application, 5) relative cost and time effectiveness, 6) can be used to assess contaminant exposure, 7)

28 can be used to assess contaminant effect, and 8) can be used for novel biomarker discovery.

Additionally, the species and class (mammal, bird, reptile, fish or amphibian) of the animals in each study was recorded in the database. A broad habitat type was also assigned to each species (marine, freshwater or terrestrial) in order to gain an overview on the representation of each species, class and habitat type within the method types.

This analysis was included as any biases towards one of these species or habitat types will be important for the analysis of method types.

To further evaluate the papers and to provide a critical analysis of the methodologies, the quality of the experimental design of each paper was assessed. In order to do this, we noted the following features for each study: total sample size, sample size of each site/treatment and species, number of sites or treatments, number of controls/reference sites, replicates per individual (e.g. several samples from the same individual over time), timing of sample collection (e.g. over several seasons or close together) and length of exposure (for experimental studies). To gain an understanding on how representative the study was of the population of the target species, data on the number of females, males and juveniles sampled in the study was also collected.

2.5 Results

2.5.1 Search results and experimental categories

A total of 82 papers identified by the search criteria were considered relevant to this review (Table 2.1). The highest proportion of papers (61%, 50) was assigned to the category of non-destructive and non-experimental, the preferred scenario for biomarker discovery in threatened wildlife (Table 2.1). The next highest category included studies that used both destructive and experimental methods (20%, 16) followed by non- destructive studies that used experimental exposure (15%, 12) and lastly, the smallest group of studies were destructive but did not use experimental exposure (12%, 10).

29 Further categorisation of these studies indicated that the source of animals was strongly related to the study type. For example, almost 100% (55) of non-experimental studies used wild animals, whereas >80% (15) of the experimental studies used laboratory animals. The use of already deceased animals was highest for studies that were non- destructive and non-experimental (25%, 15) (Table 2.2). The overall rate of confirmed biomarkers of exposure was relatively low (<50%, 42) in all of the four broad categories

(Table 2.2). However, when ‘potential’ biomarkers of exposure (those likely to be biomarkers following some validation) are included, success was much higher (>90%).

In contrast, success in confirming biomarkers of effect was much lower with an overall success rate of 7% (7) (non-destructive and experimental) (Table 2.2). Inclusion of the

‘potential’ biomarkers of effect dramatically increased the success within the broad categories, but only half as high as for biomarkers of exposure at 41% (37) (Table 2.2).

The type of sample used in most studies was blood, with 50% analysing whole blood, serum or plasma. This was followed by the sampling of internal organs (43% of studies), excreta or external samples (e.g. fur or tail clips, 26% each; external biopsy,

16%). A small number of papers collected eggs or used other ‘samples’ such as morphological or behaviour observations (4 and 6%, respectively).

2.5.2 Classification of methods and quality of experimental design

Papers using methods 1 (correlation of biomarkers in destructive and non-destructive samples), 2 (correlation of contaminants and biomarkers in non-destructive samples) and 3 (correlation of contaminated site/treatment with biomarkers in non-destructive samples) were relatively equally represented with 35, 37 and 29% of studies in these categories, respectively (Table 2.3). The remaining two methods were much less common, with 7% of papers using method 4 (validation of non-destructive sampling/biomarker detection) and 7% using method 5 (tissue or cell based in vitro studies) (Table 2.3). Method 1 was relatively equally employed in all of the broad

30 categories except by non-destructive and experimental studies, of which none used method 1 (Figure 2.1). Non-destructive and non-experimental studies predominantly employed method 2 (29%) followed by destructive and non-experimental methods (6%) and destructive and experimental methods (2%), whereas only 2% of the non- destructive and experimental studies used method 2 (Figure 2.1). Only two of the method types were represented in destructive and non-experimental studies method 1

(9%) and 2 (6%), whereas all of the method types were observed in both the non- destructive categories (Figure 2.1).

When assessed against the eight desirable traits for biomarker discovery in threatened wildlife (see section 2.4.2.3), all method types possessed at least one of these positive traits. Method 1 failed to meet many of the traits, including arguably the most important criteria, the ability to perform a method using non-destructive sampling. In contrast, method 5 included almost all of the desirable traits with the other three method types possessing a similar amount of desirable traits (Table 2.3).

Across all of the studies, there were 113 species, representing five Chordate classes as follows: 38% mammals (31), 28% birds (23), 16% reptiles (13), 14% fish (12) and 5% amphibians (4). Some studies had several species within one paper including up to 13

(Saengtienchai et al. 2014), resulting in the overall number of species being much higher than the number of studies (82). These species represented the three broad habitat categories with most species from terrestrial habitats (44%), followed by marine (33%) and aquatic (26%).

As a result of the high number and variation in species across all studies, the interaction of species with method type was not analysed further as there were not enough of a single species within each method to produce a meaningful analysis. However, it is acknowledged that the effectiveness and relevance of methods for biomarker development and validation will influence the quality and relevance of a method type,

31 so this has been carefully considered when evaluating the methods. While a review focusing on a single species or habitat type may be able to explore ideal methods for a particular species in depth, the general spread of biomarker papers overs a range of species and habitats was more conducive to a systematic review on all vertebrate species, as presented here.

The results of the quality design analysis (sample sizes, number of sites or treatments, number of controls/reference sites, replicates per individual, timing of sample collection and length of exposure) proved to be very complex and presenting these data along with the method type would result in a dilution of the overall results and prohibit a broad overview of current methods used in the literature on non-destructive biomarker discovery. For example, Sogorb et al. (2007) had an overall sample size of 102, spanning 11 different species with some species only represented by one individual and others by up to 15. It is clear that for 82 papers, the resulting data would be too diverse to allow categorisation of the papers into discreet categories based on method quality.

Therefore, for the purposes of this review these data have been ignored. However, it is acknowledged that quality of the experimental design will be an influencing factor on the results of the method analysis and this has been considered when discussing that aspect.

2.6 Discussion

2.6.1 Experimental categories

The majority of studies were assigned to the category of non-destructive and non- experimental (“ND & NE”) (Table 2.1). However, it was noted that 25% (15) of these studies relied on the use of individuals that had died of natural or other causes (i.e. not killed for the purposes of the study) (see Table S.21 in Supplementary Information). It was observed that 15% of studies using experimental methods managed to use non- destructive methods (“ND & E”), which means the methods employed by these studies

32 could also potentially be used for preferred threatened species studies, that is, non- destructive sampling without experimental exposure. However, there are still a significant number of studies (31%) that have used destructive methods for non- destructive biomarker validation, limiting their applicability to threatened wildlife.

Certainly, there is space to further develop approaches to biomarker discovery that can reduce the manipulation and destructive sampling previously utilised.

Table 2.1 Proportion of papers in each of the four categories of studies assigned to the different combinations of destructive methods and/or experimental exposure of animals to contaminants for the development of non-destructive biomarkers. Note that several papers were included multiple times as they contained both method types, therefore the total percent exceeds 100% and number of papers is greater than 82. Numbers are percentage (with the number of papers listed in parenthesis). The total number of papers categorised was 82.

Destructive?

Yes No Yes 20% (16) 15% (12) exposure?

Experimental Experimental No 12% (10) 61% (50)

While initial categorisation showed that non-destructive biomarker discovery methods were well represented in the literature, further categorisation revealed that close to a quarter (25%,15) of these studies relied on the use of already deceased animals (i.e. not killed for the purposes of the study). While this is an excellent alternative to the sacrifice of live animals, there are limitations to the dependency on an unreliable source of study material, such as waiting for the opportunity to sample stranded or dead animals. This approach is also limited in the type of samples that can be taken such as

33 histological, biochemical and cellular samples that require access to the animals within hours of death. Additionally, some studies, sampled animals that had been preserved in museums or other collections, for example those that looked at inorganic contaminants in fur (e.g. Hernout et al., 2016). This retrospective use of specimens is highly resourceful and can provide interesting information on the target species, but the data may not represent the current contaminant loads of that population. However, the use of preserved historical specimens in combination with current samples may be useful to assess changes in contaminant load over time.

The more detailed categorisation of the studies also revealed that the proportion of papers confirming biomarkers of exposure was much higher than studies confirming biomarkers of effect, although both were relatively low (<50%). This is to be expected as exposure will not always result in adverse health outcomes. It is also easier to detect exposure within an organism than it is to assign adverse health effects to exposure, particularly when measuring samples from wild animals in which health parameters can be affected by many variables. However, most importantly, this highlights the need for the application of already validated biomarkers of effect in new species, and/or further development of new methods that lead to more definitive outcomes for non-destructive biomarkers and represent negative effects of contaminant exposure. Additionally, future studies could take advantage of mathematical approaches such as computational toxicology and toxicokinetic models that can assist in establishing relationships between exposure and effects (Thomas et al. 2013). Beyond this, controlled exposure experiments may provide a robust approach for validating already established biomarkers of effect in a new species, or alternatively for identifying new biomarkers, both of which are crucial in a continually expanding and evolving chemical world.

34 Table 2.2 Detailed features (source of animals, success in confirming biomarkers of exposure and effect, and use of in vitro methods) of the studies from each category represented in Table 2.1 defined as “ND & NE” (non-destructive and non- experimental), “ND & E” (non-destructive and experimental), “D & NE” (destructive and non-experimental) and “D & E” (destructive and experimental). See Section 2.4.2.1 for more details on categorisation. The proportion of studies that possessed the features noted in the table from within each of the four categories is represented by a percent with the actual number of studies noted in parentheses. Note that some percentages for the detailed categories can add up to >100% because some papers were assigned to more than one category. The use of cell cultures is automatically non-destructive; therefore the two destructive broad categories (D&NE and D&E) were deemed ‘non- applicable’ (N/A) under the category of “use of cultured tissue or cells”.

Broad Detailed categories categories

Animals used Biomarker of Biomarker of

exposure effect

Use of

cultured Lab

Wild tissue or Potential Potential Deceased Confirmed Confirmed cells ND & NE 98% 8% 28% 40% 46% 34% 8% 6% (n=48) (49) (4) (14) (20) (23) (17) (4) (3) ND & E 50% 67% 8% 33% 75% 17% 25% 23% (n=13) (6) (8) (1) (4) (9) (2) (3) (3) D & NE 100% 0% 0% 70% 20% 50% 0% N/A (n=11) (10) (7) (2) (5) D & E (n=17) 63% 44% 0% 56% 50% 38% 0% N/A (10) (7) (9) (8) (6) TOTAL (n=82) 7% 84% 20% 18% 48% 44% 34% 7%

35

It was found that half of all studies used blood (50%), alone or in combination with other sample types. This is potentially due to the fact that blood provides arguably the most molecular information about an organism than any other tissue type (Fossi and

Leonzio 1993). Interestingly, eggs were only used by 4% of studies, likely because eggs can be destructive at a population level if apparently healthy, unhatched eggs are used.

Internal organs can also provide a lot of information about the toxicological status of an individual, especially for the examination of internal contaminant load and organ specific loads, which is perhaps why almost half (43%) utilised internal organs.

Additionally, it appears that the more difficult or the less complex/informative samples, such as fur or skin clips, excreta and morphological traits, were used less frequently. It is important to note that therefore one sample type will not always be superior to another as the usefulness of a sample type will also depend on the organism being examined and the biomarker being targeted.

2.6.2 Methodological categories

2.6.2.1 Method 1: Correlation of biomarkers in destructive and non-destructive samples

In order to monitor biomarkers of exposure and effect non-destructively in a wild population, non-destructive samples (e.g. fur, scales, toe clippings, blood) must be taken. However, for this approach to be useful, the biomarkers observed in non- destructive samples must accurately reflect contaminant load or molecular changes in the internal organs (that would usually be taken destructively). If a consistent correlation occurs then the non-destructive sample can be used for monitoring internal contaminant loads non-destructively in wildlife in the future. Therefore, one way to validate a biomarker in a non-destructive sample is to determine the correlation of this biomarker in tissues that require destructive sampling. For example, some studies that used this method looked at the concentrations of metals in fur/skin/feathers/spines in

36 relation to contaminant load in organs, such as the liver, or in fat (D'Have et al., 2005;

Dauwe et al., 2005; Fossi et al., 1996; Hernández-Moreno et al., 2013; Hernout et al.,

2016; Keller et al., 2004; Kocagoz et al., 2014; Valová et al., 2013), and many found correlations that would validate the use of non-destructive samples for future studies.

Other studies examined correlations of endogenous enzymatic markers of exposure in the same manner to determine if marker alterations in the internal organs are represented in peripheral non-destructive samples. For example, a number of papers studied B-esterase inhibition in the brain and liver in correlation with the presence of these enzymes in the blood of exposed individuals (Chahin et al., 2013; Fossi et al.,

1992; Lari et al., 1994; Sanchez et al., 1997). Additionally, correlations of biomarkers between destructive and non-destructive samples could be validated using "traditional" biological models (e.g. zebrafish, rats, Japanese quail), especially for conservative biomarkers such as GSTs, porphyrins, phase II metabolism, cholinesterase, and histology (e.g. Fossi et al., 1996; Saengtienchai et al. 2014).

The advantage of using simple correlation between biomarkers in destructive and non- destructive samples is that it is relatively straightforward and can produce strong correlative data. However, correlation does not necessarily indicate causation and, while results obtained from these methods may be a good starting point, controlled experiments may still need to be conducted to confirm that exposure is the cause of the observed effect. Hence this correlative method may not often stand alone as a tool for non-destructive biomarker discovery. Studies that employ this method for biomarker discovery can use either controlled, experimental exposure, or naturally exposed individuals and results can give an accurate, real world assessment of exposure.

Nevertheless, this method has several disadvantages when considering its use for threatened species. Most importantly, it relies on sacrificing animals, which is often unacceptable, especially in the case of threatened species. Moreover, relative to other

37 methods, this method has low cost and time effectiveness, due to the need to collect, kill and necropsy each individual. Finally, it was observed that this method generally was not used for the assessment of biomarkers of effect, especially not of new effects in the context of non-destructive biomarker discovery. Most of the studies simply correlated the presence of either contaminants or known markers between various tissue types and/or blood. This does not mean that this method could not be used to analyse markers of effect, but simply that the literature is limited in demonstrating this.

It is important to note that some studies demonstrated variations to this method that could qualify this approach as non-destructive, although with significant drawbacks.

One alternative used by 12% of studies within this method is to correlate biomarkers in destructive and non-destructive samples from naturally deceased individuals such as beached marine life or roadkill (e.g. Fossi et al., 1997a; Marsili et al., 1997), instead of sacrificing animals for the study. However, using already dead wildlife may not provide population-relevant data as the biomarker concentrations in these individuals may not be representative of healthy individuals due to a number of biases. For example, starvation or illness can cause mobilisation of fat stores that result in abnormally higher bioavailability of lipophilic contaminants compared to healthy individuals (Aguilar and

Borrell, 1994; Bengtson Nash et al., 2013), therefore not representative of the contaminant load of the population. Another interesting alternative was demonstrated in a recent study that used minor non-lethal surgery (individuals recovered and were re- released) to obtain organ biopsies (Quesada et al., 2014), which could then be used to compare biomarker levels in less invasive samples such as skin. While non-destructive sampling is always preferred over destructive sampling, in this case the limitations are significant. This method is labour-intensive, requiring specialised skills and veterinary qualifications, is highly invasive, even if it is not destructive, and is not widely applicable. Individuals need to be captured, sedated and then monitored before release

38 with the potential risk of reducing the individual’s health in the long term (Mulcahy and

Esler, 2010), and such risks cannot be taken with many species, particularly in the case of threatened populations.

Overall, correlating biomarkers between destructive and non-destructive samples is advantageous as it can provide robust correlations between exposure and effect in organs and non-destructive samples, providing accurate environmentally relevant information. This information can then result in a reduction in the need for destructive sampling as confidence in the accuracy of data obtained from non-destructive samples becomes high. However, it is relatively labour intensive, potentially costly and most importantly it requires deceased individuals or highly invasive sampling, which is generally not appropriate for threatened species.

2.6.2.2 Method 2: Correlation of multiple biomarkers in non-destructive samples only

Correlation of known biomarkers in non-destructive samples, such as contaminant load and molecular health markers, is a useful and proven method for determining the effects of contaminants in threatened or protected animals (e.g. Geens et al., 2010; Troisi et al.,

2007). This method was predominantly used by studies that were non-destructive and non-experimental (29%) and only 8% of studies in the destructive categories (Figure

2.1). Some of these studies confirmed biomarkers of chemical contaminant exposure or effect using this method in wild animals (e.g. Celis et al., 2014; Geens et al., 2010; Jara-

Carrasco et al., 2016; Marsili et al., 1998), indicating that it may be a useful tool for assessing negative effects of contaminant exposure in threatened species.

39

Figure 2.1. Percent of all studies within each broad category defined as “ND & NE”

(non-destructive and non-experimental), “ND & E” (non-destructive and experimental),

“D & NE” (destructive and non-experimental) and “D & E” (destructive and experimental) represented by bars (ND&NE, 61%; ND&E, 15%; D&NE, 12%; D&E,

20%). The proportion of papers within each broad category is divided into the five method types, described in section 2.4.2.2, represented as a percent of all papers reviewed (n=82) that used the indicated method type, and fell into the specified broad category, by patterned segments. Method 1: correlation of biomarkers in destructive and non-destructive samples; Method 2: correlation of multiple biomarkers in non- destructive samples only; Method 3: correlation between exposure (according to habitat or treatment) and detection of biomarkers in non-destructive samples (i.e. contaminant levels were not measured in the organism); Method 4: validation of the detection of biomarkers and/or of non-destructive samples, with no correlations made; and Method

5: tissue or cell-based in vitro methods. Note that the sum of all percentages is higher than 100 because several papers fall under multiple categories.

40 The presence of porphyrins in faeces was a popular technique within this method as these compounds are well established as biomarkers of exposure to persistent organic pollutants (POPs) in birds (e.g. Casini et al. 2001a; Celis et al. 2014; Jara-Carrasco et al.

2016; Rudolph et al. 2016; Fossi et al. 1996). It has also been tested as a non-destructive marker of contaminant exposure in river otters (Taylor et al., 2000) and rabbits

(Hernandez-Moreno et al., 2012). Other correlations included exposure to perfluroalkyl compounds and haematological and immune parameters, such as B-cell proliferation, in dolphin blood (Fair et al., 2013), and cadmium and lead concentrations in blood of tortoises correlated with δ-aminolevulinic acid dehydratase (δ-ALAD) activity

(Martinez-Lopez et al., 2010). Several other studies looked at contaminant load such as polycyclic aromatic hydrocarbons (PAHs), PCBs or dichlorodiphenyltrichloroethane

(DDT) and its congers in blubber in correlation with benzo(a)pyrene monooxygenase

(BPMO) activity in skin using biopsies from pinnipeds (Fossi et al., 1997b) or cetaceans

(Fossi et al., 2001).

Retinoids (vitamin A) are another well-established biomarker that was used in four studies considered in this review. These studies used blood samples (non-destructive) across several different species and identified correlations between abnormal circulating retinoid levels with the presence of contaminants (Teglia et al., 2015; Thibodeau et al.,

2012; Tornero et al., 2006; Tornero et al., 2005). Overall, these studies mostly claimed success in identifying retinoids as a marker of exposure. However, detection of abnormal retinoid levels can also be considered a marker of effect because the adverse effects associated with retinoid imbalance in mammals, birds and fish are well established (Blomhoff and Wake, 1991; Rolland, 2000; Romert et al., 1998).

These examples highlight that the use of well-established biomarkers can assist in the employment of completely non-destructive or experimental methods, which is highly beneficial for threatened species. Correlating markers in non-destructive samples is

41 advantageous over method 1 as it does not rely on the sacrificing of individuals and the collection of samples tends to be minimally invasive (e.g. blood or excreta). Also, the cost and time effectiveness is higher than the previously described method as the collection of non-destructive samples for analysis is less time and resource intensive than whole animal collection and necropsy, yet relative to method 5 (in vitro methods) sample collection is moderately costly (Table 2.3).

The two main drawbacks of this method are, firstly, that it does not largely support the discovery of novel biomarkers due to a reliance on already known markers that can be detected in non-destructive samples. Novel markers could be explored with this method by analysing the samples with non-targeted omics. However, this would be very limited as these methods are generally expensive and the required equipment and expertise are not always available to toxicology labs. Furthermore, if such a method was employed in an attempt to discover new biomarkers, a large sample size would be required as there would be a lot of ‘background interference’ within the data as wildlife samples are highly variable and non-targeted analyses produce a large volume of data (Sung et al.

2012).

Secondly, the correlation of multiple biomarkers in non-destructive samples only is not universally applicable due to the reliance on already established biomarkers, which limits the usefulness of this method in lesser-studied species. Additionally, studies using this method may miss other potentially more important effects of contaminant exposure that are not specifically targeted. In general, commonly used biomarkers are developed in model species adapted for use in the laboratory for application to wild animals (e.g.

Fossi et al., 1992). Such inter-species extrapolation is an invaluable tool for guidance when aiming to elucidate the potential effects of chemical contaminants on wildlife.

However, there are several issues with assuming biomarkers of effect will be the same across species: 1) many new chemicals are registered every year for use in agriculture or

42 industry and some of these will inevitably end up in the environment - with an ever changing cocktail of new chemicals, the effects from these chemicals and particularly their effects in combination with existing chemicals are very hard to predict (Jia et al.,

2015); and 2) contaminant exposure may have specific and varying effects in different taxa (Lauer et al., 2009; Olson et al., 2000). This is even true of individuals within a species as evidenced by a number of studies that have noted a significant difference in the relevance and accuracy of biomarkers and chemical sensitivity between species of the same class (Calabrese and Baldwin, 1994; Sogorb et al., 2007), and even between males and females of the same species (Burger et al., 2007; Fossi et al., 2007;

Hernandez-Moreno et al., 2012). Overall, widespread reliance on established biomarkers from model species may exclude the opportunity for more reliable, species- specific and informative markers of effect to be discovered (e.g. Katsiadaki et al., 2002).

Several studies explored an interesting and even less invasive alternative to tissue sampling and chemical analysis by comparing contaminant exposure with behaviour such as predator-prey interaction (Weis et al., 2001) or morphological traits (Orton et al., 2014) as a form of non-destructive biomarker of effect. Weis et al. (2001) described high variance of behaviour between individuals, making it challenging to correlate specific contaminants to behavioural patterns. However, total contaminant load was associated with adverse effects on predatory behaviour, suggesting this marker could be useful for detecting overall exposure to contaminant mixtures. Similarly, the link between morphological traits and contaminant exposure was weak or difficult to assign causality to, and therefore, Orton et al. (2014) suggested using more diagnostic molecular markers in combination with altered morphology. Moreover, this method is unlikely to be widely applicable as this kind of monitoring would not be feasible for many species, particularly larger and/or threatened vertabrates in the wild.

43 Overall, correlation of contaminant load and known biomarkers in non-destructive samples is useful when species-specific biomarkers are well established, as it provides environmentally relevant information and does not require deceased individuals.

However, it is limiting because novel biomarkers that may be more accurate and appropriate for different species are challenging to develop using this method. This method can be more reliable with the use of multiple biomarkers increasing the weight of evidence for the effects of contaminants and insight into the pathways specifically affected by contaminants exposure. The only disadvantage of using multiple markers is that more time and resources are needed for each sample analysis, however the payoff of more definitive results likely outweigh the extra costs. It is also important to note that care must be taken when predicting population level effects from biomarkers detected in individuals. This is especially important in light of the fact that almost a quarter of non- destructive (24%, 15) studies used opportunistic sampling that often use ill or deceased individuals (Table 2.2). These individuals may not represent the average contaminant load or response of the population and therefore predictive modelling based on the results may inaccurately predict the actual potential for population level impacts.

2.6.2.3 Method 3: Correlation between site/treatment and detection of biomarkers in non-destructive samples

A number of studies (29%) used a method similar to method 2 described above, however with a significant difference: these studies did not quantify contaminants in tissue or blood samples of the individuals in which biomarkers were measured. In method 3, correlation is only made between sites where chemical exposure is suspected, such as old reclaimed mining sites or heavily polluted areas (Aguilera et al., 2012; da

Silva Corrêa et al., 2016; Gauthier et al., 1999; Panti et al., 2011; e.g. Smits et al.,

2000), or where animals are experimentally exposed, but the actual uptake of contaminants is not measured (Cordi et al., 1997; Dzul-Caamal et al., 2016; Irwin,

44 2001; Jones et al., 2005; Seriani et al., 2015). This means that in these studies contaminant level in the organisms is estimated, or assumed, based on contaminant levels at the site, or from the treatment.

Due to the similarities between the two methods, method 3 has comparable advantages and disadvantages to method 2. While method 2 would always provide more accurate correlations, method 3 is beneficial when contaminant exposure in the organism cannot be measured due to limited resources or other restrictions, as long as historical pollution data exists, or exposure is controlled. Correlations with historical environmental data are, however, potentially weak as the presence of contaminants in an environment may not represent the actual exposure to the wildlife in this habitat. For example, some studies measured sediment or water in which the animals were taken from to gain a measure of exposure levels (Smits et al., 2000; Thibodeau et al., 2012). However, environmental contaminant levels do not necessarily reflect actual exposure and concentrations obtained from soil or water may under or over represent actual exposure, which can be greatly influenced by feeding habits, toxicokinetics or the physicochemical properties of the contaminants (Barber, 2008; Daley et al., 2014).

Overall, this method has a moderate time and cost efficiency for the accuracy of data obtained, and may produce only weak markers of effect due to less precise exposure data. Method 3 is mainly useful when contaminant levels in the organisms cannot be measured due to resource or other limitations.

2.6.2.4 Method 4: Validation of methods for detection of potential non-destructive biomarkers with no contaminant correlations made

A small number of studies (7%, 5) simply attempted to detect the presence of potential biomarkers in a non-destructive sample in order to validate the use of a particular non- destructive sample type. These studies make no correlation of the biomarker with contaminant exposure, site of capture or tissue concentrations of contaminants. For

45 example, one study characterised plasma cholinesterase activity in storks to provide baseline data for future studies (Oropesa et al., 2013). Other studies looked at the ability to collect a sample non-destructively, such as obtaining liver biopsies non-lethally from snakes (Quesada et al., 2014), or confirming the detection of PCBs in preen gland oil of seabirds (Yamashita et al., 2007). Several other studies focussed on validating the applicability of already known biochemical markers in new species (e.g. Jiménez et al.,

2007; Numata et al., 2008; Sogorb et al., 2007).

These studies are important and can provide valuable baseline data that contribute towards future research on biomarkers in the target species, especially when new methods of non-destructive sampling need to be established. The most significant drawback to this method is that limited information is obtained for relatively time consuming and costly experiments, therefore rendering this method less than ideal for timely discovery of biomarkers of contaminant exposure and effect in threatened wildlife.

2.6.2.5 Method 5: In vitro methods using non-destructive samples

In total, only 7% (6) of studies used in vitro methods, defined as the use of tissue explant or cultured cells. Close examination of this method led to separation of these studies into two distinct groups: 1) studies in which exposure occurred in vivo (naturally or experimentally) before tissue samples were taken and molecular responses were measured from cultured cells or tissue, and 2) studies where primary cells/tissue were cultured from an animal and then exposed to contaminants in vitro and molecular responses measured. Only one of the six studies used tissue that was exposed in vivo measuring the response associated with this exposure, and the other five studies exposed cultured cells and subsequently measured the response to this exposure. For example, one study exposed peripheral blood mononuclear cells (PBMC) from eels to contaminants before measuring changes in the protein expression (Roland et al., 2013).

46 Another study took a skin biopsy from a dolphin and exposed the tissue to contaminant mixtures in vitro before measuring alteration of a range of molecular markers in exposed tissue slices (Fossi et al., 2013). Similarly, another study exposed skin biopsy slices, obtained from a deceased dolphin, to contaminants before measuring alterations in gene expression (Lunardi et al., 2016). A unique study exposed modified rat cells to contaminant extract mixtures from wild seal livers to measure toxic potency of these environmentally relevant extracts (Nyman et al., 2003).

Of all the studies, in vitro methods appear to have the most advantages by meeting almost all of the criteria for ideal biomarker discovery in threatened species (Table 2.3).

The use of cultured cells is non-destructive, minimally invasive, and relatively high throughput (Tse et al., 2013), and therefore time and cost effective, and can be used to develop markers of exposure and, to an extent, markers of effect. Moreover, the effects of controlled exposure can be examined without the need for whole animals. Multiple assays measuring numerous endpoints can be performed on cells cultured from the target species using a single non-destructive sample making this method one of the most efficient ways to quickly obtain large amounts of data on the effect of contaminants on the target species (Finlayson et al., 2016), and enhancing the potential for the discovery of novel biomarkers (Schrattenholz et al., 2012).

Many of the studies in this review were categorised as having the ‘potential’ for discovery of biomarkers of exposure (48%) and effect (34%), with only 7% confirmed markers of effect (Table 2.2). Often this lack of confidence was due to the ambiguity of the cause of the effect that was observed in the study, because even with distinct correlation between contaminant exposure and adverse health markers, causation cannot be irrefutably implied unless the molecular cascade of events that led to the cellular effect is understood (i.e. adverse outcome pathway mapping). Controlled exposure of cells from a target species may be an ideal way to address this issue and increase the

47 success in the development of non-destructive biomarkers of chemical contaminant exposure in protected wildlife. Moreover, the use of cells is the only way to perform controlled exposure in the target species without causing potential harm to the whole organism. The use of controlled exposure in combination with non-targeted analysis such as examination of the protein expression profile in exposed cells can facilitate the discovery of novel biomarkers of effect in a non-destructive manner (Lasserre et al.,

2009; Schrattenholz et al., 2012). Two of the six studies using in vitro methods explored potential new contaminant-specific biomarkers of exposure using non-targeted omics.

Lunardi et al. (2016) identified a number of differentially regulated genes in relation to contaminant exposure, indicating specific immune and endocrine pathways were affected. Non-targeted proteomic analysis of eel white blood cells exposed to perfluorinated compounds by Roland et al. (2013) revealed altered regulation of proteins that have yet to be identified as biomarkers of exposure in any other studies.

The combination of controlled exposure, tissue from the target species and non-targeted analysis shows great promise for the discovery of new effects of contaminant exposure.

However, there are several limitations to this kind of analysis. In particular, the cost of facilities for establishing and maintaining cell culture are high and labs that are set up for this kind of research are generally focused on human health studies. Additionally, determining if the alterations in genomic or proteomic expression in cells translates to adverse health effects in a whole organism is difficult. It is apparent that this method requires further research and development to be validated as a useful tool for predicting adverse outcomes for wildlife populations.

The main disadvantage of in vitro studies is that they potentially do not represent real world effects, and linking in vitro responses to whole organism effects is complicated by toxicokinetic modifiers (Kramer et al., 2011). This issue could be addressed by the use of mathematical modelling tools along with adverse outcome pathway analysis to

48 extrapolate results from cell based studies to population level effects. However, despite the existence of such models in ecotoxicology, the use of these tools was not present in any of the six in vitro studies described here. Another limitation of in vitro studies is that the use of immortalised cell lines in place of primary cells may misrepresent this link even more (Wilkening et al., 2003). Moreover, not all effects can be seen at the cellular level as some adverse effects manifest at the organ level involving more complex interactions between cells and organs than can be observed with a single cell line. Therefore, results from in vitro studies may need to be verified in whole animals, which may be defensible following 3R principles (Zurlo et al., 1996) as the in vivo experiments would ultimately result in the reduction, refinement and replacement of animals for monitoring adverse effects of chemical contaminant exposure. However, as discussed above, destructive exposure or sampling is not an option in the case of threatened species and cannot be justified as population impacts may be too severe. A potential alternative to destructive verification could be to measure the biological responses in cells obtained from a target species exposed to a wide range of model contaminants in vitro. Following this, non-destructive samples, such as blood, could be analysed from the same target species to determine if these biomarkers identified in vitro are also produced in the whole organism. Furthermore toxicokinetic modelling and adverse outcome pathway analyses can better link these in vitro effects to whole organism effects (Andersen, 2003; Ankley et al., 2010). The combination of these non- destructive tools could potentially result in very robust biomarkers and increase efficacy and confidence in results obtained from non-destructive biomarker studies.

Overall, in vitro methods show promise for improving outcomes in non-destructive biomarker discovery, however this potential is relatively unexplored in the field of ecotoxicology of threatened species as was highlighted by the small number of papers using this method (6). The need to validate results from such studies may be filled by

49 using a multibiomarker approach that includes non-destructive samples collected from wild exposed individuals in addition to results from in vitro experiments.

2.6.3 Future directions for non-destructive biomarker discovery in threatened wildlife

Overall, the results of this systematic review highlight the limited progress that has been made in the field of non-destructive biomarker research. The paucity of studies coupled with limitations in the studies that do exist, has contributed to the hindrance of progress in this area. The classification of each study by methodology was undertaken here to provide a basis for recommendations for future research. It must be emphasised that more research in this field is required for a more substantial understanding of non- destructive biomarkers that indicate exposure and effect of contaminants in wildlife.

However, based on our assessment criteria, in vitro bioassays provide the most promising approach to non-destructive biomarker research in wildlife, although still with some limitations and best used in conjunction with other methodologies.

The five method categories that were defined and discussed above represent the techniques used in the majority of studies aimed at non-destructive biomarker discovery in threatened wildlife. All of the method types have unique advantages and disadvantages as discussed above, however, according to our criteria, in vitro methods appear to be the most ideal method for biomarker discovery in threatened wildlife as they are non-destructive, cost and time effective, universally applicable and potentially suitable for discovery of new biomarkers of exposure and effect. However, a significant shortfall of in vitro methods of biomarker discovery is that whole animal effects may not be fully represented and caution must be used in drawing conclusions from such studies. In order to address this issue, investigators could combine the use of in vitro methods for rapid assessment of the potential effects of contaminants on the target species and prioritisation of promising biochemical biomarkers before utilising non- destructive methods of whole animal sampling to confirm that these effects occur in

50 vivo. Using such methods still has limitations; however, given the understandable restrictions around working with threatened species and the urgency to find methods for the detection of contaminant effects early on, in vitro combined with non-destructive sampling has great potential to accelerate non-destructive biomarker discovery in threatened wildlife. Furthermore, it must be noted that new biomarkers will not necessarily be superior over existing markers for monitoring exposure and effect, and in some cases, well already established conserved markers (e.g. GST or ChE activity) may remain more suitable than any new markers.

It is very important to note that, as stated in section 2.4.2.2, the quality of the experimental design was not considered when assessing the outcomes of the method categories. Many studies, particularly in method categories 1 & 2 (4.2.1 and 4.2.2) were limited by small sample size or lack of reference sites or individuals (data not shown).

This is unfortunately inevitable when working with species such as large marine vertebrates, or species in which ethics is strict and destructive sampling must be kept to a minimum. Even in some laboratory based studies, individuals died before the end of the experiment as a result of exposure and therefore the sample size was reduced. The limitations in the non-destructive studies reviewed here indicates that there is a need for improvement of methods of non-destructive biomarker discovery that allows for more consistent experimental design. This further supports the notion that implementing the use of in vitro studies in combination with whole organism studies for such species would potentially improve the statistical quality of results. Cell lines can be grown and manipulated in the lab, therefore allowing for the execution of well-designed experiments that can measure a variety of endpoints in the context of toxicology and can be species specific. However, it must be noted that since contaminants go through complex toxicokinetic processes (absorption, distribution, metabolism, excretion) within functioning organisms, which are generally not reproduced in vitro, the use of

51 cell cultures may be best employed as a high throughput, robust experimental platform from which to inform whole animal studies.

In summary, there are a limited number of studies on non-destructive biomarker discovery in threatened vertebrate wildlife and, due to a low number of repeated studies on one species, there is not enough data to conclude that one method is suited to all taxa. Therefore, further research is required to determine the best method/s of non- destructive biomarker analysis in threatened species. In vitro methods were the least used methods for examining toxic effects of contaminants on threatened species, yet may allow for higher throughput and better experimental design to guide research efforts on whole organisms. A significant limiting factor for research moving towards the application of cell cultures is the severe lack of established cell lines for threatened wildlife along with the expenses of cell culture and assays for labs that are not equipped for such experiments Moving forward, we encourage more research within this field, alongside a greater use of in vitro methods including the establishment of cell lines of threatened species for use in future research.

52 Table 2.3. Matrix of eight desirable traits for each of the five methods applied for the discovery of non-destructive biomarkers in threatened wildlife identified in the studies identified in this review.

Method % of Non- Non- Real Universal Cost and Assess Assess Novel studie destructi experime world ly time exposure effect? biomarke s ve? ntal relevance applicabl effectiven ? r exposure ? e? ess discovery ? ? 1) Correlation of biomarkers 35% No Yes Yes No Low Yes Low No in destructive and non- destructive samples 2) Correlation of multiple 37% Yes Yes Yes No Medium Yes Mostly No biomarkers in non-destructive samples only 3) Correlation between 29% Yes Yes Yes Mostly Low No Yes Yes site/treatment and detection of biomarkers in non- destructive samples 4) Validation of detection of 7% Yes Yes Yes Mostly Medium No No No biomarkers and or use of non- destructive samples with no correlations made 5) In vitro detection of 7% Yes Yes Potentiall Yes High Yes Mostly Yes biomarkers resulting from in y vivo or in vitro exposure

53 2.7 References

Aguilar, A. and Borrell, A. 1994 Abnormally high polychlorinated biphenyl levels in striped dolphins (Stenella coeruleoalba) affected by the 1990-1992 Mediterranean epizootic. Sci Total Environ 154:237-247. Aguilera, C., del Pliego, P G., Alfaro, R M., Lazcano, D. and Cruz, J. 2012 Pollution biomarkers in the spiny lizard (Sceloporus spp.) from two suburban populations of Monterrey, Mexico. Ecotoxicology 21:2103-2112. Andersen, M E. 2003 Toxicokinetic modeling and its applications in chemical risk assessment. Toxicol Lett 138:9-27. Ankley, G T., Bennett, R S., Erickson, R J., Hoff, D J., Hornung, M W., Johnson, R D., Mount, D R., Nichols, J W., Russom, C L., Schmieder, P K., Serrrano, J A., Tietge, J E. and Villeneuve, D L. 2010 Adverse outcome pathways: a conceptual framework to support ecotoxicology research and risk assessment. Environ Toxicol Chem 29:730-41. Bandelj, E., van den Heuvel, M R., Leusch, F D., Shannon, N., Taylor, S. and McCarthy, L H. 2006 Determination of the androgenic potency of whole effluents using mosquitofish and trout bioassays. Aquat Toxicol 80:237-48. Barber, M C. 2008 Dietary uptake models used for modeling the bioaccumulation of organic contaminants in fish. Environ Toxicol Chem 27(4):755-777. Bengtson Nash, S M., Waugh, C A. and Schlabach, M. 2013 Metabolic concentration of lipid soluble organochlorine burdens in the blubber of Southern Hemisphere humpback whales through migration and fasting. Environ Sci Technol 47:9404- 9413. Bickham, J W., Sandhu, S., Hebert, P. D., Chikhi, L. and Athwal, R. 2000 Effects of chemical contaminants on genetic diversity in natural populations: implications for biomonitoring and ecotoxicology. Mutat Res 463:33-51. Blomhoff, R. and Wake, K. 1991 Perisinusoidal stellate cells of the liver: important roles in retinol metabolism and fibrosis. FASEB J 5:271-277. Bokkers, B G. and Slob, W. 2007 Deriving a data-based interspecies assessment factor using the NOAEL and the benchmark dose approach. Cr Rev in Toxicol 37:355- 73.

54 Booth, D T., Clayton, D H. and Block, B A. 1993 Experimental demonstration of the energetic cost of parasitism in free-ranging hosts. P Roy Soc B-Biol Sci 253:125-129. Brockmeier, E K., Scott, P D., Denslow, N D. and Leusch, F D. 2016 Transcriptomic and physiological changes in Eastern Mosquitofish (Gambusia holbrooki) after exposure to progestins and anti-progestagens. Aquat Toxicol 179:8-17. Burger, J., Fossi, C., McClellan-Green, P. and Orlando, E F. 2007 Methodologies, bioindicators, and biomarkers for assessing gender-related differences in wildlife exposed to environmental chemicals. Environ Res 104:135-152. Calabrese, E J. and Baldwin, L A. 1994 A toxicological basis to derive a generic interspecies uncertainty factor. Environ Health Persp 102:14-7. Celis, J E., Espejo, W., Gonzalez-Acuna, D., Jara, S. and Barra, R. 2014 Assessment of trace metals & porphyrins in excreta of Humboldt penguins (Spheniscus humboldti) in different locations of the northern coast of Chile. Environmental Monitoring and Assessment 186:1815-24. Chahin, A., Peiffer, J., Olry, J. C., Crepeaux, G., Schroeder, H., Rychen, G. and Guiavarc'h, Y. 2013 EROD activity induction in peripheral blood lymphocytes, liver and brain tissues of rats orally exposed to polycyclic aromatic hydrocarbons. Food Chem Toxicol 56:371-80. Cordi, B., Fossi, C. and Depledge, M. 1997 Temporal biomarker responses in wild passerine birds exposed to pesticide spray drift. Environ Toxicol Chem 16:2118. da Silva Corrêa, S A., de Souza Abessa, D M., dos Santos, L G., da Silva, E B. and Seriani, R. 2016 Differential blood counting in fish as a non-destructive biomarker of water contamination exposure. Toxicol Environ Chem 99(3):1-10. Daley, J M., Paterson, G. and Drouillard, K G. 2014 Bioamplification as a bioaccumulation mechanism for persistent organic pollutants (POPs) in wildlife. Reviews of Environmental Contamination and Toxicology (D. M. Whitacre, ed.). Springer International Publishing 227:107-155. Dauwe, T., Jaspers, V., Covaci, A., Schepens, P. and Eens, M. 2005 Feathers as a nondestructive biomonitor for persistent organic pollutants. Environ Toxicol Chem 24:442-449.

55 de la Casa-Resino, I., Hernández-Moreno, D., Navas, J M., Soler, F. and Pérez-López, M. 2013 Non-destructive multibiomarker approach in European quail (Coturnix coturnix coturnix) exposed to the herbicide atrazine. Arch Environ Con Tox 65:567-574. de Swart, R L. 1994 Impairment of immune function in harbor seals (Phoca vitulina) feeding on fish from polluted waters. Ambio 23(2):155-159. de Swart, R L., Ross, P S., Vos, J G., & Osterhaus, A D M E. 1996 Impaired immunity in harbour seals (Phoca vitulina) exposed to bioaccumulated environmental contaminants: Review of a long-term feeding study. Environmental Health Perspectives 104(4):823-828. Denslow, N D. and Sepúlveda, M S. 2007 Ecotoxicological effects of endocrine disrupting compounds on fish reproduction. In The Fish Oocyte: From Basic Studies to Biotechnological Applications (Babin, P J., Cerda, J., Lubzens, E., eds), Springer, 255-322. Desforges, J P., Sonne, C., Levin, M., Siebert, U., De Guise, S. and Dietz, R. 2016 Immunotoxic effects of environmental pollutants in marine mammals. Environ Int 86:126-139. D'Have, H., Scheirs, J., Mubiana, V K., Verhagen, R., Blust, R. and De Coen, W. 2005 Nondestructive pollution exposure assessment in the european hedgehog (Erinaceus europaeus): I. Relationships between concentrations of metals and arsenic in hair, spines, and soil. Environ Toxicol Chem 24(9):2356-2364. Dos Santos, C R., Cavalcante, A L., Hauser-Davis, R A., Lopes, R M. and Mattos Rde, C. 2016 Effects of sub-lethal and chronic lead concentrations on blood and liver ALA-D activity and hematological parameters in Nile tilapia. Ecotoxicology and Environmental Safety 129:250-256. Dzul-Caamal, R., Salazar-Coria, L., Olivares-Rubio, H F., Rocha-Gomez, M A., Giron- Perez, M I. and Vega-Lopez, A. 2016 Oxidative stress response in the skin mucus layer of Goodea gracilis (Hubbs and Turner, 1939) exposed to crude oil: A non-invasive approach. Comp Biochem Physiol A 200:9-20. Espín, S., Martínez-López, E., Jiménez, P., María-Mojica, P. and García-Fernández, A. J. 2015 Delta-aminolevulinic acid dehydratase (δALAD) activity in four free- living bird species exposed to different levels of lead under natural conditions. Environ Res 137:185-198.

56 Etzioni, R., Urban, N., Ramsey, S., McIntosh, M., Schwartz, S., Reid, B., Radich, J., Anderson, G. and Hartwell, L. 2003 The case for early detection. Nat Rev Cancer 3:243-52. Fair, P A., Romano, T., Schaefer, A M., Reif, J S., Bossart, G D., Houde, M., Muir, D., Adams, J., Rice, C. and Hulsey, T C. 2013 Associations between perfluoroalkyl compounds and immune and clinical chemistry parameters in highly exposed bottlenose dolphins (Tursiops truncatus). Environ Toxicol Chem 32 (4):736- 746. Finlayson, K A., Leusch, F D., and van de Merwe, J P. 2016 The current state and future directions of marine turtle toxicology research. Environ Int 94:113-23. Fossi, C. and Leonzio, C. 1993 Nondestructive biomarkers in vertebrates: CRC Press, Boca Raton. Fossi, C., Marsili, L., Junin, M., Castello, H., Lorenzani, J., Casini, S., Savelli, C. and Leonzio, C. 1997a Use of nondestrucitive biomarkers and residue analysis to assess the health status of endangered species of pinnipeds in the South-west Atlantic. Mar Pollut Bull 34(3):157-162. Fossi, M C., Casini, S. and Marsili, L. 1996 Porphyrins in excreta: A non-destructive biomarker for the hazard assessment of birds contaminated with PCBs. Chemosphere 33(1):29-42. Fossi, M C., Casini, S. and Marsili, L. 2007 Potential toxicological hazard due to endocrine-disrupting chemicals on Mediterranean top predators: State of art, gender differences and methodological tools. Environ Res 104:174-182. Fossi, M C., Casini, S., Maltese, S., Panti, C., Spinsanti, G. and Marsili, L. 2013 An “ex vivo” model to evaluate toxicological responses to mixtures of contaminants in cetaceans: Integumentum biopsy slices. Environ Toxicol 29:1107-1121. Fossi, M C., Casini, S., Marsili, L., Ausili, A. and Notarbartolo Di Sciara, G. 2001 Are the Mediterranean top predators exposed to toxicological risk due to endocrine disrupters? Ann NY Acad Sci 948:67-74. Fossi, M C., Leonzio, C., Massi, A., Lari, L. and Casini, S. 1992 Serum esterase inhibition in birds - a nondestructive biomarker to assess organophosphorus & carbamate contamination. Arch Environ Con Tox 23:99-104.

57 Fossi, M C., Savelli, C., Marsili, L., Casini, S., Jimenez, B., Junin, M., Castello, H. and Lorenzani, J A. 1997b. Skin biopsy as a nondestructive tool for the toxicological assessment of endangered populations of pinnipeds: Preliminary results on mixed function oxidase in Otaria flavescens. Chemosphere 35(8):1623-1635. Galloway, T. and Handy, R. 2003 Immunotoxicity of organophosphorous pesticides. Ecotoxicology 12:345-63. Gauthier, J M., Dubeau, H., Rassart, E., Jarman, W M. and Wells, R. S. 1999 Biomarkers of DNA damage in marine mammals. Mutat Res 444:427-39. Geens, A., Dauwe, T., Bervoets, L., Blust, R. and Eens, M. 2010 Haematological status of wintering great tits (Parus major) along a metal pollution gradient. Science of the Total Environment 408:1174-1179. Gilman, C I., Leusch, F D., Breckenridge, W. C. and MacLatchy, D L. 2003 Effects of a phytosterol mixture on male fish plasma lipoprotein fractions and testis P450scc activity. Gen Compa Endocr 130:172-84. Hernández-Moreno, D., de la Casa Resino, I., Fidalgo, L E., Llaneza, L., Soler Rodríguez, F., Pérez-López, M. and López-Beceiro, A. 2013 Noninvasive heavy metal pollution assessment by means of Iberian wolf (Canis lupus signatus) hair from Galicia (NW Spain): a comparison with invasive samples. Environ Monit Assess 185:10421-10430. Hernandez-Moreno, D., Perez-Lopez, M., Miguez, M P., Soler, F. and Jimenez, B. 2012. Porphyrin levels in excreta of rabbit as non-destructive biomarkers of diazinon exposure. Environ Toxicol Pharmacol 34:466-72. Hernout, B V., McClean, C J., Arnold, K E., Walls, M., Baxter, M. and Boxall, A B. 2016 Fur: A non-invasive approach to monitor metal exposure in bats. Chemosphere 147:376-381. Hillegass, M A., Waterman, J M. and Roth, J D. 2010 Parasite removal increases reproductive success in a social African ground squirrel. Behav Ecol 21:696- 700. Irwin, L K. 2001 Vitellogenin induction in painted turtle, Chrysemys picta, as biomarker of exposure to environmental levels of estradiol. Aquat Toxicol 55:49-60. Jara-Carrasco, S., Barra, R., Espejo, W., Celis, J E., González-Acuña, D., Chiang, G. and Sánchez-Hernández, J. 2016 Persistent organic pollutants and porphyrin levels in excreta of penguin colonies from the Antarctic Peninsula area. Polar Record, 53(1):1-9.

58 Jasinska, E J., Goss, G G., Gillis, P L., Van Der Kraak, G J., Matsumoto, J., de Souza Machado, A A., Giacomin, M., Moon, T. W., Massarsky, A., Gagne, F., Servos, M. R., Wilson, J., Sultana, T. and Metcalfe, C D. 2015 Assessment of biomarkers for contaminants of emerging concern on aquatic organisms downstream of a municipal wastewater discharge. Science of the Total Environment 530-531:140-53. Jia, A., Escher, B I., Leusch, F D L., Tang, J Y M., Prochazka, E., Dong, B., Snyder, E M. and Snyder, S A. 2015 In vitro bioassays to evaluate complex chemical mixtures in recycled water. Water Research 80, 1-11. Jiménez, B., Mori, G., Concejero, M A., Merino, R., Casini, S. and Fossi, M C. 2007 Vitellogenin and zona radiata proteins as biomarkers of endocrine disruption in peregrine falcon (Falco peregrinus). Chemosphere 67:S375-S378. Jin, L., Gaus, C. and Escher, B I. 2015 Bioanalytical approaches to understanding toxicological implications of mixtures of persistent organic pollutants in marine wildlife. In Comprehensive Analytical Chemistry, 67:57-84. Jones, D E., Gogal, R M., Nader, P B. and Holladay, S D. 2005 Organochlorine detection in the shed skins of snakes. Ecotox Environ Safe 60:282-287. Kang, I J., Hano, T., Oshima, Y., Yokota, H., Tsuruda, Y., Shimasaki, Y. and Honjo, T. 2006 Anti-androgen flutamide affects gonadal development and reproduction in medaka (Oryzias latipes). Marine Environmental Research 62 Suppl:S253- S257. Katsiadaki, I., Scott, A P., Hurst, M R., Matthiessen, P. and Mayer, I. 2002 Detection of environmental androgens: a novel method based on enzyme-linked immunosorbent assay of spiggin, the stickleback (Gasterosteus aculeatus) glue protein. Environ Toxicol Chem 21(9):1946-54. Keller, J M., Kucklick, J R., Harms, C A. and McClellan-Green, P D. 2004 Organochlorine contaminants in sea turtles: Correlations between whole blood and fat. Environmental Toxicology and Chemistry 23 (3), 726-738. Keller, J M., McClellan-Green, P D., Kucklick, J R., Keil, D E. and Peden-Adams, M. M. 2006 Effects of organochlorine contaminants on loggerhead sea turtle immunity: comparison of a correlative field study and in vitro exposure experiments. Environ Health Persp 114(1):70-76.

59 Kelly, B C., Ikonomou, M G., Blair, J D., Morin, A E. and Gobas, F A P C. 2007 Food web–specific biomagnification of persistent organic pollutants. Science 317:236- 239. Kendall, R J. 2016 Wildlife toxicology: Where we have been and where we are going. J Environ Analytical Toxicology 6(1):348. doi 10.4172/2161-0525.1000348. Kidd, K A., Blanchfield, P. J., Mills, K. H., Palace, V. P., Evans, R. E., Lazorchak, J. M. and Flick, R. W. 2007 Collapse of a fish population after exposure to a synthetic estrogen. P Natl A Sci 104(21): 8897-901. Kocagoz, R., Onmus, O., Onat, I., Cagdas, B., Siki, M. and Orhan, H. 2014 Environmental and biological monitoring of persistent organic pollutants in waterbirds by non-invasive versus invasive sampling. Toxicol Lett 230:208-17. Kramer, V J., Etterson, M A., Hecker, M., Murphy, C A., Roesijadi, G., Spade, D J., Spromberg, J A., Wang, M. and Ankley, G T. 2011 Adverse outcome pathways and ecological risk assessment: bridging to population-level effects. Environ Toxicol Chem 30(1):64-76. Lanctot, C., Melvin, S D., Leusch, F D., Wilson, S. and Fabbro, L. 2016 Locomotor and behavioural responses of empire gudgeons (Hypseleotris compressa) exposed to coal mine wastewater. Chemosphere 144:1560-1566. Lari, L., Massi, A., Fossi, M C., Casini, S., Leonzio, C. and Focardi, S. 1994 Evaluation of toxic effects of the organophosphorus insecticide azinphos-methyl in experimentally and naturally exposed birds. Arch Environ Con Tox 26:234-239. Lasserre, J P., Fack, F., Revets, D., Planchon, S., Renaut, J., Hoffmann, L., Gutleb, A C., Muller, C P. and Bohn, T. 2009 Effects of the endocrine disruptors atrazine and PCB 153 on the protein expression of MCF-7 human cells. J Proteome Res 8:5485-96. Lauer, B., Tuschl, G., Kling, M. and Mueller, S O. 2009 Species-specific toxicity of diclofenac and troglitazone in primary human and rat hepatocytes. Chem-Biol Interact 179:17-24. Leusch, F D., Chapman, H F., Kay, G W., Gooneratne, S R. and Tremblay, L A. 2006 Anal fin morphology and gonadal histopathology in mosquitofish (Gambusia holbrooki) exposed to treated municipal sewage effluent. Arch Environ Con Tox 50:562-574.

60 Leusch, F D L., van den Heuvel, M R., Laurie, A D., Chapman, H F., Gooneratne, S R. and Tremblay, L A. 2005 Quantification of vitellogenin mRNA induction in mosquitofish (Gambusia affinis) by reverse transcription real-time polymerase chain reaction (RT-PCR). Biomarkers 10(6):429-438. Lunardi, D., Abelli, L., Panti, C., Marsili, L., Fossi, M C. and Mancia, A. 2016 Transcriptomic analysis of bottlenose dolphin (Tursiops truncatus) skin biopsies to assess the effects of emerging contaminants. Mar Environ Res 114:74-9. Mahy, B W J., Barrett, T., Evans, S., Anderson, E C. and Bostock, C J. 1988 Characterization of a seal morbillivirus. Nature 336(10):115. Marcogliese, D J. and Pietrock, M. 2011 Combined effects of parasites and contaminants on animal health: parasites do matter. Trends Parasitol 27(3):123- 130. Marsili, L., Fossi, M C., Casini, S., Savelli, C., Jimenez, B., Junin, M. and Castello, H. 1997 Fingerprint of polycyclic aromatic hydrocarbons in two populations of southern sea lions (Otaria flavescens). Chemosphere 34(4):759-770. Marsili, L., Fossi, M C., Notarbartolo Di Sciara, G., Zanardelli, M., Nani, B., Panigada, S. and Focardi, S. 1998 Relationship between organochlorine contaminants and mixed function oxidase activity in skin biopsy specimens of Mediterranean fin whales (Balaenoptera physalus). Chemosphere 37(8):1501-1510. Martinez-Lopez, E., Sousa, A R., Maria-Mojica, P., Gomez-Ramirez, P., Guilhermino, L. and Garcia-Fernandez, A. J. 2010. Blood [delta]-ALAD, lead and cadmium concentrations in spur-thighed tortoises from Southeastern Spain and Northern Africa. Ecotoxicology 19:670-677. Martyniuk, C J. and Simmons, D B. 2016 Spotlight on environmental omics and toxicology: a long way in a short time. Comp Biochem Physiol D 19:97-101. Melvin, S D. 2016 Short-term exposure to municipal wastewater influences energy, growth, and swimming performance in juvenile Empire Gudgeons (Hypseleotris compressa). Aquat Toxicol 170:271-278. Melvin, S D. and Wilson, S P. 2013 The utility of behavioural studies for aquatic toxicology testing: a meta-analysis. Chemosphere 93:2217-2223. Mulcahy, D M. and Esler, D. 2010 Survival of captive and free-ranging Harlequin Ducks (Histrionicus histrionicus) following surgical liver biopsy. J Wildlife Dis 46(4):1325-1329.

61 Numata, M., Fawcett, J P. and Rosengren, R J. 2008 Induction of in vitro EROD activity and in vivo caffeine metabolism in two species of New Zealand birds. Environ Toxicol Phar 25:358-364. Nyman, A M., Schirmer, K. and Ashauer, R. 2014 Importance of toxicokinetics for interspecies variation in sensitivity to chemicals. Envir Sci Tech 48:5946-5954. Nyman, M., Bergknut, M., Fant, M L., Raunio, H., Jestoi, M., Bengs, C., Murk, A J., Koistinen, J., B., Pelkonen, O., Tysklind, M., Hirvi, T. and Helle, E. 2003 Contaminant exposure and effects in Baltic ringed & grey seals as assessed by biomarkers. Mar Environ Res 55:73-99. Oaks, J L., Gilbert, M., Virani, M Z., Watson, R T., Meteyer, C U., Rideout, B A., Shivaprasad, H L., Ahmed, S., Chaudhry, M J I., Arshad, M., Mahmood, S., Ali, A. and Khan, A A. 2004 Diclofenac residues as the cause of vulture population decline in Pakistan. Nature 427:630-633. Olson, H., Betton, G., Robinson, D., Thomas, K., Monro, A., Kolaja, G., Lilly, P., Sanders, J., Sipes, G., Bracken, W., Dorato, M., Van Deun, K., Smith, P., Berger, B. and Heller, A. 2000 Concordance of the toxicity of pharmaceuticals in humans and in animals. Regul Toxicol Pharm 32:56-67. Oropesa, A L., Gravato, C., Sánchez, S. and Soler, F. 2013 Characterization of plasma cholinesterase from the White stork (Ciconia ciconia) and its in vitro inhibition by anticholinesterase pesticides. Ecotox Environ Safe 97:131-138. Orton, F., Baynes, A., Clare, F., Duffus, A. L., Larroze, S., Scholze, M. and Garner, T. W. 2014 Body size, nuptial pad size and hormone levels: potential non- destructive biomarkers of reproductive health in wild toads (Bufo bufo). Ecotoxicology 23:1359-1365. Osterhaus, A. D. and Vedder, E. J. 1988 Identification of virus causing recent seal deaths. Nature 335:20. Panti, C., Spinsanti, G., Marsili, L., Casini, S., Frati, F. and Fossi, M C. 2011. Ecotoxicological diagnosis of striped dolphin (Stenella coeruleoalba) from the Mediterranean basin by skin biopsy and gene expression approach. Ecotoxicology 20:1791-1800. Pickering, C. and Byrne, J. 2014 The benefits of publishing systematic quantitative literature reviews for PhD candidates and other early-career researchers. High Educ Res Dev 33(3):534-548.

62 Porte, C., Janer, G., Lorusso, L C., Ortiz-Zarragoitia, M., Cajaraville, M P., Fossi, M C. and Canesi, L. 2006 Endocrine disruptors in marine organisms: Approaches and perspectives. Comp Biochem Physiol C 143:303-315. Quesada, R J., McCleary, R J R., Heard, D J. and Lillywhite, H B. 2014 Non-lethal sampling of liver tissue for toxicologic evaluation of Florida cottonmouths snakes, Agkistrodon piscivorus conanti. Ecotoxicology 23:33-37. Relyea, R A. and Diecks, N. 2008 An unforseen chain of events: lethal effects of pesticides on frogs at sublethal concentrations. Ecol Appl 18(7):1728-1742. Roland, K., Kestemont, P., Hénuset, L., Pierrard, M. A., Raes, M., Dieu, M. and Silvestre, F. 2013 Proteomic responses of peripheral blood mononuclear cells in the European eel (Anguilla anguilla) after perfluorooctane sulfonate exposure. Aquat Toxicol 128–129:43-52. Rolland, R M. 2000. A review of chemically-induced alterations in thyroid and vitamin A status from field studies of wildlife and fish. J Wildlife Dis 36(4):615-635. Romert, A., Tuvendal, P., Simon, A., Dencker, L., and Eriksson, U. 1998 The identification of a 9-cis retinol dehydrogenase in the mouse embryo reveals a pathway for synthesis of 9-cis retinoic acid. P Natl A Sci 95(8):4404-4409. Ross, P S., Deswart, R L., Reijnders, P J H., Vanloveren, H., Vos, J G. and Osterhaus, A D M E. 1995 Contaminant-related suppression of delayed-type hypersensitivity and antibody-responses in harbor seals fed herring from the Baltic sea. Environ Health Persp 103:162-167. Saengtienchai, A., Ikenaka, Y., Nakayama, S M M., Mizukawa, H., Kakehi, M., Bortey‐ Sam, N., Darwish, W S., Tsubota, T., Terasaki, M. and Poapolathep, A. 2014. Identification of interspecific differences in phase II reactions: determination of metabolites in the urine of 16 mammalian species exposed to environmental pyrene. Environ Toxicol Chem 33(9):2062-2069. Sanchez, J C., Fossi, M C. and Focardi, S. 1997 Serum ''B'' esterases as a nondestructive biomarker for monitoring the exposure of reptiles to organophosphorus insecticides. Ecotox Environ Safe 37:45-52. Schrattenholz, A., Soskic, V., Schopf, R., Poznanovic, S., Klemm-Manns, M. and Groebe, K. 2012. Protein biomarkers for in vitro testing of toxicology. Mutat Res 746:113-123.

63 Scott, P D., Bartkow, M., Blockwell, S J., Coleman, H M., Khan, S J., Lim, R., McDonald, J A., Nice, H., Nugegoda, D., Pettigrove, V., Tremblay, L A., Warne, M S J. and Leusch, F D L. 2014 An assessment of endocrine activity in Australian rivers using chemical and in vitro analyses. Environ Sci Pollut R 21:12951-12967. Seriani, R., França, J G., Lombardi, J V., Brito, J M. and Ranzani-Paiva, M J T. 2015. Hematological changes and cytogenotoxicity in the tilapia Oreochromis niloticus caused by sub-chronic exposures to mercury and selenium. Fish Physiol Biochem 41:311-322. Smits, J E., Wayland, M E., Miller, M J., Liber, K. and Trudeau, S. 2000 Reproductive, immune, and physiological end points in tree swallows on reclaimed oil sands mine sites. Environ Toxicol Chem 19(12):2951-2960. Soffker, M. and Tyler, C R. 2012. Endocrine disrupting chemicals and sexual behaviours in fish--a critical review on effects & possible consequences. Crit Rev Toxicol 42:653-668. Sogorb, M A., Ganga, R., Vilanova, E. and Soler, F. 2007 Plasma phenylacetate and 1- naphthyl acetate hydrolyzing activities of wild birds as possible non-invasive biomarkers of exposure to organophosphorus and carbamate insecticides. Toxicol Lett 168:278-285. Solomon, K R., Carr, J A., Du Preez, L H., Giesy, J P., Kendall, R J., Smith, E E. and Van Der Kraak, G J. 2008 Effects of atrazine on fish, amphibians, and aquatic reptiles: a critical review. Criti Rev Toxicol 38:721-772. Spurgeon, D J., Jones, O A H., Dorne, J L C M., Svendsen, C., Swain, S. and Stürzenbaum, S. R. 2010 Systems toxicology approaches for understanding the joint effects of environmental chemical mixtures. Sci Total Environ 408:3725- 3734. Sumpter, J P. and Jobling, S. 1995 Vitellogenesis as a biomarker for estrogenic contamination of the aquatic environment. Environ Health Persp 103(Suppl 7):173-178. Sung, J., Wang, Y., Chandrasekaran, S., Witten, D M. and Price, N D. 2012 Molecular signatures from omics data: from chaos to consensus. Biotechnol J 7(8):946-957. Taylor, C., Duffy, L. K., Terry Bowyer, R. and Blundell, G M. 2000 Profiles of fecal porphyrins in river otters following the Exxon Valdez oil spill. Mar Pollut Bull 40(12):1132-1138.

64 Teglia, C M., Attademo, A M., Peltzer, P M., Goicoechea, H C. and Lajmanovich, R C. 2015 Plasma retinoids concentration in Leptodactylus chaquensis (Amphibia: Leptodactylidae) from rice agroecosystems, Santa Fe province, Argentina. Chemosphere 135:24-30. Thibodeau, J., Filion, S., Spear, P., Paquin, J. and Boily, M. 2012 Oxidation of retinoic acids in hepatic microsomes of wild bullfrogs Lithobates catesbeianus environmentally-exposed to a gradient of agricultural contamination. Ecotoxicology 21:1358-1370. Thomas, R S., Philbert, M A., Auerbach, S S., Wetmore, B A., Devito, M J., Cote, I., Rowlands, J C., Whelan, M P., Hays, S M. and Andersen, M E. 2013 Incorporating new technologies into toxicity testing and risk assessment: moving from 21st century vision to a data-driven framework. Toxicol Sci 136(1):4-18. Tornero, V., Borrell, A., Aguilar, A., Forcada, J. and Lockyer, C. 2006 Organochlorine contaminant and retinoid levels in blubber of common dolphins (Delphinus delphis) off northwestern Spain. Environ Pollut 140:312-321. Troisi, G., Borjesson, L., Bexton, S. and Robinson, I. 2007 Biomarkers of polycyclic aromatic hydrocarbon (PAH)-associated hemolytic anemia in oiled wildlife. Environ Res 105:324-329. Tse, A C K., Lau, K Y T., Ge, W. and Wu, R S S. 2013 A rapid screening test for endocrine disrupting chemicals using primary cell culture of the marine medaka. Aquat Toxicol 144-145:50-58. Vaidya, V S. and Bonventre, J. V. 2010 Biomarkers: in medicine, drug discovery, and environmental health. John Wiley & Sons, Hoboken, New Jersey. Valová, Z., Hudcová, H., Roche, K., Svobodová, J., Bernardová, I. and Jurajda, P. 2013 No relationship found between mercury and lead concentrations in muscle and scales of chub Squalius cephalus L. Environ Monit Assess 185:3359-3368. van der Oost, R., Beyer, J. and Vermeulen, N P. 2003 Fish bioaccumulation and biomarkers in environmental risk assessment: a review. Environ Toxicol Pharm 13:57-149. Van Loveren, H., Ross, P S., Osterhaus, A D M E. and Vos, J. G. 2000 Contaminant- induced immunosuppression and mass mortalities among harbor seals. Toxicol Lett 112-113:319-324.

65 Wang, Z J., Liu, X H., Jin, L., Pu, D Y., Huang, J. and Zhang, Y G. 2016 Transcriptome profiling analysis of rare minnow (Gobiocypris rarus) gills after waterborne cadmium exposure. Comp Biochem Physiol D 19:120-128. Weis, J S., Samson, J., Zhou, T., Skurnick, J. and Weis, P. 2001 Prey capture ability of mummichogs (Fundulus heteroclitus) as a behavioral biomarker for contaminants in estuarine systems. Can J Fish Aquat Sci 58(7):1442-1452. Wilkening, S., Stahl, F. and Bader, A. 2003 Comparison of primary human hepatocytes & hepatoma cell line Hepg2 with regard to their biotransformation properties. Drug Metab Dispos 31(8):1035-1042. Yamashita, R., Takada, H., Murakami, M., Fukuwaka, M. A. and Watanuki, Y. 2007 Evaluation of noninvasive approach for monitoring PCB pollution of seabirds using preen gland oil. Environ Sci Technol 41, 4901-4906. Zurlo, J., Rudacille, D. and Goldberg, A M. (1996) The three Rs: the way forward. Environ Health Persp 104(8): 878-880.

66 Supporting Information for Chapter 2

Table S2.1: List of all papers included in the database after a systematic search of the literature was performed. The broad experimental categories, source of test animals used, confirmed or potential biomarkers of exposure and effect, and the method type that each study was categorised into are denoted by a grey circle.

67 Broad experimental Animals used Biomarker of Biomarker of Method category category exposure effect References ND& Decease Confirm Confirm ND&E D&NE D&E Wild Lab Potential Potential 1 2 3 4 5 NE d ed ed Aguilera, C., del Pliego, P. G., Alfaro, R. M., Lazcano, D., & Cruz, J. (2012) Pollution biomarkers in the spiny lizard (Sceloporus spp.) from two suburban populations of Monterrey, Mexico. Ecotoxicology, 21(8), 2103- 2112. Anzolin, D. G., Sarkis, J. E., Diaz, E., Soares, D. G., Serrano, I. L., Borges, J. C., . . . Carvalho, P. S. (2012) Contaminant concentrations, biochemical and hematological biomarkers in blood of West Indian manatees Trichechus manatus from Brazil. Marine Pollution Bulletin, 64(7), 1402- 1408. Baillon, L., Pierron, F., Oses, J., Pannetier, P., Normandeau, E., Couture, P., . . . Baudrimont, M. (2016) Detecting the exposure to Cd and PCBs by means of a non-invasive transcriptomic approach in laboratory and wild contaminated European eels (Anguilla anguilla ). Environmental Science and Pollution Research, 23(6), 5431-5441. Barata, C., Fabregat, M. C., Cotin, J., Huertas, D., Sole, M., Quiros, L., . . . Pina, B. (2010) Blood biomarkers and contaminant levels in feathers and eggs to assess environmental hazards in heron nestlings from impacted sites in Ebro basin (NE Spain). Environmental Pollution, 158(3), 704-710. Casini, S., Fossi, M. C., Cavallaro, K., Marsili, L., & Lorenzani, J. (2002) The use of porphyrins as a non-destructive biomarker of exposure to contaminants in two sea lion populations. Marine Environmental Research, 54(3-5), 769- 773. Casini, S., Fossi, M. C., Gavilan, J. F., Barra, R., Parra, O., Leonzio, C., & Focardi, S. (2001) Porphyrin levels in excreta of sea birds of the Chilean coasts as nondestructive biomarker of exposure to environmental pollutants. Archives of Environmental Contamination and Toxicology, 41(1), 65-72. Celis, J. E., Espejo, W., Gonzalez-Acuna, D., Jara, S., & Barra, R. (2014) Assessment of trace metals and porphyrins in excreta of Humboldt penguins

(Spheniscus humboldti ) in different locations of the northern coast of Chile. Environmental Monitoring and Assessment, 186(3), 1815- 1824. Chahin, A., Peiffer, J., Olry, J. C., Crepeaux, G., Schroeder, H., Rychen, G., & Guiavarc'h, Y. (2013) EROD activity induction in peripheral blood lymphocytes, liver and brain tissues of rats orally exposed to polycyclic aromatic hydrocarbons. Food and Chemical Toxicology, 56, 371-380.

68 Champoux, L., Rodrigue, J., Desgranges, J. L., Trudeau, S., Hontela, A., Boily, M., & Spear, P. (2002) Assessment of contamination and biomarker responses in two species of herons on the St. Lawrence river. Environmental Monitoring and Assessment, 79(2), 193-215. Ciliberti, A., Martin, S., Ferrandez, E., Belluco, S., Rannou, B., Dussart, C., . . . de Buffrenil, V. (2013) Experimental exposure of juvenile savannah monitors (Varanus exanthematicus) to an environmentally relevant mixture of three contaminants: effects and accumulation in tissues. Environmental Science and Pollution Research, 20(5), 3107-3114. Clapp, J. B., Bevan, R. M., & Singleton, I. (2012) Avian urine: Its potential as a non-invasive biomonitor of environmental metal exposure in birds. Water Air and Soil Pollution, 223(7), 3923-3938. Cordi, B., Fossi, C., & Depledge, M. (1997) Temporal biomarker responses in wild passerine birds exposed to pesticide spray drift. Environmental Toxicology & Chemistry, 16(10), 2118. D'Have, H., Scheirs, J., Mubiana, V. K., Verhagen, R., Blust, R., & De Coen, W. (2005) Nondestructive pollution exposure assessment in the european hedgehog (Erinaceus europaeus): I. Relationships between concentrations of metals and arsenic in hair, spines, and soil. Environmental Toxicology and Chemistry, 24(9), 2356-2364. D'Havé, H., Scheirs, J., Mubiana, V. K., Verhagen, R., Blust, R., & De Coen, W. (2006). Non-destructive pollution exposure assessment in the European hedgehog (Erinaceus europaeus): II. Hair and spines as indicators of endogenous metal and As concentrations. Environmental Pollution, 142(3), 438-448. da Silva, C. C., Klein, R. D., Barcarolli, I. F., & Bianchini, A. (2016) Metal contamination as a possible etiology of fibropapillomatosis in juvenile female green sea turtles Chelonia mydas from the southern Atlantic Ocean. Aquatic Toxicology, 170, 42-51. da Silva Corrêa, S. A., de Souza Abessa, D. M., dos Santos, L. G., da Silva, E. B., & Seriani, R. (2016) Differential blood counting in fish as a non- destructive biomarker of water contamination exposure. Toxicological and Environmental Chemistry, 99(3), 1-10. Dauwe, T., Jaspers, V., Covaci, A., Schepens, P., & Eens, M. (2005). Feathers as a nondestructive biomonitor for persistent organic pollutants. Environmental Toxicology and Chemistry, 24(2). de la Casa-Resino, I., Hernández-Moreno, D., Navas, J. M., Soler, F., & Pérez-López, M. (2013) Non-destructive multibiomarker approach in European quail (Coturnix coturnix coturnix ) exposed to the herbicide atrazine. Archives of Environmental Contamination and Toxicology, 65(3), 567-574.

69 Długaszek, M., & Kopczyński, K. (2014). Correlations between elements in the fur of wild animals. Bulletin of Environmental Contamination & Toxicology, 93(1), 25-30. Dos Santos, C. R., Cavalcante, A. L., Hauser-Davis, R. A., Lopes, R. M., & Mattos Rde, C. (2016) Effects of sub-lethal and chronic lead concentrations on blood and liver ALA-D activity and hematological parameters in Nile tilapia. Ecotoxicology and environmental safety, 129, 250-256. Dzul-Caamal, R., Salazar-Coria, L., Olivares-Rubio, H. F., Rocha-Gomez, M. A., Giron-Perez, M. I., & Vega-Lopez, A. (2016) Oxidative stress response in the skin mucus layer of Goodea gracilis (Hubbs and Turner, 1939) exposed to crude oil: A non-invasive approach. Comparative Biochemistry and Physiology, Part A, 200, 9-20. Eissa, B. L., Ossana, N. A., Ferrari, L., & Salibián, A. (2010) Quantitative behavioral parameters as toxicity biomarkers: fish responses to waterborne cadmium. Archives of Environmental Contamination and Toxicology, 58(4), 1032-1039. Fair, P. A., Romano, T., Schaefer, A. M., Reif, J. S., Bossart, G. D., Houde, M., . . . Hulsey, T. C. (2013) Associations between perfluoroalkyl compounds and immune and clinical chemistry parameters in highly exposed bottlenose dolphins (Tursiops truncatus ). Environmental Toxicology and Chemistry, 32(4), 736-746. Fossi, C., Marsili, L., Junin, M., Castello, H., Lorenzani, J., Casini, S., . . . Leonzio, C. (1997). Use of nondestrucitive biomarkers and residue analysis to assess the health status of endangered species of pinnipeds in the South- west atlantic. Marine Pollution Bulletin, 34(3), 157-162. Fossi, M. C., Casini, S., Maltese, S., Panti, C., Spinsanti, G., & Marsili, L. (2013) An “ex vivo” model to evaluate toxicological responses to mixtures of contaminants in cetaceans: Integumentum biopsy slices. Environmental Toxicology, 29(10), 1107-1121. Fossi, M. C., Casini, S., & Marsili, L. (1996) Porphyrins in excreta: A nondestructive biomarker for the hazard assessment of birds contaminated with PCBs. Chemosphere, 33(1), 29-42. Fossi, M. C., Casini, S., & Marsili, L. (2007) Potential toxicological hazard due to endocrine-disrupting chemicals on Mediterranean top predators: State of art, gender differences and methodological tools. Environmental Research, 104(1), 174-182. Fossi, M. C., Casini, S., Marsili, L., Ausili, A., & Notarbartolo Di Sciara, G. (2001) Are the Mediterranean top predators exposed to toxicological risk due to endocrine disrupters? Annals of the New York Academy of Sciences, 948(1), 67-74. Fossi, M. C., Leonzio, C., Massi, A., Lari, L., & Casini, S. (1992) Serum esterase inhibition in birds - a nondestructive biomarker to assess

70 organophosphorus and carbamate contamination. Archives of Environmental Contamination and Toxicology, 23(1), 99-104. Fossi, M. C., Savelli, C., Marsili, L., Casini, S., Jimenez, B., Junin, M., . . . Lorenzani, J. A. (1997) Skin biopsy as a nondestructive tool for the toxicological assessment of endangered populations of pinnipeds: Preliminary results on mixed function oxidase in Otaria flavescens . Chemosphere, 35(8), 1623-1635. Gauthier, J. M., Dubeau, H., Rassart, E., Jarman, W. M., & Wells, R. S. (1999). Biomarkers of DNA damage in marine mammals. Mutation Research, 444(2), 427-439. Geens, A., Dauwe, T., Bervoets, L., Blust, R., & Eens, M. (2010) Haematological status of wintering great tits (Parus major) along a metal pollution gradient. Science of the Total Environment, 408(5), 1174-1179. Hall, A. J., Green, N. J. L., Jones, K. C., Pomeroy, P. P., & Harwood, J. (1998) Thyroid hormones as biomarkers in grey seals. Marine Pollution Bulletin, 36(6), 424-428. Henriksen, E. O., Gabrielsen, G. W., & Skaare, J. U. (1998) Validation of the use of blood samples to assess tissue concentrations of organochlorines in glaucous gulls, Larus hyperboreus . Chemosphere, 37(13), 2627-2643. Hernández-Moreno, D., de la Casa Resino, I., Fidalgo, L. E., Llaneza, L., Soler Rodríguez, F., Pérez-López, M., & López-Beceiro, A. (2013) Noninvasive heavy metal pollution assessment by means of Iberian wolf

(Canis lupus signatus ) hair from Galicia (NW Spain): a comparison with invasive samples. Environmental Monitoring and Assessment, 185(12), 10421-10430. Hernandez-Moreno, D., Perez-Lopez, M., Miguez, M. P., Soler, F., & Jimenez, B. (2012). Porphyrin levels in excreta of rabbit as non-destructive biomarkers of diazinon exposure. Environmental Toxicology and Pharmacology, 34(2), 466-472. doi:10.1016/j.etap.2012.06.009 Hernout, B. V., McClean, C. J., Arnold, K. E., Walls, M., Baxter, M., & Boxall, A. B. (2016). Fur: A non-invasive approach to monitor metal exposure in bats. Chemosphere, 147, 376-381. doi:10.1016/j.chemosphere.2015.12.104 Hopkins, W. A., Roe, J. H., Snodgrass, J. W., Jackson, B. P., Kling, D. E., Rowe, C. L., & Congdon, J. D. (2001) Nondestructive indices of trace element exposure in squamate reptiles. Environmental Pollution, 115(1), 1- 7. Ilizaliturri-Hernández, C. A., González-Mille, D. J., Mejía-Saavedra, J., Espinosa-Reyes, G., Torres-Dosal, A., & Pérez-Maldonado, I. (2013)Blood lead levels, δ-ALAD inhibition, and hemoglobin content in blood of giant toad (Rhinella marina ) to asses lead exposure in three areas surrounding an industrial complex in Coatzacoalcos, Veracruz, Mexico. Environmental Monitoring and Assessment, 185(2), 1685-1698.

71 Jara-Carrasco, S., Barra, R., Espejo, W., Celis, J. E., González-Acuña, D., Chiang, G., & Sánchez-Hernández, J. (2016). Persistent organic pollutants and porphyrin levels in excreta of penguin colonies from the Antarctic Peninsula area. Polar Record, 1-9. doi:10.1017/s0032247416000607 Jiménez, B., Mori, G., Concejero, M. A., Merino, R., Casini, S., & Fossi, M. C. (2007) Vitellogenin and zona radiata proteins as biomarkers of endocrine disruption in peregrine falcon (Falco peregrinus ). Chemosphere, 67(9), S375- S378. Jones, D. E., Gogal, R. M., Nader, P. B., & Holladay, S. D. (2005) Organochlorine detection in the shed skins of snakes. Ecotoxicology and Environmental safety, 60(3), 282-287. Jones, D. E., Magnin-Bissel, G., & Holladay, S. D. (2009) Detection of polycyclic aromatic hydrocarbons in the shed skins of corn snakes (Elaphe guttata ). Ecotoxicology and Environmental Safety, 72(7), 2033-2035. Keller, J. M., Kucklick, J. R., Harms, C. A., & McClellan-Green, P. D. (2004) Organochlorine contaminants in sea turtles: Correlations between whole blood and fat. Environmental Toxicology and Chemistry, 23(3), 726- 738. Kocagoz, R., Onmus, O., Onat, I., Cagdas, B., Siki, M., & Orhan, H. (2014) Environmental and biological monitoring of persistent organic pollutants in waterbirds by non-invasive versus invasive sampling. Toxicology Letters, 230(2), 208-217. Lari, L., Massi, A., Fossi, M. C., Casini, S., Leonzio, C., & Focardi, S. (1994) Evaluation of toxic effects of the organophosphorus insecticide azinphos- methyl in experimentally and naturally exposed birds. Archives of Environmental Contamination and Toxicology, 26(2), 234-239. Lunardi, D., Abelli, L., Panti, C., Marsili, L., Fossi, M. C., & Mancia, A. (2016) Transcriptomic analysis of bottlenose dolphin (Tursiops truncatus ) skin biopsies to assess the effects of emerging contaminants. Marine Environmental Research, 114, 74-79. Marsili, L., Fossi, M. C., Casini, S., & Focardi, S. (1996) PCB levels in bird blood and relationship to MFO responses. Chemosphere, 33(4), 699- 710. Marsili, L., Fossi, M. C., Casini, S., Savelli, C., Jimenez, B., Junin, M., & Castello, H. (1997) Fingerprint of polycyclic aromatic hydrocarbons in two populations of southern sea lions (Otaria flavescens ). Chemosphere, 34(4), 759-770. Marsili, L., Fossi, M. C., Notarbartolo Di Sciara, G., Zanardelli, M., Nani, B., Panigada, S., & Focardi, S. (1998) Relationship between organochlorine contaminants and mixed function oxidase activity in skin biopsy specimens of Mediterranean fin whales (Balaenoptera physalus ). Chemosphere, 37(8), 1501-1510. Martínez-López, E., Sousa, A. R., María-Mojica, P., Gómez-Ramírez, P.,

Guilhermino, L., & García-Fernández, A. J. (2010) Blood δ-ALAD, lead and

72 cadmium concentrations in spur-thighed tortoises (Testudo graeca ) from Southeastern Spain and Northern Africa. Ecotoxicology, 19(4), 670- 677. Numata, M., Fawcett, J. P., & Rosengren, R. J. (2008) Induction of in vitro EROD activity and in vivo caffeine metabolism in two species of New Zealand birds. Environmental Toxicology and Pharmacology, 25(3), 358- 364. Nyman, M., Bergknut, M., Fant, M. L., Raunio, H., Jestoi, M., Bengs, C., . . . Helle, E. (2003) Contaminant exposure and effects in Baltic ringed and grey seals as assessed by biomarkers. Marine Environmental Research, 55, 26. Oropesa, A. L., Gravato, C., Sánchez, S., & Soler, F. (2013) Characterization of plasma cholinesterase from the White stork (Ciconia ciconia ) and its in vitro inhibition by anticholinesterase pesticides. Ecotoxicology and Environmental Safety, 97, 131-138. Oropesa, A. L., Pérez-López, M., & Soler, F. (2014) Characterization of plasma cholinesterase in rabbit and evaluation of the inhibitory potential of diazinon. Ecotoxicology and Environmental Safety, 100, 39-43. Orton, F., Baynes, A., Clare, F., Duffus, A. L., Larroze, S., Scholze, M., & Garner, T. W. (2014) Body size, nuptial pad size and hormone levels: potential non-destructive biomarkers of reproductive health in wild toads (Bufo bufo ). Ecotoxicology, 23(7), 1359-1365. Panti, C., Spinsanti, G., Marsili, L., Casini, S., Frati, F., & Fossi, M. C. (2011) Ecotoxicological diagnosis of striped dolphin (Stenella coeruleoalba ) from the Mediterranean basin by skin biopsy and gene expression approach. Ecotoxicology, 20(8), 1791-1800. Quesada, R. J., McCleary, R. J. R., Heard, D. J., & Lillywhite, H. B. (2014) Non-lethal sampling of liver tissue for toxicologic evaluation of Florida cottonmouths snakes, Agkistrodon piscivorus conanti . Ecotoxicology, 23(1), 33-37. Rainwater, T. R., Wu, T. H., Finger, A. G., Canas, J. E., Yu, L., Reynolds, K. D., . . . McMurry, S. T. (2007) Metals and organochlorine pesticides in caudal scutes of crocodiles from Belize and Costa Rica. Science of the Total Environment, 373(1), 146-156. Rainwater, T. v., Selcer, K. W., Nespoli, L. M., Finger, A. G., Ray, D. A., Platt, S. G., . . . McMurry, S. T. (2008) Plasma vitellogenin in Morelet's crocodiles from contaminated habitats in northern Belize. Environmental Pollution, 153(1), 101-109. Rees, C. B., McCormick, S. D., & Li, W. (2005) A non-lethal method to estimate CYP1A expression in laboratory and wild Atlantic salmon (Salmo salar ). Comparative Biochemistry and Physiology, Part C, 141(3), 217- 224. Rey Vázquez, G., Meijide, F. J., Da Cuña, R. H., Lo Nostro, F. L., Piazza, Y. G., Babay, P. A., . . . Guerrero, G. A. (2009) Exposure to waterborne 4- tert- octylphenol induces vitellogenin synthesis and disrupts testis morphology in the South American freshwater fish Cichlasoma dimerus (Teleostei,

73 Perciformes ). Comparative Biochemistry and Physiology, Part C, 150(2), 298- 306. Roland, K., Kestemont, P., Hénuset, L., Pierrard, M. A., Raes, M., Dieu, M., & Silvestre, F. (2013) Proteomic responses of peripheral blood mononuclear cells in the European eel (Anguilla anguilla ) after perfluorooctane sulfonate exposure. Aquatic Toxicology, 128–129, 43-52. Rosa, C., O'Hara, T. M., Hoekstra, P. F., Refsal, K. R., & Blake, J. E. (2007) Serum thyroid hormone concentrations and thyroid histomorphology as biomarkers in bowhead whales (Balaena mysticetus ). Canadian Journal of Zoology, 85(5), 609-618. Rudolph, I., Chiang, G., Galban-Malagon, C., Mendoza, R., Martinez, M., Gonzalez, C., . . . Barra, R. (2016) Persistent organic pollutants and porphyrins biomarkers in penguin faeces from Kopaitic Island and Antarctic Peninsula. Science of the Total Environment, 573, 1390-1396. Saengtienchai, A., Ikenaka, Y., Nakayama, S. M. M., Mizukawa, H., Kakehi, M., Bortey‐Sam, N., . . . Poapolathep, A. (2014) Identification of interspecific differences in phase II reactions: determination of metabolites in the urine of 16 mammalian species exposed to environmental pyrene. Environmental Toxicology and Chemistry, 33(9), 2062-2069. Sánchez-Chardi, A., Marques, C. C., Gabriel, S. I., Capela-Silva, F., Cabrita, A. S., López-Fuster, M. J., . . . Mathias, M. D. L. (2008) Haematology, genotoxicity, enzymatic activity and histopathology as biomarkers of metal pollution in the shrew Crocidura russula . Environmental Pollution, 156(3), 1332-1339. Sanchez, J. C., Fossi, M. C., & Focardi, S. (1997) Serum ''B'' esterases as a nondestructive biomarker for monitoring the exposure of reptiles to organophosphorus insecticides. Ecotoxicology and Environmental Safety, 38(1), 45-52. Seriani, R., Abessa, D. M., Moreira, L. B., Cabrera, J. P., Sanches, J. Q., Silva, C. L., . . . Ranzani-Paiva, M. J. (2015) In vitro mucus transportability, cytogenotoxicity, and hematological changes as non- destructive physiological biomarkers in fish chronically exposed to metals. Ecotoxicology and Environmental Safety, 112, 162-168. Seriani, R., França, J. G., Lombardi, J. V., Brito, J. M., & Ranzani-Paiva, M. J. T. (2015) Hematological changes and cytogenotoxicity in the tilapia

Oreochromis niloticus caused by sub-chronic exposures to mercury and selenium. Fish Physiology and Biochemistry, 41(1), 311-322. Sogorb, M. A., Ganga, R., Vilanova, E., & Soler, F. (2007) Plasma phenylacetate and 1-naphthyl acetate hydrolyzing activities of wild birds as possible non-invasive biomarkers of exposure to organophosphorus and carbamate insecticides. Toxicology Letters, 168(3), 278-285.

74 Taylor, C., Duffy, L. K., Terry Bowyer, R., & Blundell, G. M. (2000) Profiles of fecal porphyrins in river otters following the Exxon Valdez oil spill. Marine Pollution Bulletin, 40(12), 1132-1138. Teglia, C. M., Attademo, A. M., Peltzer, P. M., Goicoechea, H. C., & Lajmanovich, R. C. (2015) Plasma retinoids concentration in Leptodactylus chaquensis (Amphibia: Leptodactylidae ) from rice agroecosystems, Santa Fe province, Argentina. Chemosphere, 135, 24-30. Thibodeau, J., Filion, S., Spear, P., Paquin, J., & Boily, M. (2012) Oxidation of retinoic acids in hepatic microsomes of wild bullfrogs Lithobates catesbeianus environmentally-exposed to a gradient of agricultural contamination. Ecotoxicology, 21(5), 1358-1370. Tornero, V., Borrell, A., Aguilar, A., Forcada, J., & Lockyer, C. (2006) Organochlorine contaminant and retinoid levels in blubber of common dolphins (Delphinus delphis ) off northwestern Spain. Environmental Pollution, 140(2), 312-321. Tornero, V., Borrell, A., Aguilar, A., Wells, R. S., Forcada, J., Rowles, T. K., & Reijnders, P. J. H. (2005) Effect of organochlorine contaminants and individual biological traits on blubber retinoid concentrations in bottlenose dolphins (Tursiops truncatus ). Journal of Environmental Monitoring, 7(2), 109-114. Troisi, G., Borjesson, L., Bexton, S., & Robinson, I. (2007) Biomarkers of polycyclic aromatic hydrocarbon (PAH)-associated hemolytic anemia in oiled wildlife. Environmental Research, 105(3), 324-329. Troisi, G. M. (2013). T-cell Responses in oiled guillemots and swans in a rehabilitation setting. Archives of Environmental Contamination and Toxicology, 65(1), 142-148. Valová, Z., Hudcová, H., Roche, K., Svobodová, J., Bernardová, I., & Jurajda, P. (2013) No relationship found between mercury and lead concentrations in muscle and scales of chub Squalius cephalus L. Environmental Monitoring and Assessment, 185(4), 3359-3368. Weis, J. S., Samson, J., Zhou, T., Skurnick, J., & Weis, P. (2001) Prey capture ability of mummichogs (Fundulus heteroclitus ) as a behavioral biomarker for contaminants in estuarine systems. Canadian Journal of Fisheries and Aquatic Sciences, 58(7), 1442-1452. Wilson, J. Y., Wells, R. S., Aguilar, A., Borrell, A., Tornero, V., Reijnders, P., . . . Stegeman, J. J. (2007) Correlates of cytochrome P450 1A1 expression in bottlenose dolphin (Tursiops truncatus ) integument biopsies. Toxicological Sciences, 97(1), 111-119. Yamashita, R., Takada, H., Murakami, M., Fukuwaka, M. A., & Watanuki, Y. (2007) Evaluation of noninvasive approach for monitoring PCB pollution of seabirds using preen gland oil. Environmental Science and Technology, 41(14), 4901-4906.

75

76 Chapter 3: Influence of concentration and exposure duration on global protein expression of green sea turtle primary skin cells exposed to polychlorinated biphenyl (PCB) 153 and perfluorononanoic acid (PFNA)

Statement of Contribution for Chapter 3

This chapter includes a co-authored article that has been submitted to Environmental

Science & Technology for consideration for publication. The bibliographic details of the co-authored paper are:

Stephanie Chaousis, Frederic DL Leusch, Amanda Nouwens, Steven D Melvin and

Jason P van de Merwe (under review) Influence of concentration and exposure duration on global protein expression of green sea turtle primary skin cells exposed to polychlorinated biphenyl (PCB) 153 and perfluorononanoic acid (PFNA). Environ Sci

Technol.

My contribution to the paper include:

• the conceptualisation of the design of the experiment,

• undertaking all lab work,

• data analysis, and

• writing of the manuscript.

My co-authors provided input into the design of the study, interpretation of the data and final polishing of the writing.

This chapter is formatted in the style of the journal of publication.

77 (Signed) ______(Date) 24/04/2019 Stephanie Chaousis

(Countersigned) ______(Date) 23/04/2019 Jason van de Merwe

78 3.1 Abstract

Adverse effects of chemical exposure in threatened wildlife are often poorly understood, despite the known risks that pollutants pose to vulnerable populations. The consequences of chemical exposure are particularly challenging to explore in large marine species such as sea turtles, cetaceans, and pinnipeds due to various practical and ethical constraints that preclude traditional toxicology research on these animals. In this study, we aim to address some of these limitations by investigating the usefulness of an ethical and higher throughput cell-based approach to elucidate molecular level effects of contaminants on sea turtles. To examine functional effects at the molecular level, we measured changes in cellular protein expression resulting from contaminant exposure using a primary sea turtle cell line in vitro and validated the relevance of the results by looking for dysregulation of known in vivo markers of contaminant exposure. Our experimental design aimed to address basic questions necessary for moving forward with cell-based proteomics, specifically, we explored the influence of time and concentration on overall protein expression in sea turtle skin cells. Primary green turtle cells were exposed to polychlorinated biphenyl (PCB) 153 and perfluorononanoic acid

(PFNA) for 24 and 48 h at three sub-lethal environmentally relevant concentrations (1,

10 and 100 µg/L). The global protein expression of each sample was determined using sequential window acquisition of all theoretical mass spectra (SWATH-MS) with over

1000 proteins identified within a 1% false discovery rate (FDR) threshold. Each treatment produced 90 to 400 significantly differentially expressed proteins compared to control cells (p-value <0.0001). Overall, 24 h exposure resulted in a greater number of identified dysregulated proteins compared to 48 h exposure, for both contaminants across all three concentrations. Variation was lower between replicates, both in terms of overall total protein expression and the proportion of up and down-regulated proteins, at

24 h compared to 48 h. There was no statistically significant concentration dependent

79 response relationship observed for the number of differentially expressed proteins with either contaminant at the level of individual protein expression. Differential expression did not present as a concentration dependent response relationship for either contaminant. Two known in vivo markers of contaminant exposure, superoxide dismutase and glutathione S-transferase, were significantly dysregulated at all concentrations of 24 h PCB153 exposure. This study therefore provides an optimised framework for conducting cell-based exposure studies in wildlife proteomics studies, and highlights that proteins detected in vitro could act as biomarkers of chemical exposure and effect in vivo.

3.2 Keywords

Proteomics; in vitro bioassay; sea turtle; PCB; PFC; sub-lethal; toxicity

3.3 Introduction

Oceans were once viewed as a final frontier safe from human disturbance, but it is now clear that human activities introduce a diversity of stressors including chemical pollutants into marine environments (Johnston et al. 2015, Tornero and Hanke 2016).

Chemical exposure can impose physiological burdens on marine wildlife, which may exacerbate the effects of other anthropogenic pressures such as overfishing, habitat degradation and climate change (Noyes et al. 2009, Kendall 2010). Reported adverse effects of persistent pollutants in large marine vertebrates include endocrine disruption

(Tabuchi et al. 2006, Fossi et al. 2007, Routti et al. 2010), immunotoxicity (Van

Loveren et al. 2000, Keller et al. 2006), and adverse developmental effects (Vos et al.

2003, Baccarelli and Bollati 2009). These types of sub-lethal effects are important to identify and understand since, as the Adverse Outcome Pathway framework shows, lower-level initiating events can ultimately have dramatic higher-level consequences

(Ankley et al. 2010). For example, interference with hormone receptor signalling can

80 negatively impact endocrine systems, significantly inhibit reproductive outputs, and this in turn can lead to population decline (Kidd et al. 2007). Large marine vertebrates like sea turtles, pinnipeds and cetaceans, are known to be vulnerable to sub-lethal adverse effects resulting from chronic exposure to persistent chemical pollutants, since they are generally long lived, fat-storing and coastal feeders (Lutz et al. 2002, Bengtson Nash et al. 2013). Unfortunately, sub-lethal adverse outcomes are often inconspicuous in large marine vertebrates and associated risks are therefore often not recognized until consequences become evident at the population level. Exploring responses to contaminants in marine megafauna under controlled conditions is also extremely challenging due to logistical and ethical constraints, and research has therefore been limited (Chaousis et al. 2018). Considerable benefits are thus likely to arise from establishment of robust in vitro methods to explore toxicity risks to marine megafauna.

Due to various practical and ethical factors, most toxicology research on large threatened marine species has been field-based and reliant on non-destructive tissue and blood sampling (Chaousis et al. 2018). There is at present major international interest and considerable benefits to be gained from establishing ethical alternatives that can be used to explore toxicological effects associated with chemical pollutants with respect to large marine vertebrates. Species-specific cellular bioassays offer a promising option for enhancing field-based research aimed at elucidating mechanistic toxicity information, including the potential to identify specific pathways that are vulnerable to dysregulation from contaminant exposure (Roland et al. 2013, Lunardi et al. 2016). In vitro bioassays measuring discrete endpoints have indeed represented a major breakthrough in ethical toxicity testing because they are robust, high throughput, and provide quantifiable species-specific toxicological data (Leusch et al. 2010, Oropesa et al. 2013).

Notwithstanding, traditional single-endpoint approaches are limited to providing information on specific modes of action and offer little in the way of generating new

81 hypotheses regarding other cellular responses that might represent key initiating events leading to adverse outcomes.

Recent technological advancements in non-targeted analysis, such as global protein expression, have contributed tremendously to our ability to explore the mechanistic underpinnings of toxicological injury at the organism level (Kennedy 2002, Rabilloud and Lescuyer 2015). Surprisingly, such techniques have received limited attention as complementary tools for in vitro bioassays (Schrattenholz et al. 2012). Considering the advantages of untargeted techniques and ethical concerns preventing robust toxicological studies in marine megafauna, we propose that non-targeted proteomic analysis of species-specific in vitro cell lines could offer a powerful approach for directing toxicity research at the whole-organism level, by identifying chemicals and cellular effects that warrant attention. Non-targeted quantitative proteomics is still a relatively new technique in toxicology and particularly for studies relevant to large marine vertebrates, despite the fundamental techniques having existed for several decades (Patterson and Aebersold 2003). Proteins are ideal molecular markers because they are involved in almost all metabolic and regulatory pathways, form the structural basis of cells, and are predicted to be less transient than gene expression patterns

(Denslow et al. 2012). Proteins are also highly suitable as cross platform markers since they are detected in cells, tissues, and blood samples collected from whole-organisms, increasing the likelihood that biomarkers identified at the cellular level in vitro will have relevance at the organism level (Benninghoff 2006).

Biomarkers currently associated with PCB and PFNA exposure in green sea turtles include cytochrome P450 (CYP)1A, CYP2K1, glutathione-S-transferase, superoxide dismutase (SOD), and catalase (CAT) (Valdivia et al. 2007, Richardson et al. 2009,

Richardson et al. 2010, Labrada-Martagón et al. 2011). Since these markers and other proteins related to adverse effect pathways manifest at the cellular level, they are logical

82 endpoints to validate the relevance of in vitro responses for predicting in vivo effects.

However, to the best of our knowledge there are currently only four studies that have examined such responses based on in vitro exposure of cells to chemical contaminants

(Lasserre et al. 2009, Roland et al. 2013, Lin et al. 2014, Vuong et al. 2016). None of these studies has utilised a data independenent acquisition mass spectrometric methods known as sequential window acquisition of all theoretical mass spectra (SWATH-MS) analysis, which is a novel and powerful method for identification and relative quantification of proteins within complex samples (Gillet et al. 2012).

As a model species for this research, green sea turtles (Chelonia mydas) are ideal as they share traits similar to other large marine vertebrates that make them to susceptible to adverse effects of contaminant exposure. Green sea turtles are long lived (Chaloupka et al. 2004, Limpus and Fien 2009), fat-storing (Hamann et al., 2002) and coastal feeders with strong foraging site fidelity (Limpus et al. 1992). As is the case for other large marine vertebrates, biomarker discovery and mechanistic research in sea turtles is a relatively new field, with the earliest study originating in 2003 (Keller 2003) and very few studies since. In a recent review, Finlayson et al. (2016) reported just 13 studies addressing biomarkers of contaminant exposure in sea turtles, highlighting a need for research exploring responses to chemical in these organisms. Evidently, sea turtles are an excellent candidate species for developing methodologies for pairing data intensive untargeted proteomics techniques with species-specific in vitro cellular bioassays. As with any omics approach, it is important to explore potential sources of variance that might be relevant considerations for such research (Simmons et al. 2015). In short, a basic level of scrutiny is necessary to explore how methodological choices influence omics outcomes and subsequent interpretations from toxicity experiments (Fent and

Sumpter 2011).

83 The present study describes an approach for optimising untargeted proteomic analysis of green sea turtle cells exposed to contaminants in vitro. The primary aim was to understand what influence experimental design has on observed variations in overall protein dysregulation. Specifically, we measured 1) the total number of proteins identified, 2) differences in the number of differentially expressed proteins and 3) the proportion of up and down-regulated proteins; all in relation to different exposure durations (24- and 48 h), concentrations (1, 10 and 100 μg/L) and toxicants

(polychlorinated biphenyls (PCBs) and perfluoroalkyl compounds (PFCs). We also aimed to look for cytochrome P450 (CYP)1A, CYP2K1, glutathione-S-transferase, superoxide dismutases (SODs), and catalase (CAT) due to their reported use as in vivo biomarkers of exposure (Valavanidis et al. 2006).

3.4 Methods

3.4.1. Chemicals

Polychlorinated biphenyls (PCBs) and perfluoroalkyl compounds (PFCs) were chosen as focal contaminants since both have been detected in tissue, blood and eggs of

C.mydas, (van de Merwe et al. 2009, Richardson et al. 2010, van de Merwe et al.

2010b) and sub-lethal adverse effects and poor health have been attributed to these chemicals (van de Merwe et al. 2010b, Komoroske et al. 2011, Keller et al. 2014,

Zychowski and Godard-Codding 2017). PCBs are among the most prevalent contaminants in sea turtles with PCBs detected up to ~ 271 µg/L in C.mydas blood (van de Merwe et al. 2010a). PFCs are similarly prevalent in the environment and are contaminants of emerging concern (Wang et al. 2017), and detected in sea turtles at relatively high concentrations of up to 106 µg/L total PFCs in blood (Keller et al. 2005).

Exposure concentrations of 1, 10 and 100 µg/L for each contaminant were chosen 1) because these concentrations have been detected in sea turtles, 2) facilitate comparison

84 with previous studies of a similar nature using comparable concentrations (Lasserre et al. 2009, Roland et al. 2013) and 3) were verified as sub-cytotoxic (see sections 3.4.3 and 3.5.1) and therefore support evaluation of mechanistic responses.

Perfluorononanoic acid (PFNA) >97% pure was purchased from Sigma-Aldrich (Castle

Hill; product number 394459) and 2,2’,4’,4’,5,5’-hexachlorobiphenyl (PCB153) was purchased from Novachem (Heidelberg; product number C-153N). Stock solutions (10 mg/mL) were made in HPLC grade methanol and stored at -20°C until use. Working solutions of 100 mg/L (100 ppm) were then prepared and used to achieve the concentration range for in vitro experiments.

3.4.2. Primary cell culture

Primary skin fibroblasts from a C. mydas sub-adult (hereafter referred to as GT12s-p) were previously established by Finlayson and colleagues (Finlayson et al. 2019b). The

GT12s-p cells were thawed from cryopreserved stocks and cultured according to conditions optimised for these cells by Finlayson (et al. 2019b). Cells were grown in

RPMI medium without L-glutamine (Sigma-Aldrich), and supplemented with 10%

Foetal Bovine Serum (FBS) (LifeTechnologies), 1% penicillin/streptomycin

(LifeTechnologies) and 1% amphotericin. Cultures were grown in an incubator at 30°C and 5% CO2 in a humidified atmosphere. Cells were cultured for a minimum of two passages before being used in bioassays, and cultures did not exceed 8 passages.

3.4.3. Validation that exposure concentrations are not cytotoxic

A resazurin cytotoxicity assay was conducted in duplicate on separate days to determine the cytotoxicity of PFNA and PCB153 in the GT12s-p cell line. Cells were seeded in culture medium at a density of 3.5 × 104 cells/cm2 in a 96 well flat bottom plate and incubated at 30°C and 5% CO2 overnight to achieve adherence to the growth surface.

The following day, culture medium was removed, cells were gently washed once with

85 warm Phosphate Buffered Saline (PBS), and 90 µL serum free RPMI media was added to each well. Following this, 10 µL of 10× PFNA or PCB153 stocks (in methanol) was added in triplicate to the highest concentration well then serially diluted (2-fold, 12 point) at 12 concentrations ranging from 0.08 µM - 170 µM for PCB153 and 0.5 µM to

800 µM for PFNA. The highest concentration of methanol (10%) in PBS was used as a negative control and 0.1% Triton-X was used as a positive control. Cells were exposed for either 24 or 48 h, after which the exposure was ended by replacement of the treatment media with fresh untreated media. To measure changes in cell viability, 20 µL of 0.15 mg/mL sterile filtered resazurin dye made up in PBS was added to each well.

After 24 h of incubation with resazurin dye at culture conditions, plates were read at λex

= 544 nm and λem = 590 nm using a FLUOstar Omega plate reader (BMG Labtech) and the relative cell viability (compared to controls) was calculated.

3.4.4. 24 and 48h exposure bioassays for proteomic analysis

GT12s-p cells were seeded at a density of 3.5 × 104 cells/cm2 in 25 cm2 tissue culture flasks in culture medium and incubated at 30°C and 5% CO2 conditions overnight to ensure adherence of the cells to the flask surface. The following day, cells were washed twice with PBS before 3 mL of serum free media with either 0.01% methanol

(negative/solvent control), or 1, 10 or 100 µg/L PFNA or PCB153 was added. Each treatment was performed in triplicate and for both 24 and 48 h exposure durations. After the relevant incubation time, cells were examined for morphological changes under a microscope before removing the media and washing the cells twice with PBS.

3.4.5. Protein extraction

Cells were scraped from the surface of the culture flasks with a cell scraper in 2 mL

PBS and transferred into Eppendorf tubes before being pelleted by centrifugation at

2,500 ×g for 10 min at 4°C. The supernatant was discarded, and pellets were resuspended in 100 µL of Mammalian Protein Extraction Reagent (M-PER) (Thermo),

86 spiked with 1× protease inhibitor cocktail (Gibco), for lysis and extraction of total cell protein content. After 10 min incubation with gentle shaking, tubes were centrifuged at

14,000 ×g for 15 min at 4°C to pellet cellular debris. The supernatant containing the protein solution was transferred to a clean Eppendorf tube and the cell debris pellet discarded. Proteins were extracted from the solution by 15% v/v trichloroacetic acid

(TCA) precipitation, performed overnight at -20°C. Following precipitation, pellets were separated from the supernatant by centrifugation at 10,000 ×g, 4°C for 30 min.

The supernatant was discarded, and pellets were washed with 90% ice cold acetone before being immediately resuspended in 200 µl of 7 M urea with 4 M thiourea solution and sonicated at 20°C for 30 min. After total pellet solubilisation, pH was increased to ~

8.5 by addition of 40 mM tris pH 9.5 and urea concentration decreased to 2 M by the addition of ultrapure water to each sample.

3.4.6. Protein quantification and digestion

A modified Bradford Lowry assay (Bradford 1976) was used to quantify total protein content in each sample. Briefly, 5.5 µL of each sample or 5.5 µL bovine serum albumin

(BSA) were added to 250 µL of Coomassie Brilliant Blue dye (Sigma-Aldrich) in a 96 well clear flat-bottomed plate, in duplicate. BSA was added at eight concentrations ranging from 2 mg/mL to 0.125 mg/mL in order to produce a standard curve. Plates were incubated for 1 h in the dark at room temperature before reading absorbance at 595 nm in a microplate reader (BMG Flurostar). The BSA standard curve was used to calculate total protein concentration of each sample in mg/mL. Duplicates of each sample containing 5 µg of protein were digested using ThermoFisher in-solution tryptic digestion kit. In short, proteins were reduced by incubation at 95°C for 5 min in the presence of 50 mM ammonium bicarbonate and 10 mM DTT. Alkylation was performed by adding 15 mM of iodoacetamide (IAA) to samples and incubating for 20 min in the dark. Lastly, proteins were digested in the presence of approximately 5 ng/µL

87 trypsin (1:50 enzyme:protein ) at 37°C for 3 h with addition of more trypsin (increasing the total concentration to 10 ng/µL) at 30°C overnight. Samples were filtered using C18

Ziptips (Merck) with 5 µg duplicates pooled and eluted in 60% acetonitrile (ACN) and

0.1% trifluoracetic acid (TFA). Solvents were evaporated by centrifuging the samples in a vacuum concentrator (Eppendorf) at 45°C for 1 h before resuspending the dried peptides in 100 µL of 0.1% formic acid.

3.4.7. LC-MS/MS

Samples were fractionated using reversed-phase chromatography on a Shimadzu

Prominence nanoLC system. Using a flow rate of 30 µL/min, samples were desalted on an Agilent C18 trap (0.3 × 5 mm, 5 µm) for 3 min, followed by separation on a Vydac

Everest C18 (300 Å, 5 µm, 150 mm × 150 µm) column at a flow rate of 1 µL/min. A gradient of 10-60% buffer B over 45 min where buffer A = 1% ACN / 0.1% FA and buffer B = 80% ACN / 0.1% FA was used to separate peptides. Eluted peptides were directly analysed on a TripleTof 5600 instrument (ABSciex) using a Nanospray III interface. Gas and voltage settings were adjusted as required. Mass spectrometric time of flight (MS TOF) scan across 350-1800 m/z was performed for 0.5 sec followed by information dependent acquisition of up to 20 peptides with intensity greater than 100 counts, across 40-1800 m/z (0.05 sec per spectra) using collision energy (CE) of 40 +/-

15 V. For SWATH analyses, MS scans across 350-1800 m/z were performed (0.5 sec), followed by high sensitivity data independent acquisition (DIA) mode using 26 m/z isolation windows for 0.1 sec, across 400-1250 m/z. CE values for SWATH samples were automatically assigned by Analyst software based on m/z mass windows.

3.4.8. Data analysis

To analyse the cytotoxicity of PCB153 and PFNA, exposure concentration (log) for each chemical was plotted against cell viability and the concentration-effect relationship and EC10 values were analysed in GraphPad Prism using the hill slope equation.

88 For the purpose of obtaining protein identifications in order to match identification data with spectra obtained from DIA MS, an ion library was generated using data dependent acquisition (DDA) MS files obtained from injections of combined protein digests from all treatments (Controls, PCB153, and PFNA) at 24 h and repeated for combined 48 h samples. The ion library was searched in ProteinPilot v5.01 against the National Center for Biotechnology Information (NCBI) C. mydas proteome (derived from the full draft genome) (accessed on the 2nd of February 2017). Search parameters included digestion with trypsin, cys alkylation set to iodoacetamide, and thorough search with false discovery rate (FDR) analysis providing counts of total identified proteins and peptides for all treatments combined at each exposure interval within the 1, 5 and 10% global

FDR brackets. The ScieX SWATH 2.0 data processing software calculates the FDR using both chromatograms generated from fragment ion data as well as spectral components to calculate FDR. Multiple chromatographic (peak height, retention time, consistency of peaks) and spectral (mass accuracy, contiguous ion series) variables are used to make up the score. Then a pseudo-reverse sequence is generated for each peptide and analysed in the same manner. Following this, the target and decoy sequences are scored as above and ranked based on this score. The FDR is computed by

2 X the number of decoys in the list and an FDR assigned to each peptide. The peak heights from the chromatographic outputs generated from fragment ion data are used for relative quantification of peptides between samples, actual quantification is not performed.

In order to obtain quantitative information, protein data as obtained by DIA (or

SWATH-MS) was matched against the ion library using PeakView (ABS Sciex) software. The resulting data set was formatted for further analysis using the python script developed by Kerr et al., 2018) before further analysis in R. To determine

89 changes in global protein expression related to contaminant exposure, the abundance of each discrete protein detected in our samples was compared between the relevant control and treatment by performing a HighNil comparison in the MSStats package

(Choi et al. 2014) in R version 2.2.7. The fold change (positive or negative) in protein abundance relative to the control, along with the significance and standard error of the mean (SEM) of this change, was reported. Results of the HighNiL comparison were filtered and all proteins that were not significantly dysregulated from the control

(p>0.0001) were excluded from further analysis.

In order to compare the overall differential expression between each treatment, the number of significantly (p<0.0001) differentially expressed proteins were counted for each replicate of each treatment and the mean and SEM calculated in GraphPad Prism.

Full Factorial Generalised Linear Models (GLMs) were used to analyse for differences in the overall number of differentially expressed proteins for each chemical individually, using JMP 9.0.1 (SAS Institute Inc.). The GLMs followed a Log-Poisson distribution and included an Over-dispersion Test to account for the limited number of samples and were supported by comparing second order Akaike Information Criterion

(AICc) scores for alternative model parameters. The coefficient of variation of the number of differentially expressed proteins at each treatment and exposure time was calculated by dividing the standard deviation by the mean.

The numbers of up and down-regulated proteins within each treatment were counted and their proportions calculated by dividing the number of up- or down-regulated proteins by the total number of significantly differentially expressed proteins within that treatment. These proportion values were then arcsine square root transformed. For further analysis, treatment groups were combined since no significant difference was observed between exposure concentrations for PCB153 or PFNA (see Section 3.5.2).

Two-way ANOVAs were performed for the contaminants separately, to explore the

90 influence of directionality (i.e., up vs. down regulation) in relation exposure duration

(i.e., 24 vs. 48 h) and whether there was an interaction between these factors.

With the aim of understanding if altered protein expression was a specific response elicited by each compound or a shared or general toxic response, the data was filtered for proteins that were exclusively significantly (p<0.0001) differentially expressed at any exposure concentration of one chemical. The proportion of exclusively expressed proteins for each contaminant within each exposure time was calculated and presented as a percentage.

To compare the expression of known in vivo markers with our in vitro results, the names of several known biomarkers (and their isoforms) of contaminant exposure in sea turtles, namely cytochrome P450 (CYP)1A, CYP2K1, glutathione-S-transferases, superoxide dismutases (SOD), and catalase (CAT) were searched within the entire data set. If these markers were present amongst the significantly differentially expressed proteins, the fold change for each compound and treatment along with SEM was plotted against treatment in GraphPad prism. Note that the purpose here was not statistical comparison, but visualisations, to facilitate discussion of how in vitro proteomic responses align, qualitatively, with expectations in wild-caught turtles.

3.5 Results and Discussion

3.5.1 Cytotoxicity results and cell morphology

Concentration-effect curves for both contaminants were comparable following 24 and

48 h exposure but, as expected, cytotoxicity was slightly higher following longer exposure (Figure 3.1). Cytotoxicity was also observed to increase with exposure concentration for both PCB153 and PFNA (Figure 3.1), however, an EC10 could only be calculated for PFNA (18.42 µM) at 24 h exposure as cell viability did not drop below

90% for PCB153 (Figure 3.1a). The cytotoxicity results contributed to our decision to

91 use a maximum exposure concentration of 100 µg/L (0.27 µM PCB153; 0.21 µM

PFNA) for both compounds, since this falls well below (i.e., 50 to 80 × less than) calculated EC10 values for PCB153 (47.70 µM 48 h) and PFNA (18.43 µM 24 h; 11.60

µM 48 h) in GT12s-p cells (Table S3.1). While we didn’t reach a high enough concentration to accurately determine EC50, it is interesting to note that a recent study from our lab found an EC50 of 480 µM of PCB153 after 24-h exposure of healthy primary green sea turtle skin cells (Finlayson et al. 2019a).

Figure 3.1. Percent survival of G12s-p (C.mydas skin) cells exposed to A; 0.08 µM –

170 µM PCB 153, and B; 0.5 µM – 800 µM PFNA, over 24 and 48 hours. Error bars represent standard error of the mean (SEM) (n = 6).

92 Examination of the cells under a microscope following exposures for proteomic analysis revealed no apparent morphological changes for most treatments, with all cells appearing qualitatively similar to the controls before and after exposure (Figure S3.1).

The only exception was 48 h exposure to 100 µg/L for PCB153, which appeared to cause rounding of the cells (Figure S3.1). It is unclear why this occurred since the exposure concentration was well below cytotoxic levels (Table S3.1), however, it is interesting to note that 100 µg/L PCB153 exposure caused similar morphological changes in human breast cancer cells (Lasserre et al., 2009) as well as alterations to the cell cycle (Venkatesha et al. 2010). This observed morphological change may be reflected in the protein expression at this concentration and warrants further investigation. We are not aware of any studies attempting to correlate morphological changes with effects on protein expression in C. mydas cells, so we can only speculate at this time.

3.5.2 Effect of concentration, exposure time and contaminant on total number of dysregulated proteins

We chose 24 and 48 h exposures since these are standard timeframes commonly used in cellular assays of this nature (Lasserre et al. 2009, Roland et al. 2013, Lin et al. 2014,

Vuong et al. 2016). We rationalised that the results would be useful for comparison, and that the methodology would be feasible and allow us to explore how small differences in exposure duration might influence proteomic responses. We note that longer exposures may be of interest, however it is difficult to maintain primary cells for longer than 48 h without supplementing FBS and hence a maximum exposure time of 48 h was considered appropriate here.

There were significantly fewer differentially expressed proteins observed at 48 compared to 24 h for both PCB153 (p<0.0001) and PFNA (p<0.0001). However, there was no difference related to exposure concentration (p = 0.829) nor was there an

93 interaction between time and concentration (p = 0.382) for PCB153 (Figure 3.2a). We observed a slight concentration-dependent increase in the number of differentially expressed proteins in cells exposed to PFNA at 24 h, and an apparent concentration- dependent decrease at 48 h. However, the GLM similarly revealed no significant difference between concentrations (p = 0.475) and no interaction (p = 0.190) for PFNA

(Figure 3.2b). While not significant, the number of differentially expressed proteins appeared to correlate with changes in chemical concentration for PFNA, and it is interesting that the correlation was positive at 24 h and negative at 48 h (Figure 3.2b).

The difference related to time corresponded with higher variance (coefficient of variation; CoV) between replicates after 48 h compared to 24 h exposure (Table 3.1).

It must be noted that it is possible that the concentration range used here was too high to reveal a concentration dependent change, and therefore an experimental exposure with a lower concentration would be useful in confirming the lack of observable concentration- dependent protein expression changes. Furthermore, the compounds used for experimental exposure have been shown to strongly adhere to laboratory plastics. While care was taken to avoid the use of plastic where possible, the assays were conducted in plastic cell culture flasks. It is possible that adherence of some of the compound to the flask would have lowered the exposure concentration from the intended concentration.

This may have also been a factor influencing the difference in result between 24 and 48 hours. Future studies of this nature would benefit from measurement of the actual exposure dose (chemical present in media and cells) so as to assist in interpreting the results.

94 1 μg/L a) 500 PCB 153 10 μg/L 100μg/L 400 a b 300

200

100

0 24 48

b) 500 PFNA a 400 b

300 No. differentially expressed proteins 200

100

0 24 48

Exposure duration (h)

Figure 3.2. The total number of proteins differentially expressed (p-value <0.0001) in cells exposed to a) PCB 153 and b) PFNA compared to controls. Error bars represent standard error of the mean (SEM) (n=3).

95 Table 3.1. Comparison of the coefficient of variation (CoV, calculated as standard deviation / mean) between 24 and 48 h exposure of the number of overall proteins significantly (p<0.0001) differentially expressed at each treatment derived from the data presented in Figure 3.2.

Concentration n Coefficient of variation Chemical (%) 24 h 48 h PCB 153 1 µg/L 3 0.07 0.37 PCB 153 10 µg/L 3 0.20 0.55 PCB 153 100 µg/L 3 0.04 0.28 PFNA 1 µg/L 3 0.20 0.88 PFNA 10 µg/L 3 0.05 0.46 PFNA 100 µg/L 3 0.08 0.53

There are several possible explanations for an overall reduction in the number of differentially expressed proteins, and corresponding increased CoV following longer exposure duration. The first being that cells were starved of FBS for the duration of the exposures and media was not replenished. While this is standard practice to ensure consistent exposure conditions with in vitro experiments, after 48 h of serum starvation it is possible that the cells may have started entering a different cell cycle (Khammanit et al. 2008). Reaching confluence in the flask may also have caused the cells to enter a different cell cycle (Weiser et al. 1990), or to de-differentiate, as primary cells have been shown to do in culture (Craven et al., 2006), which might likewise explain this trend. Indeed, it has previously been shown that gene and protein expression vary at different stages of the cell cycle (Dolatabadi et al. 2017). Alternatively, it is possible that the cells became better adapted to the exposure or initiated compensatory mechanisms at 48 h aimed at re-establishing cellular homeostasis. In this case, a reduction in the number of differentially expressed proteins would be expected due to digestion of proteins no longer being expressed, by naturally present proteases

(Goldberg and St. John 1976). Finally, it is also conceivable that cellular metabolism

96 was reduced after 48 h of serum starvation, which would similarly be reflected by a reduction in overall protein expression and partial degradation by natural proteases.

Regardless of the specific cause, we hypothesise that the observed increase in CoV reflects a transitionary phase where some cells were still exhibiting the full extent of differential protein expression at 48 h, whereas others were responding through one or more of the aforementioned processes. This is an important outcome, since it suggests that fundamental choices in experimental design can introduce considerable variance in protein expression (Simmons et al. 2015). There are currently no specific guidelines indicating the most appropriate duration for in vitro exposures aimed at measuring proteomic responses, but we suggest that such research is an important validation step necessary for future cell-based proteomic research.

3.5.3 Practical aspects of the methodology used for cellular bioassay proteomic analysis

The use of in-solution digestion, as opposed to gel-based methods commonly used in non-targeted studies, proved advantageous in that over 1000 proteins were identified within the global 1% FDR bracket for both exposure lengths (Table 3.2). This rate of identification is considerably greater than that of similar in vitro toxicological studies that prepared samples using 2- dimensional differential gel electrophoresis (2D-DIGE).

When preparing samples with 2-DIGE, Vuong et al., (2016) identified 180 proteins,

Roland et al. (2013) identified 164 proteins, and Lasserre et al. (2009) only observed 66 different dysregulated proteins (total identified proteins not reported). The omission of prefractionation techniques such as gel electrophoresis and the performance of direct in- solution digest greatly increases sensitivity allowing for the detection of lower abundance proteins (McDonald et al., 2002), as evident in this study with 5 × more proteins identified than in similar studies that employed gel-based methods.

97 This indicates considerable improvement from gel-based digestion in terms of the diagnostic power and ability to unravel mechanistic pathways, but there are also practical advantages for experimental design. Specifically, previous studies needed to use 75 cm2 or larger tissue culture flasks to obtain sufficient material, which physically limits the number of concentrations that can be included in one experiment. Compared to gel-based sample separation, in-solution digested samples require less material for mass-spectrometry proteomic analysis, facilitating the use of smaller 25 cm2 tissue culture flasks and thus a greater number of replicates and treatments can be tested. This fundamental improvement in experimental design, and the resultant greater number of positively identified proteins, supports the assertion that in-solution digestion is superior for obtaining a comprehensive overview of protein expression in complex samples

(Bunai and Yamane 2005).

Table 3.2. The protein identification yields and number of distinct peptides within each false discovery rate (FDR) threshold for all treatments (all concentrations of PCB153 and PFNA) combined at 24 and then 48 h exposure derived from the Protein Pilot search conducted on the IDA ion library.

Global 24 h 48 h FDR Number of Number Number of Number threshold identified of distinct identified of distinct proteins peptides proteins peptides 1% 1088 6659 1124 7312 5% 1191 7661 1233 8394 10% 1223 8490 1233 9230

3.5.4 Effect of treatments on up and down regulation and compound-specific protein dysregulation

There was no significant difference in the proportion of proteins that were up- versus down-regulated following 24 or 48 h exposure, for either PCB153 (Figure 3.3a; two- way ANOVA, p<0.05) or PFNA (Figure 3.3b; two-way ANOVA, p<0.05). We did

98 observe an apparent reduction in the proportion of down-regulated proteins in cells exposed for 48 h (most notably for PFNA) that is consistent with the overall lower number of differentially expressed proteins identified after this timeframe, but the higher variance following longer exposure resulted in this being non-significant. These results are in contrast to those reported by Lasserre et al. (2009), who showed dramatic differences in the percent of up- and down-regulated proteins of in vitro exposed cells with 88% of dysregulated proteins down-regulated for atrazine treatment and 75% of dysregulated proteins up-regulated as a result of PCB153 exposure. A similar, more distinct difference in the proportion of up- vs down-regulated proteins was also observed by Vuong et al., (2016). These differences may be due a number of factors, such as cell type, since Lassere (et al. 2009) used immortal human cells and Vuong (et al., 2016) used human lung cells while our study used GT12s-p primary turtle cells.

Furthermore, the different digestion techniques (in-gel vs in-solution) or other minor variations in sample handling could also play a part in this disparity (Timms et al.,

2007). In summary, these contrasting results highlight a need for exploring possible sources of such differences, particularly including the distinction between biological and technical sources of variation in omics research.

99 a) PCB153 Upregulated 1.5 Downregulated

1.0

0.5

0.0 24 48

b) 1.5 PFNA

1.0 Proportion of up and down regulated proteins

0.5

0.0 24 48

Exposure duration (h) Figure 3.3 Proportion (calculated from the total number of differentially expressed proteins within treatments) of proteins that were up-regulated and down-regulated in cells exposed to a) PCB 153 or b) PFNA. Error bars represent standard error of the mean (SEM) (n = 3).

Our search for chemical-specific protein dysregulation revealed just 11.13% and

12.13% of differentially expressed proteins unique to PCB153 and PFNA, respectively, after 24 h exposure. This could be taken as evidence that most (~90%) of the changes in protein expression at 24 h for both compounds were general stress responses, or alternatively that the two compounds exert their effects through similar mechanisms and/or pathways (Table S3.2). These groups of compound specific response proteins would be of particular interest when attempting to identify biomarker targets of exposure as the more specific a measured response is to a marker as unambiguity is one of the most important traits of a good biomarker (Rifai et al. 2006).

100 Interestingly, approximately double the proportion of dysregulated proteins was unique to each compound after 48 h exposure (PCB153 25.81%; PFNA 24.73%), which is consistent with our hypothesis that the earlier responses at 24 h reflect general stress responses (Table S3.2). To clarify, we suggest that the reduction in compound-specific dysregulation of proteins at 48 h compared to 24 h reflects the cells ability to compensate for general stress responses to return to equilibrium. As such, later time points may prove more informative when unravelling the mechanistic underpinnings of toxicological responses with proteomics.

Considering the marked increase in variation following a marginally longer exposure duration (i.e., 48 h compared to 24 h) and potential for this to mask and confound proteomic responses, there is a clear need for standardisation of cell-based proteomics methods to achieve optimal reproducibility and validity from these studies (Sauer et al.

2017). Reproducibility of omics studies appears to be somewhat elusive even when biological variation is minimised by using cell lines and in vitro exposure, but it has been argued that a different definition of reproducibility may be required for proteomics. Specifically, it has been suggested that rather than focussing on reproducibility of changes to individual proteins, a focus on functional changes relevant for specific pathways, such as oxidative stress, is more appropriate (Ning and Lo 2010).

3.5.5 In vitro expression of in vivo biomarkers expressed by cells exposed to PCB153 for 24 h.

Two potential biomarkers of PCB exposure in C. mydas, superoxide dismutase (SOD) and glutathione S-transferases (GSTs) (Labrada-Martagón et al. 2011), were significantly dysregulated in cells exposed to PCB153 compared to control cells (Figure

3.4) after 24 h exposure. While the detection of in vivo biomarkers is a promising indicator that our results point to markers at the whole organism level, it is important to carefully examine how our results compare to those of in vivo studies. Blood

101 concentrations of SOD and GSTs have been observed to increase in correlation with

PCB exposure in sea turtles (Labrada-Martagón et al. 2011). However, our in vitro results showed the opposite effect with a marked down-regulation of SODs and GSTs by cells as a result of PCB153 exposure (Figure 3.4). This may be as a result of our targeted focus on intracellular and membrane-bound proteins and the exclusion of excreted proteins in the media. Furthermore, the form of SOD detected here is the cytosolic copper-zinc (Cu-Zn) containing SOD, which is an important enzyme for eliminating superoxide radicals from within the cell (Hu et al. 2005). It is then logical to assume that measures of SODs and GSTs in extracellular fluid such as blood would only include excreted and not intracellular enzymes. If these enzymes are excreted from the cell as a response to contaminant exposure then it is likely that in our study, both of the markers were excreted into the media in response to PCB exposure resulting in the decrease in cytosolic concentrations observed herein. Alternatively, the observation of a reduction in these proteins resulting from chemical exposure may simply indicate overt toxicity rather than a specific toxic response such as oxidative stress. It would be valuable for future experiments targeting these markers to measure both intra- and extra-cellular proteins to confirm is the movement of these enzymes in response to exposure.

102 a) 0.0 c) 0.0 * * Superoxide dismutase -0.5 -0.5 [Cu-Zn] isoform X2

-1.0 -1.0 Fold change Fold change -1.5 -1.5

-2.0 -2.0 1 10 100 1 10 100

PFNA concentration (µg/L)

b) 0.0 Glutathione S-transferase

-0.5

-1.0 Fold change

-1.5 1 10 100

PCB153 concentration (µg/L)

Figure 3.4. Fold change of a & c) Cu-Zn superoxide dismutase (SOD) and b) glutathione S-transferase compared to the control samples, in cells exposed three concentrations of PCB153 and PFNA for 24 h. Error bars represent standard error of the mean (SEM) (n = 3). Data for GSTs from PFNA exposure not shown as it was not significantly dysregulated compared to controls at any of the exposure concentrations.

*SODs were not dysregulated at these concentrations.

Another potential reason for this conflicting result is that the tissue type used in our experiment (skin) is not a great source of these enzymes within the whole organism and therefore the results do not fully reflect in vivo effects. Valdiva (et al. 2007) reported varying baseline levels of SOD activity between sea turtle tissue types with liver producing the most SOD activity of several analysed tissues. Here we chose skin tissue as it is the most accessible type of tissue to obtain from wildlife and therefore the ideal tissue type to use when dealing with protected species. In saying that, future cell-based

103 experiments on non-targeted protein expression could examine differences in protein- level responses between cells derived from various tissue types.

The observation of this contrast between in vivo and in vitro data as seen here has implications for future studies applying proteomics techniques with cellular bioassays.

Our perception of measured effects must be carefully interpreted with consideration of the context and the applied methodology, especially variations in tissue-specific responses. Furthermore, these results point to the importance of examining overall affected pathways (e.g. oxidative stress) rather than focusing in on changes in single proteins. Additionally, as the proteome of C. mydas used for our search here was largely based on genome sequence translation rather than intact protein characterisation, all of the isoforms of these enzymes and their functions have not been fully elucidated in this species. Therefore, there are limitations in the extent to which the results can be interpreted. However, as proteome data is constantly being updated and refined as a result of molecular studies, the ability of future investigations in this field to interpret the data will deepen.

3.6 Conclusion

In summary, this study presents for the first-time effects of contaminant exposure on the global protein expression of C. mydas cells at several concentrations, times and to different toxicants. Exposure time had the strongest influence on overall protein expression with 24 h giving the most consistent results This research advances both the knowledge specific to the field of green sea turtle toxicology while simultaneously providing a framework for other wildlife toxicologists to use in investigating effects of contaminant exposure on the target species.

104 These data indicate that a wealth of information can be obtained from non-targeted protein expression analysis; however, technical sources of variation, such as length of exposure time, can greatly influence the results and possibly mask biological responses.

This supports the sentiment detailed in several recent reviews that despite the rapid advancement of proteomics technologies in the last decade, the application of proteomics to toxicology has lagged. This lag is mainly attributed to the difficulty of accurately identifying the biological response to exposure amongst the high level of

‘noise’ from technical and biological variation that is visible at the molecular level.

Therefore it is vital that before toxicological interpretation of data resulting from such experiments is undertaken that the sources of variation are well understood.

Our results suggest that an exposure duration of 24 h provides greater insights into potential effects and cellular responses, less impacted by compensatory mechanisms and cell culture artefacts. Future studies examining in vitro global proteomic responses to exposure should aim to confirm reproducibility of their methods and distinguish the influence of experimental parameters on the observed changes in protein expression from actual biological responses. Observed responses should be carefully examined from a functional and pathway perspective rather than only focusing changes of single proteins. Furthermore, it would be ideal if in vitro proteomics toxicology methods are standardised and a best practise established, in view of realising the value proteomics technologies in an ecotoxicological context.

3.7 References

Ankley, G T., Bennett, R S., Erickson, R J., Hoff, D J., Hornung, M W., Johnson, R D., Mount, D R., Nichols, J W., Russom, C L., Schmieder, P K. and Serrrano, J A. 2010. Adverse outcome pathways: a conceptual framework to support ecotoxicology research and risk assessment. Environ Toxicol Chem 29:730-741.

105 Baccarelli, A. and Bollati, V. 2009. Epigenetics and environmental chemicals. Curr Opin Pediatr 21:243-251. Bengtson Nash, S M., Waugh, C A. and Schlabach, M. 2013. Metabolic concentration of lipid soluble organochlorine burdens in the blubber of Southern Hemisphere humpback whales through migration and fasting. Environ Sci Technol 47:9404- 9413. Benninghoff, A D. 2006. Toxicoproteomics--The Next Step in the Evolution of Environmental Biomarkers? Toxicol Sci 95:1-4. Bradford, M M. 1976. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 72:248-254. Bunai, K. and Yamane, K. 2005. Effectiveness and limitation of two-dimensional gel electrophoresis in bacterial membrane protein proteomics and perspectives. J Chromatogr B Analyt Technol Biomed Life Sci 815:227-236. Chaloupka, M., Limpus, C. and Miller, J. 2004. Green turtle somatic growth dynamics in a spatially disjunct Great Barrier Reef metapopulation. Coral Reefs 23:325- 335. Chaousis, S., Leusch, F D L. and van de Merwe, J P. 2018. Charting a path towards non-destructive biomarkers in threatened wildlife: A systematic quantitative literature review. Environ Pollut 234:59-70. Choi, M., Chang, C Y., Clough, T., Broudy, D., Killeen, T., MacLean, B. and Vitek, O. 2014. MSstats: an R package for statistical analysis of quantitative mass spectrometry-based proteomic experiments. Bioinformatics 30:25242526. Craven, R A., Stanley, A J., Hanrahan, S., Dods, J., Unwin, R., Totty, N., Harnden, P., Eardley, I., Selby, P J. and Banks, R E. 2006. Proteomic analysis of primary cell lines identifies protein changes present in renal cell carcinoma. Proteomics 6:2853-2864. Cseh, T., Sanschagrin, S., Hawari, J. and Samson, R. 1989. Adsorption-desorption characteristics of polychlorinated biphenyls on various polymers commonly found in laboratories. Appl. Environ. Microbiol., 55(12), 3150-315 Denslow, N D., Griffitt, R J. and Martyniuk, C J. 2012. Advancing the Omics in aquatic toxicology: SETAC North America 31st Annual Meeting. Ecotoxicol Environ Saf 76:1-2.

106 Dolatabadi, S., Candia, J., Akrap, N., Vannas, C., Tesan Tomic, T., Losert, W., Landberg, G., Aman, P. and Stahlberg, A. 2017. Cell Cycle and Cell Size Dependent Gene Expression Reveals Distinct Subpopulations at Single-Cell Level. Front Genet 8:1. Fent, K. and J P., Sumpter. 2011. Progress and promises in toxicogenomics in aquatic toxicology: is technical innovation driving scientific innovation? Aquat Toxicol 105:25-39. Finlayson, K A., Leusch, F D L. and van de Merwe, J P. 2019a. Cytotoxicity of organic and inorganic compounds to primary cell cultures established from internal tissues of Chelonia mydas. Sci Total Environ 664:958-967. Finlayson, K A., Leusch, F D L. and van de Merwe, J P. 2019b. Primary green turtle (Chelonia mydas) skin fibroblasts as an in vitro model for assessing genotoxicity and oxidative stress. Aquat Toxicol 207:13-18. Finlayson, K A., Leusch F. D. L., and van de Merwe, J. P. 2016. The current state and future directions of marine turtle toxicology research. Environ Int 94:113-123. Fossi, M C., Casini, S. and Marsili, L. 2007. Potential toxicological hazard due to endocrine-disrupting chemicals on Mediterranean top predators: State of art, gender differences and methodological tools. Environ Res 104:174-182. Gillet, L C., Navarro, P., Tate, S., Rost, H., Selevsek, N., Reiter, L., Bonner, R. and Aebersold, R. 2012. Targeted data extraction of the MS/MS spectra generated by data-independent acquisition: a new concept for consistent and accurate proteome analysis. Mol Cell Proteomics 11:O111 016717. Goldberg, A L. and St. John, A. 1976. Intracellular protein degradation in mammalian and bacterial cells: Part 2. Annu Rev Biochem 45:747-804. McDonald, W H. and Yates, J. R. 2002. Shotgun proteomics and biomarker discovery. Dis Markers, 18: 99-105. Hamann, M., Limpus, C. and Whittier, J. 2002. Patterns of lipid storage and mobilisation in the female green sea turtle (Chelonia mydas). J Comp Physiol B, 172:485-493. Hu, Y., Rosen, D G., Zhou, Y., Feng, L., Yang, G., Liu, J. and Huang, P. 2005. Mitochondrial manganese-superoxide dismutase expression in ovarian cancer: role in cell proliferation and response to oxidative stress. J Biol Chem 280:39485-39492.

107 Johnston, E L., Mayer-Pinto, M., Crowe, T P. and Frid, C. 2015. Review: Chemical contaminant effects on marine ecosystem functioning. J Applied Ecology 52:140-149. Keller, J M. 2003. Occurence and effects of organochlorine contaminants in sea turtles. Duke University, North Carolina, USA. Keller, J M., Balazs, G H., Nilsen, F., Rice, M., Work, T M. and Jensen, B A. 2014. Investigating the potential role of persistent organic pollutants in Hawaiian green sea turtle fibropapillomatosis. Environ Sci Technol 48:7807-7816. Keller, J M., Kannan, K., Taniyasu, S., Yamashita, N., Day, R D., Arendt, M D., Segars, A L. and Kucklick, J R. 2005. Perfluorinated compounds in the plasma of loggerhead and Kemp's ridley sea turtles from the southeastern coast of the United States. Environ Sci Technol 39:9101-9108. Keller, J M., McClellan-Green, P D., Kucklick, J R., Keil, D E. and Peden-Adams, M M. 2006. Effects of organochlorine contaminants on loggerhead sea turtle immunity: comparison of a correlative field study and in vitro exposure experiments. Environ Health Persp 114:70-76. Kendall, R J. 2010. Introduction and Overview.in Kendall, R J., Lacher, T E., Cobb, G P., and Cox, S B., editors. Wildlife Toxicology: Emerging Contaminant and Biodiversity Issues CRC Press, USA. Kennedy, S. 2002. The role of proteomics in toxicology: identification of biomarkers of toxicity by protein expression analysis. Biomarkers 7:269-290. Kerr, E D., Phung, T K., Fox, G P., Platz, G J. and Schulz, B L. 2018. The intrinsic and regulated proteomes of barley seeds in response to fungal infection. bioRxiv, 445916. Khammanit, R., Chantakru, S., Kitiyanant, Y. and Saikhun, J. 2008. Effect of serum starvation and chemical inhibitors on cell cycle synchronization of canine dermal fibroblasts. Theriogenology 70:27-34. Kidd, K A., Blanchfield, P J., Mills, K H., Palace, V P., Evans, R E., Lazorchak, J M. and Flick, R W. 2007. Collapse of a fish population after exposure to a synthetic estrogen. Proc Natl Acad Sci USA 104:8897-8901. Komoroske, L M., Lewison, R L., Seminoff, J A., Deheyn, D D. and Dutton, P H. 2011. Pollutants and the health of green sea turtles resident to an urbanized estuary in San Diego, CA. Chemosphere 84:544-552.

108 Labrada-Martagón, V., Rodríguez, P A T., Méndez-Rodríguez, L C. and Zenteno-Savín, T. 2011. Oxidative stress indicators and chemical contaminants in East Pacific green turtles (Chelonia mydas) inhabiting two foraging coastal lagoons in the Baja California peninsula. Comp Biochem Physiol C 154:65-75. Lasserre, J P., Fack, F., Revets, D., Planchon, S., Renaut, J., Hoffmann, L., Gutleb, A C., Muller, C P. and Bohn, T. 2009. Effects of the endocrine disruptors atrazine and PCB 153 on the protein expression of MCF-7 human cells. J Proteome Res 8:5485-5496. Leung, K M. 2018. Joining the dots between omics and environmental management. Integr Environ Assess Manag 14:169-173. Leusch, F D L., de Jager, C., Levi, Y., Lim, R., Puijker, L., Sacher, F., Tremblay, L A., Wilson, V S. and Chapman, H F. 2010. Comparison of five in vitro bioassays to measure estrogenic activity in environmental waters. Environ Sci Technol 44:3853-3860. Limpus, C., J. Miller, C. Paramenter, D. Reimer, McLachlan, N. and Webb, R. 1992. Migration of green (Chelonia mydas) and loggerhead (Caretta caretta) turtles to and from eastern Australian rookeries. Wildlife Res 19:347-357. Limpus, C J. and L. Fien. 2009. A biological review of Australian marine turtles. Environmental Protection Agency. Lin, Y R., Kuo, C J., Lin, H Y., Wu, C J. and Liang, S S. 2014. A proteomics analysis to evaluate cytotoxicity in NRK-52E cells caused by unmodified Nano- Fe(3)O(4). Sci World J 2014:754721. Lunardi, D., Abelli, L., Panti, C., Marsili, L., Fossi, M C. and Mancia, A. 2016. Transcriptomic analysis of bottlenose dolphin (Tursiops truncatus) skin biopsies to assess the effects of emerging contaminants. Mar Environ Res 114:74-79. Lutz, P L., Musick, J A. and Wyneken, J. 2002. The biology of sea turtles. CRC press. Ning, M., and Lo, E H. 2010. Opportunities and challenges in omics. Transl Stroke Res 1:233-237. Noyes, P D., McElwee, M K., Miller, H D., Clark, B W., Van Tiem, L A., Walcott, K C., Erwin, K N. and Levin, E D. 2009. The toxicology of climate change: environmental contaminants in a warming world. Environ Int 35:971-986. Oropesa, A-L., Gravato, C., Sánchez, S. and Soler, F. 2013. Characterization of plasma cholinesterase from the White stork (Ciconia ciconia) and its in vitro inhibition by anticholinesterase pesticides. Ecotox Environ Safe 97:131-138.

109 Patterson, S D. and Aebersold, R H. 2003. Proteomics: the first decade and beyond. Nat Genet 33 Suppl:311-323. Rabilloud, T. and Lescuyer, P. 2015. Proteomics in mechanistic toxicology: history, concepts, achievements, caveats, and potential. Proteomics 15:1051-1074. Rifai, N., Gillette, M A. and Carr, S A. 2006. Protein biomarker discovery and validation: the long and uncertain path to clinical utility. Nat Biotechnol 24:971- 983. Richardson, K L., Gold-Bouchot, G. and Schlenk, D. 2009. The characterization of cytosolic glutathione transferase from four species of sea turtles: loggerhead (Caretta caretta), green (Chelonia mydas), olive ridley (Lepidochelys olivacea), and hawksbill (Eretmochelys imbricata). Comp Biochem Physiol C 150:279- 284. Richardson, K L., Lopez Castro, M., Gardner, S C. and Schlenk, D. 2010. Polychlorinated biphenyls and biotransformation enzymes in three species of sea turtles from the baja california peninsula of mexico. Arch Environ Con Tox 58:183-193. Roland, K., Kestemont, P., Hénuset, L., Pierrard, M A., Raes, M., Dieu, M. and Silvestre, F. 2013. Proteomic responses of peripheral blood mononuclear cells in the European eel (Anguilla anguilla) after perfluorooctane sulfonate exposure. Aquat Toxicol 128–129:43-52. Routti, H., Arukwe, A., Jenssen, B M., Letcher, R J., Nyman, M., Backman, C. and Gabrielsen, G W. 2010. Comparative endocrine disruptive effects of contaminants in ringed seals (Phoca hispida) from Svalbard and the Baltic Sea. Comp Biochem Physiol C 152:306-312. Sauer, U G., Deferme, L., Gribaldo, L., Hackermuller, J., Tralau, T., van Ravenzwaay, B., Yauk, C., Poole, A., Tong, W. and Gant, T W. 2017. The challenge of the application of 'omics technologies in chemicals risk assessment: Background and outlook. Regul Toxicol Pharmacol 91(Suppl 1):S14-S26. Schrattenholz, A., Soskic, V., Schopf, R., Poznanovic, S., Klemm-Manns, M. and Groebe, K. 2012. Protein biomarkers for in vitro testing of toxicology. Mutat Res 746:113-123.

110 Simmons, D B., Benskin, J P., Cosgrove, J R., Duncker, B P., Ekman, D R., Martyniuk, C J. and Sherry, J P. 2015. Omics for aquatic ecotoxicology: control of extraneous variability to enhance the analysis of environmental effects. Environ Toxicol Chem 34:1693-1704. Tabuchi, M., Veldhoen, N., Dangerfield, N., Jeffries, S., Helbing, C C. and Ross, P S. 2006. PCB-related alteration of thyroid hormones and thyroid hormone receptor gene expression in free-ranging harbor seals (Phoca vitulina). Environ Health Persp 114:10241031. Timms, J F., Arslan-Low, E., Gentry-Maharaj, A., Luo, Z., T’Jampens, D., Podust, V N., Ford, J.,Fung, E T., Gammerman, A., Jacobs, I. and Menon, U., 2007. Preanalytic influence of sample handling on SELDI-TOF serum protein profiles. Clin Chem, 53(4):645-656. Tornero, V. and G. Hanke. 2016. Chemical contaminants entering the marine environment from sea-based sources: A review with a focus on European seas. Mar Pollut Bull 112:17-38. Valdivia, P A., Zenteno-Savín, T., Gardner, S C. and Aguirre, A A. 2007. Basic oxidative stress metabolites in eastern Pacific green turtles (Chelonia mydas agassizii). Comp Biochem Physiol C 146:111-117. van de Merwe, J P., Hodge, M., Olszowy, H A., Whittier, J M. and Lee, S Y. 2010a. Using blood samples to estimate persistent organic pollutants and metals in green sea turtles (Chelonia mydas). Mar Pollut Bull 60:579-588. van de Merwe, J P., Hodge, M., Whittier, J M., Ibrahim, K. and Lee, S Y. 2010b. Persistent organic pollutants in the green sea turtle Chelonia mydas: nesting population variation, maternal transfer, and effects on development. Mar Ecol Prog Ser 403:269-278. van de Merwe, J P., Hodge, M., Whittier, J M. and Lee, S Y. 2009. Analysing persistent organic pollutants in eggs, blood and tissue of the green sea turtle (Chelonia mydas) using gas chromatography with tandem mass spectrometry (GC- MS/MS). Anal Bioanal Chem 393:1719-1731. Valavanidis, A., Vlahogianni, T., Dassenakis, M. and Scoullos, M. 2006. Molecular biomarkers of oxidative stress in aquatic organisms in relation to toxic environmental pollutants. Ecotoxicol Environ Saf 64:178-189.

111 Van Loveren, H., Ross, P., Osterhaus, A. and Vos, J. 2000. Contaminant-induced immunosuppression and mass mortalities among harbor seals. Toxicol Lett 112:319-324. Venkatesha, V A., Kalen, A L., Sarsour, E H. and Goswami, P C. 2010. PCB-153 exposure coordinates cell cycle progression and cellular metabolism in human mammary epithelial cells. Toxicol Lett 196:110-116. Vos, J G., Bossart, G., Fournier, M. and O'Shea, T. 2003. Toxicology of marine mammals. CRC Press. Vuong, N Q., Goegan, P., Mohottalage, S., Breznan, D., Ariganello, M., Williams, A., Elisma, F., Karthikeyan, S., Vincent, R. and Kumarathasan, P. 2016. Proteomic changes in human lung epithelial cells (A549) in response to carbon black and titanium dioxide exposures. J Proteomics 149:53-63. Wang, Z., DeWitt, J C., Higgins, C P. and Cousins, I T. 2017. A never-ending story of per- and polyfluoroalkyl substances (PFASs)? Environ Sci Technol 51:2508- 2518. Wieser, R J., Renauer, D., Schäfer, A., Heck, R., Engel, R., Schütz, S. and Oesch, F. 1990.Growth control in mammalian cells by cell-cell contacts. Environ Health Persp, 88:251. Zychowski, G V. and Godard-Codding, C A. 2017. Reptilian exposure to Polycyclic aromatic hydrocarbons and associated effects. Environ Toxicol Chem 36:25-35.

112 Supporting Information for Chapter 3

Table S3.1: Calculated EC10 for cell viability for each compound at 24 and 48 hours of

exposure expressed in µM and the highest exposure concentration of 100 µg/L

expressed in µM for each compound. ‘N/A’ indicates that EC10 could not be calculated

due to an insufficient change in cell viability.

Compound EC10 24 hr (µM) EC10 48 hr Highest exposure (µM) concentration (µM) PCB 153 N/A 47.70 0.554 PFNA 18.43 11.60 0.217

a)

Control 1 before Control 2 before Control 3 before

Control 1 after Control 2 after Control 3 after

b)

PCB153 PCB153 PCB153 100 µg/L 1 before 100 µg/L 2 before 100 µg/L 3 before

PCB153 PCB153 PCB153 100 µg/L 1 after 100 µg/L 2 after 100 µg/L 3 after

Figure S3.1. Images showing (a) no morphological changes to control cells after 48 h A incubation compared to (b) changes seen in cells after 48 h exposure to PCB153 100

µg/L. Before images were taken of cells between 18-24 hours after seeding in 25cm2

C D 113 tissue culture flasks before contaminant exposure and after images were taken immediately after 48 h exposure.

Table S3.2. Table indicating the percent of all proteins significantly differentially expressed (p-value<0.0001) at each exposure time that are dysregulated exclusively within that chemical treatment (and not in the other) after exposure at 24 or 48 h.

Exposure time PCB 153 PFNA 24h 11.13% 12.13% 48h 25.81% 24.73%

114 Chapter 4: Changes in global protein expression of five sea turtle (Chelonia mydas) tissue types exposed to PCB 153, PFNA and phenanthrene indicate potential new biomarkers of exposure

Statement of contribution for Chapter 4

This chapter includes a co-authored article that has been submitted to Ecotoxicology and Environmental Safety for consideration for publication. The bibliographic details of the co-authored paper are:

Stephanie Chaousis, Frederic DL Leusch, Amanda Nouwens, Steven D Melvin and

Jason P van de Merwe (under review) Changes in global protein expression of five sea turtle (Chelonia mydas) tissue types exposed to PCB 153, PFNA and phenanthrene indicate potential new biomarkers of exposure. Ecotox Environ Safe.

My contribution to the paper include:

• the conceptualisation of the design of the experiment,

• undertaking all lab work,

• data analysis, and

• writing of the manuscript.

My co-authors provided input into the design of the study, interpretation of the data and final polishing of the writing.

This chapter is formatted in the style of the journal of publication.

115 (Signed) ______(Date) 24/04/2019

Stephanie Chaousis

(Countersigned) ______(Date) 23/04/2019

Jason van de Merwe

116 4.1 Abstract

Non-targeted protein expression at the cellular level can provide insight into mechanistic effects of contaminants in wildlife and hence new and potentially more accurate biomarkers of exposure and effect. However, this technique has been relatively unexplored in the realm of in vitro biomarker discovery in threatened wildlife despite the vulnerability of this group of animals to adverse sublethal effects of contaminant exposure. Here we examined the usefulness of this method for biomarker discovery by investigating differences in the response of primary cells from five different tissue types of green sea turtles (Chelonia mydas) exposed to three contaminants known to accumulate in this species. Cells derived from C. mydas skin, liver, kidney, ovary and small intestine were exposed to 100 µg/L of either polychlorinated biphenyl 153

(PCB153), perfluorononanoic acid (PFNA) or phenanthrene for 24 h. The global protein expression was then quantitatively evaluated using sequential window acquisition of all theoretical mass spectra (SWATH-MS). Comparison of the molecular response between cell types revealed that while a majority of proteins were mutually expressed in controls, the response to exposure was significantly different between tissue types.

Furthermore, a comparison to known markers of exposure in sea turtles from the literature indicated that in vitro response does reflect current known in vivo response to an extent. However, while it proved promising and yielded novel biomarker candidates worth pursuing, a deeper understanding of the connection between the response at these two levels is needed to confirm the accuracy of in vitro protein expression as a tool for biomarker discovery in wildlife.

4.2 Keywords

In vitro; proteomics; sea turtle; PCBs; PAHs, phenanthrene

117 4.3 Introduction

Proteomics in ecotoxicology is an emerging technique that is still being explored for its contribution towards understanding the mechanistic basis of how contaminants affect living organisms (Dowling and Sheehan, 2006; Sanchez et al., 2011). Besides unravelling how contaminants affect threatened species, one of the main aims of many proteomics studies performed on wildlife from an ecotoxicology perspective is the identification of biomarkers of chemical exposure (Kennedy, 2002). Biomarkers are used to detect if organisms have been exposed to specific contaminants and provide information on the health status of these individuals and are typically non-destructive in their measurement (i.e. from blood). Proteins are ideal candidates for biomarkers as they are present in all tissues and blood, and have specific functions (George et al., 2010), which can provide unambiguous information on meaningful physiological effects in the organism (Hoeng et al., 2014). Recent technological advances in the field of proteomics paired with greater accessibility of both expertise and equipment have provided new opportunities to explore the use of proteomics in wildlife toxicology. Examining how organisms respond to contaminant exposure at the protein level is expected to provide improved functional knowledge regarding the adverse effects of chemical exposure in threatened wildlife and a better assessment and management of chemical hazards.

Biomarkers in animals are usually measured either experimentally through controlled laboratory studies, or in wild animals from samples collected in the field (Monserrat et al., 2007). Regardless of the experimental paradigm, collecting samples from animals has certain intrinsic limitations. For example, sampling animals can often be labour intensive and low-throughput making it difficult to obtain sufficient replicates for robust analysis. There are also ethical considerations, particularly when investigating threatened species such as sea turtles, in which case the animals cannot be experimented

118 on in a controlled setting or sacrificed to obtain tissue or organ samples. With these limitations in mind, cell cultures offer many advantages over in vivo sample collection, most notably since they are more ethical and inherently high-throughput (Leusch et al.,

2010). An additional potential advantage where threatened species are concerned is the ability to obtain mechanistic information from in vitro studies using cell cultures derived from the species of interest. In contrast, in vivo studies typically rely on data obtained from related species or extrapolation from existing data, which both increase uncertainty (Ankley et al., 2010). Considering the many perceived benefits of using cellular bioassays in ecotoxicology, research is needed to determine whether meaningful mechanistic information can be obtained by using species-specific cell lines, and to establish the methodology for moving forward with this goal.

Despite the widespread use of biomarkers in ecotoxicology, limitations around the experimentation on threatened species have meant that research in this area has been restricted. The application of biomarkers is generally targeted and focussed on screening for a small number of established endpoints, many of which are associated with oxidative stress or activation of antioxidant defence pathways (Cortes-Gomez et al.,

2018; Dzul-Caamal et al., 2016; Jasinska et al., 2015; Labrada-Martagón et al., 2011;

Salmón et al., 2018). However, with modern analytical capabilities including proteomics facilitating broad-scale non-targeted analysis, it is possible to obtain a comprehensive overview of effects occurring at the molecular level and therefore the potential to discover new pathways and biomarkers that are activated by contaminant exposure (Apraiz et al., 2006; Benninghoff, 2006). Combining non-targeted proteomics analysis with in vitro exposure stands to offer an ethical starting place for discovery of novel biomarkers but has only been investigated in four toxicological studies to date

(Lasserre et al., 2009; Lin et al., 2014; Roland et al., 2013; Vuong et al., 2016).

119 In working towards the meaningful application of cell-based proteomics for toxicity assessment and biomarker discovery, one important question that has so far not been addressed is the optimal tissue to use. Studies on human cells have shown that in vitro protein expression can differ significantly between tissue types (Wang et al., 2019).

Liver cells, for example, might seem like an obvious choice for toxicology studies as the liver is where most detoxification and conjugation occurs (Gu and Manautou, 2012).

However, in aiming to develop a method that can be applied to all wildlife including protected species that cannot be sacrificed for tissue collection, we also need to consider the accessibility of the tissue type from a practical standpoint (e.g. skin can be collected non-destructively, but liver cannot).

Here we explore global protein expression in five green sea turtle (Chelonia mydas) primary cell cultures derived from five different tissues (skin, kidney, liver, ovary and small intestine) after exposure to three environmentally relevant contaminants: polychlorinated biphenyl 153 (PCB153), perfluorononanoic acid (PFNA) and phenanthrene. Changes in global protein expression between the tissue types are compared with each other as well as with protein biomarkers of exposure in sea turtles reported in the literature. Results were interpreted with the overall aim of examining the usefulness of this method, and each tissue type, as a new tool for biomarker discovery in sea turtles.

4.4 Methods

4.4.1. Chemicals Three chemical contaminants known to accumulate in sea turtles were selected for the exposure experiments. 2,2’,4’,4’,5,5’-Hexachlorobiphenyl (PCB 153; 99.9%

Novachem, Heidelberg; product no. C-153N), perfluorononanoic acid (PFNA 97%;

Sigma-Aldrich, product no. 394459) and phenanthrene (99.5% Sigma-Aldrich product no. 695114) have all been detected in blood and tissue of sea turtles (Camacho et al.,

120 2013; Cocci et al., 2018; Keller, 2003; Keller et al., 2005; Keller et al., 2004a; Keller et al., 2012). Stock solutions (10 mg/mL) of each chemical were prepared in methanol and stored at -20ºC until use.

While the production and use of PCBs has been banned or restricted globally since 2001

(Stockholm Convention, 2001) , PCBs are lipophilic and resistant to degradation and thus still commonly detected in wildlife (Ashraf, 2017; Wania, 2003).

Perfluorononanoic acid (PFNA) is one of the more ubiquitous perfluorinated compounds (PFCs). PFCs have been used widely in industrial and personal products and consequently have been dispersed within the natural environment and found to accumulate in wildlife (Houde et al., 2011). PFNA can cause developmental toxicity in zebrafish (Zheng et al., 2011) and has been shown to induce acute immunotoxicity at low doses in mice (Rockwell et al., 2013). Polyaromatic hydrocarbons (PAHs) such as phenanthrene have been detected in sea turtle eggs and blood (Alam and Brim, 2000;

Camacho et al., 2013). A recent study by Cocci et al. (2018) examined gene biomarkers of PAH exposure in sea turtles, and detected 1.17 ng/mL phenanthrene (and up to 80 ng/mL total PAHs) and observed correlation of three gene biomarkers (HSP60, CYP1A and ERa) and indication of DNA damage in correlation with PAH exposure.

4.4.2. Primary cell culture

Primary cells were established in our lab from C. mydas kidney fibroblasts, liver fibroblast/epithelial cells (mixed culture), ovary epithelial cells, skin fibroblasts, and small intestine fibroblasts (hereafter referred to as GT14k-p, GT10l-p, GT12o-p,

GT12s-p, GT12i-p, respectively) using the transplant method described in Finlayson et al. (2019b) were used for this study. The cells were thawed from cryopreserved stocks and cultured in Roswell Park Memorial Institute (RPMI) medium without L-glutamine

(Sigma-Aldrich, product no. R0883) supplemented with 10% Foetal Bovine Serum

(FBS) (Life Technologies, product no. 10099-141), 1% penicillin/streptomycin (Life

121 Technologies, product no. 15140-122) and 1% amphotericin (Sigma-Aldrich, product no. A9528). Cultures were grown in an incubator at 30°C and 5% CO2 in a humidified atmosphere. Cells were cultured for a minimum of two passages before being used in exposure experiments, and cultures did not exceed eight passages.

4.4.3 In vitro exposure to chemical contaminants

Cells were seeded in tissue culture 25 cm2 flasks at a density of 3.5×104 cells per cm2, and incubated at culture conditions overnight to ensure attachment and acclimatisation of cells. The following day, cells were washed twice with phosphate buffered saline

(PBS) before 3 mL of un-supplemented RPMI culture media was added to each flask containing 100 µg/L of either PCB 153, PFNA or phenanthrene or solvent control

(0.01% methanol) in quadruplicate. Cells were exposed to contaminants for 24 h under usual culture conditions as described in section 4.4.2.

4.4.4 Extraction of total intracellular and membrane-bound protein content

At the completion of the 24-h exposure period, cells were washed in the flask two times with PBS before removal of the cells by manual scraping. Cells were then transferred to

2 mL tubes in 1.5 mL of PBS. A final wash step was performed with the samples centrifuged at 2,500 ×g for 10 min at 4°C and the supernatant discarded. To extract total protein content, each sample was resuspended in 100 µL of mammalian protein extraction reagent (M-PER) (ThermoFischer Scientific, product no. 78503) supplemented with protease inhibitor (Sigma-Aldrich, product no. P8340). The samples were shaken on ice for 10 min before being sonicated in ice cold water for 10 min to maximise extraction of intracellular and membrane bound proteins. Samples were then centrifuged at 16,000 ×g for 15 min to remove cellular debris from the supernatant. The supernatant containing the protein extract was then transferred to a fresh tube in preparation for quantification and clean up.

122 4.4.5 Protein quantification and digestion by filter aided sample preparation (FASP)

In order to prepare samples for identification and quantification, the protein content of each sample (mg/mL) was determined using a modified Bradford-Lowry assay. Briefly,

5 µL of each sample or bovine serum albumin (BSA) was added to a 96-well flat bottom plate. Bradford Coomassie reagent (Sigma-Aldrich, product no. B6916) was added to each well (250 µL) and the plate incubated in the dark for 20 min. The plate was then read at 595 nm in a spectrophotometer (FLUOstar Omega, BMG Labtech,

Germany) and a linear regression of the BSA standard curve was produced to calculate sample protein concentration. Filter aided sample preparation (FASP) was performed as per Wisniewski et al. (2009) with modifications in order to remove any lipids, salts and reagent from the sample before in-solution digestion of the proteins into peptides.

Approximately 50 µg of protein from each sample was transferred to the top of a 10 kDa molecular weight cut off filter insert in a 1.5 mL nanosep tube (PALL, product no.

ODO10C33) before being diluted with 400 µL of 8 M urea. The samples were then centrifuged at 14,000 ×g for 25 min at 20°C to allow for the solution to pass through the filter membrane. Then, 200 µL of 8 M urea with 50 mM of ammonium bicarbonate

(ABC) was added on top of the filter and samples centrifuged again. To reduce disulfide bonds, 200 µL of a solution containing 50 mM of dithiothreitol (DTT), 8 M urea and, 50 mM ABC was added to the top of the nanosep filter and samples incubated at 56°C in a water bath for 30 min. Following this, alkylation was performed with the addition of 50 mM iodoacetamide (IAA) to each sample and further incubation at room temperature in the dark for 30 min. DTT was added to samples after incubation to quench excess IAA before the filtrate was passed through the membrane by centrifugation. The filtrate was discarded and 100 µL of 50 mM ABC added to top of tube in addition to 1 µg of trypsin protease (MS grade-Pierce-ThermoFischer, product no. 90057). Samples were incubated at 37°C before the filter containing the digested proteins was transferred to a

123 new collection tube and centrifuged at 14,000 ×g for 25 min at 20°C. The filtrate containing the digested peptides was zip-tipped using a C18 Merck Millipore 10 µL zip tip and the peptides eluted in 60 % acetonitrile (ACN) and 0.1% formic acid (FA).

Solvents were evaporated by centrifuging samples in a vacuum concentrator

(Eppendorf) at 45°C for 1 h before resuspending the dried peptides in 100 µL of 0.1 %

FA.

4.4.6 LC-MS/MS

Reverse phased liquid chromatography (LC) was performed on a Shimadzu Prominence nanoLC system to separate the samples. Samples were desalted on an Agilent C18 trap

(0.3 × 5 mm, 5 µm) using a flow rate of 30 µL/min before separation on a Vydac

Everest C18 (300 A, 5 µm, 150 mm × 150 µm) column with a flow rate of 1µL/min.

Peptides were separated using a gradient of 10-60 % buffer B over 45 min where buffer

A = 1 % ACN / 0.1 % FA and buffer B = 80 % ACN / 0.1 % FA. Peptides were then analysed on a TripleTof 5600 instrument (ABSciex) using a Nanospray III interface with spray voltage set at 2700 V, gas 1 = 10 L / min and curtain gas = 30 L / min. Time of flight mass spectrometry (TOF MS) scans across 350-1800 m/z was performed for

0.5 sec followed by information dependent acquisition of up to 20 peptides with intensity greater than 100 counts, across 40-1800 m/z (0.05 sec per spectra) using collision energy (CE) of 40 +/- 15 V. For sequential window acquisition of all theoretical mass spectra (SWATH-MS) analyses, MS scans across 350-1800 m/z were performed (0.5 sec), followed by high sensitivity data independent acquisition (DIA) mode using 26 m/z isolation windows for 0.1 sec, across 400-1250 m/z. Collision energy (CE) values for SWATH samples were automatically assigned by Analyst software based on m/z mass windows.

124 4.4.7 Search of the literature

To compare our results obtained from in vitro exposure to known markers of in vivo exposure in turtles, the literature was searched. Google Scholar and SciVerse were used to conduct the search with the terms ‘sea turtle’ or ‘marine turtle’, ‘biomarker’ and

‘exposure’ combined with one of the following terms; ‘PCB153’, ‘PFNA’

‘Phenanthrene’, PCBs’, ‘PAHs’ and ‘PFCs’. Only papers written in English from scientific journals were included in the results and only those that had examined protein biomarkers of exposure to the above listed contaminants or classes of chemicals were included in the resulting biomarker database.

4.4.8 Data analysis

To create a database of identified proteins, an ion library was generated using data dependent acquisition (DDA) MS data files searched in ProteinPilot v5.01 against the national center for biotechnology information (NCBI) C. mydas C. mydas proteome

(derived from the full draft genome) (accessed on the 2nd of February 2017). The search results were refined by specification of; the digestive enzyme (trypsin), cys alkylation

(iodoacetamide), and inclusion of a thorough search with false discovery rate (FDR) analysis.The ScieX SWATH 2.0 data processing software calculates the FDR using both chromatograms generated from fragment ion data as well as spectral components to calculate FDR. Multiple chromatographic (peak height, retention time, consistency of peaks) and spectral (mass accuracy, contiguous ion series) variables are used to make up the score. Then a pseudo-reverse sequence is generated for each peptide and analysed in the same manner. Following this, the target and decoy sequences are scored as above and ranked based on this score. The FDR is computed by 2 X the number of decoys in the list and an FDR assigned to each peptide. The peak heights from the chromatographic outputs generated from fragment ion data are used for relative quantification of peptides between samples, actual quantification is not performed.

125 To obtain relative quantitative data for each protein set, data as obtained by DIA or

SWATH-MS was matched against the ion library using PeakView (ABS Sciex) software. The resulting dataset was formatted for further analysis using python script developed by Kerr et al. (2018) before further analysis in R statistical analysis software. To determine changes in global protein expression related to contaminant exposure, the abundance of each discrete protein detected in the samples was compared between the relevant control and treatment by performing a HighNil comparison in the

MSStats package (Choi et al., 2014) in R version 2.2.7. The fold change (positive or negative) of protein abundance relative to the control along with the significance of this change was reported.

To investigate the overall protein profiles of cells derived from different cells (i.e. unexposed cells), a Venn diagram was created using Jvenn (Bardou et al., 2014) comparing lists of identified proteins from each set of unexposed cells. Lists comprised of all identified proteins (but without quantification data) were inserted into the software and the number of shared and unique proteins between each tissue type were noted.

To determine which known biomarkers obtained from the literature search (described in section 4.4.7) were present in our data, a heat map of quantitative data of exposed cells

(protein ID and relative fold change from controls) was produced using Tableau data visualisation software. The heatmap was filtered for protein names that matched those identified in the literature search and all other proteins were excluded from the list. This data was then graphed for each compound of exposure and tissue type to visualise the relative fold change of each biomarker in response to exposure.

With the aim of extracting potential new biomarkers of exposure, we filtered the results of each tissue type and exposure compound for the most highly upregulated and most

126 strongly downregulated proteins and extracted this data to a table. These proteins were then further investigated using gene ontology (GO) annotations to group expressed protein by molecular function and biological processes.

4.5 Results and Discussion

Overall, we found that while the different tissue types largely have the same protein expression profile, the response to contaminant exposure at the protein level was significantly different between tissues. Furthermore, changes in protein expression resulting from in vivo exposure was reflected in our results, however, we found other proteins were (up to 10×) more strongly dysregulated than the known markers. In the context of biomarker discovery for wildlife, these results suggest that 1) the type of tissue that primary cells are derived from is important as molecular response to exposure varies greatly between types and, 2) the most accessible tissue (i.e. skin) may not be the most representative of molecular changes of exposure in the whole organism.

4.5.1 Overall protein profiles of unexposed cells derived from different tissues

Overall, we found that a large proportion of the identified proteins (>89%) in unexposed cells were shared between two or more tissue types with each tissue expressing only

10%, or less, unique proteins (Figure 4.1). We observed 621 proteins shared amongst all cell types, suggesting that any of these cell types could potentially be used to screen for contaminants that exert their effects through mechanisms related to these proteins

(Figure 4.1). However, this high level of congruency in protein expression between tissues does not affirm that a similar molecular response will be observed between tissue types as variations between dominant pathways and sensitivity will likely occur.

Geiger et al. (2012) compared global protein expression of 11 human cell lines and found 10,361 proteins per cell line and high similarity (97%) of expressed proteins

127 between at least two cell lines for each proteome. This was similar to the results found by Wang et al. (2019) who observed that very few proteins show truly tissue specific expression when comparing the proteome of 29 different human tissues. However, while these studies illustrated a global similarity in the presence of proteins, the level of expression of these proteins (the amount of protein present in the cell) differed between cell types, with up to 67% of shared proteins significantly differentially expressed between the tissue types.

128 Figure 4.1. Venn diagram indicating the shared and unique proteins identified by data dependent acquisition (DDA) mass spectrometry of cell lines derived from green sea turtle (Chelonia mydas) liver, kidney, ovary, small intestine and skin tissue. The vertical bar graph below the venn diagram indicates how many proteins were in each list and the horizontal bar scale indicated how many proteins were shared between 1, 2, 3, 4 and all tissue groups. (Made using jvenn online software (Bardou et al., 2014)).

In our study, cells derived from kidney, ovary and small intestine possessed 102, 90 and

73 unique proteins not identified in other types of cells, respectively (Figure 4.1).

Interestingly, liver and skin fibroblasts, which are arguably the most commonly applied in cellular ecotoxicology, had no unique proteins (Figure 4.1). Ovary and kidney shared the most proteins with 66 uniquely in common, liver and skin shared 38, liver and kidney 34, and kidney and small intestine 28. All other combinations of tissues didn’t have any common proteins.

This information is interesting because it suggests that cell lines that are less commonly applied in toxicology research may contain important biomarker proteins, and associated pathways sensitive to contaminant exposure, that have yet to be fully characterised. It is important to emphasize that these cells are primary cells and susceptible to dedifferentiation in culture and it is likely that tissue-specific metabolic activity measured at the protein level may be altered after several passages of in vitro culture (Craven et al., 2006). While outside of the scope of this study, it should be noted that this proteomic data may be useful in identifying/characterising primary cells and their metabolic activity for studies in which this is the focus (Pan et al., 2009).

When interpreting the data from this study it is important to take into account that our examination of the cellular proteome was not as deep (with <1000 proteins identified) in contrast to other studies due to the methodology selected here. A deeper proteomic

129 analysis requires the use of techniques such as multiple prefractionation steps prior to

MS analysis which we did not employ here due to limitations of our methods, namely the small amount of starting material (cell pellets). Studies that have analysed the complete proteome observe >10,000 proteins in one cell line (Geiger et al., 2012; Wang et al., 2019), which is closer to the complete proteome as estimated from transcriptomic data (Wilhelm et al., 2014). It highly likely that there may be more unique proteins that distinguish these cells and their metabolic activity at a lower abundance, but our data does not include such proteins as the focus of the study was not cell characterisation.

4.5.2 Overall number of differentially expressed proteins between cell types

There were several interesting observations when we considered only those proteins that were differentially expressed with a greater than two fold-change (up- or down- regulation) response to PCB153, PFNA or phenanthrene exposure. The first is that liver cells produced a much higher number of differentially expressed proteins for all treatments than cells derived from any other tissue type, with almost 3× more dysregulated proteins (73) than kidney cells (27), which were the next highest (Table

4.1). This suggests that liver cells exhibit a greater response to the contaminants used in this study and is perhaps not surprising considering the role the liver plays in chemical detoxification. It is interesting to note that the majority of the proteins dysregulated in liver cells were up-regulated (83.6%), whereas kidney cells responded mostly with downregulated proteins (88.9%). Secondly, skin cells exhibited the smallest number of differentially expressed proteins, with only 12 total dysregulated proteins falling in this fold-change threshold. This indicates that skin cells may be less sensitive to contaminants and therefore in some respects may not be an ideal tissue for biomarker discovery. However, these results are in contrast to a recent study where toxicity to different sea turtle cells indicated that skin was among the most sensitive tissue type and therefore a good candidate for in vitro toxicity studies (Finlayson et al., 2019a). This

130 disparity in results only further emphasises how single endpoint and non-targeted molecular studies can differ greatly in the information that is obtained about toxic response of a tissue or cell. While a comparatively lower proteome response of skin cells compared to other tissue types in this study indicates that skin may not be an ideal target for mechanistic studies, it is important to note the ease of access to this type of tissue. While liver appears to be one of the more sensitive and responsive cells lines when examining protein expression, it is much more challenging to non-destructively collect liver tissue. However, it is possible to perform laparoscopies and obtain a portion of liver without harming the animal and results such as those obtained here can help inform future studies on whether this kind of more extreme tissue collection is beneficial to the species.

131 Table 4.1. Table showing the number of differentially expressed proteins for each tissue type and treatment that had over a fold change of ≥ 2 for up-regulated proteins and £ -2 for down-regulated proteins.

Tissue Upregulated Downregulated Total up & down PCB 153 PFNA Phenanthrene Total PCB 153 PFNA Phenanthrene Total

Kidney 1 1 1 3 12 8 4 24 27

Liver 19 25 17 61 2 5 5 12 73

SI 2 1 8 11 9 2 1 12 23

Skin 2 5 2 9 1 2 0 3 12

Ovary 1 12 2 15 0 6 1 7 22

Total 25 44 30 99 24 23 11 58 157

132 4.5.3 Identifying biomarkers previously reported in sea turtles

Our survey of the literature identified proteins reported as useful biomarkers of contaminant exposure in sea turtles: heat shock protein (HSP) 60, glutathione S- transferases (GSTs) and superoxide dismutases (SODs), cytochrome P450s and catalase

(Table 4.2). Of these, many were up- and/or down- regulated in our study in cells from different tissue types following exposure to PCB153, PFNA and phenanthrene (Figure

4.2). This analysis revealed that relatively few protein biomarkers have been established for sea turtles, highlighting the need to broadly screen for potentially important candidates that have yet to be identified (see section 4.5.4).

133 Table 4.2. Table showing results of a literature search for molecular markers of contaminant (PCB, PFNA (or PFC), Phenanthrene (or PAHs)) exposure in marine (a) and freshwater (b) turtles. (a) Marine turtles

Controlled Location of What effect is Blood or In vivo or Physiological Biomarker Contaminant Species exposure (c) or in exposure (if in indicated by Reference/s tissue type in vitro function of marker situ exposure (n) situ) marker

Superoxide *study only collected Chelonia Tissue; liver, N/A *study only Baja California, Oxidative stress Antioxidant enzyme Valdivia et al. dismutase (SOD) baseline data on oxidative mydas muscle, collected baseline Mexico (2007) metabolism in sea turtles lung, kidney data on oxidative metabolism in ST Catalase (CAT) *study only collected Chelonia Tissue; liver, N/A *study only Baja California, Oxidative stress Antioxidant enzyme Valdivia et al. baseline data on oxidative mydas muscle, collected baseline Mexico (2007) metabolism in sea turtles lung, kidney data on oxidative metabolism in ST Glutathione S- *study only collected Chelonia Tissue; liver, N/A *study only Baja California, Oxidative stress Detoxification Valdivia et al. transferase baseline data on oxidative mydas muscle, collected baseline Mexico enzyme (2007) (GSTs) metabolism in sea turtles lung, kidney data on oxidative metabolism in ST Gene biomarker PAHs Caretta Whole blood In vivo In situ Mediterranean Cocci et al. expression caretta basin (2018) (CYP1A, HSP60, ERa)

134 (b) Freshwater turtles Controlled Location of What effect is Blood or In vivo or Physiological Biomarker Contaminant Species exposure or in situ exposure (if indicated by Reference/s tissue type in vitro function of marker exposure in situ) marker CYP4501A PCBs (arochlor) Chrysemys: picta liver In vivo Controlled/lab and n/a Exposure Biotransformation Yawetz et al. picta, picta elgans in situ of hydrophobic (1997) and Mauremys chemicals capsica rivulata

Cytochrome Doesn’t specify Chrysemys picta liver In vivo In situ Cape Cod, N/A N/A Rie (2000) P4501A, PCBs but mentions MA, USA GSTs that these markers are known PCB markers

135 There were several interesting observations when evaluating changes in these known protein biomarkers in our cell cultures. The first observation was that HSP60 is fairly consistently down regulated in all tissues and treatments (Figure 4.2) suggesting that it may be a good generic marker of chemical exposure but does not appear to be chemical- specific. This finding agrees with the literature as heat shock proteins appear to be a better indicator of a generalised stress response to exposure than markers of a particular effect (Gupta et al., 2010). However, we observed a downregulation of HSP60 where upregulation would be an expected response to stress. This change may indicate overt toxicity, or that H6P60 was excreted and therefore would have been measured at increased levels in the media (if this analysis had been performed). Secondly, CYP450 was only dysregulated in liver cells exposed to PFNA (downregulated) and in small intestine exposed to phenanthrene (upregulated) (Figure 4.2). CYP450 and its enzymes have been shown to increase in liver in response to PFC exposure (Cheng and Klaassen,

2008) in contrast to our results here where PFNA caused a decrease in liver CYP450.

The upregulation of CYP450 by small intestine exposed to phenanthrene in our study agrees with the literature that illustrates phenanthrene may undergo biotransformation via cytochrome P450 (Schober et al., 2010). This indicates that CYP450 upregulation in small intestine sea turtle cells may be a useful marker of phenanthrene exposure as it was not dysregulated with any other exposures and it was significantly dysregulated

(2.026 fold change) in small intestine (Figure 4.2). Thirdly, it is interesting to observe that the general proteomic response to the same chemical of exposure was different between all tissues however, kidney, skin and ovary appeared to share similar responses when looking at these selected markers (Figure S4.1). On the other hand, we observed that liver and small intestine more often than not showed a similar response to exposure; for example, SOD was downregulated after PFNA exposure for these two tissue types whereas this protein was upregulated in the other three (Figure S4.1). Furthermore,

136 SOD was always upregulated after exposure to each of the compounds for kidney skin and ovary but was always downregulated for liver and small intestine. This indicates that the different cell types have unique responses to contaminant exposure, and therefore, further comparison of these different cell types as tools for biomarker discovery is important.

Interestingly, while liver cells exhibited the greatest number of dysregulated proteins overall (Table 4.1), relatively few of the known biomarkers (Table 4.2) were amongst the most dramatically altered in our exposure experiments when considering fold change (Figure 4.2). This suggests that non-traditional protein biomarkers may represent a novel (and perhaps more sensitive) option for monitoring wildlife exposed to contaminants. Finally, although all expected biomarkers showed up in at least one of the cell line/contaminant exposure scenarios, several (e.g., catalase, cytochrome P450 and GST 2) showed very little expression (low fold change) in any sample, and are perhaps therefore comparatively poorer candidates for biomonitoring.

An important consideration when investigating the potential use of proteins as biomarkers of chemical exposure is that protein biomarkers described in the literature are often referred to in general terms, which may not reflect the diversity of related proteins that fall under these broad protein types. For example, when searching the data for activation of traditional biomarkers in our samples, there were multiple results obtained for the proteins glutathione S-transferases (GSTs) and superoxide dismutases

(SODs). Responsiveness of these related proteins was quite disparate, even for the same cell line and contaminant exposure. For example, three GST isoforms were identified

(GST, GST 2 and GST Mu 1), and while GSTs were generally up-regulated following exposure to all three contaminants, GST 2 and GST Mu 1 were much less influenced by chemical exposure (Figure 4.2). Similarly, while SODs were often up- or down- regulated following exposure to all contaminants, SOD Mn was rarely dysregulated.

137 This is an important observation for studies evaluating proteomic biomarkers because it demonstrates the importance of specificity for the correct protein. Furthermore, as the proteome of C. mydas used for our search here was largely based on genome sequence translation rather than intact protein characterisation, all of the isoforms of these enzymes and their functions have not been fully elucidated in this species. Therefore there are limitations to the extent in which the results can be interpreted however, as proteome data is constantly being updated and refined as a result of molecular studies, the ability of future investigations in this field to interpret the data will deepen.

Overall, these results suggest that proteins can be useful biomarkers of chemical exposure, both generally and for specific chemicals. Certainly, they offer an improvement on current limited assessments of adverse health effects such as haematological parameters (Camacho et al., 2013; Keller et al., 2004b), although care must be taken to ensure the correct proteins are identified, especially when variants of particular protein groups exist.

Non-targeted proteomic analysis on cell lines is a very new approach to cell based- bioassays with studies comparing global protein expression between human tissue and cells first conducted within this decade (Burkard et al., 2011; Geiger, et al., 2012). To our knowledge, global protein expression of primary sea turtle cell lines has not previously been explored in addition to the comparison of changes in this expression in response to contaminant exposure. In light of this, it is important that conclusions about mechanistic toxicological effects of contaminants be cautious and that all variables that may influence these sensitive measures be considered. In particular, reproducibility of this method and consistency of the proteomic response to exposure should be confirmed with repeated studies of this nature in the future.

138

Figure 4.2. Heatmap displaying fold change of significantly (p<0.01) differentially expressed known protein biomarkers of PCB, PFNA or phenanthrene exposure in marine and freshwater turtles as identified from the literature (see Table 4.2). A fold change of one indicates double the expression of a protein relative to the controls. Standard error is <0.17 for all values. OS=Organism Species; GN=Gene Name.

139 4.5.4 Potential new candidate biomarkers

When looking at non-targeted protein expression data to identify potential new biomarkers of chemical exposure, we need to consider markers that are robust, distinct, sensitive and specific to chemical exposure. This means that the proteins should be consistently dysregulated as a result of exposure (robust), as well as highly dysregulated in order to stand out from the baseline ‘noise’ of normal protein expression which naturally fluctuates (distinct). It is interesting to note that while all biomarker proteins identified in the literature (Table 4.2) were differentially expressed after exposure to

PCB153, PFNA and phenanthrene, the most strongly up- and down- regulated proteins

(Table 4.3) have not previously been suggested as potential biomarkers. The proteomic response patterns suggest that some of these proteins could potentially be used as new biomarkers of chemical exposure in sea turtles. For example, all three contaminants induced the same highest upregulated protein in liver, carbonyl reductase [NADPH] 1, and in small intestine, poly(RC)-binding protein 2. These proteins could therefore potentially be used as biomarkers of general chemical stress when measured in vivo.

Sets of markers could also be used to indicate specific exposure, for example up- regulation of carbonyl reductase [NADPH] 1 in combination with the down-regulation of phosphate carrier protein in liver (Table 4.3) could be a specific indicator of phenanthrene exposure in liver, while up-regulation of lipoma-preferred partner like protein in combination with down-regulation of LAG longevity assurance like protein 1 in kidney could indicate PCB153 exposure (Table 4.3).

140 Table 4.3. A table indicating the most strongly up- (a) and down- (b) regulated proteins for each tissue type and treatment relative to controls

(p<0.01) along with fold change (FC) values and standard error (SE).

(a) Up-regulated proteins Tissue PCB 153 FC SE PFNA FC ±SE Phenanthrene FC SE Kidney Lipoma-preferred partner like protein 2.27 ±0.21 Lipoma-preferred partner like 2.55 ±0.24 Signal peptidase complex 2.25 ±0.22 protein subunit 2 Liver Carbonyl reductase [NADPH] 1 4.59 ±0.25 Carbonyl reductase [NADPH] 1 5.43 ±0.24 Carbonyl reductase [NADPH] 4.48 ±0.24 1 SI Poly(RC)-binding protein 2 2.16 ±0.05 Poly(RC)-binding protein 2 2.09 ±0.05 Poly(RC)-binding protein 2 2.84 ±0.06 Skin Cytochrome c oxidase subunit 2 2.51 ±0.07 Brain acid soluble protein 1 like 3.23 ±0.04 Cytochrome c oxidase subunit 2.92 ±0.68 protein 2 Ovary Uncharacterised protein 2.04 ±0.06 Prefoldin subunit 2 (Fragment) 5.20 ±0.20 Sulfurtransferase 3.32 ±0.11 (tr|M7BW30|M7BW30_CHEMY)

(b) Down-regulated proteins

Tissue PCB 153 FC ±SE PFNA FC ±SE Phenanthrene FC ±SE Kidney LAG longevity assurance like protein 1 -5.57 ±0.22 Actin, cytoplasmic type 5 -3.34 ±0.13 Stomatin-like protein 2 -2.56 ±0.12 Liver Coiled-coil domain-containing protein 47 -4.11 ±0.38 Eukaryotic translation -2.53 ±0.10 Phosphate carrier protein -9.72 ±0.14 initiation factor 3 subunit M SI Heterogeneous nuclear ribonuclear protein U -2.04 ±0.20 Annexin (fragment) -12.7 ±0.41 Thrombospondin-1 -9.18 ±0.33 (fragment) Skin Protein Noxp20 -1.05 ±0.02 Myelin P2 protein -9.05 ±0.27 Myelin P2 protein -1.95 ±0.06 Ovary Calumenin -3.63 ±0.17 Single-stranded DNA- -3.27 ±0.26 Lymphocyte function- -2.19 ±0.09 binding protein associated antigen 3

141 One of the most strongly dysregulated proteins was annexin that was downregulated 12.7 fold in small intestine cells exposed to PFNA (Table 4.3). Several other in vitro studies have also observed a downregulation of annexin after PFNA exposure (Wei et al., 2009; Zhang et al., 2012) indicating this may be a candidate biomarker of exposure. Annexin is involved in

Ca2+ signalling and is therefore vital in many cellular processes (Gerke et al., 2005).

Suppression of this protein could cause or indicate adverse effects to the organism, meaning that it could serve as both a marker of exposure and effect. Another significantly dysregulated protein was lymphocyte function-associated antigen 3, which was strongly down-regulated in ovary cells after exposure to phenanthrene. This information is not only interesting when considering new biomarkers of exposure but given the important role of this protein in immune system functioning (Anikeeva et al., 2005) may provide insight into mechanistic effects of phenanthrene on sea turtles.

For the proteins identified here to be potentially used as new biomarkers of chemical exposure these proteins must also be able to be measured in vivo. It would be logical to assume that markers that are highly dysregulated in vitro as a response to exposure are good in vivo candidates, especially if these markers indicate important effects such as oxidative stress, or immunotoxicity. However, it is possible that these markers do not appear in vivo.

Proteins may be degraded in vivo by physiological processes, or their expression limited by negative feedback loops that only occur within the organism and aren’t replicated in vitro.

When considering which tissue type is best, it is not entirely clear from our results if one is better than another simply by looking at the expression of molecular markers as all tissue types expressed known in vivo markers as well as unique proteins that could be new biomarker candidates. It is important that the advantages of one tissue type in terms of molecular usefulness be balanced with the practical aspects of obtaining the tissue. For example, small intestine cells appeared to be the most sensitive as it had the strongest

142 response (based on fold change) compared to the other tissue types with proteins such as annexin that are known in vivo markers. However, the establishment of small intestine primary cells would require operating on or sacrificing an animal, which is not possible for threatened species. On the other hand, skin cells can be obtained relatively easily without the need for sacrificing or harming an individual (through small biopsy punch) and these cells also appear useful for new biomarker analyses. While many more studies are needed to carefully examine the pros and cons of combining non-targeted proteomics analysis with in vitro cell-based contaminant exposure, this study presents a framework for discovering novel specific and non-specific biomarkers that can potentially be used as non-destructive indicators of exposure to chemical contaminants and effects in threatened wildlife.

4.6 References

Amiard-Triquet, C., Amiard J C. and Rainbow, P S. 2013. Ecological Biomarkers: Indicators Ecotoxicological Effects. CRC Press, Taylor & Francis, Boca Raton, USA. Alam, S K., and Brim, M S. 2000. Organochlorine, PCB, PAH, and metal concentrations in eggs of loggerhead sea turtles (Caretta caretta) from northwest Florida, USA. J Environ Sci Heal B. 35:705-724. Anikeeva, N., Somersalo, K., Sims, T N., Thomas, V K., Dustin, M L. and Sykulev, Y. 2005. Distinct role of lymphocyte function-associated antigen-1 in mediating effective cytolytic activity by cytotoxic T lymphocytes. Proc Natl Acad Sci U S A. 102:6437- 6442. Ankley, G T., Bennett, R S., Erickson, R J., Hoff, D J., Hornung, M W., Johnson, R D., Mount, D R., Nichols, J W., Russom, C L., Schmieder, P K., Serrrano, J A., Tietge, J E. and Villeneuve, D L. 2010. Adverse outcome pathways: a conceptual framework to support ecotoxicology research and risk assessment. Environ Toxicol Chem. 29:730- 741. Apraiz, I., Mi, J. and Cristobal, S. 2006. Identification of proteomic signatures of exposure to marine pollutants in mussels (Mytilus edulis). Mol Cell Proteomics. 5:1274-1285. Ashraf, M A. 2017. Persistent organic pollutants (POPs): a global issue, a global challenge. Environ Sci Pollut Res Int. 24:4223-4227.

143 Bardou, P., Mariette, J., Escudie, F., Djemiel, C., and Klopp, C. 2014. jvenn: an interactive Venn diagram viewer. Bmc Bioinformatics 15. Benninghoff, A D. 2006. Toxicoproteomics--The Next Step in the Evolution of Environmental Biomarkers? Toxicol Sci. 95:1-4. Camacho, M., Luzardo, O P., Boada, L D., Lopez Jurado, L F., Medina, M., Zumbado, M. and Oros, J. 2013. Potential adverse health effects of persistent organic pollutants on sea turtles: evidences from a cross-sectional study on Cape Verde loggerhead sea turtles. Sci Total Environ. 458-460:283-289. Cheng, X. and Klaassen, C D. 2008. Perfluorocarboxylic acids induce cytochrome p450 enzymes in mouse liver through activation of ppar-α and car transcription factors. Toxicol Sci. 106:29-36. Choi, M., Chang, C Y., Clough, T., Broudy, D., Killeen, T., MacLean, B. and Vitek, O. 2014. MSstats: an R package for statistical analysis of quantitative mass spectrometry-based proteomic experiments. Bioinformatics. 30:2524-2526. Cocci, P., Mosconi, G., Bracchetti, L., Nalocca, J M., Frapiccini, E., Marini, M., Caprioli, G., Sagratini, G. and Palermo, F A. 2018. Investigating the potential impact of polycyclic aromatic hydrocarbons (PAHs) and polychlorinated biphenyls (PCBs) on gene biomarker expression and global DNA methylation in loggerhead sea turtles (Caretta caretta) from the Adriatic Sea. Sci Total Environ. 619-620:49-57. Convention, S. 2001. Stockholm convention on persistent organic pollutants. In Stockholm Convention, Geneva, Switzerland. Cortes-Gomez, A A., Morcillo, P., Guardiola, F A., Espinosa, C., Esteban, M A., Cuesta, A., Girondot, M. and Romero, D. 2018. Molecular oxidative stress markers in olive ridley turtles (Lepidochelys olivacea) and their relation to metal concentrations in wild populations. Environ Pollut. 233:156-167. Craven, R A., Stanley, A J., Hanrahan, S., Dods, J., Unwin, R., Totty, N., Harnden, P., Eardley, I., Selby, P J. and Banks, R E. 2006. Proteomic analysis of primary cell lines identifies protein changes present in renal cell carcinoma. Proteomics. 6:2853-2864. Dowling, V A. and Sheehan, D. 2006. Proteomics as a route to identification of toxicity targets in environmental toxicology. Proteomics. 6:5597-5604. Dzul-Caamal, R., Salazar-Coria, L., Olivares-Rubio, H F., Rocha-Gomez, M A., Giron-Perez, M I. and Vega-Lopez, A. 2016. Oxidative stress response in the skin mucus layer of Goodea gracilis (Hubbs and Turner, 1939) exposed to crude oil: A non-invasive approach. Comp Biochem Physiol A. 200:9-20.

144 Finlayson, K A., Leusch, F D L., and van de Merwe, J P. 2019a. Cytotoxicity of organic and inorganic compounds to primary cell cultures established from internal tissues of Chelonia mydas. Sci Total Environ. 664:958-967. Finlayson, K A., Leusch, F D L. and van de Merwe, J P. 2019b. Primary green turtle (Chelonia mydas) skin fibroblasts as an in vitro model for assessing genotoxicity and oxidative stress. Aquat Toxicol 207:13-18. Geiger, T., Wehner, A., Schaab, C., Cox, J. and Mann, M. 2012. Comparative proteomic analysis of eleven common cell lines reveals ubiquitous but varying expression of most proteins. Mol Cell Proteomics 11(3):M111-014050. George, J., Singh, R., Mahmood, Z. and Y. Shukla. 2010. Toxicoproteomics: New paradigms in toxicology research. Toxicol Mech Method. 20:415-423. Gerke, V., Creutz, C E. and Moss, S E. 2005. Annexins: linking Ca2+ signalling to membrane dynamics. Nat Rev Mol Cell Bio. 6:449-461. Gu, X. and J.E. Manautou. 2012. Molecular mechanisms underlying chemical liver injury. Expert Rev Mol Med 14. Gupta, S C., Sharma, A., Mishra, M., Mishra, R K. and Chowdhuri, D K. 2010. Heat shock proteins in toxicology: How close and how far? Life Sciences. 86:377-384. Hoeng, J., Talikka, M., Martin, F., Ansari, S., Drubin, D., Elamin, A., Gebel, S., Ivanov, N V., Deehan, R., Kogel, U., Mathis, C., Schlage, W K., Sewer, A., Sierro, N., Thomson, T. and Meitsch, M C. 2014. Toxicopanomics Applications of Genomics, Transcriptomics, Proteomics, and Lipidomics in Predictive Mechanistic Toxicology. Hayes' Principles and Methods of Toxicology, 6th Edition:295-332. Houde, M., De Silva, A O., Muir, D C. and Letcher, R J. 2011. Monitoring of perfluorinated compounds in aquatic biota: an updated review. Environ Sci Technol. 45:7962-7973. Jasinska, E J., Goss, G G., Gillis, P L., Van Der Kraak, G J., Matsumoto, J., de Souza Machado, A A., Giacomin, M., Moon, T W., Massarsky, A., Gagne, F., Servos, M R., Wilson, J., Sultana, T. and Metcalfe, C D. 2015. Assessment of biomarkers for contaminants of emerging concern on aquatic organisms downstream of a municipal wastewater discharge. Sci Total Environ. 530-531:140-153. Keller, J M. 2003. Occurence and effects of organochlorine contaminants in sea turtles. In Nicolas School of The Environent and Earth Sciences. Doctor of Philosophy. Duke University, North Carolina, USA. 366.

145 Keller, J M., Kannan, K., Taniyasu, S., Yamashita, N., Day, R D., Arendt, M D., Segars, A L. and Kucklick, J R. 2005. Perfluorinated compounds in the plasma of loggerhead and Kemp's ridley sea turtles from the southeastern coast of the United States. Environ Sci Technol. 39:9101-9108. Keller, J M., Kucklick, J R. and McClellan-Green, P D. 2004a. Organochlorine contaminants in loggerhead sea turtle blood: Extraction techniques and distribution among plasma and red blood cells. Arch Environ Con Tox. 46:254-264. Keller, J M., Kucklick, J R., Stamper, M A., Harms, C A. and McClellan-Green, P D. 2004b. Associations between organochlorine contaminant concentrations and clinical health parameters in loggerhead sea turtles from North Carolina, USA. Environ Health Persp. 112:1074-1079. Keller, J M., Ngai, L., Braun McNeill, J., Wood, L D., Stewart, K R., O'Connell, S G. and Kucklick, J R. 2012. Perfluoroalkyl contaminants in plasma of five sea turtle species: comparisons in concentration and potential health risks. Environ Toxicol Chem. 31:1223-1230. Kennedy, S. 2002. The role of proteomics in toxicology: identification of biomarkers of toxicity by protein expression analysis. Biomarkers. 7:269-290. Kerr, E D., Phung, T K., Fox, G P., Platz, G J. and Schulz, B L. 2018. The intrinsic and regulated proteomes of barley seeds in response to fungal infection. bioRxiv, 449516. Labrada-Martagón, V., Tenorio Rodríguez, P A., Méndez-Rodríguez, L C., and Zenteno- Savín, T. 2011. Oxidative stress indicators and chemical contaminants in East Pacific green turtles (Chelonia mydas) inhabiting two foraging coastal lagoons in the Baja California peninsula. Comp Biochem Physiol C 154:65-75. Lasserre, J P., Fack, F., Revets, D., Planchon, S., Renaut, J., Hoffmann, L., Gutleb, A C., Muller, C P., and Bohn, T. 2009. Effects of the endocrine disruptors atrazine and PCB 153 on the protein expression of MCF-7 human cells. J Proteome Res. 8:5485-5496. Leusch, F D L., de Jager, C., Levi, Y., Lim, R., Puijker, L., Sacher, F., Tremblay, L A., Wilson, V S. and Chapman, H F. 2010. Comparison of five in vitro bioassays to measure estrogenic activity in environmental waters. Environ Sci Technol. 44:3853- 3860. Lin, Y R., Kuo, C J., Lin, H Y., Wu, C J. and Liang, S S. 2014. A proteomics analysis to evaluate cytotoxicity in NRK-52E cells caused by unmodified Nano-Fe(3)O(4). Sci World J 2014:754721.

146 Monserrat, J M., Martinez, P E., Geracitano, L A., Amado, L L., Martins, C M., Pinho, G L., Chaves, I S., Ferreira-Cravo, M., Ventura-Lima, J. and Bianchini, A. 2007. Pollution biomarkers in estuarine animals: critical review and new perspectives. Comp Biochem Physiol C. 146:221-234. Pan, C., Kumar, C., Bohl, S., Klingmueller, U. and Mann, M. 2009. Comparative Proteomic Phenotyping of Cell Lines and Primary Cells to Assess Preservation of Cell Type- specific Functions. Mol Cell Proteomics. 8:443-450. Rie, M T., Lendas, K A. and Wood, B R. 2000. Hepatic biotransformation enzymes in a sentinel species, the painted turtle (Chrysemys picta), from Cape Cod, Massachusetts: seasonal-, sex- and location related differences. Biomarkers. 5:382-394. Rockwell, C E., Turley, A E., Cheng, X., Fields, P E. and Klaassen, C D. 2013. Acute Immunotoxic Effects of Perfluorononanoic Acid (PFNA) in C57BL/6 Mice. Clin Exp Pharmacol. (2161-1459) Suppl 4:S4-002. Roland, K., Kestemont, P., Hénuset, L., Pierrard, M-A., Raes, M., Dieu, M. and Silvestre, F. 2013. Proteomic responses of peripheral blood mononuclear cells in the European eel (Anguilla anguilla) after perfluorooctane sulfonate exposure. Aquat Toxicol. 128– 129:43-52. Salmón, P., Stroh, E., Herrera-Dueñas, A., von Post, M. and Isaksson, C. 2018. Oxidative stress in birds along a NOx and urbanisation gradient: An interspecific approach. Sci Total Environ. 622-623:635-643. Sanchez, B.C., Ralston-Hooper, K. and Sepulveda, M S. 2011. Review of recent proteomic applications in aquatic toxicology. Environ Toxicol Chem. 30:274-282. Schober, W., Pusch, G., Oeder, S., Reindl, H., Behrendt, H. and Buters, J T M. 2010. Metabolic activation of phenanthrene by human and mouse cytochromes P450 and pharmacokinetics in CYP1A2 knockout mice. Chem-Biol Interact. 183:57-66. Valdivia, P A., Zenteno-Savín, T., Gardner, S C. and Alonso Aguirre, A. 2007. Basic oxidative stress metabolites in eastern Pacific green turtles (Chelonia mydas agassizii). Comp Biochem Phys C. 146:111-117. Vuong, N Q., Goegan, P., Mohottalage, S., Breznan, D., Ariganello, M., Williams, A, Elisma, F., Karthikeyan, S., Vincent, R. and Kumarathasan, P. 2016. Proteomic changes in human lung epithelial cells (A549) in response to carbon black and titanium dioxide exposures. J Proteomics 149:53-63.

147 Wang, D., Eraslan, B., Wieland, T., Hallstrom, B M., Hopf, T., Zolg, D P., Zecha, J., Asplund, A., Li, L-H., Meng, C., Frejno, M., Schmidt, T., Schnatbaum, K., Wilhelm, M., Ponten, F., Uhlen, M., Gagneur, J., Hahne, H. and Kuster, B. 2019. A deep proteome and transcriptome abundance atlas of 29 healthy human tissues. Mol Syst Biol. 15(2):e8503. Wania, F. 2003. Assessing the potential of persistent organic chemicals for long-range transport and accumulation in polar regions. Environ Sci Tech. 37:1344-1351. Wei, Y., Shi, X., Zhang, H., Wang, J., Zhou, B. and Dai, J. 2009. Combined effects of polyfluorinated and perfluorinated compounds on primary cultured hepatocytes from rare minnow (Gobiocypris rarus) using toxicogenomic analysis. Aquat Toxicol 95:27-36. Wilhelm, M., Schlegl, J., Hahne, H., Gholami, A M., Lieberenz, M., Savitski, M M., Ziegler, E., Butzmann, L., Gessulat, S., Marx, H., and Mathieson T,. 2014. Mass spectrometry based draft of the human proteome. Nature. 509(7052):582. Wisniewski, J R., Zougman, A., Nagaraj, N. and Mann, M. 2009. Universal sample preparation method for proteome analysis. Nat Methods. 6:359-362. Yawetz, A., BenedekSegal, M. and Woodin, B. 1997. Cytochrome P4501A immunoassay in freshwater turtles and exposure to PCBs and environmental pollutants. Environ Toxicol Chem. 16:1802-1806. Zhang, W., Liu, Y., Zhang, H. and Dai, J. 2012. Proteomic analysis of male zebrafish livers chronically exposed to perfluorononanoic acid. Environ Int. 42:20-30. Zheng, X M., Liu, H L., Shi, W., Wei, S., Giesy, J P. and Yu, H X. 2011. Effects of perfluorinated compounds on development of zebrafish embryos. Environ Sci Pollut Res Int. 19:2498-2505.

148 Chapter 5: Non-targeted proteomic analysis of blood from three distinct populations of green sea turtle (Chelonia mydas) reveals significantly altered immune states between populations.

Statement of Contribution for Chapter 5 This chapter includes a co-authored article that has been submitted to Ecotoxicology and

Environmental Safety for consideration for publication. The bibliographic details of the co- authored paper are:

Stephanie Chaousis, Frederic DL Leusch, Col Limpus, Amanda Nouwens, Liesbeth Weijs,

Antonia Weltmeyer, Adrian Covaci and Jason P van de Merwe (to be submitted) Non- targeted proteomic analysis of blood from three distinct populations of green sea turtle

(Chelonia mydas) reveals significantly altered immune states between populations. Sci Tot

Environ.

My contribution to the paper include:

• the conceptualisation of the design of the experiment,

• undertaking all lab work,

• data analysis, and

• writing of the manuscript.

My co-authors provided input into the design of the study, interpretation of the data and final polishing of the writing. This chapter is formatted in the style of the journal of publication.

(Signed) ______(Date) 24/04/2019 Stephanie Chaousis

149

(Countersigned) ______(Date) 23/04/2019 Jason van de Merwe

150 5.1 Abstract

All seven species of sea turtle are facing increasing pressures from human activities that are impacting their health. Changes in circulating blood proteins can provide an early indicator of adverse health outcomes for an organism or population. Non-targeted proteomics is a powerful tool for detecting changes in global protein expression and for discovering biomarkers of chemical exposure and effect. However, up until now this approach has been used to limited extent in the context of wildlife toxicology. In this study, we present for the first time a non-targeted examination of the protein expression in sea turtle plasma obtained from three foraging populations of green turtles (Chelonia mydas) on the Queensland coast.

Relative changes in protein expression between sites were compared and potential markers of exposure were investigated. Blood plasma protein profiles were distinct between populations from all sites, with 85 out of the 116 identified proteins differentially expressed between the sites (p<0.001). The most strongly dysregulated proteins were predominantly acute phase proteins (APP), suggestive of differing immune status between the populations, with the

Moreton Bay turtles displaying the highest upregulation of known markers of immunotoxicity, such as haptoglobin and fibrinogen. Fourty-five different organohalogens were also measured in green turtle plasma samples as exposure to some organohalogens (e.g., polychlorinated biphenyls, PCBs) has previously been identified as a cause for immune dysregulation in marine animals. The few detected organohalogens were at very low pg/mL concentrations at all sites, and are unlikely to be the cause of the proteome differences observed. However, a number of dysregulated proteins indicated that these populations may be exposed to different concentrations of other contaminants, such as metals and fluorinated flame retardants. The results of this study provide important information about differences in protein expression between different populations of foraging green turtles, and form the basis

151 for future studies aiming to identify biomarkers of chemical exposure and effect in green sea turtles.

5.2 Keywords

Biomarker; contaminants; wildlife

5.3 Introduction

All seven species of sea turtle are listed as vulnerable, endangered or critically endangered by the International Union for Conservation of Nature (IUCN) Red List Assessments (IUCN,

2019). It is critical that such vulnerable species are monitored closely for threats that may ultimately lead to population decline and biodiversity loss. Aside from readily observable causes of mortality due to human activity, such as destructive fishing practices and recreational boating (Casale, 2010), exposure to anthropogenic contaminants is among the numerous environmental pressures that can trigger adverse health effects (Hutchinson et al.

2013). Monitoring physiological changes occurring in individuals can alert conservation biologists to negative health effects arising as a result of chemical exposure, and early recognition of physiological changes due to contaminant exposure can allow for mitigation of population decline through timely intervention. For example, a mass mortality event was triggered in a population of Baltic seals in 1988 when tens of thousands of individuals died within the year as a result of a morbillivirus (Osterhaus et al. 1989). Such a dramatic mass mortality event had not previously been reported and investigations later revealed that immunosuppression caused by exposure to anthropogenic contaminants including polychlorinated biphenyls (PCBs), dioxins and furans exacerbated the sensitivity of the animals to the virus (de Swart et al. 1996, Van Loveren et al. 2000).

Organohalogens are one of the most commonly investigated and probably best-known class of contaminants detected in sea turtle tissue and blood. PCB153 for example has been

152 reported in several species of sea turtle (Camacho et al., 2013; Camacho et al., 2014; Keller et al., 2004a; Ragland et al., 2011; van de Merwe et al., 2010a; van de Merwe et al., 2010b) at concentrations up to 530 pg/mL in blood (Camacho et al. 2014) and 87 ng/g w.w. in fat

(Corsolini et al. 2000). While these concentrations are lower than those in other marine species, studies have found that low-level PCB exposure can correlate with immunotoxicity

(Keller et al. 2004, Keller et al. 2006, Rousselet et al. 2017) and changes to blood biochemistry in sea turtles (Camacho et al. 2013), and adverse reproductive and developmental effects in freshwater turtles (Bergeron et al. 1994, Eisenreich et al. 2009,

Ming-Ch'eng Adams et al. 2016). The evidence of sublethal effects of PCB exposure, particularly immune modulation, is concerning for these vulnerable populations. Indeed, detrimental population-level effects of immune suppression in otherwise healthy individuals has been observed in other species of large marine vertebrates. A similar event would be devastating for an already threatened species, and therefore the development of tools to monitor health status of these species is important.

Biomarkers can be an effective way to non-destructively monitor the health of wildlife and detect early signs of adverse sublethal effects of contaminant exposure (Fossi 1994, Chaousis et al. 2018). Biomarkers are usually molecular markers measured in bodily fluids (or sometimes tissue biopsies) and are ideally specific to a particular measure of health functioning (Atkinson Jr et al. 2001). Biomarkers can indicate general poor health such as alterations in blood biochemistry (Geens et al. 2010, Ilizaliturri-Hernandez et al. 2013, Dos

Santos et al. 2016), or can reveal a specific adverse effect such as endocrine disruption (Zota et al. 2018). Connecting biomarkers to contaminant exposure in this manner is powerful but challenging, and the discovery of new biomarkers in threatened wildlife is particularly challenging due to limitations around experimentation and handling of threatened species

(Chaousis et al. 2018). However, recent advances in molecular techniques have opened up

153 proteomics as a new avenue for researching potential new biomarkers of chemical exposure in non-destructive samples (Martyniuk and Simmons 2016).

Proteomics, specifically mass spectrometry based non-targeted proteomics, is a relatively new technique in toxicology (George et al. 2010, Martyniuk and Simmons 2016). Its potential for biomarker discovery has yet to be explored for threatened wildlife toxicology , however the benefits of this approach has been clearly demonstrated with lab-based experimental studies on wildlife (Dondero et al. 2010, Yadetie et al. 2017). Blood plasma is arguably the most useful sample type for protein expression analysis in the context of wildlife toxicology as it is relatively simple to collect and theoretically representative of protein expression in the organism (Lee et al. 2006). The ability to identify and quantify all detectable proteins in a sample provides valuable insights into species or even population specific effects of contaminant exposure and enhances the opportunity to discover valuable biomarkers for monitoring the health of populations of threatened marine species.

In the present study, we investigated the blood plasma proteome of three green sea turtle

(Chelonia mydas) foraging populations, and potential correlations of differentially expressed proteins with contaminant exposure. More specifically, our aims were to 1) compare the blood plasma protein profiles of green turtles from different foraging sites along the

Queensland coast, 2) investigate the identity and function of the most highly up- and down- regulated proteins, and 3) quantify and compare the concentrations of several classes of organohalogens in these populations to investigate any influence these chemicals may have on protein expression in sea turtles.

154 5.4 Methods

5.4.1 Collection of turtle blood plasma

5.4.1.1 Site selection

Three green sea turtle (Chelonia mydas) foraging sites along the south eastern Queensland coast were selected based on having different chemical profiles present in the environment or, where data were not available, differences in known sources of contaminants in the catchments. The three sites were, the Port of Gladstone (PG; Lat -23.822335 Long

151.272543), Hervey Bay (HB; Lat -25.363082, Long 152.915703) and Moreton Bay (MB;

Lat -27.279373, Long 153.344700). The catchment of PG is characterised by large coal mines and heavy industry (LNG plants, coal terminals, and chemical processing plants), and chemical profiling of sediment and seagrasses in coastal waters have previously detected elevated concentrations of trace metals, but low concentrations of organic chemicals (Beale et al. 2017). HB is located near the coast of an intensive agricultural area, and monitoring has previously detected pesticides and organic chemical contaminants commonly associated with farming practices (McMahon et al. 2005). Finally, MB is located in coastal waters surrounded by a much larger human population than the other two sites with 425,000 residents as of 2016 (ABS, 2016) in the immediate area, not to mention that the MB catchment supports the third most populous city in Australia, Brisbane, with over 1.9 million residents (ABS, 2016). Relative to this the populations of the other two sites are very small with 33,000 residents in PG and 52,288 in HB (ABS, 2016). MB also has significant industrial activity (Port of Brisbane) as well as agriculture and aquaculture inputs (e.g.,

Logan River). Both organic and non-organic contaminants have been detected at notable concentrations in sediment (Morelli and Gasparon, 2012) and marine megafauna inhabiting

Moreton Bay (Vijayasarathy et al, 2019; Weijs et al, 2016). Therefore, the contaminant exposure profiles of the foraging populations of C.mydas in these three regions will be

155 distinct from each other, and likely be dominated by mining and industrial activity at PG, agricultural activity at HB and a mix of urban, industrial and agricultural activities at MB

Site Sources of chemical Contaminants detected at site References contamination

Hervey Bay Agricultural activity Pesticides, organic contaminants McMahon et al. 2005 (diuron, atrazine, ametryn)

Port of Industrial and mining Trace metals (Cr, Cu, Pb and Ni) Beale et al. 2017 Gladstone activity

Urban, industrial and Organic contaminants (PCBs and Morelli and Gasparon, Moreton Bay agricultural activity DDXs, PCDDs) and trace metals 2014; Vijayasarathy et (Cr, Ni, Cu, Zn, Cd and Co) al, 2019. Weijs et al 2016 (Table 5.1).

Table 5.1. Table indicating the source of chemical contaminants at each study site along with classes of contaminants previously detected at each site.

PCB = Polychlorinated Biphenyls, DDX = Dichlorodiphenyltrichloroethane (and associated isomers and metabolites), PCDDs = Polychlorinated dibenzodioxins.

5.4.1.2. Sample collection

Free-ranging subadult C.mydas with a curved carapace length (CCL) between 65 and 85 cm were hand-captured from MB between November 2015 and July 2016 (n=23), from HB between October/November 2015 and September/October 2016 (n=25) and from PG between

May and October 2016 (n=24). Turtles from MB and PG were captured using rodeo techniques (Limpus, 1978), while HB turtles were collected while they were basking on land amongst mangrove forests at night. Blood (50 mL) was drawn from the dorsocervical sinus using 21G needles and 50 mL syringes, and immediately dispensed into 10 mL lithium- heparin vacutainers. Four mL of each blood sample were then transferred to 2 mL tubes and centrifuged at 2,500 ×g for 5 min. The resulting plasma supernatant was then transferred into new 1.5 mL tubes, ensuring that no red blood cells were transferred from the pellet. These samples were stored at -20°C until further use.

156 5.4.2 Non-targeted proteomic analysis

5.4.2.1 Protein quantification and digestion by Filter Aided Sample Preparation (FASP)

Two mL of plasma were thawed on ice before high speed centrifugation at 15,000 ×g to remove any debris or insolubilized material. The supernatant was then transferred to a fresh tube and the protein content of each sample quantified using a modified Bradford-Lowry assay before the equivalent of 50 µg of protein from each sample was added to the top of a 10 kD molecular weight cut off filter insert in a 1.5 mL nanosep tube (PALL). Filter Aided

Sample Preparation (FASP) was then performed as described by Wisniewski et al. (2009) with modifications. Briefly, samples were washed by the addition of 400 µL of 8 M urea and centrifuged at 14,000 ×g at 20°C for 20 min or until all sample had moved through the filter.

Following this, 200 µL of 8 M urea with 50 mM of ammonium bicarbonate (ABC) was washed through the column by centrifugation at 14,000 ×g for 20 min. The flow-through was discarded and 200 µL of 8 M urea, along with 50 mM ABC and 50 mM dithiothreitol (DTT; for reduction of sulphide bonds) was added to the tube and then incubated at 56°C for 30 min in a water bath. Samples were then alkylated with the addition of 50 mM iodoacetamide

(IAA) to each sample and incubated for a further 30 min at room temperature (approximately

25°C) in the dark. DTT was then added to the sample to quench excess IAA before the filtrate was passed through the membrane by centrifugation. The filtrate was discarded and

100 µL of 50 mM ABC with 1 µg of trypsin protease was added to the sample before incubation overnight at 37°C. The following day, the filter containing the digested proteins was transferred to a new tube and eluted from the filter by centrifugation at 14,000 ×g, 20°C for 20 min. The filtrate containing the digested peptides was zip-tipped using a C18 Merck

Millipore 10 µL zip tip and the peptides eluted in 60% acetonitrile (ACN) and 0.1% formic acid (FA). Solvents were evaporated by centrifuging samples in a vacuum concentrator at

45°C for 1 h before resuspending the dried peptides in 100 µL of 0.1% FA.

157 5.4.2.2 Liquid Chromatography-tandem Mass Spectrometry (LC-MS/MS) analysis

Reverse phased liquid chromatography was performed on a Shimadzu Prominence nanoLC system to separate the samples. Samples were desalted on an Agilent C18 trap (0.3 mm × 5 mm × 5 µm) using a flow rate of 30 µL/min before separation on a Vydac Everest C18 (300

A, 5 µm, 150 mm × 150 µm) column with a flow rate of 1 µL/min. Peptides were separated using a gradient of 10-60% buffer B over 45 min where buffer A = 1% ACN / 0.1% FA and buffer B = 80% ACN / 0.1% FA. Peptides were then analysed on a TripleTof 5600 instrument (ABSciex) using a Nanospray III interface with voltage and gas settings adjusted as required. MS TOF scans across 350-1800 m/z was performed for 0.5 sec followed by information dependent acquisition of up to 20 peptides with intensity greater than 100 counts, across 40-1800 m/z (0.05 sec per spectra) using collision energy (CE) of 40 ± 15 V. For sequential window acquisition of all theoretical mass spectra (SWATH-MS) analyses, MS scans across 350-1800 m/z were performed (0.5 sec), followed by high sensitivity DIA mode using 26 m/z isolation windows for 0.1 sec, across 400-1250 m/z. CE values for SWATH samples were automatically assigned by Analyst software based on m/z mass windows.

5.5 Organohalogen analysis

5.5.1 Samples, chemicals and target compounds.

Plasma samples (n = 72) of green turtles were analysed for a range of organohalogen compounds. In all samples, 25 polychlorinated biphenyl (PCB) congeners (IUPAC numbers:

CB 28, 49, 52, 74, 95, 99, 101, 105, 118, 128, 138, 146, 153, 156, 170, 171, 177, 180, 183,

187, 194, 196/203, 199, 206, 209), seven polybrominated diphenyl ethers (PBDEs) (IUPAC numbers: BDE 28, 47, 99, 100, 153, 154, 183), DDXs (p,p’-DDE, p,p’-DDT), chlordanes

(oxychlorane-OxC, trans-nonachlor-TN, cis-nonachlor-CN, trans-chlordane-TC, cis- chlordane-CC), hexachlorobenzene (HCB), hexachlorocyclohexanes (HCHs; α-, β-, γ-) and two naturally-produced methoxylated PBDEs (2’-MeO-BDE 68 and 6-MeO-BDE 47) were

158 targeted. Standards were purchased from Wellington Laboratories (PBDEs and MeO-PBDEs) and Accustandard (PCBs and organochlorine pesticides (OCPs)). All solvents used (acetone, n-hexane, dichloromethane (DCM), iso-octane) were SupraSolv grade (Merck).

5.5.2 Sample preparation.

The method used for the sample preparation and clean-up is detailed below. Exactly 1 mL of plasma was extracted by liquid/liquid extraction with 50 µL internal standard (25 pg/µL 13C-

HCB + 100 pg/µL PCB 143 + 20 pg/µL ε-HCH + 10 pg/µL BDE 77 in acetone), 1 mL

MilliQ water, 200 µL formic acid (96% purity), and 4 mL n-hexane/DCM (4:1, v/v). After centrifugation (5 min, 3000 ×g), the supernatant was collected. The centrifugation step was repeated two times by adding an additional 4 mL of n-hexane/DCM (4:1, v/v). The supernatants were combined, evaporated to near dryness under a gentle nitrogen stream and re-solubilized in 0.5 mL of n-hexane. Extracts were cleaned up (removal of lipids) on 1 g of acidified silica (44%, sulphuric acid 98% purity) with 10 mL of n-hexane/DCM (1:1, v/v).

Eluates were evaporated to near dryness and reconstituted in 100 µL recovery standard (50 pg/µL PCB 207). Extracts were refrigerated at 4°C until further analysis.

5.5.3 Gas Chromatography-Mass Spectrometry (GC-MS) analysis.

All analytes (PCBs, DDTs, HCB, chlordanes, HCHs, PBDEs and MeO-PBDEs) were measured with an Agilent 6890 gas chromatograph coupled with a 5973 mass spectrometer system (GC-MS). The GC was equipped with a 30 m × 0.25 mm × 0.25 µm DB-5 capillary column. The MS operated in electron capture negative ionisation (ECNI) mode and was used in the selected ion-monitoring (SIM) mode with two ions monitored for each analyte or homologue group.

5.5.4 Quality assurance/quality control (QA/QC).

Mean recoveries (SD; both expressed in %) for internal standards were 74 (9), 80 (9), 69 (7), and 75 (11) for PCB 143, 13C-HCB, ε-HCH, and BDE 77, respectively. For each analyte, the

159 mean procedural blank value was used for subtraction. After blank subtraction, the limit of quantification (LOQ) was set at three times the standard deviation of the procedural blank, which ensures > 99 % certainty that the reported value is originating from the sample. For analytes that were not detected in procedural blanks, LOQs were calculated for a ratio S/N equal to 10. LOQs depended on the sample intake and on the analyte and ranged between 1 and 4 pg/mL serum. QC was performed by regular analyses of procedural blanks, calibration standards and solvent blanks.

5.6 Data analysis

5.6.1 Protein identification and quantification

To obtain protein identifications from MS data, an ion library was generated using DDA MS data files, searched in ProteinPilot v5.01 against the NCBI C. mydas C. mydas proteome

(derived from the full draft genome) (accessed on the 2nd of February 2017). Search parameters included digestion with trypsin, cys alkylation set to iodoacetamide, and thorough search with false discovery rate (FDR) analysis. FDR was calculated at both the global and local level with the global calculations taking into account the whole identification list and local calculations based on the rate of incorrect identifications with a similar ‘N’ ranking as each protein within the database. Quantitative protein data as obtained by DIA or SWATH-

MS was matched against the ion library using PeakView (ABS Sciex) software. The resulting dataset was formatted for further analysis using a python script developed by Kerr et al.

(2018) before further analysis in R v2.2.7. To assess if our digestion and clean-up method was adequate and to compare protein identifications to the literature, the number of proteins and peptides identified at the 1% and 5% FDR threshold were reported (see Table 5.2). The

ScieX SWATH 2.0 data processing software calculates the FDR using both chromatograms generated from fragment ion data as well as spectral components to calculate FDR. Multiple chromatographic (peak height, retention time, consistency of peaks) and spectral (mass

160 accuracy, contiguous ion series) variables are used to make up the score. Then a pseudo- reverse sequence is generated for each peptide and analysed in the same manner. Following this, the target and decoy sequences are scored as above and ranked based on this score. The

FDR is computed by 2 X the number of decoys in the list and an FDR assigned to each peptide. The peak heights from the chromatographic outputs generated from fragment ion data are used for relative quantification of peptides between samples, actual quantification is not performed.

5.6.2 Comparison of protein profiles between sites

To investigate whether there was different overall protein expression across sites, Principal

Component Analysis (PCA) was performed using the PCA() function in the FactoMineR 1.41 package for R 3.5.3. The results of the PCA analysis were plotted using the factoextra 1.0.5 package.

To investigate if turtles from each site expressed unique protein profiles compared to the other sites, we grouped all individuals from each site and compared relative protein expression between all combinations of sites. In order to do this, a HighNil comparison in the

MSStats package (Choi et al. 2014) in R v2.2.7 was performed in which typically protein expression is quantified from test samples or sites relative to a control sample or site. As there was no genuine control site, we simply compared all combinations of sites to each other to obtain relative protein expression. Moreton Bay (MB) was compared against HB resulting in fold change of proteins within HB turtles relative to MB and the same was done for PG against HB and finally PG against MB. The fold change (positive or negative) of protein abundance relative to the comparison site along with the significance of this change was reported. Heatmaps were made to compare the expression profiles of significantly (p<0.0001) differentially expressed proteins between each combination of sites.

161 5.6.3 Identifying potential markers through analysis of protein function expression

In order to identify potential chemical biomarker candidates, the list of differentially expressed proteins was refined by including only the top ten most strongly dysregulated (five up and five down) proteins from each site comparison for further analysis. The common name of protein in this list was then searched in Google Scholar along with the terms ‘turtle’

‘biomarker’ ‘exposure’ in order to determine if they had previously been recognised as potential biomarkers. If there was no associated turtle biomarker, the protein name was then searched alongside the terms ‘biomarker’ and ‘exposure’ to see if these proteins had been associated with contaminant exposure in any other species. Finally, in order to gain insight into the physiological systems that were associated with the differential protein expression, each protein was searched in the literature for its reported biological function and these data were recorded (Table 5.4).

5.6.4 Organohalogen concentrations in sea turtle plasma

Shapiro-Wilk normality tests were performed to test whether the data were normally distributed. As this was not the case and the number of samples with concentrations above

LOQ were relatively low, non-parametric tests (Kruskal-Wallis) were used to investigate differences in concentrations between the three sites. Data were analysed using GraphPad

Prism v7.0 and the level of significance was set at α = 0.05.

5.7 Results and Discussion

5.7.1 Non-targeted protein analysis of sea turtle plasma

In these green turtle plasma samples, we identified 116 and 133 proteins within the 1% and

5% FDR threshold, respectively, calculated at the global level (Table 5.2). The number of proteins detected in these samples was within the range of what would be expected from crude plasma protein extracts. Without immunoaffinity depletion of high abundance proteins, non-targeted MS of blood plasma will typically yield 150-200 protein identifications at a 5%

162 FDR threshold (Tu et al. 2010), which is low when compared to the hundreds typically seen from tissue or cell extractions. While the use of more complex prefractionation techniques may yield greater protein identifications, it is preferable to maintain a simple protocol for biomarker discovery in wildlife. Therefore, the methods employed here provide a sufficient range of proteins to explore for biomarker discovery without the need for additional sample processing steps.

Table 5.2. Number of proteins and peptides identified at 1% and 5% false discovery rate thresholds calculated at both the local and global level.

Protein 1% Peptide 1% Protein 5% Peptide 5%

Local 108 1631 108 1935

Global 116 1990 133 2354

5.7.2 Protein expression profile between sites

Combining all the data in a PCA analysis did not produce dramatically different groups

(Figure S5.1) despite clear inter-site variability of protein expression evident by number of proteins that were significantly (p<0.01) differentially expressed between the three populations (Figure S5.2). The clearest separation of the three sites was produced when comparing PC1 and PC4 (Figure 5.1). This comparison revealed that one of the strongest driving factors of variation between the sites was the differential expression of immune- related proteins, in particular, antigen wc1.1., immunoglobulins, complement-system proteins, and alpha-fetoprotein (Table S5.1). For example, antigen wc1.1 is a known marker of inflammation in cows (Table 5.4) and may have a similar role in sea turtles; alpha- fetoprotein plays an important role in modulation immunity in humans (Kong, et al., 2012)

163 and complement-proteins are involved in the clearing of microbes and support innate and adaptive immunity (Carroll, 2004). Additionally, the HighNil comparison (Figure 5.2) showed that immunoglobulin j chain, a protein that assists the secretion of antibodies into mucosa, was more highly expressed at MB than the other two sites (Figure 5.2) and is known to be a first line of defence against pathogens in humans (Johansen et al. 2000). This would indicate that MB turtles overall have higher levels of inflammation as this marker is more strongly expressed at this site compared to both of the other sites (Figure 5.2). Haemoglobin and fibrinogen were also drivers of PC4 and could be indicative of poor health. Haemoglobin is typically very stable in green sea turtles and the plasma concentration doesn’t tend to change, regardless of breeding status or diet (Alkindi & Mahmoud, 2002), and hence fluctuations in the plasma concentration may indicate a general change in health status

(Oliveira-Junior, et al., 2009). Additionally, fibrinogen has been proposed as a marker of health along with other haematological parameters in sea and freshwater turtles, however it hasn’t shown much promise as yet (Li, et al., 2015; Moore, et al., 2015).

164

Figure 5.1. Principal component analysis (PCA) plotting PC1 (36.5% of the variance) against

PC4 (4.2% of the variance). The PC1 axis is indicative of overall protein expression (from low on the left to high on the right), while the PC4 axis is driven mostly by proteins associated with immune responses (from low expression at the bottom to high at the top).

Ellipses indicates the region within 1 SD of the group average (68% probability). PG = Port of Gladstone (n=25), HB = Hervey Bay (n=24), MB = Moreton Bay (n=24).

165

Figure 5.2. Heatmap comparing the protein expression profiles of each site indicating the five mostly strongly up-regulated and five most strongly down-regulated proteins for each site comparison. A positive number (orange) indicates that expression at the second site was higher than at the first site (e.g., adiponectin-like protein was 1.368× more expressed at MB than at HB). A negative number (blue) indicates a lower expression at the second site compared to the first site. A blank cell indicates that expression of this protein was not significantly different between the two sites. Several of the proteins were most strongly up or down regulated for more than one comparison, hence there are only 23 proteins listed. In addition, as a number of these proteins have multiple forms (e.g., alpha-2-macroglobulin), and there are two uncharacterised proteins, this list effectively includes 15 distinct and identifiable proteins whose function is described in Table 5.4. HB = Hervey Bay (n=24), PG = Port of Gladstone (n=24) and MB = Moreton Bay (n=24).

166 Table 5.3. Summary of the total number of significantly (p<0.0001) differentially expressed proteins between sites and the lowest and highest value of fold change for each comparison

(derived from Figure S5.2). FC = fold change.

Total # proteins Value of highest FC Value of lowest FC

HB vs. MB 62 3.576 -2.646

HB vs. PG 57 2.216 -2.636

MB vs. PG 64 3.491 -2.685

To obtain relative quantitative information on protein expression at each site, we compared the protein expression profiles of each population to each other (Figure 5.2). We observed more than 73% of the identified proteins (85 out of 116) were differentially expressed between the sites: 62 between HB-MB, 57 between HB-PG and 64 between MB-PG (Figure

S5.1). Interestingly, MB produced the highest number of differentially expressed proteins when compared to the other two sites, with the most pronounced fold change in both up- and down-regulation (Table 5.3; Figure S5.2). Given that the composition of blood plasma proteins can vary greatly even between genetically identical individuals (Liu et al. 2015) it is notable that population level protein expression was not masked by individual variability. It is well known that the presence and concentration of individual proteins within the blood is influenced by many factors such as diet, stress, movement, sleep cycle and many other behavioural and environmental influences. Considering this, it would be logical to assume that protein expression profiles would vary greatly between individuals within the same population and that this variability would mask any commonly over or under expressed proteins within a sample set. Yet significant differences in protein expression between sites were found in our results. This seems to suggest that the differentially expressed proteins identified here may represent a physiological response to powerful and differing environmental factors at each of the sites.

167 Delving deeper into the identity and function of the most highly dysregulated proteins between sites yielded some interesting insights. Indeed, a majority (11/17) of the most highly dysregulated proteins are involved in immune regulatory pathways in vertebrates, with seven being acute phase proteins (APP) (Table 5.3). This agrees with the results of the PCA analysis comparison (Figure 5.1 and Table S5.1), which indicated that most of the proteins that were different between sites were related to immune function. More specifically, the level of APP alpha-2-macroglobulin-like protein was significantly higher in MB and HB turtles compared to PG (Figure 5.2). An increase in circulating APPs is a well-established indicator of inflammation associated with an acute immune response within an organism

(Gruys et al. 2005). Furthermore, two isoforms of fibrinogen, another important marker of inflammation (Davalos and Akassoglou 2012), were elevated in MB turtles compared to the other sites (Figure 5.2). In addition, pentraxin fusion protein was highly up regulated in the

MB population relative to HB (Figure 5.2). Pentraxin proteins have been identified as regulators of macrophage activity and act to mitigate inflammation (Shiraki et al. 2016). The green turtles at MB and HB may therefore be in an increased state of compromised immune function. It is well known that organohalogen (particularly PCB) exposure can cause immune system modulation in marine and freshwater turtles (Keller et al. 2006, Yu et al. 2012,

Rousselet et al. 2017). Particularly, haptoglobin, a known marker of organohalogen exposure in canines (Sonne et al. 2007), was upregulated in turtles from the MB site, which was also the site of the highest and most widespread PCB153 exposure (Figure S5.3). There is therefore some evidence that the up-regulation of these immune related proteins is related to contaminant exposure.

168 Table 5.4. Functional role of the 10 most strongly up and down regulated from each site comparison (from Figure 5.2) as described in previous studies. APP = acute phase protein. Only 15 of the 17 proteins are described here as two of the proteins listed in Figure 5.2 are uncharacterized and therefore no functional information currently exists.

Protein Function Potential marker for Species Reference

Adiponectin Immune Multiple diseases (it is anti-inflammatory, antiatherogenic, Human Sun et al. (2009) antidiabetic). Alpha-1- Balance action of Plasma marker for effects of oil (PAH and alkyl phenols) and produced Atlantic Bohne-Kjersem et antitrypsin (APP) protease enzymes in water in fish. cod al. (2009) the lung Alpha-2- Proteinase inhibitor Wide variety of functions helps bind a number of different cytokines Human Rehman et al. macroglobulin and inhibit almost all proteinase. Said to protect the body against a (2013) like (APP) number of infections and can be used as a marker for a number of diseases. Antigen wc1.1 Immune/inflammation Proinflammatory. Circulating T-cells that produce chemokine Bovine McGill et al. macrophage inflammatory proteins. (2013) Apolipoprotein Bind lipids Differentially regulated in various diseases states including aging and Human Soares et al. age-related diseases (Alzheimer’s) markers of neurological disorders? (2012); Mayeux (2004) Ceruloplasmin Inflammation Marker for chronic inflammation and stress in pets an acute phase Pets Cray (2013) (APP) protein (app) a part of the acute phase response which is an (dogs) inflammatory reaction of the immune system. M Complement c4 Inflammation Complement system part of the immune system that helps antibodies Human Finehout et al. (APP) and phagocytic cells to clear microbes and damages cells from an (2005) organism. Marker for neurodegenerative disease Complement Inflammation Complement system includes over 30 circulating blood and cell bound Human Markiewski and factor (APP) proteins that acts as a defence against microbial pathogens, removes Lambris (2007) immune complexes, apoptotic cells and the products of inflammation.

169 Fibrinogen alpha Blood clotting cascade Large role in the formation of blood clots. Human Tousoulis et al. chain (APP) (2011); Mosesson (2003) Haptoglobin Inflammation Increases in plasma as an inflammation initiation event-is anti- Human Wang et al. (2001); (APP) inflammatory. Decreased in exposed sled dogs to PCBs, possibly a Sonne et al. marker of OH exposure. OH/PCB exposure associated with adverse (2007); Beckmen immune effects. et al. (2003) Immunoglobulin j Immune Immunoglobulin J chain assists IgM or IgA to be secreted into mucosa. Human Johansen et al. chain Secretory antibodies are the first line of defence against pathogens and (2000) noxious substances that favour mucosa as their entry. Pentraxin fusion Immune Family of proteins consisting of CRP, serum amyloid. Many functions Human Gewurz et al. protein and they interact with complement system (phagocytes and (1995) leukocytes). Plasma protease Cancer/inflammation Plays a role in regulating physiological pathways such as complement Human Davis et al. (2008) c1 inhibitor activation, blood coagulation, fibrinolysis and the generation of kinins as well as the suppression of inflammation. Regenerating islet Cancer/inflammation Over expressed in drug resistant colon cancer cells. Mainly in gastro Human Parikh et al. (2012) derived protein 4 of the gut intestinal tract and upregulated during inflammation of the gut. Serum albumin Transport Primarily a carrier protein in blood plasma however some evidence for Human Anraku et al. (APP-negative) antioxidant activity. (2001)

170 5.7.3 Contaminant exposure and differential protein expression

Of all compounds targeted in the analysis of organohalogens, only five PCB congeners, two chlordanes, HCB, one PBDE congener and two MeO-PBDEs were detected (Table

5.5). When detected, the concentrations of the measured organohalogens (particularly

PCBs) were relatively low compared to previous studies on marine turtles in this region

(van de Merwe et al. 2010) and globally (Camacho et al. 2014; Keller, et al. 2004). The concentrations of MeO-PBDEs were highest, followed by PCBs and HCB (Table S5.2,

Figure S5.3). In terms of differences in concentrations between sites, MeO-PBDE, TN and BDE47 were the only three compounds in which statistically significant (p<0.01) variation in concentrations were present (p<0.0001; p=0.0071; p=0.0079, respectively)

(Table S5.2a). However, posthoc analyses only differentiated significant differences between sites in MeO-PBDEs with MB having significantly higher concentrations than

PG (p<0.0001) or HB (p<0.001) (Table S5.2b). MeO-PBDEs are naturally occurring

PBDE analogues produced by algae or sponges and are therefore not considered anthropogenic POPs. It is interesting to note, however, that the highest concentration of

TN was at MB, and TN was only detected in turtles from this site (Table 5.5). While the concentrations are very low and therefore unlikely to be causing adverse health effects at this stage, these results do indicate that MB turtles are more highly exposed to contaminants than the other two populations.

Within the PCBs, CB 153 and 138 were present at the highest concentrations and together accounted for 67-70% of all PCBs (Table 5.5), but these concentrations were very low compared to the values reported by previous studies on turtles from different regions. For example, van de Merwe et al. (2010) reported an average of 137 ± 31.8 pg/mL for 16 green turtles in Queensland; Camacho et al. (2013) reported an average of

47.0 ± 60.0 pg/mL for 50 loggerhead turtles in Cape Verde; and Ragland et al. (2011) reported an average of 3580 ± 5300 pg/mL in loggerhead turtles in NC, USA (note that 171 values that were originally reported as pg/g wet weight have converted to pg/mL using a density of 1025 kg/m3 for blood). This is likely due to the fact that PCB concentrations have continued to decrease in the environment over the last decade following strict regulation on manufacture and use (Wang et al. 2016), and are therefore currently less of a threat to wildlife.

172

Table 5.5. Concentrations (pg/mL) of organohalogens detected in green sea turtle plasma collected from Hervey Bay (n=25), Port of Gladstone (n=24) and Moreton Bay

(n=23). SD = standard deviation, *indicates that this compound was only detected in one replicate at this site hence there is no reported SD. ND = not detected. Note that the following compounds were measured but not detected in any of the plasma samples

(with detection limits ranging from 1-4 pg/mL): CB 28, 49, 52, 74, 95, 99, 101, 105,

118, 128, 146, 156, 170, 171, 177, 183, 194, 199, 206 and 209; BDE 28, 99, 100, 153,

154 and 183), p,p’-DDE, p,p’-DDT, chlordanes (oxychlordane, cis-nonachlor, cis- chlordane) and HCH (α-, β-, γ-).

Organohalogen Hervey Bay Port of Gladstone Moreton Bay Mean ± SD (n=25) Mean ± SD (n=24) Mean ± SD (n=23)

PCB 138 4.34 ± 2.56 3.75 ± 2.13 4.91 ± 1.78 PCB 153 4.33 ± 3.87 4.14 ± 3.24 6.21 ± 2.80 PCB 180 2.47 ± 1.94 1.94 ± 1.32 2.50 ± 1.24 PCB 187 1.30 ± 0.11 1.52 ± 0.49 1.59 ± 0.60 PCB 196/203 ND ND 1.19* TN 1.83 ± 0.19 0.92 ± 0.03 2.93 ± 0.84 TC 1.47* ND 1.13* HCB 3.07 ± 2.07 4.56 ± 3.34 3.12 ± 2.32 BDE 47 ND ND 5.05 ± 3.38 2-MeO-BDE68 56.5 ± 14.5 5.83 ± 5.46 29.4 ± 32.9 6-MeO-BDE47 47.9 ± 131.5 3.77 ± 1.75 ND

PCB=polychlorinated biphenyl; TN=trans-nonachlor; TC=trans-chlordane;

HCB=hexachlornobenzene; BDE=brominated diphenyl ethers

The low concentrations and similar levels between sites suggest that organohalogens are unlikely to significantly contribute to the differentially expressed proteins observed here in the different C.mydas populations. However, there are other contaminants including other organic chemicals and metals potentially at these sites, which could be

173 contributing to the differences in plasma protein expression observed. The presence of some of these contaminants can be inferred by land use and previous environmental monitoring at these sites. The Port of Gladstone is a fast growing industrialised and urbanised harbour with significant trace metal concentration in the harbour, including cadmium, copper and nickel (Angel et al. 2010). Hervey Bay is less developed than the other two sites but the land is heavily used for agriculture and the predominant contaminants are herbicides such as diuron and atrazine, which have been measured in the seagrass sediment in the surrounding waters (McMahon et al. 2005). Moreton Bay is the most developed coastal region of the three sites and studies have reported high concentrations of both organic and non-organic contaminants in the region either measured in local marine wildlife (Hermanussen et al. 2008, Weijs et al. 2016) and sediments (Morelli et al. 2012). Hence the differing contaminant profiles at the three sites could be influencing the variance in plasma protein expression profiles between populations.

Interestingly, ceruloplasmin, a protein more highly expressed in PG turtles compared to the other two sites, has been identified as a biomarker of copper exposure in humans

(Saha et al. 2008). Seeing as PG has a heavier mining usage in its catchment, this may indicate that turtles in this region are being more highly exposed to copper. Indeed, a

2012 study by Gaus et al. reported higher than normal levels of copper in sea turtles inhabiting PG, supporting our findings here. This indicates that the population level observation of increased ceruloplasmin may be a potential marker for copper exposure in sea turtles and could be further explored as such. Similarly, another differentially expressed protein here, adiponectin, has been positively correlated with serum concentrations of perfluorinated compounds (PFCs) in humans (Lin et al. 2011). This protein is more highly expressed at both MB and PG (Figure 5.2) and could potentially

174 indicate a higher presence of PFCs at these two sites, which is likely, particularly for

MB, given the release of PFCs from the major airport in this area (Aylward, 2018).

In addition, many other factors such as diet, behaviour, external stress (e.g., increased boating activity) are bound to have an effect on the physiology of the turtles and therefore the plasma proteome. The clear difference in protein expression between sites is an indicator that non-targeted proteomics is a useful tool to observe population level biomarkers of exposure and effect. However, it is imperative that a multifaceted data collection approach is utilised in the future to allow for more concrete correlations between protein expression and environmental factors to be established.

5.8 Conclusions

Our study examined the sea turtle plasma proteome for the first time and revealed that the expression of circulating can be distinctly different between small subsets of populations of wild sea turtles despite intra-individual variability. This is the first indication that non-targeted proteomic analysis of blood plasma may be useful for monitoring the health of wild sea turtle populations. Differences of proteins between sites were evident and these proteins appeared to be important indicators of health

(immune function). While immune function depression has previously been linked to

PCB exposure in marine animals, organohalogen (including PCB) concentrations were low in turtles from all sites, suggesting that these contaminants are unlikely to be the cause of the difference in proteins related to immune function observed here. The differences could be caused by a number of other factors, including other contaminants or pathogens. This study also identified a number of proteins that could be explored further as biomarkers of chemical exposure, including ceruloplasmin and adiponectin, potential markers for Cu and PFC exposure, respectively.

Future studies on these green turtle populations should probe deeper into causes of the differential immune function protein expression observed. Additionally, future wildlife 175 toxicology research could utilise non-targeted analysis to examine the plasma proteome of wild populations. Non-targeted plasma protein profiling is not widely used in wildlife but appears to be a useful tool for revealing physiological differences occurring at the population level that may be attributed to anthropogenic pressures on these species.

5.9 References

Alkindi, A Y A. and Mahmoud, I Y. 2002. Hematological survey in two species of sea turtles in the Arabian Sea during nesting season. Pakistan J Biol. Sci, 5(3):359- 361. Angel, B M., Hales, L T., Simpson, S L., Apte, S C., Chariton, A A., Shearer, D A. and Jolley, D F. 2010. Spatial variability of cadmium, copper, manganese, nickel and zinc in the Port Curtis Estuary, Queensland, Australia. Mar Freshwater Res 61(2):170-183. Anraku, M., Yamasaki, K., Maruyama, T., Kragh-Hansen, U. and Otagiri, M. 2001. Effect of oxidative stress on the structure and function of human serum albumin. Pharm Res 18:632-639. Atkinson Jr, A. J., W. A., Colburn, V G., DeGruttola, D L., DeMets, G J., Downing, D., Hoth, F., Oates, J A., Peck, C C., Schooley, R T A. and Spilker B A. 2001. Biomarkers and surrogate endpoints: Preferred definitions and conceptual framework. Clin Pharmacol Ther. 69(3):89-85. Beale, D J., Crosswell, J., Karpe, A V., Ahmed, W., Williams, M., Morrison, P D., Metcalfe, S., Staley, C., Sadowsky, M J., Palombo, E A. and Steven, A D L. 2017. A multi-omics based ecological analysis of coastal marine sediments from Gladstone, in Australia's Central Queensland, and Heron Island, a nearby fringing platform reef. Sci Total Environ 609:842-853. Beckmen, K B.,. Blake, J E., Ylitalo, G M., Stott, J L. and O’Hara, T M. 2003. Organochlorine contaminant exposure and associations with hematological and humoral immune functional assays with dam age as a factor in free-ranging northern fur seal pups (Callorhinus ursinus). Mar Pollut Bull 46:594-606. Bergeron, J M., Crews, D. and Mclachlan, J A. 1994. PCBs as environmental estrogens - turtle sex determination as a biomarker of environmental contamination. Environ health Persp 102:780-781.

176

Bohne-Kjersem, A., Skadsheim, A., Goksoyr, A. and Grosvik, B E. 2009. Candidate biomarker discovery in plasma of juvenile cod (Gadus morhua) exposed to crude North Sea oil, alkyl phenols and polycyclic aromatic hydrocarbons (PAHs). Mar Environ Res 68:268-277. Camacho, M., Boada, L D., Oros, J., Lopez, J., Zumbado, M., Almeida-Gonzalez, M. and Luzardo O P. 2014. Monitoring organic and inorganic pollutants in juvenile live sea turtles: results from a study of Chelonia mydas and Eretmochelys imbricata in Cape Verde. Sci Total Environ 481:303-310. Camacho, M., Luzardo, O P., Boada, L D., Lopez Jurado, L F., Medina, M., Zumbado, M. and Oros, J. 2013. Potential adverse health effects of persistent organic pollutants on sea turtles: evidences from a cross-sectional study on Cape Verde loggerhead sea turtles. Sci Total Environ 458-460:283-289. Chaousis, S., F. D L Leusch, and J P. van de Merwe. 2018. Charting a path towards non-destructive biomarkers in threatened wildlife: A systematic quantitative literature review. Environ Pollut 234:59-70. Choi, M., Chang, C Y., Clough, T., Broudy, D., Killeen, T., MacLean, B. and Vitek, O. 2014. MSstats: an R package for statistical analysis of quantitative mass spectrometry-based proteomic experiments. Bioinformatics 30:2524-2526. Corsolini, S., Aurigi, S. and Focardi, S. 2000. Presence of polychlorobiphenyls (PCBs) and coplanar congeners in the tissues of the mediterranean loggerhead turtle Caretta caretta. Mar Pollut Bull 40:952-960. Cray, C. 2013. Biomarkers of inflammation in exotic pets. J Exot Pet Med 22:245-250. Davalos, D., and Akassoglou, K. 2012. Fibrinogen as a key regulator of inflammation in disease. Semin Immunopathol 34:43-62. Davis, A E., Mejia 3rd, P. and Lu, F. 2008. Biological activities of C1 inhibitor. Mol Immunol 45:4057-4063. de Swart, R L., Ross, P S., Vos, J G. and Osterhaus, A D M E. 1996. Impaired immunity in harbour seals (Phoca vitulina) exposed to bioaccumulated environmental contaminants: review of a long-term feeding study. Environ Health Persp 104:5.

Dondero, F., Negri, A., Boatti, L., Marsano, F., Mignone, F. and Viarengo, A. 2010. Transcriptomic and proteomic effects of a neonicotinoid insecticide mixture in the marine mussel (Mytilus galloprovincialis, Lam.). Sci Total Environ 408(18):3775-3786.

177

Dos Santos, C R., Cavalcante, A L., Hauser-Davis, R A., Lopes, R M. and Mattos Rde, C. 2016. Effects of sub-lethal and chronic lead concentrations on blood and liver ALA-D activity and hematological parameters in Nile tilapia. Ecotoxicol Environ Saf 129:250-256. Eisenreich, K M., Kelly, S M. and Rowe, C L. 2009. Latent mortality of juvenile snapping turtles from the Upper Hudson River, New York, exposed maternally and via the diet to polychlorinated biphenyls (PCBs). Environ Sci Technol 43:6052-6057. Finehout, E J., Franck, Z. and Lee, K H. 2005. Complement protein isoforms in CSF as possible biomarkers for neurodegenerative disease. Dis Markers 21:93-101. Fossi, M C. 1994. Nondestructive biomarkers in ecotoxicology. Environ Health Perspect 102 Suppl 12:49-54. Geens, A., Dauwe, T., Bervoets, L., Blust, R. and Eens, M. 2010. Haematological status of wintering great tits (Parus major) along a metal pollution gradient. Sci Total Environ 408:1174-1179. George, J., Singh, R., Mahmood, Z. and Shukla, Y. 2010. Toxicoproteomics: New paradigms in toxicology research. Toxicol Mech and Method 20:415-423. Gewurz, H., Zhang, X H. and Lint, T F. 1995. Structure and function of the pentraxins. Curr Opin Immunol 7:54-64. Gruys, E., Toussaint, M J., Niewold, T A. and Koopmans, S J. 2005. Acute phase reaction and acute phase proteins. J Zhejiang Univ Sci B 6:1045-1056. Hathout, Y., Brody, E., Clemens, P R., Cripe, L., DeLisle, R K., Furlong, P., Gordish- Dressman, H., Hache, L., Henricson, E., Hoffman, E P., Kobayashi, Y M., Lorts, A., Mah, J K., McDonald, C., Mehler, B., Nelson, S., Nikrad, M., Singer, B., Steele, F., Sterling, D., Sweeney, H L., Williams, S. and L. Gold. 2015. Large- scale serum protein biomarker discovery in Duchenne muscular dystrophy. Proc Natl Acad Sci USA 112:7153-7158. Hermanussen, S., Matthews, V., Papke, O., Limpus, C J. and Gaus, C. 2008. Flame retardants (PBDEs) in marine turtles, dugongs and seafood from Queensland, Australia. Mar Pollut Bull 57:409-418. Hutchinson, T H., Lyons, B P., Thain, J E. and Law, R J. 2013. Evaluating legacy contaminants and emerging chemicals in marine environments using adverse outcome pathways and biological effects-directed analysis. Mar Pollut Bull 74:517-525.

178

Ilizaliturri-Hernandez, C A., Gonzalez-Mille, D J., Mejia-Saavedra, J., Espinosa-Reyes, G., Torres-Dosal, A., and Perez-Maldonado, I. 2013. Blood lead levels, [delta]- ALAD inhibition, and hemoglobin content in blood of giant toad to asses lead exposure in three areas surrounding an industrial complex in Coatzacoalcos, Veracruz, Mexico. Environ Monit Assess 185:1685. Johansen, F E., Braathen, R. and Brandtzaeg, P. 2000. Role of J chain in secretory immunoglobulin formation. Scand J Immunol 52:240-248. Keller, J M., Kucklick, J R., Stamper, M A., Harms, C A. and McClellan-Green, P D. 2004. Associations between organochlorine contaminant concentrations and clinical health parameters in loggerhead sea turtles from North Carolina, USA. Environ Health Persp 112:1074-1079. Keller, J M., McClellan-Green, P D., Kucklick, J R., Keil, D E. and Peden-Adams, M M. 2006. Effects of organochlorine contaminants on loggerhead sea turtle immunity: comparison of a correlative field study and in vitro exposure experiments. Environmental Health Perspectives 114:70-76. Kerr, E D., Phung, T K., Fox, G P., Platz, G J. and Schulz, B L. 2018. The intrinsic and regulated proteomes of barley seeds in response to fungal infection. Analytical Biochemistry. Lee, H J., Lee, E Y., Kwon, M S. and Paik, Y K. 2006. Biomarker discovery from the plasma proteome using multidimensional fractionation proteomics. Curr Opin Chem Biol 10:42-49. Lin, C Y., Wen, L L., Lin, L Y., Wen, T W., Lien, G W., Chen, C Y., Hsu, S H., Chien, K L., Sung, F C., Chen, P C. and Su, T C. 2011. Associations between levels of serum perfluorinated chemicals and adiponectin in a young hypertension cohort in Taiwan. Environ Sci Technol 45:10691-10698. Liu, Y., Buil, A., Collins, B C., Gillet, L C., Blum, L C., Cheng, L Y., Vitek, O., Mouritsen, J., Lachance, G., Spector, T D., Dermitzakis, E T. and Aebersold, R. 2015. Quantitative variability of 342 plasma proteins in a human twin population. Mol Syst Biol 11:786. Markiewski, M M. and Lambris, J D. 2007. The role of complement in inflammatory diseases from behind the scenes into the spotlight. Am J Pathol 171:715-727. Martyniuk, C J., and Simmons, D B. 2016. Spotlight on environmental omics and toxicology: a long way in a short time. Comp Biochem Physiol Part D 19:97- 101.

179

Mayeux, R. 2004. Biomarkers: potential uses and limitations. NeuroRx 1:182-188. McGill, J L., Nonnecke, B J., Lippolis,J D., Reinhardt, and T A., Sacco, R E. 2013. Differential chemokine and cytokine production by neonatal bovine gammadelta T-cell subsets in response to viral toll-like receptor agonists and in vivo respiratory syncytial virus infection. Immunology 139:227-244. McMahon, K., Bengtson Nash, S., Eaglesham, G., Muller, J F., Duke, N C. and Winderlich, S. 2005. Herbicide contamination and the potential impact to seagrass meadows in Hervey Bay, Queensland, Australia. Mar Pollut Bull 51:325-334. Ming-Ch'eng Adams, C I., Baker, J E. and Kjellerup, B V. 2016. Toxicological effects of polychlorinated biphenyls (PCBs) on freshwater turtles in the United States. Chemosphere 154:148-154. Morelli, G., Gasparon, M., Fierro, D., Hu, W P. and Zawadzki, A. 2012. Historical trends in trace metal and sediment accumulation in intertidal sediments of Moreton Bay, southeast Queensland, Australia. Chem Geol 300-301:152-164. Mosesson, M W. 2003. Fibrinogen gamma chain functions. J Thromb Haemost 1:231- 238. Osterhaus, A D M E., Groen, J., UytdeHaag, F G C M., Visser, I K G., Bildt, M., Bergman, A., and Klingeborn, B. 1989. Distemper virus in Baikal seals. Nature 338. Parikh, A., Stephan, A F. and Tzanakakis, E S. 2012. Regenerating proteins and their expression, regulation and signaling. Biomol Concepts 3:57-70. Ragland, J M., Arendt, M D., Kucklick, J R. and Keller, J M. 2011. Persistent organic pollutants in blood plasma of satellite-tracked adult male loggerhead sea turtles (Caretta caretta). Environ Toxicol Chem 30:1549-1556. Rehman, A A., H. Ahsan, and F H. Khan. 2013. alpha-2-Macroglobulin: a physiological guardian. J Cell Physiol 228:1665-1675. Rousselet, E., Levin, M., Gebhard, E., Higgins, B M., DeGuise, S. and Godard- Codding, C A J. 2017. Polychlorinated biphenyls (PCBs) modulate both phagocytosis and NK cell activity in vitro in juvenile loggerhead sea turtles (Caretta caretta). J Toxicol Environ Health A 80:556-561. Saha, A., Karnik, A., Sathawara, N., Kulkarni, P. and Singh, V. 2008. Ceruloplasmin as a marker of occupational copper exposure. J Expo Sci Env Epid 18:332-337.

180

Shiraki, A., Kotooka, N., Komoda, H., Hirase, T., Oyama, J I. and Node, K. 2016. Pentraxin-3 regulates the inflammatory activity of macrophages. Biochem Biophys Rep 5:290-295. Soares, H D., Potter, W Z., Pickering, E., Kuhn, M., Immermann, F W., Shera, D M., Ferm, M., Dean, R A., Simon, A J., Swenson, F. and Siuciak, J A. Biomarkers consortium alzheimer's disease plasma proteomics project. 2012. Plasma biomarkers associated with the apolipoprotein E genotype and Alzheimer disease. Arch Neurol 69:1310-1317. Sonne, C., Fonfara, S., Dietz, R., Kirkegaard, M., Letcher, R J., Shahmiri, S., Andersen, S., and Moller, P. 2007. Multiple cytokine and acute-phase protein gene transcription in West Greenland sledge dogs (Canis familiaris) dietary exposed to organic environmental pollutants. Arch Environ Contam Toxicol 53:110-118. Stastna, M. and Van Eyk, J E. 2012. Secreted proteins as a fundamental source for biomarker discovery. Proteomics 12:722-735. Sun, Y., Xun, K., Wang, C., Zhao, H., Bi, H., Chen, X. and Wang, Y. 2009. Adiponectin, an unlocking adipocytokine. Cardiovasc Ther 27:59-75. Tousoulis, D., Papageorgiou, N., Androulakis, E., Briasoulis, A., Antoniades, C. and Stefanadis, C. 2011. Fibrinogen and cardiovascular disease: genetics and biomarkers. Blood Rev 25:239-245. Tu, C., Rudnick, P A., Martinez, M Y., Cheek, K L., Stein, S E., Slebos, R J. and Liebler, D C. 2010. Depletion of abundant plasma proteins and limitations of plasma proteomics. J Proteome Res 9:4982-4991. van de Merwe, J P., Hodge, M., Olszowy, H A., Whittier, J M. and Lee, S Y. 2010. Using blood samples to estimate persistent organic pollutants and metals in green sea turtles (Chelonia mydas). Mar Pollut Bull 60:579-588. Van Loveren, H., Ross, P., Osterhaus, A. and Vos, J. 2000. Contaminant-induced immunosuppression and mass mortalities among harbor seals. Toxicol Lett. 112:319-324. Vijayasarathy, S., Weijs, L., Grant, S., Gallen, M. and Gaus, C. 2019. PCDD/F and PCB levels in different tissues from dugongs (Dugong dugon) inhabiting the Queensland coastline. Mar Pollut Bull, 139:23-31.

181

Wang, X., Thai, P K., Li, Y., Li, Q., Wainwright, D., Hawker, D W. and Mueller, J F. 2016. Changes in atmospheric concentrations of polycyclic aromatic hydrocarbons and polychlorinated biphenyls between the 1990s and 2010s in an Australian city and the role of bushfires as a source. Environ Pollut 213:223- 231. Wang, Y., Kinzie, E., Berger, F G., Lim, S K. and Baumann, H. 2001. Haptoglobin, an inflammation-inducible plasma protein. Redox Rep 6:379-385. Weijs, L., Vijayasarathy, S., Villa, C A., Neugebauer, F., Meager, J J. and Gaus, C. 2016. Screening of organic and metal contaminants in Australian humpback dolphins (Sousa sahulensis) inhabiting an urbanised embayment. Chemosphere 151:253-262. Wisniewski, J R., Zougman, A., Nagaraj, N. and Mann, M. 2009. Universal sample preparation method for proteome analysis. Nat Methods 6:359-362. Yadetie, F., Oveland, E., Døskeland, A., Berven, F., Goksøyr, A., and Karlsen, O A. 2017. Quantitative proteomics analysis reveals perturbation of lipid metabolic pathways in the liver of Atlantic cod (Gadus morhua) treated with PCB 153. Aquat Toxicol, 185, 19-28. Yu, S., Halbrook, R S. and Sparling, D W. 2012. Accumulation of polychlorinated biphenyls (PCBs) and evaluation of hematological and immunological effects of PCB exposure on turtles. Bull Environ Contam Toxicol 88:823-827. Zota, A R., Geller, R J., Romano, L E., Coleman-Phox, K., Adler, N E., Parry, E., Wang, M., Park, J S., Elmi, A F., Laraia, B A. and Epel, E S. 2018. Association between persistent endocrine-disrupting chemicals (PBDEs, OH-PBDEs, PCBs, and PFASs) and biomarkers of inflammation and cellular aging during pregnancy and postpartum. Environ Int 115:9-20.

182

Supporting Information for Chapter 5

a)

b) c)

d) e)

Figure S5.1. Results of the PCA analysis, showing: (a) Scree plot of the top 25 dimensions; (b) PC1 vs PC2; (c) PC1 vs PC3; (d) PC1 vs PC4; and (e) PC1 vs PC5.

183

Figure S5.2 Heatmap of protein expression comparisons between the three sites with all proteins significantly (p<0.0001) differentially expressed included. A positive number

(blue) indicates that expression at the second site was higher than at the first site (e.g., adiponectin-like protein was 1.368× more expressed at MB than at HB). A negative number (orange) indicates a lower expression at the second site compared to the first site. A blank cell indicates that expression of this protein was not significantly different between the two sites. HB = Hervey Bay (n=24), PG = Port of Gladstone (n=24) and

MB = Moreton Bay (n=24).

184

15 15

10 10

5 5

HCB (pg/mL) PCB153 (pg/mL)

0 0 PG HB MB PG HB MB

15 8

6 10

4

5

CB 138 (pg/mL) CB 180 (pg/mL) 2

0 0 PG HB MB PG HB MB

150 3

2 100

1 50

CB 187 (pg/mL)

2-MeO-BDE68 (pg/mL)

0 0 PG HB MB PG HB MB

Figure S5.3. Average concentration (pg/mL) of organohalogen grouped by site: Port of

Gladstone (n=24), Hervey Bay (n=25) and Moreton Bay (n= 23) with error bars representing SD. The only compounds with significantly different means across sites were 2-MeO-BDE68 (p<0.0001), TN (p=0.0071) and BDE47 (p=0.0079). However, a multiple comparisons test only distinguished significantly different concentrations of 2-

MeO-BDE68 between sites (Table S5.2).

185

Table S5.1. Results of the principal component analysis (PCA). Dim. = Dimension.

Dim. variability Proteins contributing more than 3% to this Dim. explained by this Dim. PC1 36.5% alpha-1-antitrypsin-like: 3.50% alpha-1-antiproteinase 2-like: 3.30% venom factor-like: 3.01% PC2 7.17% fibrinogen beta chain isoform X2: 6.36% fibrinogen gamma chain: 4.64% fibrinogen alpha chain: 4.52% serum albumin: 4.38% alpha-fetoprotein: 3.36% alpha-1-antitrypsin-like: 3.22% haptoglobin: 3.17% immunoglobulin lambda-like polypeptide 5-like: 3.03% PC3 5.45% alpha-2-macroglobulin-like: 12.3% fetuin-B: 5.42% regenerating islet-derived protein 4-like: 5.40% acetylcholinesterase-like: 5.38% apolipoprotein Eb-like: 4.95% apolipoprotein C-II-like: 3.84% ceruloplasmin isoform X1: 3.81% carboxypeptidase B2: 3.36% apolipoprotein C-III: 3.28% PC4 4.22% antigen WC1.1-like: 9.59% fibrinogen gamma chain: 6.25% fibrinogen alpha chain: 5.63% immunoglobulin lambda-like polypeptide 5-like: 5.02% hemoglobin subunit alpha-A-like: 4.87% alpha-fetoprotein: 4.37% hemoglobin subunit beta-like: 4.19% uncharacterized protein LOC102934447: 3.79% complement component C7: 3.75% complement factor H-like: 3.09% PC5 3.69% putative V-set and immunoglobulin domain-containing-like protein IGHV4OR15-8-like: 9.98% complement C3-like: 8.74% tetranectin: 7.46% apolipoprotein Eb-like: 5.03% apolipoprotein C-III: 4.58% inter-alpha-trypsin inhibitor heavy chain H2: 4.24% alpha-2-HS-glycoprotein: 3.80% cytosolic phospholipase A2 gamma-like: 3.22%

186

Table S5.2: Results of non-parametric one-way ANOVA (Kruskal-Wallis) comparing concentrations of selected OH between sites (a), and results of between site comparisons (b) when ANOVA indicated statistical significance (i.e., for 2-MeO-

BDE68, TN and BDE 47). Significant values (p<0.01) are highlighted in bold.

a) b) 2-MeO-BDE68 p-value = Compound p-value = PG v HB 0.5447 PCB138 0.0963 PG V MB <0.0001 PCB153 0.0362 HB V MB <0.001 PCB180 0.2539 PCB187 0.7689 PCB187 0.7689 TN p-value = PG v HB 0.2402 PCB 196/203 0.1429 TN 0.0071 PG V MB 0.0214 HB V MB >0.9999 TC 0.333 HCB 0.4037 BDE 47 p-value = BDE 47 0.0079 PG v HB 0.113 2-MeO-BDE68 <0.0001 PG V MB >0.999 6-MeO-BDE47 0.2263 HB V MB 0.0553

187

Chapter 6: Thesis Summary, Key Findings and Future

Research

6.1 Thesis summary

The overall aim of this thesis was to investigate new non-destructive methods for protein biomarker discovery in threatened wildlife, using the green sea turtle (Chelonia mydas) as a model species. More specifically, in vitro and in vivo approaches, combined with a novel non-targeted proteomic profiling technique were used to investigate: 1) the influence of time and exposure concentration on the global cellular protein expression of in vitro chemical exposure experiments, 2) the potential of in vitro experiments to reveal new biomarkers of chemical exposure, 3) the influence of tissue type on the baseline cellular global protein production and the response to exposure to known environmental contaminants, and 4) differences in in vivo global protein expression in wild-caught turtles from different foraging locations. The application of novel molecular analyses in combination with in vitro and in vivo techniques is still very new to wildlife toxicology, and this thesis therefore provided important advances in this field of science.

In the early stages of this thesis, a systematic quantitative literature review was conducted (Chapter 2) that highlighted the current limitations impeding biomarker discovery in threatened wildlife. This review is the first of its kind and provides valuable insights into the current state of threatened wildlife biomarker discovery methodology. The main method employed by previous studies was a correlative approach in which concentrations of contaminants, either in animal blood and tissue or the surrounding environment, were correlated with the presence of markers measured in vivo. However, the results of this review confirmed the notion that to link changes in blood and tissue biomarkers with contaminant exposure without controlled exposure

188 experiments is difficult. The quantitative assessment of the literature also confirmed that the use of in vitro exposure studies along with non-targeted proteomic analysis has been minimally explored in this field. Given that such methodologies have been successfully applied in the field of biomarker discovery in human disease, it seemed logical to investigate the potential of these tools to enhance studies on contaminant exposure in wildlife. However, molecular based studies come with their own sets of limitations, such as a much higher sensitivity to technical sources of variation, and therefore careful optimisation of these methods is required.

In light of the findings of the literature review, the usefulness of molecular level changes measured in cells exposed to known environmental contaminants as a tool for non-destructive biomarker discovery was assessed. As an important first step in this process, to understand the sources of technical variation in molecular studies, the aim was to test the effects of several experimental parameters on changes in cellular protein expression while also interpreting the toxicological implications of these results

(Chapter 3). In this chapter, a non-targeted analysis of global protein expression in primary sea turtle skin cells was presented, following exposure to three environmentally relevant contaminants. The main aim of this experiment was to optimise methods for non-targeted proteomic analysis of primary sea turtle cells exposed in vitro while exploring the effects of experimental design on the observed protein expression. The results indicated that length of exposure time had a stronger effect on global protein expression than even a 100-fold change in exposure concentration. These results supported the notion that observed changes at the molecular level can be strongly influenced by technical variations, and provided optimal exposure time (24 hr) and chemical concentration (100 µg/L) to use in future in vitro studies of this kind. Chapter

3 also revealed some congruence between protein dysregulation resulting from in vitro exposure with several oxidative stress biomarkers known to be dysregulated in turtles 189 exposed in vivo to PCB153 and PFNA. Furthermore, some proteins not previously reported as biomarkers of chemical exposure were even more strongly dysregulated in these in vitro exposures. The results of this chapter clearly illustrated that in vitro exposure experiments are worth exploring further as a tool for wildlife biomarker discovery.

In follow-up to the findings of Chapter 3, it was important to explore the differences in protein expression of cell cultures from different tissues. The rationalisation for this chapter was that future studies on wildlife would benefit from working with a single tissue type, as it is generally not feasible, due to practical and economic limitations, to use a large range of tissue types in chemical exposure experiments. Therefore, it is important to be aware of the influence of different tissue types on the molecular response to in vitro exposure when aiming to develop a method for high throughput biomarker discovery in threatened wildlife. In Chapter 4, the baseline (control) proteomes were compared between cells from five different tissues (skin, ovary, small intestine, kidney and liver), as well as in response to exposure to known environmental contaminants. Interestingly, the results demonstrated that while control cell protein expression was similar between tissues, changes in the protein expression in response to chemical exposure was different between the tissue types. There was quite a disparity between the molecular response of the different cells types to chemical exposure, and it was not clear which tissue type was best for representing the molecular level effects.

However, it was noted that typically skin cells are easiest to obtain and are often the only reliably non-destructive option. This chapter also revealed that the most strongly dysregulated proteins were not previously identified as markers of exposure in the literature, indicating these highly dysregulated proteins could be new candidate biomarkers. This was the first time that global protein expression of green sea turtle cells for these five tissue types has been reported in addition to changes in the response 190 of these cells to contaminant exposure. The methods presented in this chapter provide a framework for future studies of this nature and the resulting novel protein data for these cells can act as a competitive database for further research on turtle cell proteomics.

Considering the success in observing molecular effects through non-targeted proteomic profiling on cells, and because this method has not been widely performed on wildlife, the next chapter detailed the analysis of global protein expression in blood plasma of sea turtles captured at three different foraging grounds in Southeast Queensland

(Chapter 5). Additionally, the presence of selected organohalogens in the same plasma samples was measured with the view of correlating chemical exposure with in vivo protein expression. The data presented in Chapter 5 revealed that southeast Queensland turtles have relatively low blood organohalogen concentrations. Comparison of plasma proteins showed statistically meaningful differences between protein expression profiles of these populations despite the fact that in vivo protein expression varies greatly between individuals. In addition, significant differences in biomarkers known to indicate altered immune states between the populations was observed. However, due to the low concentrations of organohalogens within the blood these changes could not be correlated with exposure from these particular compounds. There may, however, be other chemical contaminants and/or pathogens not measured that are contributing to the detected changes in plasma protein levels. The observed population level differences in plasma protein expression in this chapter indicated that non-targeted proteomic analysis is a useful tool to explore in vivo exposure markers in wildlife.

6.2 Key findings from this thesis

There were several key findings from this thesis (Table 6.1).

191

Table 6.1. Summary of the objectives and main findings from this thesis.

Objective Chapter Key finding

1) Assess state of non-destructive biomarker discovery methods in 2 • In vitro methods and non-targeted molecular analysis are promising,

wildlife but underutilised, techniques for the purpose of non-destructive

biomarker discovery

2a) Optimise parameters for biomarker discovery experiments 3 • Duration of in vitro exposure had a significant effect on observed

using in vitro exposures coupled with non-targeted proteomics global protein expression

analysis • Chemical concentration did not have a significant influence on

protein expression

• 24 hr exposure and 100 µg/L exposures are optimal conditions for

these types of experiments

3a) Investigate the impact of biological variation (tissue type) on 4 • Cellular protein expression in response to in vitro exposure varies

in vitro protein expression greatly between cells derived from different tissue types

2b) and 3b) Assess how well the observed cellular level effects of 3 & 4 • Changes in cellular protein expression resulting from in vitro

exposure represent those identified from in vivo exposure exposure can indicate whole organism effects (known in vivo markers

were expressed in vitro)

192

3c) Identify potential biomarkers of exposure and effect in sea 4 • Proteins that have not previously been described as markers were turtles more strongly dysregulated than known markers, indicating potential

new biomarker candidates

4) Assess the usefulness of non-targeted proteomics on wild- 5 • Non-targeted analysis of green turtle plasma proteome revealed caught turtle samples for identifying differences in marker population level differences in protein expression expression between populations with differing chemical exposure • Multiple dysregulated proteins indicated altered immune states profiles between the populations

193

Firstly, the literature review (Chapter 2) revealed that in vitro exposure and non-targeted proteomics has not been fully explored as a tool for biomarker discovery. There are currently only four studies that have measured changes in global cellular protein expression following in vitro exposure to a contaminant (Lasserre et al. 2009, Roland et al. 2013, Lin et al. 2014, Vuong et al. 2016). This study used a slightly different method wherein non-targeted mass spectrometry of non-pre fractionated samples were analysed.

This yielded a lot more proteins than were identified in previous studies, indicating that the method employed here is preferable for biomarker studies.

Secondly, our exploration of this method in Chapter 3 revealed that the length of exposure time had a significant effect on overall protein expression, much more so than chemical concentration. This chapter also indicated that 24 hr exposure and 100 µg/L exposures are optimal conditions for in vitro biomarker discovery experiments. As no studies to our knowledge have previously compared these two elements for this method, it is difficult to make comparisons. However, it is interesting to note that a similar study by (Roland et al. 2013) found that a 100-fold increase in concentration resulted in 9-fold more differentially expressed proteins. This may be due to the fact that a less sensitive method for non-targeted proteomics was used compared to our study and highlights the influence that experimental parameters can have on protein expression.

Thirdly, changes in the expression of known in vivo markers of exposure in the cells after in vitro exposure to known environmental contaminants was seen, which indicates that this method can produce results that are relevant to whole organism responses

(Chapter 3). Similarly, Lasserre et al. (2009) found that dysregulated proteins resulting from in vitro atrazine exposure of breast cancer cells represented similar pathways as those activated by in vivo exposure, and likewise Roland et al. (2013) found in vivo markers expressed in vitro. This is a good indicator that in vitro proteomic results can represent in vivo effects, and that molecular changes at this level are relevant for 194 biomarker discovery. It is interesting then to note that there were a number of proteins more strongly dysregulated than these known markers in response to chemical exposure, and that these markers could indicate new previously unknown pathways that are activated as a result of exposure. These proteins may be better candidates for in vivo biomarkers in this species as the response was stronger than that of traditional markers, and future studies could further examine this potential.

Fourthly, we discovered that while cells derived from different tissue types share similar baseline protein expression profiles, the response to contaminant exposure at the protein level varies distinctly (Chapter 4). Again, as this kind of study has not been reported in the literature before, it is difficult to make comparisons. There have, however, been several studies comparing the proteome of tissues and cells derived from different organs of the human body (Ponten et al. 2009, Geiger et al. 2012). Both these studies found that there was minimal (~2%) difference between the protein profile of the tissue types (i.e., minimal qualitative differences), however, the level of protein expression between cells was distinct (i.e., quantitative differences). This may indicate the different pathways that are emphasised over others in certain tissue types, such as greater detoxification in liver cells. It is therefore not surprising that when stimulated by contaminant exposure, we saw tissue-specific variation in the strength of the different activated pathways as represented by changes in protein expression. Overall, as Perkins et al. (2011) highlights, this method of reverse engineering adverse outcome pathways, beginning at the molecular level, is ambitious as there are many complex interactions occurring at the cellular, organ and whole organism levels that are challenging to unravel.

Lastly, it was observed that non-targeted protein expression analysis can be useful for distinguishing between health status of distinct wild sea turtle populations (Chapter 5). 195

Examining the global protein expression in the blood plasma of three distinct foraging populations of green turtles (Chelonia mydas) in southeast Queensland revealed population level differences in protein expression. This was an interesting find given that protein expression in blood plasma is influenced by a multitude of factors, can change rapidly and therefore varies greatly between individuals (Liu et al. 2015). In light of this, it is very significant that some proteins were overall more highly expressed within one population compared to the other. Most notably we identified that immune and inflammation-related markers were higher at the Moreton Bay site compared to the other two sites. These changes were not correlated with contaminant exposure as only low concentrations of organohalogens were detected, but it is possible that this immune alteration is due to exposure to one, or more, of the various other contaminants

(chemical and/or pathogens) that these animals are exposed to. Future studies could further probe the causes of the population level differences in protein expression observed here in relation to contaminant exposure.

6.3 Future research

The novel use of cell cultures and non-targeted molecular analysis for biomarker discovery presented in this thesis has provided invaluable information that can inform future studies of this nature. Non-targeted protein expression appears to provide a wealth of information at both the cellular and whole organism level, and more research into relating the changes in expression resulting from contaminant exposure could be done with the view of harnessing this technology for the improvement of outcomes of biomarker discovery research in threatened species. We anticipate that methods described here will provide a framework for future biomarker research in threatened wildlife. Furthermore, the molecular data collected in this study will provide a baseline for future research on biomarker discovery in sea turtles.

196

Future research could focus on confirming the reproducibility of changes in protein expression observed after in vitro exposure with an emphasis on clarifying other sources of variation not investigated here. An independent repetition of the same experiment here would provide further understanding into the reproducibility of quantitative protein expression data under differing experimental parameters. One of the most important traits of a good biomarker is consistency of expression as a response to exposure.

Therefore, studies focusing on identifying the biomarker candidates which are most consistently dysregulated across repeated experimental trials would be of great value to biomarker discovery in this context.

When attempting to translate in vitro results to whole organisms, it is vital to be aware of the variability that can occur due to biological factors. Here we have tested the differences in various tissue types, however there are many other factors that may produce variable protein responses to in vitro exposure. For example, future studies could test how cells derived from different individuals of differing sex, breeding status, age and health respond to in vitro exposure. Beyond this, examining the molecular response of cells derived from different species of the same genus and furthermore different species all together would provide a greater understanding of the specificity of particular biomarker candidates identified in such experiments. In addition to this, the effects of different chemicals and mixtures could be explored to determine which markers are truly chemical specific and whether identified markers remain the same for chemicals in mixture. In vitro studies could also examine the proteome of extracellular fluid after exposure in an attempt to detect biomarkers that are excreted from the cells. It is logical that proteins detected in blood plasma of whole animals are largely excreted from cells rather than in the intracellular protein matrix which is what was examined in chapters 3 and 4.

197

Furthermore, future toxicological studies on green sea turtle populations should focus on exploring possible causes of the altered immune state of Moreton Bay turtles observed here. As discussed in Chapter 5, there are many potential causes for population level differences in plasma protein levels and future whole organism studies could aim to collect more data on the environment of the target organism in order to gain better insight into potential causes of effects such as altered immune states as observed here. Additionally, further validation of potential in vivo markers of chemical exposure identified here could be investigated through more targeted approaches such as immuno-based assays to confirm the usefulness of these markers.

6.4 Final remarks

This thesis provides valuable knowledge to the field of non-destructive biomarker discovery in threatened wildlife. In summary, the limitations surrounding biomarker discovery in threatened wildlife need to be addressed for conservation efforts to move forward. The research presented here demonstrates that non-targeted proteomic profiling and in vitro exposure experiments shows great promise as a combined tool to enhance biomarker discovery in threatened wildlife. Future research should be directed towards further refining this method and expanding its application to hopefully advance efforts towards the protection of threatened wildlife.

6.5 References

Addison, R. F. 1989. Organochlorines and Marine Mammal Reproduction. Can J Fish Aquat Sci 46:360-368. Ankley, G T., Bennett, R S., Erickson, R J., Hoff, D J., M. W. Hornung, M W., Johnson, R D., Mount, D R., Nichols, J W., Russom, C L., Schmieder, P K., Serrrano, J A., Tietge, J E. and Villeneuve, D L. 2010. Adverse outcome pathways: a conceptual framework to support ecotoxicology research and risk assessment. Environ Toxicol Chem 29:730-741.

198

Antunes, S C., Marques, S M., Pereira, R., Goncalves, F. and Nunes, B. 2010. Testing procedures for the determination of several biomarkers in different species, for environmental assessment of pollution. J Environ Monit 12:1625-1630. Baccarelli, A. and Bollati, V. 2009. Epigenetics and environmental chemicals. Curr Opin Pediatr 21:243-251. Bengtson Nash, S M., Waugh, C A. and Schlabach, M. 2013. Metabolic concentration of lipid soluble organochlorine burdens in the blubber of Southern Hemisphere humpback whales through migration and fasting. Environ Sci Technol 47:9404- 9413. Benninghoff, A D. 2006. Toxicoproteomics--The Next Step in the Evolution of Environmental Biomarkers? Toxicol. Sci. 95:1-4. Bickham, J W., Sandhu, S., Hebert, P D., Chikhi, L. and Athwal, R. 2000. Effects of chemical contaminants on genetic diversity in natural populations: implications for biomonitoring and ecotoxicology. Mutat Res 463:33-51. Booth, D T., Clayton, D H. and Block, B A. 1993. Experimental Demonstration of the Energetic Cost of Parasitism in Free-Ranging Hosts. P R Soc B 253:125-129. Cardellicchio, N., Giandomenico, S., Ragone, P. and Di Leo, A. 2000. Tissue distribution of metals in striped dolphins (Stenella coeruleoalba) from the Apulian coasts, Southern Italy. Mar Environ Res 49:55-66. Daszak, P., Cunningham, A A. and Hyatt, A D. 2001. Anthropogenic environmental change and the emergence of infectious diseases in wildlife. Acta Trop 78:103- 116. de Swart, R L., Ross, P S., Vos, J G. and Osterhaus, A D M E. 1996. Impaired Immunity in Harbour Seals (Phoca vitulina) Exposed to Bioaccumulated Environmental Contaminants: Review of a Long-term Feeding Study. Environ Health Persp 104:5. Etzioni, R., Urban, N., Ramsey, S., McIntosh, M., Schwartz, S., Reid, B., Radich, J., Anderson, G. and Hartwell, L. 2003. The case for early detection. Nat Rev Cancer 3:243-252. Fair, P A., Romano, T., Schaefer, A M., Reif, J S., Bossart, G D., Houde, M., Muir, D., Adams, J., Rice, C. and Hulsey, T C. 2013. Associations between perfluoroalkyl compounds and immune and clinical chemistry parameters in highly exposed bottlenose dolphins (Tursiops truncatus). Environ Toxicol Chem 32:736-746.

199

Finlayson, K A., Leusch, F D. and van de Merwe, J P. 2016. The current state and future directions of marine turtle toxicology research. Environ Int 94:113-123. Gardner, S C., Fitzgerald, S L., Vargas, B A. and Rodríguez, L M. 2006. Heavy Metal Accumulation in Four Species of Sea Turtles from the Baja California Peninsula, Mexico. BioMetals 19:91-99. Geiger, T., Wehner, A., Schaab, C., Cox, J. and Mann, M. 2012. Comparative proteomic analysis of eleven common cell lines reveals ubiquitous but varying expression of most proteins. Mol Cell Prot 11(3):M111-014050. Grønborg, M., Kristiansen, T Z., Iwahori, A., Chang, R., Reddy, R., Sato, N., Molina, H., Jensen, O N., Hruban, R H. and Goggins, M G. 2006. Biomarker discovery from pancreatic cancer secretome using a differential proteomic approach. Mol Cell Proteomics 5:157-171. Guillette, L J. and Gunderson, M P. 2001. Alterations in development of reproductive and endocrine systems of wildlife populations exposed to endocrine-disrupting contaminants. Reproduction 122:857-864. Guillette, L J., Pickford, D B., Crain, D A., Rooney, A A. and Percival, H F. 1996. Reduction in penis size and plasma testosterone concentrations in juvenile alligators living in a contaminated environment. Gen Comp Endocr 101:32-42. Halpern, B S., Walbridge, S., Selkoe, K A., Kappel, C V., Micheli, F., D'Agrosa, C., Bruno, J F., Casey, K S., Ebert, C. and Fox, H E. 2008. A global map of human impact on marine ecosystems. Science 319:948-952. Harvell, D., Aronson, R., Baron, N., Connell, J., Dobson, A., Ellner, S., Gerber, L., Kim, K., Kuris, A., McCallum, H., Lafferty, K., McKay, B., Porter, J., Pascual, M., Smith, G., Sutherland, K. and Ward, J. 2004. The rising tide of ocean diseases: unsolved problems and research priorities. Front Ecol Environ 2:375- 382. Ikemoto, T., Kunito, T., Watanabe, I., Yasunaga, G., Baba, H., Miyazaki, N., Petrov, E A. and Tanabe, S. 2004. Comparison of trace element accumulation in Baikal seals (Pusa sibirica), Caspian seals (Pusa caspica) and northern fur seals (Callorhinus ursinus). Environ Pollut 127:83-97. Islam, M S. and Tanaka, M. 2004. Impacts of pollution on coastal and marine ecosystems including coastal and marine fisheries and approach for management: a review and synthesis. Mar Pollut Bull 48:624-649.

200

Jemec, A., Drobne, D., Tisler, T. and Sepcic, K. 2010. Biochemical biomarkers in environmental studies--lessons learnt from enzymes catalase, glutathione S- transferase and cholinesterase in two crustacean species. Environ Sci Pollut Res Int 17:571-581. Keller, J M. 2003. Occurence and effects of organochlorine contaminants in sea turtles. Duke University, North Carolina, USA. Keller, J M., McClellan-Green, P D., Kucklick, J R., Keil, D E. and Peden-Adams, M M. 2006. Effects of organochlorine contaminants on loggerhead sea turtle immunity: comparison of a correlative field study and in vitro exposure experiments. Environ Health Persp 114:70-76. Kelly, B C., Ikonomou, M G., Blair, J D., Morin, A E. and Gobas, F A. 2007. Food web–specific biomagnification of persistent organic pollutants. Science 317:236-239. Kendall, R J. 2010. Introduction and Overview.in Kendall, R J., Lacher, T E., Cobb, G P. and Cox, S B., editors. Wildlife Toxicology: Emerging Contaminant and Biodiversity Issues CRC Press, USA. Kidd, K A., Blanchfield, P J., Mills, K H., Palace, V P., Evans, R E., Lazorchak, J M. and Flick, R W.. 2007. Collapse of a fish population after exposure to a synthetic estrogen. Proc Natl Acad Sci USA 104:8897-8901. Lasserre, J P., Fack, F., Revets, D., Planchon, S., Renaut, J., Hoffmann, L., Gutleb, A C., Muller, C P. and Bohn, T. 2009. Effects of the endocrine disruptors atrazine and PCB 153 on the protein expression of MCF-7 human cells. J Proteome Res 8:5485-5496. Leusch, F D L, Chapman, H F., Kay, G W., Gooneratne, S R. and Tremblay, L A. 2006. Anal Fin Morphology and Gonadal Histopathology in Mosquitofish (Gambusia holbrooki) Exposed to Treated Municipal Sewage Effluent. Arch Environ Con Tox 50:562-574. Limpus, C., Miller, J., Paramenter, C., Reimer, D., McLachlan, N. and Webb, R. 1992. Migration of green (Chelonia mydas) and loggerhead (Caretta caretta) turtles to and from eastern Australian rookeries. Wildlife Res. 19:347-357. Lin, Y R., Kuo, C J., Lin, H Y., Wu, C J. and Liang, S S. 2014. A proteomics analysis to evaluate cytotoxicity in NRK-52E cells caused by unmodified Nano- Fe(3)O(4). Sci World J 2014:754721.

201

Liu, Y., Buil, A., Collins, B C., Gillet, L C., Blum, L C., Cheng, L Y., Vitek, O., Mouritsen, J., Lachance, G., Spector, T D., Dermitzakis, E T. and Aebersold, R. 2015. Quantitative variability of 342 plasma proteins in a human twin population. Mol Syst Biol 11:786. Manikkam, M., Guerrero-Bosagna, C., Tracey, R., Haque, M M. and Skinner, M K. 2012. Transgenerational actions of environmental compounds on reproductive disease and identification of epigenetic biomarkers of ancestral exposures. PLoS ONE 7:e31901. Matta, M B., Linse, J., Cairncross, C., Francendese, L. and Kocan, R M. 2001. Reproductive and transgenerational effects of methylmercury or Aroclor 1268 on Fundulus heteroclitus. Environ Toxicol Chem 20:327-335. Melvin, S D. and Wilson, S P. 2013. The utility of behavioral studies for aquatic toxicology testing: a meta-analysis. Chemosphere 93:2217-2223. Muir, D., Braune, B., DeMarch, B., Norstrom, R., Wagemann, R., Lockhart, L., Hargrave, B., Bright, D., Addison, R. and Payne, J. 1999. Spatial and temporal trends and effects of contaminants in the Canadian Arctic marine ecosystem: a review. Sci Total Environ 230:83-144. Nielsen, J B., Nielsen, F., Jorgensen, P J. and Grandjean, P. 2000. Toxic metals and selenium in blood from pilot whales (Globicephala melas) and sperm whales (Physeter catodon). Mar Pollut Bull 40:348-351. Noyes, P D., McElwee, M K., Miller, H D., Clark, B W., Van Tiem, L A., Walcott, K C., Erwin, K N. and Levin, E D. 2009. The toxicology of climate change: environmental contaminants in a warming world. Environ Int 35:971-986. Oaks, J L., Gilbert, M., Virani, M Z., Watson, R T., Meteyer, C U., Rideout, B A., Shivaprasad, H L., Ahmed, S., Chaudhry, M J I., Arshad, M., Mahmood, S., Ali, A. and Khan, A A. 2004. Diclofenac residues as the cause of vulture population decline in Pakistan. Nature 427:630-633. Pavlou, M P. and Diamandis, E P. 2010. The cancer cell secretome: a good source for discovering biomarkers? J Proteomics 73:1896-1906. Peakall, D B. 1994. The Role of Biomarkers in Environmental Assessment .1. Introduction. Ecotoxicology 3:157-160. Perkins, E J., Chipman, J K., Edwards, S., Habib, T., Falciani, F., Taylor, R., Van Aggelen, G., Vulpe, C., Antczak, P. and Loguinov, A. 2011. Reverse engineering adverse outcome pathways. Environ Toxicol Chem 30:22-38.

202

Ponten, F., Gry, M., Fagerberg, L., Lundberg, E., Asplund, A., Berglund, L., Oksvold, P., Bjorling, E., Hober, S., Kampf, C., Navani, S., Nilsson, P., Ottosson, J., Persson, A., Wernerus, H., Wester, K. and Uhlen, M. 2009. A global view of protein expression in human cells, tissues, and organs. Mol Syst Biol 5:337. Porte, C., Janer, G., Lorusso, L C., Ortiz-Zarragoitia, M., Cajaraville, M P., Fossi, M C. and Canesi, L. 2006. Endocrine disruptors in marine organisms: Approaches and perspectives. Comp Biochem Phys C 143:303-315. Reijnders, P J. 1986. Reproductive failure in common seals feeding on fish from polluted coastal waters. Nature 324:456-457. Relyea, R A. and Diecks, N. 2008. An unforseen chain of events: lethal effects of pesticides on frogs at sublethal concentrations. Ecol Appl 18:14. Roland, K., Kestemont, P., Hénuset, L., Pierrard, M A., Raes, M., Dieu, M. and Silvestre, F. 2013. Proteomic responses of peripheral blood mononuclear cells in the European eel (Anguilla anguilla) after perfluorooctane sulfonate exposure. Aquat Toxicol 128–129:43-52. Ross, P S., De Swart, R L., Van Loveren, H., Osterhaus, A D. and Vos, J G. 1996. The immunotoxicity of environmental contaminants to marine wildlife: a review. Annu Rev Fish Dis 6:151-165. Schwarzenbach, R P., Escher, B I., Fenner, K., Hofstetter, T B., Johnson, C A., von Gunten,U. and Wehrli, B.2006. The challenge of micropollutants in aquatic systems. Science 313:1072-1077. Skalnikova, H., Motlik, J., Gadher, S J. and Kovarova, H. 2011. Mapping of the secretome of primary isolates of mammalian cells, stem cells and derived cell lines. Proteomics 11:691-708. Spurgeon, D J., Jones, O A., Dorne, J L C., Svendsen, C., Swain, S. and Stürzenbaum, S R. 2010. Systems toxicology approaches for understanding the joint effects of environmental chemical mixtures. Sci Total Environ 408:3725-3734. Stastna, M., and Van Eyk, J E. 2012. Secreted proteins as a fundamental source for biomarker discovery. Proteomics 12:722-735. Tanabe, S. 2002. Contamination and toxic effects of persistent endocrine disrupters in marine mammals and birds. Mar Pollut Bull 45:69-77. Tornero, V., Borrell, A., Aguilar, A., Forcada, J. and Lockyer, C. 2006. Organochlorine contaminant and retinoid levels in blubber of common dolphins (Delphinus delphis) off northwestern Spain. Environ Pollut 140:312-321.

203

Vaidya, V S. and Bonventre, J V. 2010. Biomarkers: in medicine, drug discovery, and environmental health. John Wiley & Sons. Valavanidis, A., Vlahogianni, T., Dassenakis, M. and Scoullos, M. 2006. Molecular biomarkers of oxidative stress in aquatic organisms in relation to toxic environmental pollutants. Ecotoxicol Environ Saf 64:178-189. Van Bressem, M F., Raga, J A., Di Guardo, G., Jepson, P D., Duignan, P J., Siebert, U., Barrett, T., Santos, M., Moreno, I B. and Siciliano, S. 2009. Emerging infectious diseases in cetaceans worldwide and the possible role of environmental stressors. Dis Aquat Organ 86:143-157. van de Merwe, J P., Hodge, M., Olszowy, H A., Whittier, J M. and Lee, Y. 2010a. Using blood samples to estimate persistent organic pollutants and metals in green sea turtles (Chelonia mydas). Mar Pollut Bull 60:579-588. van de Merwe, J P., Hodge, M., Whittier, J M., Ibrahim, K., and Lee. S Y. 2010b. Persistent organic pollutants in the green sea turtle Chelonia mydas: nesting population variation, maternal transfer, and effects on development. Mar. Ecol. Prog. Ser. 403:269-278. Van Loveren, H., Ross, P., Osterhaus, A., and Vos, J. 2000. Contaminant-induced immunosuppression and mass mortalities among harbor seals. Toxicol. Lett. 112:319-324. Vos, J G., Dybing, E., Greim, H A., Ladefoged, O., Lambre, C., Tarazona, J V., Brandt, I. and Vethaak. A D. 2000. Health effects of endocrine-disrupting chemicals on wildlife, with special reference to the European situation. Crit Rev Toxicol 30:71-133. Vuong, N Q., Goegan, P., Mohottalage, S., Breznan, D., Ariganello, M., Williams, A., Elisma, F., Karthikeyan, S., Vincent, R. and Kumarathasan, P. 2016. Proteomic changes in human lung epithelial cells (A549) in response to carbon black and titanium dioxide exposures. J Proteomics 149:53-63. Waugh, C A., Nichols, P D., Schlabach, M., Noad, M. and Bengtson Nash, S. 2014. Vertical distribution of lipids, fatty acids and organochlorine contaminants in the blubber of southern hemisphere humpback whales (Megaptera novaeangliae). Mar Environ Res 94:24-31. Wei, Y., Chan, L L., Wang, D., Zhang, H., Wang, J. and Dai, J. 2008. Proteomic analysis of hepatic protein profiles in rare minnow (Gobiocypris rarus) exposed to perfluoroctanoic acid. J Proteome Res 7:1729-1739.

204