Genetic Risk Factors for PTSD: A -Set Analysis of Neurotransmitter Receptors

Michael Lewis

Dissertation submitted to the faculty of the Virginia Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy In Psychology

Russell T. Jones Bruce H. Scarpa-Friedman Margaret T. Davis Rachel A. Diana

May 12, 2020 Blacksburg, VA Keywords: PTSD, Gene-set analysis, GSA-SNP2, Neurotransmitter receptors, serotonin, glutamate, intracellular signaling

Genetic Risk Factors for PTSD: A Gene-Set Analysis of Neurotransmitter Receptors Michael Lewis Abstract (Academic)

PTSD is a moderately heritable disorder that causes intense and chronic suffering in afflicted individuals. The pathogenesis of PTSD is not well understood, and genetic mechanisms are particularly elusive. Neurotransmitter systems are thought to contribute to PTSD etiology and are the targets of most pharmacotherapies used to treat PTSD, including the only two FDA approved options and a wide array of off-label options. However, the degree to which variations in which encode for and regulate neurotransmitter receptors increase risk of developing PTSD is unclear. Recently, large collaborative groups of PTSD genetics researchers have completed genome-wide association studies (GWAS) using massive sample sizes and have made summary statistics available for public use. In 2018, a new technique for high-powered analysis of GWAS summary statistics called GSA-SNP2 was introduced. In order to explore the relationship between PTSD and genetic variants in widely theorized molecular targets, this study applied GSA-SNP2 to manually curated neurotransmitter gene-sets. Curated gene-sets included nine total “neurotransmitter receptor group” gene-sets and 45 total “receptor subtype” gene-sets. Each “neurotransmitter receptor group” gene-sets was designed to capture concentration of genetic risk factors for PTSD within genes which encode for all receptor subtypes that are activated by a given neurotransmitter. In contrast, “receptor subtype” gene-sets focused on specific subtypes and also accounted for intracellular signaling; each was designed to capture concentration of genetic risk factors for PTSD within genes which encode for specific receptor subtypes and the intracellular signaling through which they exert their effects. Due to practical considerations, this work used summary statistics derived from a GWAS with far fewer participants (2,424 cases; 7,113 controls) than initially planned (23,212 cases; 151,447 controls). Prior to controlling for multiple comparisons, 7 of the investigated gene-sets reached statistical significance at the p ≤ .05 level. However, after controlling for multiple comparisons, none of the investigated gene-sets reached statistical significance. Due to limited statistical power of the current work, these results should be interpreted very cautiously. The current study is best interpreted as a preliminary study and is most informative in relation to refining study design. Implications for next steps are emphasized in discussion and nominally significant results are synthesized with the literature to demonstrate the types of research questions that might be addressed by applying a refined version of this study design to a larger sample.

Genetic Risk Factors for PTSD: A Gene-Set Analysis of Neurotransmitter Receptors Michael Lewis General Audience Abstract

Though nearly all individuals will be exposed to a potentially traumatic event in their lifetime, only a small percentage will experience PTSD, which is a severe psychological disorder. Though genetics are known contribute to an individual’s level of risk for developing PTSD, relatively little is known about which particular genetic differences are key. Neurotransmitter receptors are thought to contribute to the risk for PTSD and are a key aspect of medications for PTSD. However, little is known about whether genetic differences in neurotransmitter receptors contribute to risk for developing PTSD. Recently, large collaborative groups of PTSD genetics researchers have completed studies which investigate genetic risk factors from across the genome using massive sample sizes and have made the statistical output of these studies available to the public. In 2018, a new technique called GSA-SNP2 was created to help assist with efforts to analyze aspects of that statistical output that have not been previously analyzed. This study used GSA-SNP2 to analyze the degree to which groups of neurotransmitter receptor genes contribute to the risk of developing PTSD. Due to the coronavirus pandemic, the researcher did not have access to the computing power needed to analyze the initially planned data which included 23,212 individuals with PTSD and 151,447 individuals without PTSD. As a substitute, the current work is an analysis using statistical output data from a study which included 2,424 individuals with PTSD and 7,113 individuals without PTSD. Based on a level of statistical significance that is typically used in most psychological studies, seven of the investigated gene-sets contribute highly to the risk for PTSD. However, it was necessary to use a different threshold for statistical significance due to the testing of many different groups of genes. After making that adjustment, none of the investigated gene-sets reached statistical significance. Due to limited statistical power of the current work, these results should be interpreted very cautiously. The current study is best interpreted as a preliminary study and is most informative in relation to refining study design. Implications for next steps are emphasized in discussion and nominally significant results are synthesized with the literature to demonstrate the types of research questions that might be addressed by applying a refined version of this study design to a larger sample.

iv

DEDICATION To my father, Richard Allen Lewis. Thank you for everything you did to instill in me the work ethic needed to confront the challenges of a Ph.D. More importantly, thank you for showing me the power of love and kindness. I strive every day to put into practice the wisdom and caring you have bestowed upon me to. Though you are no longer here in person, you are always present in spirit and in my memory. I will forever cherish the countless happy memories of our times together. To my mother, Cynthia Theresa Lewis. Thank you for your immeasurable sacrifices and support, during both good times and bad. Had you not been there to encourage me, I do not believe that I would have attempted a Ph.D., let alone reached this particular milestone. Your continued presence in my live is a blessing each day and I am grateful for the countless lessons you have taught me, memories we have made together, and memories we continue to make. To my brother, Patrick Terrence Coyle. Thank you for the inspiration and advice you provided in paving the way for me to begin my career at Virginia Tech. As our career paths diverge, I remain grateful to you for guiding me to the start of the trail and for your continued advice, support, and friendship. To my partner, Anne Stewart Deekens. Thank you for adding immense joy to my life and for helping me become a better and more balanced person. Of the many blessings my time in Blacksburg have bestowed upon me, your presence in my life is by far my favorite. I cannot wait to begin our life together in Boston. Finally, thank you to the friends and lab mates whose support and advice has been invaluable throughout these years. In particular, I would like to thank Benjamin Biermeier- Hanson and Connor Sullivan.

v

ACKNOWLEDGMENTS To my advisor, Russell T. Jones, thank you for your guidance and mentorship throughout graduate school. I cannot tell you how grateful I am that you believed in me and supported my ambition to seek a Ph.D. in biological psychology. I feel truly grateful to have had an advisor who not only supports me as an academic, but also as a person. I look forward to remaining friends for years to come. To Bruce H. Friedman, thank you for opening the door for me to biological research by adopting me into the Mind Body lab. Your mentorship and support were critical in successfully transitioning to biological psychology and in acquiring skills in psychophysiology that I will apply in postdoc and beyond. To Margaret T. Davis, thank you for your guidance and mentorship. Your encouragement gave me the necessary confidence to pursue molecular research. You have imparted lessons and shaped skills that will have an imprint on my work throughout my career. I cannot tell you how grateful I am that you adopted me as a mentee. To Rachel A. Diana, thank you for helping to spark my interest in neuroscience. The assignments and discussions in your class were the beginning of my journey into molecular research and provided me with a model to emulate when I began teaching. I am forever grateful for the wisdom and knowledge each of you have bestowed upon me and I look forward to continuing to grow my relationship with each of you. Finally, thank you to Maddy Ryan and Marcus Nguyen for your important contributions to data organization and to Aiden England for your invaluable contribution to data cleaning in Java.

vi

Table of Contents

1.0 – Introduction ...... 1

1.1 – PTSD: Public Health Impact and Importance of Genetic Risk Factors ...... 3

1.2 – Genetics of PTSD: Brief History and Foundational Knowledge……………...……...…..5

1.3 – Population Structure: Challenges and Strategies………………………….…………….12

1.4 – Psychiatric Genomics Consortium and Million Veteran Program………………………15

1.5 – Genetic Pathways Analysis: From Summary Statistics to Novel Insights ...... 23

1.6 – Neurotransmitter Receptors and Curated Gene-sets: The Case For a Hypothesis-Driven

Gene-set Analysis ...... 31

2.0 – Hypotheses ...... 40

3.0 – Method ...... 42

3.1 – PGC Methodology ...... 42

3.2 – Data Acquisition ...... 54

3.3 – Data Reduction ...... 55

3.4 – Data Analysis ...... 58

3.5 – Post-Hoc Analysis ...... 63

4.0 – Results……………………………………………………………………………………...65

4.1 – Planned Analysis ...... 65

4.2 – Post-Hoc Analysis ...... 69

5.0 –Discussion…………………………………………………………………………………..72 vii

5.1 - Effect of Gene-set Size Minimum on Hypothesis Testability …..………………………72

5.2 - Preliminary Results: Lessons Learned and Next Steps …….…………………………....73

5.3 – Limitations…...………………………………………………………………………….76

5.4 – Future Directions…..…………………………………………………………………….81

5.5 - Conclusion………………………………………………………………………………85

Glossary ...... 87

References ...... 89

viii

List of Tables and Figures

Table 1. Results ...... 112

Table 2. Post-Hoc Results...... 113

Figure 1. Gene-sets ...... 114

Figure 2. Post-Hoc Gene-sets ...... 119

Supplemental Figure S1. Gene-scores for planned gene-sets ...... 122

Supplemental Figure S2. Gene-scores for post-hoc gene-sets ...... 129

Genetic Risk Factors for PTSD….. 1

1.0 Introduction

Though trauma exposure is a necessary component of the diagnostic criteria for posttraumatic stress disorder (PTSD), just an estimated 13.6% of individuals who are exposed to a potentially traumatic event develop chronic PTSD (Atwoli, Stein, Koenen, & McLaughlin,

2015). Genetic risk factors appear to play a large role in shaping vulnerability to PTSD, with most heritability estimates falling in the 30-40% range (Koenen et al., 2005; Sartor et al., 2011;

M. B. Stein, Jang, Taylor, Vernon, & Livesley, 2002; True et al., 1993; Xian et al., 2000) and gender-specific estimates for females reaching as high as 71% (Sartor et al., 2011). Thus, elucidation of genetic risk factors of PTSD may contribute to efforts to understand the pathogenesis of the disorder.

Though the pathogenesis of PTSD remains largely mysterious (Breen et al., 2018; Zoladz &

Diamond, 2013), it is a near consensus that neurotransmitter systems play an important role in the disorder (Richter-Levin, Stork, & Schmidt, 2018; Zoladz & Diamond, 2013). Relatedly, nearly all pharmacotherapies for PTSD target neurotransmitter systems (Hoskins et al., 2015;

Krystal et al., 2017). In spite of this, the degree to which specific neurotransmitter systems are particularly influential in the pathogenesis of PTSD is unclear (Hoskins et al., 2015; Krystal et al., 2017). Further, to the extent that neurotransmitter systems do play a role in PTSD, the role of genetics is unclear (Banerjee, Morrison, & Ressler, 2017; Duncan, Cooper, & Shen, 2018;

Duncan, Ratanatharathorn, et al., 2018). For example, relatively little is known about the degree to which variation in genes which encode for neurotransmitter receptors and their downstream signaling mechanisms contributes to risk for PTSD (Duncan, Cooper, et al., 2018). Though a small body of literature suggests that individual variants in neurotransmitter receptor genes may Genetic Risk Factors for PTSD….. 2 be correlated with PTSD (D. Mehta & Binder, 2012), these variants and the receptors which they affect do not operate in isolation (Holmes, Girgenti, Davis, Pietrzak, & DellaGioia, 2017;

Nichols & Nichols, 2008). Thus, studies which examine the association of PTSD with genetic differences in biologically related groups of neurotransmitter receptors may provide useful information.

Thanks to recent advances in genome-wide analysis, genetic understanding of PTSD has advanced in recent years and may be on the precipice of major breakthroughs (Banerjee et al.,

2017; Duncan, Ratanatharathorn, et al., 2018; Nievergelt et al., 2019). Already, genome-wide association studies (GWAS) have shed light on general principles of complex genetic phenotypes, a category which includes PTSD (Banerjee et al., 2017; Duncan, Ratanatharathorn, et al., 2018; Nievergelt et al., 2019). Given that PTSD is a complex genetic phenotype, analyses of groups of genes may be helpful in providing a more complete picture than is obtained through analyses of individual variants alone (Holmans et al., 2009; Skelton, Ressler, Norrholm,

Jovanovic, & Bradley-Davino, 2012). Recently, a large-scale collaborative data collection effort called the Psychiatric Genomics Consortium has conducted genome-wide analysis on a large enough sample (over 30,000 PTSD cases) to enable analysis of groups of genes (aka gene-sets)

(Nievergelt et al., 2019).

Using gene-set analysis, it is possible to study the degree to which genetic variants which are correlated with the diagnosis of PTSD are concentrated within particular biologically related groups of genes (Wang, Li, & Hakonarson, 2010). In order to make use of freely available genetic data, gene-set analysis methods which use GWAS summary statistics as input have been developed (De Leeuw, Neale, Heskes, & Posthuma, 2016; Holmes et al., 2017). A recently developed gene-set analysis technique called GSA-SNP2 has shown statistical properties which Genetic Risk Factors for PTSD….. 3 suggest that it might be well suited for exploratory gene-set analysis using summary statistics from a recent GWAS that was performed by the Psychiatric Genomics Consortium (Yoon et al.,

2018a). Using GSA-SNP2, it is possible to tailor analysis toward topics of particular clinical relevance such as neurotransmitter receptors (Yoon et al., 2018a). Given the importance of elucidating the role of specific neurotransmitter systems in the pathogenesis of PTSD, an analysis of the relationship between PTSD and neurotransmitter receptor gene-sets may be informative. In this work, the researcher used GSA-SNP2 to conduct a preliminary analysis of the degree to which genetic variants which are correlated with PTSD are highly concentrated in neurotransmitter receptor gene-sets. Though the current study is preliminary, findings might inform next steps in this line of research.

1.1 PTSD: Public Health Impact and Importance of Genetic Risk Factors

PTSD is a disorder of extreme and enduring stress reactions to a traumatic event (Gentes et al., 2014). Though only a subset of individuals who are exposed to a potentially traumatic event develop chronic PTSD (Briggs‐Gowan et al., 2010), PTSD is a major public health concern (Benjet et al., 2016). PTSD symptoms are often severe (Pietrzak, Goldstein, Southwick,

& Grant, 2011), chronic (Cassiers et al., 2018), and treatment resistant (Association, 2017;

Krystal et al., 2017). The impairing nature of PTSD is reflected in the estimated $3 billion per year of lost productivity caused by PTSD worldwide (Atwoli et al., 2015) and in the multitude of studies demonstrating a negative impact of PTSD on interpersonal relationships (McFarlane &

Bookless, 2001). The intransigence of PTSD symptoms is reflected in the relatively low remission rates of individuals receiving current first-line treatments (Krystal et al., 2017). The commonality of potentially traumatic event exposure is indexed by prevalence estimates of 70% Genetic Risk Factors for PTSD….. 4 worldwide and as high as 90% in the United States (Banerjee et al., 2017; Ekblad, Jaranson,

Boris Drožđek. Broken Spirits. The treatment of traumatized asylum seekers, & Pp, 2004;

Kilpatrick et al., 2013). PTSD is a significant public health concern at the global level, with an estimated lifetime prevalence of 10.4% in women and 5% in men (Ekblad et al., 2004; Tol et al.,

2014). In order to effectively address this public health concern, it is important to identify risk and resilience factors which contribute to the observed variability in vulnerability to PTSD

(Zoladz & Diamond, 2016).

Epidemiological studies estimate that a subset of between 9% and 37% of individuals who are exposed to potentially traumatic events develop chronic PTSD (Banerjee et al., 2017;

Ekblad et al., 2004; Kilpatrick et al., 2013). Though differences in the type and degree of trauma exposure partially explain differences in outcome, they do not fully account for differences in outcome (D. Mehta & Binder, 2012; C. Sullivan, Jones, Hauenstein, & White, 2017). Studies in both humans and rodents suggest that the development of PTSD symptoms and PTSD phenotypes varies considerably even among individuals exposed to the same type of potentially traumatic event (Richter-Levin et al., 2018). Thus, a holistic understanding of the pathophysiology of PTSD requires an understanding of the role of pre-existing risk and resilience factors, biological and otherwise (Koenen et al., 2003; Richter-Levin et al., 2018). Pre- existing biological risk and resilience factors are difficult to study in PTSD, due to the scarcity of pre-trauma data (Grant, Beck, Marques, Palyo, & Clapp, 2008; Yamamoto et al., 2009). In contrast with many putative neurobiological mechanisms of PTSD, genetic differences inherently exist prior to experience (Gillespie, Phifer, Bradley, & Ressler, 2009). Thus, even using cross-sectional data, researchers can be confident that these differences preceded trauma exposure (Gillespie et al., 2009). Though this does not guarantee that they are causal, it does rule Genetic Risk Factors for PTSD….. 5 out the possibility of being effects of trauma exposure. Further, estimates from twin studies suggest that PTSD is a moderately heritable disorder (Koenen et al., 2005; Koenen et al., 2003;

M. B. Stein et al., 2002; True et al., 1993; Xian et al., 2000). Thus, in the long run, investigations of the genetic architecture of PTSD may have potential to help explain a clinically meaningful portion of the overall variability in individual risk for PTSD (Duncan, Cooper, et al., 2018; C.

Nievergelt et al., 2018; Nievergelt et al., 2019). Further, recent advances in collaborative data collection and bioinformatics may allow for high-powered gene-set analysis of PTSD using only

GWAS summary statistics (Yoon et al., 2018a). Though gene-set analysis is a relatively new approach to psychiatric genetics research, it is built upon a decades-old foundation of genetic knowledge (Holmans et al., 2009).

1.2 Genetics of PTSD: Brief History and Foundational Knowledge

Prior to the development of methodology for measuring molecular aspects of the genetics of disorders such as PTSD, twin studies provided evidence that a meaningful portion of risk for

PTSD is inherited at birth (Afifi, Asmundson, Taylor, & Jang, 2010; Chantarujikapong et al.,

2001). Twin studies index the macro-level contribution of genomic differences to risk for PTSD by studying monozygotic twins (100% shared genetics) as well as dizygotic twins (average of

50% shared genetics) (Afifi et al., 2010; Middeldorp, Cath, Van Dyck, & Boomsma, 2005).

Since same-household monozygotic twins and same-household dizygotic twins are assumed to have equal amounts of shared environmental influence, these studies act as natural experiments and are able to control for environment (Afifi et al., 2010; Middeldorp et al., 2005). Twin studies of PTSD control for trauma in a number of ways. One common approach is to only include twin pairs for which both twins have been exposed to Criterion A trauma in heritability estimates Genetic Risk Factors for PTSD….. 6

(Koenen et al., 2005; Sartor et al., 2011; M. B. Stein et al., 2002; True et al., 1993; Xian et al.,

2000). Some studies use additional controls such as separating heritability estimates into assaultive and non-assaultive trauma (M. B. Stein et al., 2002) or separately modeling heritability in twin pairs whose trauma exposure includes combat and those whose exposure does not (True et al., 1993). While the temporal relationship between genetics and PTSD is clear, this does not imply that all genetic associations represent increased vulnerability to developing PTSD in the wake of trauma (Middeldorp et al., 2005). For example, due to the effects of genetics on risk taking behaviors and social traits, some proportion of PTSD heritability is attributable to heritability of the likelihood of experiencing potentially traumatic events (Afifi et al., 2010;

Sartor et al., 2012). Thus, it is important to disentangle genetic risk due to increased vulnerability from the effects of potentially traumatic events from genetic risk due to increased likelihood of potentially traumatic event exposure (Afifi et al., 2010). Many studies do this by modeling the heritability of trauma exposure and statistically controlling for those effects (Sartor et al., 2011).

Based on twin studies, that the heritability of PTSD is generally asserted to be in the 30-

40% range (Koenen et al., 2005; M. B. Stein et al., 2002; Xian et al., 2000). However, genetic analyses should be interpreted both holistically and cautiously (Banerjee et al., 2017; P. F.

Sullivan, 2007). Though there is a general consensus that twin studies are the gold standard for estimating heritability (Duncan, Cooper, et al., 2018), heritability estimates are subject to methodological biases and are highly particular to the specific sample from which they are calculated (Johnson, Turkheimer, Gottesman, & Bouchard Jr, 2009). The influence of participant characteristics and study methodology on heritability estimates is reflected in the high variability of those estimates (Johnson et al., 2009). For PTSD, estimates have ranged from 23.5% in one study using an all-male sample (True et al., 1993) to 71% in one study using an all-female Genetic Risk Factors for PTSD….. 7 sample (Sartor et al., 2011). Though the exact cause of these discrepancies is unclear, studies offered alternative hypotheses (e.g. possible underreporting of trauma leading to heritability being estimated for only the most severe cases) (Sartor et al., 2011). Though twin studies have been valuable in showing that psychological traits depend on both genes and environment (and continue to be useful in controlling for genetic effects) (Duncan, Cooper, et al., 2018), investigation of specific genetic mechanisms requires molecular DNA sequencing (Duncan,

Cooper, et al., 2018; True et al., 1993).

The advent of early DNA sequencing methods in the 1970s and “second-generation” sequencing methods in the 1990s paved the way for molecular genetics-based investigations

(Duncan, Cooper, et al., 2018). Studies which seek to elucidate molecular genetic mechanisms have generally focused on single-nucleotide polymorphisms (SNPs) (Duncan, Cooper, et al.,

2018; Syvänen, 2001). Each SNP represents one (i.e. specific location on the genome) for which one of two possible nucleotide variants, known as alleles, may appear (Syvänen, 2001). So far, over 100 million distinct SNPs have been identified in humans (NIMH, 2018). Of note, many genes have multiple SNPs (De Leeuw et al., 2016). For example, to date, 9 different SNPs have been identified for the HTR1A gene, which encodes the serotonin-1A receptor (David et al.,

2005). Additionally, SNPs which fall just outside of the coding regions for a receptor are often included as part of that gene’s regulatory structure, especially if they are directly involved in regulating expression of the gene (Huckins et al., 2019; Mooney & Wilmot, 2015; Pickrell et al.,

2010; Yoon et al., 2018a). Further, since genes often encode proteins that work together in coordinated biological systems, genes can be organized into gene-sets (De Leeuw et al., 2016).

Gene-sets can be derived from statistical analysis or manually curated based on a review of the literature (Mooney & Wilmot, 2015). For example, a collection of all 18 genes which encode the Genetic Risk Factors for PTSD….. 8

14 known serotonin receptors (one gene per , five for the ionotropic serotonin-3 receptor) could possibly represent a gene-set for serotonin receptors (Mooney &

Wilmot, 2015; Nichols & Nichols, 2008). Additionally, a collection of the five genes which encode the serotonin-3 receptor could comprise a serotonin-3 receptor gene-set (Nichols &

Nichols, 2008; Pickrell et al., 2010). Due to the statistical properties of gene-set analysis, manually curated gene-sets can be tailored to nearly any level of specificity of interest to the researcher (De Leeuw et al., 2016). Additionally, because they test gene-sets one at a time, gene- set analysis studies can (and often do) include tests of multiple gene-sets which include some of the same genes (De Leeuw et al., 2016). Though many gene-set analyses utilize an agnostic approach and test thousands of distinct but overlapping gene-sets, hypothesis-driven approaches can increase statistical power and interpretability (De Leeuw et al., 2016; Holmes et al., 2017;

Mooney & Wilmot, 2015; Skelton et al., 2012). In order to understand the importance of analyses at the level of genes and gene-sets, a brief overview of findings from studies which focus on SNPs is informative.

One of the most widely used methodologies for studying the relationship between SNPs and PTSD is the candidate gene approach (Duncan, Cooper, et al., 2018; T. J. Jorgensen et al.,

2009). Candidate gene studies apply a case-control study design and seek to estimate population- level correlations between one or several SNPs of interest and PTSD (Miller et al., 2018; Zhu &

Zhao, 2007). Though candidate gene studies dominated early attempts to identify specific genetic variants involved in PTSD (Duncan, Cooper, et al., 2018), the methodology has been increasingly scrutinized (Bosker et al., 2011; Duncan, Cooper, et al., 2018; P. F. Sullivan, 2007;

Wei, Tang, & Li, 2012). Major critiques include: too few variants tested (Bosker et al., 2011;

Duncan, Cooper, et al., 2018; García-Campos, Espinal-Enríquez, & Hernández-Lemus, 2015; P. Genetic Risk Factors for PTSD….. 9

F. Sullivan, 2007; Wei et al., 2012), lack of control for genetic confounds such as linkage disequilibrium (Bosker et al., 2011; Duncan, Cooper, et al., 2018; García-Campos et al., 2015; P.

F. Sullivan, 2007; Wei et al., 2012), and a poor replication record (Bosker et al., 2011; Duncan,

Cooper, et al., 2018; P. F. Sullivan, 2007). Though findings from candidate gene studies should be viewed as preliminary (García-Campos et al., 2015), a brief overview of findings from these studies is informative.

As they pertain to PTSD, a wide array of specific SNPs have been found to be correlated, though none have consistently replicated (Bosker et al., 2011; Duncan, Cooper, et al., 2018).

However, a number of variants in genes which encode and directly regulate expression of neurotransmitter receptors have been identified as possibly correlated with PTSD (Banerjee et al., 2017; Smoller, 2016; Zoellner, Pruitt, Farach, Jun, & anxiety, 2014). For example, a meta- analysis reported that the minor allele (i.e. less common allele) of the SNP rs1800497 was correlated with PTSD (Li et al., 2016). Rs1800497 is found within the DRD2 gene, which encodes the D-2 (Li et al., 2016). Of note, authors stated that future studies with larger sample sizes were needed in order to validate the findings (Li et al., 2016). Further, individual candidate gene studies have identified associations of PTSD with SNPs within genes which encode for a variety of receptors including serotonin-1A (G. M. Sullivan et al., 2013),

GABA receptor B3 (Feusner et al., 2001), cholinergic receptor A5 (Jooyeon & N, 2011), and beta-2 (Liberzon et al., 2014) among others (Skelton et al., 2012). Broadly speaking, significant findings from candidate gene studies have primarily identified variants in neurotransmitter receptors and glucocorticoids as possible risk loci (Banerjee et al., 2017;

Smoller, 2016; Zoellner et al., 2014). Though these results should be viewed as exploratory, they Genetic Risk Factors for PTSD….. 10 support the general theory that variants in genes which encode for neurotransmitter receptors may contribute to the risk for PTSD.

Recent innovations in high-throughput data analytic techniques led to the advent of genome wide association studies (GWAS), which enables the concurrent testing of genetic variants across the entire genome in relation to PTSD (Duncan, Cooper, et al., 2018). Early

PTSD GWAS efforts included samples in the thousands and identified several correlations of

SNPs and PTSD (Duncan, Cooper, et al., 2018). In contrast with candidate gene studies, most small GWAS studies have not identified statistically significant SNPs within coding regions of canonical substrates of PTSD such as neurotransmitter receptor (Duncan, Cooper, et al., 2018).

However, “small GWAS studies” (i.e. studies with samples in the thousands) have been criticized as having insufficient statistical power to test for millions of SNPs and for often failing to replicate (Duncan, Cooper, et al., 2018; Faye, Sun, Dimitromanolakis, & Bull, 2011). Though the limitations of small GWAS studies and candidate gene studies are thoroughly reviewed throughout the literature (Bosker et al., 2011; Duncan, Cooper, et al., 2018; García-Campos et al., 2015; P. F. Sullivan, 2007; Wei et al., 2012), small GWAS and candidate gene studies played a crucial role in elucidating general principles of the genetics of PTSD (Duncan, Cooper, et al.,

2018).

Evidence from small GWAS and candidate gene studies informed the now-consensus view that PTSD (like other major psychiatric disorders) is a “complex genetic phenotype”

(Belmont & Leal, 2005; Duncan, Cooper, et al., 2018; Marian, 2012). In other words, PTSD arises from the additive and/or combined effects of numerous genetic variants and environmental influences (Belmont & Leal, 2005; Duncan, Cooper, et al., 2018; Marian, 2012; D. Mehta &

Binder, 2012). Additional examples of complex genetic phenotypes include height, personality, Genetic Risk Factors for PTSD….. 11 intelligence, diabetes, and many medical disorders (Beauchaine & Constantino, 2017; Belmont

& Leal, 2005). All available evidence suggests that the heritability of a complex genetic phenotypes such as PTSD is the end product of thousands of small-effect genetic variants

(Boyle, Li, & Pritchard, 2017; Marian, 2012; Skelton et al., 2012). Of note, factors such as gene x environment interactions, trauma type, chronicity of exposure, and age of exposure require investigation but are beyond the scope of available data for this study (C. Nievergelt et al., 2018).

One possible confound that has received considerable attention in the literature is population structure, also known as population stratification (i.e. group-level differences in average allelic frequencies between ancestral groups which may confound GWAS findings) (Bouaziz,

Ambroise, & Guedj, 2011; Hellwege et al., 2017; Marchini, Cardon, Phillips, & Donnelly, 2004).

One general finding from small GWS studies is that, due to population stratification, ancestry may be an important variable to account for when conducting GWAS (C. Nievergelt et al., 2018). Though available evidence suggests that most causal variants of complex traits are shared across ancestral groups, discrepant GWAS findings do commonly occur, likely due to confounding (Marigorta & Navarro, 2013). Two small GWAS studies compared results between ancestral groups and found that statistically significant SNP correlations with PTSD did not overlap (Ashley-Koch et al., 2015; Xie et al., 2013). For example, one study examined the top

100 SNPs for African Americans and European Americans and found no overlap (Xie et al.,

2013). However, it should be noted that small GWAS studies of PTSD have had a poor replication record, regardless of ancestry (Duncan, Cooper, et al., 2018). Though small GWAS studies of PTSD highlight the confounding effects of a number of variables (including but not limited to population stratification), findings from these studies also suggest that it may be possible to mitigate these effects. For example, one study identified a genome-wide significant Genetic Risk Factors for PTSD….. 12

GWAS SNP in an African American sample which was replicated at the nominally significant level (p<.05) in a European American sample using a candidate gene approach (Guffanti et al.,

2013). Further, one study using an all-male sample conducted stratified GWAS analysis in each of European Ancestry, African Ancestry, and Hispanic/Native American Ancestry groups, followed by a trans-ancestral meta-analysis (Nievergelt et al., 2015). In the transethnic meta- analysis, this study identified 12 genome-wide significant SNPs (Nievergelt et al., 2015). Based on findings from small GWAS studies, population stratification is an important confound to address. Strategies include independent replication and stratification with meta-analysis. Though addressing population stratification is considered to be a major statistical challenge, confronting this challenge is critical in assuring that genetic discoveries generalize to all populations

(Hellwege et al., 2017).

1.3 Population Structure: Challenges and Strategies

In addition to the statistical challenges associated with population structure, scientists hold a social obligation to be particularly conscientious in their use of language and thorough in their explication of their reasoning when communicating methodological decisions related to race (Yudell, Roberts, DeSalle, & Tishkoff, 2016). Though specific definitions and uses of the words race and ancestry vary, there is a general agreement that these are different concepts

(Caulfield et al., 2009). While race is generally described as a social construct that is used to place individuals into intersubjectively imagined groups, ancestry is generally thought to describe an individual’s genealogical lineage (Caulfield et al., 2009; Yudell et al., 2016). In spite of this, conflation of these concepts has been shown to perpetuate conscious and subconscious racial discrimination in a variety of settings, including in medical care (Yudell et al., 2016). For Genetic Risk Factors for PTSD….. 13 example, the finding that individuals with cystic fibrosis are more likely to have genetic characteristics suggestive of an ancestral lineage that is largely traceable to Europe (Mateu,

Calafell, Ramos, Casals, & Bertranpetit, 2002) has contributed to the mistaken perception among many doctors that the disorder is a “white” disorder (Spencer, Venkataraman, Higgins,

Stevenson, & Weller, 1994). Relatedly, cystic fibrosis is underdiagnosed and carries a higher average disease burden among African Americans and other racial minority populations

(Quittner et al., 2010; Spencer et al., 1994). Though it is not a panacea, precise language and detailed explanation of methodological reasoning are essential in order to reduce the pernicious effects of conflation of race and ancestry (Yudell et al., 2016).

In this work, the term race is used to refer to a socially constructed concept which has historically been used to place individuals into vaguely defined hierarchical categories based on a preconceived notion that cosmetic phenotypic traits, geographical ancestry, and individual characteristics covary. In contrast to the use of race to describe a socially constructed and intersubjective variable, this work uses the term geographic ancestry group to describe a genetically and statistically defined group-level variable that is specific to the study within which the population is defined. In this work, the term geographic ancestral group refers to a group of participants in a specific study that has been defined using genetic cutoff criteria specific to that study and is comprised of individuals who have been estimated to meet cutoff criteria for that group based on the statistical methods used to estimate genealogical history in that study.

Though group-level differences between geographic ancestral populations are estimated to account for just 3 to 5 % of overall genetic variation among humans (Rosenberg et al., 2002), simulation studies demonstrate that population structure can increase rates of both type-I and type-II error in GWAS studies (Marchini et al., 2004). Importantly, these confounding effects do Genetic Risk Factors for PTSD….. 14 not imply between-group differences in which variants are causal variants or in their biological function (Marigorta & Navarro, 2013). Additionally, the underlying statistical properties which contribute to the effects of population structure on GWAS results applies to any type of subpopulation within a study that varies from other study subpopulations with respect to both genetics and with respect to the phenotype of interest (Marchini et al., 2004). In short, due to its reliance on correlation, GWAS is prone to confounding whenever two latent populations differ with respect to the frequency of a given allele while also differing with respect to the frequency of PTSD for reasons other than differential frequency of that specific allele (Marchini et al.,

2004).

The most common ways of addressing population structure are with the use of genomic control, principal components analysis adjustment using Eigenstrat, principal components adjusted regression, and stratified analyses with meta-analysis (Bouaziz et al., 2011). Briefly, genomic control entails first calculating an inflation facto, then dividing the uncorrected GWAS test statistic (the chi-square) by that quantity (Devlin & Roeder, 1999). Principal components analysis adjustment using Eigenstrat entails modeling continuous axes of variation in ancestry and covariance of ancestry with case/control status; genotype and phenotype variables are adjusted by amounts attributable to ancestry prior to testing for association of SNPs with the phenotype of interest (Price et al., 2006). Principal components adjusted regression entails performing a principal components analysis on all SNPs to reduce dimensionality of genotype data and using top principal components as covariates in GWAS (rather than using them to adjust genotype and phenotype variables prior to GWAS as in Eigenstrat) (Bouaziz et al., 2011).

Stratified analyses with meta-analysis entails first assigning participants to relatively homogeneous genetically determined ancestry group, then performing separate GWAS on each Genetic Risk Factors for PTSD….. 15 ancestry group, then performing meta-analysis of those separate GWAS results (Bouaziz et al.,

2011). Numerous simulation studies have compared these techniques and results suggest that each can be effective but that none are without their limitations (Zhao, Mitra, Kanetsky,

Nathanson, & Rebbeck, 2018). Numerous parameters influence the degree to which a particular strategy is appropriate, especially the degree of population structure present, the type of population structure present, and the degree to which cases/control status systematically varies with ancestry (Zhao et al., 2018). Based on one simulation study, principal components analysis adjusted regression has the strongest evidence of being effective across various levels of population structure (Bouaziz et al., 2011). Additionally, stratified analysis and meta-analysis can be a useful technique for large collaborative meta-analyses, which often combine data from various studies with different methodologies (C. M. Nievergelt et al., 2018). One advantage of this approach is that it allows for analyses to be conducted both separately within each geographic ancestry group and also in a trans-geographic ancestry group sample (C. M.

Nievergelt et al., 2018). As GWAS studies transition from smaller GWAS to new big data approaches, techniques that enable analysis in diverse samples will be critical (Logue et al.,

2015).

1.4 Psychiatric Genomics Consortium and Million Veteran Program

With increasing knowledge of the nature of highly complex genetic phenotypes such as

PTSD, a consensus has grown that extremely large sample sizes (e.g. tens of thousands of cases) are needed in order to achieve adequate statistical power for GWAS (Duncan, Cooper, et al.,

2018; Duncan, Ratanatharathorn, et al., 2018). The potential for large multi-lab GWAS studies to yield replicable results is demonstrated in other disorders (Coleman, Gaspar, Bryois, & Breen, Genetic Risk Factors for PTSD….. 16

2019; Lencz & Malhotra, 2015; Logue et al., 2015). In order to elucidate the genetic architecture of psychiatric disorders, including PTSD, large-scale collaborative efforts to collect and share data have been formed (Duncan, Cooper, et al., 2018; Gelernter et al., 2019). The Psychiatric

Genomics Consortium and Million Veteran Program are among the largest such efforts (Duncan,

Cooper, et al., 2018; Gelernter et al., 2019). The Psychiatric Genomics Consortium PTSD workgroup has made strides toward being the first large GWAS group to conduct large-scale

GWAS in diverse samples in any disorder (Logue et al., 2015). In order to address population stratification, Psychiatric Genomics Consortium and Million Veteran Program studies have made use of a combined strategy of meta-analysis and principal components adjusted regression

(Duncan, Ratanatharathorn, et al., 2018; Gelernter et al., 2019; Nievergelt et al., 2019).

Additionally, these studies used stratification and meta-analysis to control for sex and to address various methodological differences between small GWAS studies which contributed data for these large-scale studies (Duncan, Ratanatharathorn, et al., 2018; Gelernter et al., 2019;

Nievergelt et al., 2019).

In April 2017, the Psychiatric Genomics Consortium published its first GWAS of PTSD, which included over 5,000 PTSD cases and over 15,000 controls (Duncan, Ratanatharathorn, et al., 2018). Though the April 2017 Psychiatric Genomics Consortium study did not detect any genome-wide significant SNPs (i.e. SNPs that were statistically significant after correcting for multiple testing), this finding was in line with expectations for a GWAS of that size (Duncan,

Ratanatharathorn, et al., 2018). In July 2019, the Million Veteran Program published a GWAS investigating genetic risk factors of PTSD reexperiencing symptoms in a sample of 165,000 veterans (Gelernter et al., 2019). The researchers of that study increased statistical power relative the previous Psychiatric Genomics Consortium GWAS by using a large sample size and Genetic Risk Factors for PTSD….. 17 narrowing its focus to DSM-IV reexperiencing symptoms, perhaps decreasing statistical noise by reducing phenotypic heterogeneity (Gelernter et al., 2019). In line with expectations for a substantially higher-powered study, the Million Veteran Program identified eight genome-wide significant SNPs as well as a number of genome-wide significant genes and gene-sets (Gelernter et al., 2019). Though the Million Veteran Program study focused on a narrower aspect of PTSD rather than the full syndrome, these findings add to current understanding of the genetics of

PTSD. Most recently, in October 2019, the Psychiatric Genomics Consortium published the largest GWAS of the full syndrome of PTSD to date, including over 30,000 PTSD cases and over 170,000 controls (Nievergelt et al., 2019). The Psychiatric Genomics Consortium reported six genome-wide significant SNPs. Though Million Veteran Program and Psychiatric Genomics

Consortium are still collecting more data in order to reach sample sizes needed to identify larger numbers of robust SNPs, recent findings have yielded important insights (Duncan,

Ratanatharathorn, et al., 2018; Gelernter et al., 2019; Nievergelt et al., 2019).

The October 2019 Psychiatric Genomics Consortium study performed stratified meta- analysis based on both biological sex and geographic ancestry group (Nievergelt et al., 2019). A total of nine different stratified groups were examined. Stratified geographic ancestry groups in the study were termed “European and European Americans” and “African and African-

Americans.” Additionally, a group termed “trans-ethnic” was examined which included both of the aforementioned geographic ancestral groups as well as a group of participants comprised of groups referred to as “Latino” and “Native Americans,” each of which were too small to be analyzed separately from the “trans-ethnic” meta-analysis (Nievergelt et al., 2019). For each of the three geographic ancestry groups, stratified analyses were performed on a male sample, a female sample, and a combined sex sample (Nievergelt et al., 2019). No analyses performed in Genetic Risk Factors for PTSD….. 18 the “trans-ethnic” group returned genome-wide significant results. In contrast, analyses of the

“European and European Americans” group and “African and African-Americans” group yielded a total of six genome-wide significant SNPs for PTSD, none of which overlapped with the eight

Million Veteran Program SNPs (Nievergelt et al., 2019).

In the combined sex analyses, the Psychiatric Genomics Consortium identified two significant SNPs in the “European and European Americans” sample and one significant SNP was identified in the “African and African-Americans” sample (Nievergelt et al., 2019). When stratifying by both race and sex, three more significant SNPs were found in males only (two in

“European and European Americans” male, one in “African and African-Americans” males, none in females of either geographic ancestry group) (Nievergelt et al., 2019). There was zero overlap between significant SNPs in “European and European Americans” and “African and

African-Americans” samples (Nievergelt et al., 2019). Consistent with many other large GWAS results, the identified genome-wide significant SNPs were not identified as risk loci for PTSD in previous small GWAS or candidate gene studies (Logue et al., 2015; Nievergelt et al., 2019).

While the identification of novel correlations is a major goal of GWAS, the ultimate goal is to identify causal variants (i.e. SNPs which have a causal role). Due to the ubiquity of linkage disequilibrium (i.e. the tendency for SNPs to have nonrandom correlations with one another), complex genetic phenotypes such as PTSD tend to have many more correlated SNPs than they do causal variants (i.e. SNPs which truly affect disease risk) and it is advisable to perform additional analysis in order to interpret results (De Leeuw et al., 2016; P. F. Sullivan, 2007).

In order to aid interpretation of SNP results, Functional Mapping and Annotation

(FUMA) was used to investigate likely functional roles of SNP hits (i.e. SNPs identified as correlated with PTSD) using statistical gene prioritization from gene function data drawn from Genetic Risk Factors for PTSD….. 19

18 biological data repositories (Nievergelt et al., 2019). Using FUMA, the six associated SNPs were mapped to 10 total genes; some intergenic SNPs had multiple genes within their loci, allowing some SNPs to map to multiple genes (Nievergelt et al., 2019). In addition to a lack of common variants in the two geographic ancestry groups, there was zero overlap in genes which fell within the loci of significant SNPs of the two groups (Nievergelt et al., 2019). This suggests that the genome-wide significant SNPs identified in those particular samples may impact different biological processes.

The four significant SNPs from “European and European Americans” samples

(rs34517852, rs14875732, rs9364611, rs571848662) were mapped to five genes (Nievergelt et al., 2019). Evidence for a functional role of these SNPs was either inconclusive or suggestive of no direct functional impact on core biological processes implicated in PTSD (Nievergelt et al.,

2019). The locus for rs34517852 included the gene ZDHHC14, which encodes a mitochondrial expressed in the brain (Nievergelt et al., 2019). Of note, rs34517852 was shown to be in high linkage disequilibrium with the functional variant rs35262389 and the Combined

Annotation-Dependent Depletion score (i.e. a machine-learning based estimation of deleteriousness based on simulated estimates of selective pressure) suggested that rs35262389 is more likely the causal variant (Nievergelt et al., 2019). However, little is known about the biological function of rs35262389 (Nievergelt et al., 2019). No active chromatin (i.e. chromatin from which RNA can be transcribed) was found within the locus (i.e. genome location) for rs14875732, which included the genes KAZN and TMEM51-AS1, indicating a spurious correlation (Nievergelt et al., 2019). Thus, all available evidence suggests that rs34517852 and rs14875732 had no true effects on PTSD (Nievergelt et al., 2019). The rs9364611 locus may have weak effects on and included the gene PARK2, which has a possible indirect Genetic Risk Factors for PTSD….. 20 role in dopaminergic neuron (Nievergelt et al., 2019). However, the functional implications of these weak effects on transcription and indirect effect on dopamine are unclear

(Nievergelt et al., 2019). There was only weak transcription in chromatin of rs571848662 locus, which was mapped to the zinc finger coding gene ZNF813 (Nievergelt et al., 2019). It is unclear whether rs9364611 or rs571848662 has a functional role in PTSD. Due to the small world nature of networks, including genetic and protein networks, the omnigenic model suggests that any variant with regulatory effects in at least one disease-relevant tissue will have some

(likely very small) effect on disease risk (Boyle et al., 2017). Thus, it may be that some of the aforementioned active SNPs have small, nontrivial indirect effects on neurotransmitter receptor expression and/or PTSD through interactions within protein networks (Boyle et al., 2017). What is clear is that none of these SNPs represents a direct causal variant of PTSD nor directly alter neurotransmitter receptor encoding or expression (Nievergelt et al., 2019).

In contrast with findings from “European and European Americans” samples, the two

SNPs from “African and African-Americans” samples (rs115539978, rs142174523) were mapped to a total of five genes (Nievergelt et al., 2019). The SNP rs115539978 included three genes within its risk locus (LINC02335, MIR5007, TUC338). Though little is known about the function of those genes, rs115539978 interacts with stress to influence non-coding RNA for

LINC00458 (promotes neuron differentiation (Miller et al., 2018)) and TUC338 (may contribute to neuron differentiation (Wen et al., 2018)), suggesting a possible indirect role in mediating the impact of stress on neurodevelopment. Though some non-coding RNA play an indirect role in the genetic expression of neurotransmitter receptors, these effects are typically diffuse, rather than being directed at a single causal pathway (Pickrell et al., 2010). Integration of neuroimaging and psychophysiology data, which were available for a subset of participants (neuroimaging n = Genetic Risk Factors for PTSD….. 21

87 and psychophysiology n = 299) who were genotyped at rs1155539978, suggested that T allele carriers have increased amygdala volume and decreased fear potentiated startle habituation, perhaps identifying downstream effects of rs115539978 mutations (Nievergelt et al., 2019).

Additionally, preliminary evidence suggested a possible functional role for the other risk SNP, rs142174523. Within the risk locus were two genes which impact immune functioning via regulatory effects on transcription factor binding, LINC0257 and HLA-B (Nievergelt et al.,

2019). Expression quantitative trait loci analysis (analysis of which loci in the genome best explain variation in levels of a given mRNA) suggested that rs142174523 alleles are differentially association with expression of at least ten genes, most of which primarily act as regulators of other genes and proteins (Nievergelt et al., 2019). Thus, “African and African-

Americans” SNPs appeared to increase risk for PTSD through wide-reaching interactive networks, rather than direct effects on core PTSD mechanisms (Boyle et al., 2017). This finding is consistent with the omnigenic model, which states that a majority of the heritability of complex genetic phenotypes stems from “peripheral genes” that do not encode for core disease molecules nor directly impact their expression (Boyle et al., 2017). Peripheral genes are thought to exert small but nontrivial effects on a broad group of biological processes through indirect effects (Boyle et al., 2017). In contrast, “core genes” are thought to exert concentrated effects on a particular biological mechanism by either encoding or directly regulating expression of molecular targets (Boyle et al., 2017). Robust evidence for PTSD core genes has been elusive, but a recent study by Million Veteran Program may have identified at least one core gene of

PTSD re-experiencing symptoms (Gelernter et al., 2019).

In contrast with SNPs discussed so far, the recent Million Veteran Program study identified SNPs which may impact core genes in PTSD re-experiencing symptoms. For example, Genetic Risk Factors for PTSD….. 22 one SNP was mapped to CRHR1, which directly encodes the corticotropic releasing hormone receptor 1 (CRHR1) (Gelernter et al., 2019). Importantly, CRHR1 has long been theorized as an important molecular target within the canonical PTSD mechanism of dysregulated HPA-axis stress response (Boyle et al., 2017; Gelernter et al., 2019). Additionally, gene-based testing identified CRHR1 as a top gene and that result was replicated in the UK Biobank cohort

(Gelernter et al., 2019). Further, follow-up analyses convincingly supported CRHR1 as the causal variant (Gelernter et al., 2019). Another significant SNP was mapped to HSD17B11, which directly encodes an enzyme involved in catalyzing synthesis of steroid hormones such as testosterone and androgen (Gelernter et al., 2019). Thus, two out of eight SNPs appear to exert biological effects which align with current conceptualizations of stress-related psychopathology and are concentrated within the neuroendocrine stress response (Gelernter et al., 2019).

Additionally, both of the two SNPs were nominally significant in a smaller replication study in separate cohort with far lower military combat trauma exposure (Gelernter et al., 2019).

However, a majority of genome wide significant SNPs in the study may be peripheral genes.

In contrast with the two aforementioned SNPs from the Million Veteran Program study, the other significant SNPs in the study were linked to genes with diffuse mechanisms as well as pleiotropy (i.e. the influence of one gene on multiple theoretically nonoverlapping phenotypes), suggesting that these may have been peripheral genes (Gelernter et al., 2019). These findings fit well with the omnigenic model, which suggests that most of the genes which contribute to any given complex disorder impact various phenotypes through indirect effects on many biological processes, including a small group of core genes for each phenotype (Boyle et al., 2017).

Confidence in Million Veteran Program findings is increased by replication using a candidate gene approach in a separate sample; five of eight SNPs replicated at the nominal significance Genetic Risk Factors for PTSD….. 23 level (p < .05), one SNP replicated after BonFerroni correction, and all eight SNPs showed concordant direction of change (Gelernter et al., 2019). Million Veteran Program findings increase confidence in the possibility of identifying core genes for PTSD, but it is unclear whether core genes exist for the entire heterogenous syndrome that is PTSD (Boyle et al., 2017;

Galatzer-Levy & Bryant, 2013). Gene-set analysis has been proposed as a possible method for identifying groups of genes that collectively are core to complex genetic disorders such as PTSD

(Holmans et al., 2009; Mooney & Wilmot, 2015).

1.5 Genetic Pathways Analysis: From Summary Statistics to Novel Insights Due to the small effects of individual SNPs and high polygenicity of PTSD, extremely high-powered studies are needed for robust findings at the level of SNPs (De Leeuw et al.,

2016). Further, due to the stringent statistical corrections required for multiple testing of millions of SNPs, even large GWAS studies with tens of thousands of participants likely have high type-

II error rates (Wang et al., 2010). Based on estimates that the heritable portion of PTSD risk stems from thousands of risk variants, archival GWAS data may contain many additional insights awaiting discovery (Skelton et al., 2012). However, due to the high potential for secondary analyses to be biased, any attempts to create new knowledge from old data must be carried out carefully (Kenakin & Christopoulos, 2013). Though the current study is underpowered due to small sample size, preliminary studies such as the current work can be helpful in guiding careful design of future higher-powered studies (Lee, Whitehead, Jacques, &

Julious, 2014). First, an overview of the statistical and conceptual bases of currently used methods is informative. Genetic Risk Factors for PTSD….. 24

In order to detect a true signal that was previously missed, increased ability to more effectively model the phenomenon of interest and increased statistical power are both important

(Greenland et al., 2016). While GWAS analysis is tailored toward separating out the effects of many individual SNPs, some groups of SNPs aggregate within functionally distinct regions of

DNA and act synchronously (along with other nucleotides in the region) in order to encode and regulate the synthesis of a particular RNA or protein (Sartor et al., 2012; P. F. Sullivan, 2007).

The combined effects of these distinct groups of nucleotides (aka genes), are aggregated into one common biological endpoint (De Leeuw et al., 2016). By statistically aggregating SNP effects into gene effects, gene-based analysis increases statistical power while potentially creating units of analysis that are more suited to detecting significant effects from genes with multiple causal

SNPs (Mooney & Wilmot, 2015). Due to these properties, gene-based analysis is a useful complementary approach for GWAS (Plomin, DeFries, Knopik, & Neiderhiser, 2016; Wang et al., 2010). Findings from the most recent Psychiatric Genomics Consortium study illustrate the potential for well-conceived complementary or secondary analyses to detect signals that GWAS misses due to being underpowered and/or tailored to a different level of analysis (Nievergelt et al., 2019).

Using complementary gene-based analysis, the recent Psychiatric Genomics Consortium study identified two enriched genes (i.e. genes that were statistically significant after controlling for multiple comparisons) in the “European and European Americans” sample but none in the

“African and African-Americans” sample (Nievergelt et al., 2019). Of note, the “trans-ethnic” sample was not tested at the gene level due to the general principal that GWAS analysis that does not yield a genome-wide significant SNP is not expected to yield any enriched genes (Nievergelt et al., 2019). Demonstrating the possibility for gene-based analysis to detect signals that GWAS Genetic Risk Factors for PTSD….. 25 may not, the two identified genes in the “European and European Americans” sample did not overlap with the loci of any of statistically significant GWAS SNPs (Nievergelt et al., 2019).

More specifically, in the “European and European Americans” combined sex sample, the genes

SH3RF3 and PODXL were found to be correlated with PTSD (Nievergelt et al., 2019).

Consistent with findings from other disorders for which gene-based analysis have identified novel candidate mechanisms (Holmans et al., 2009), the identified genes were not previously examined as candidate genes for PTSD (Nievergelt et al., 2019). Consistent with the omnigenic hypothesis, SH3RF3 contributes to a wide range of biological functions that do not necessarily directly target core processes of PTSD (Boyle et al., 2017; Safran et al., 2010). Processes impacted by SH3RF3 include protein-protein interactions, metal ion binding, ubiquitin protein activity, and activity (Nievergelt et al., 2019). Similarly, PODXL aligns with the omnigenic model as it is a highly pleiotropic gene that is implicated in alcoholism,

Alzheimer’s disease, , epilepsy, and schizophrenia (Boyle et al., 2017; Nievergelt et al., 2019). Biologically, PODXL is expressed in microvessals and may impact blood brain barrier function, which in turn may impact selective transport of critical molecules such as glucose, water, and amino acids (Shishkina, Kalinina, Berezova, & Dygalo, 2012). The observation that these genes may be best classified as peripheral genes adds credence to the omnigenic model. While the identification of novel genes demonstrates the capacity for gene- based analysis to supplement GWAS findings and to identify possible genetic risk factors that have not been hypothesized, gene-based analysis can also be used to corroborate and increase confidence in GWAS findings.

Demonstrating the potential for gene-based analysis to both supplement and support

GWAS findings, the Million Veteran Program study used the same gene-based technique to Genetic Risk Factors for PTSD….. 26 identify 30 enriched genes, including the aforementioned corticotrophin releasing hormone receptor 1 coding gene, CRHR1 (Gelernter et al., 2019). Of the 30 identified genes, zero overlapped with Psychiatric Genomics Consortium-identified PTSD risk genes from either

GWAS mapping or gene-based analysis (Nievergelt et al., 2019). The reasons for this are unclear, but one possible reason is that the Million Veteran Program study is focused on the

DSM-IV reexperiencing symptom cluster, which represents just one aspect of PTSD, and therefore might be a more precisely specified phenotype with its own unique core genes (T. P.

Sullivan, Fehon, Andres‐Hyman, Lipschitz, & Grilo, 2006). However, given that these studies both detected only a small fraction of the estimated thousands of nontrivial causal variants, it is important not to assume that a lack of correlation in GWAS implies a lack of causal influence

(De Leeuw et al., 2016). In order to detect true signals within putative causal pathways, especially pathways that might be better conceptualized at a higher level of analysis than genes, analysis of biologically related gene-sets may yield novel insights.

In order to increase the probability of detecting effects that are not detected by gene- based analysis, further consolidation of data into biologically related groups is desirable (De

Leeuw et al., 2016). Further, in addition to increasing statistical power, the further consolidation of genes into gene-sets allows one to obtain a more holistic picture of PTSD genetics by modeling the biological pathways within which genes exert their effects (Belmont & Leal, 2005;

Boyle et al., 2017; De Leeuw et al., 2016). In support of that goal, gene-set analysis allows one to tailor analysis to pertinent research questions (Yoon et al., 2018b). For example, more than

10,000 different publicly available gene-sets are available for download through databases such as MsigDB (Mooney & Wilmot, 2015; Yoon et al., 2018a). Accessing these databases, researchers can either test all available (or many available) gene-sets in order to implement a Genetic Risk Factors for PTSD….. 27 hypothesis-blind agnostic approach or can hand pick gene-sets suited to their interests (Mooney

& Wilmot, 2015). Additionally, researchers may create manually curated gene-sets based on their own literature review or may create or download statistically generated datasets in order to implement a data-driven approach (Mooney & Wilmot, 2015). The Psychiatric Genomics

Consortium study applied gene-set analysis to a broad group of 10,894 different gene-sets downloaded from MSigDB(Nievergelt et al., 2019). Findings support the commonly stated belief among gene-set analysis pioneers that analysis at the level of gene-sets may yield insights which are otherwise missed.

Using a gene-set analysis technique known as MAGMA, the recent Psychiatric Genomics

Consortium study identified four enriched gene-sets (Nievergelt et al., 2019). In gene-set analysis, the term enriched gene-set is used to describe a gene-set which, after adjusting the threshold for statistical significance to account for multiple comparisons, has been identified as more strongly correlated with the disorder than would be expected by chance (De Leeuw et al.,

2016). Enrichment is tested using a competitive hypothesis tests, which compares the level of correlation of genes within the gene-set to the average level of correlation of all other genes in the genome (De Leeuw et al., 2016). In the Psychiatric Genomics Consortium study, two enriched gene-sets were identified within the “European and European Americans” sample and two enriched gene-sets were found in the “African and African-Americans” sample (Nievergelt et al., 2019). As was the case during gene-level analysis, the “trans-ethnic” sample was not tested at the gene-set level due to the general principal that GWAS analysis that does not yield a genome-wide significant SNP is not expected to yield any enriched gene-sets (Nievergelt et al.,

2019). Genetic Risk Factors for PTSD….. 28

In contrast with significant results at the level of SNPs and genes, enriched gene-sets in the “European and European Americans” sample were relatively easy to interpret in relation to core biological substrates of PTSD (Nievergelt et al., 2019). More specifically, the one enriched gene-set from the “European and European Americans” combined-sex sample was a publicly available curated gene-set of 57 genes which were described as “Any process that activates or increases the frequency, rate or extent of tumor necrosis factor superfamily cytokine production”

(Nievergelt et al., 2019). A burgeoning body of literature, including a systematic review from

2017 which included 27 studies, suggest that Tumor Necrosis Factor (TNF) may be a promising potential biomarker of PTSD (A. K. Smith et al., 2011). Additionally, pre-clinical research demonstrates that acute stress can lead to elevated levels of TNF in the hippocampus, which in turn can impair memory (Ohgidani et al., 2016). Further, one pre-clinical study investigated individual differences in TNF response to stress (measured in the spleen) and found that rodents with higher TNF stress response exhibited higher levels of depression-like behavior (Gómez-

Lázaro et al., 2011). Based on the known biological functions of genes within the enriched TNF gene-set (G. S. E. A. Database, 2020), it is possible that variation in these genes may contribute to individual differences in susceptibility to developing PTSD through differential modulation of

TNF stress response. This demonstrates that there is a theoretically coherent explanation for the connection between the TNF gene-set and PTSD. Of note, when viewed in isolation, many of the individual genes within the enriched TNF gene-set have diffuse effects on many biological processes (G. S. E. A. Database, 2020) and would therefore fit the profile of a peripheral gene

(Boyle et al., 2017). Based on this evidence, gene-set analysis may enable researchers to make sense of the processes through which the combined effects of many genes contribute to specific complex genetic phenotypes. The gene-set analysis performed in the combined-sex “European Genetic Risk Factors for PTSD….. 29 and European Americans” sample provides corroborating evidence and perhaps broadens this theory.

The one enriched gene-set in the male-only “European and European Americans” sample included 47 genes and the function was described as “the appearance of interleukin-1 beta due to biosynthesis or secretion following a cellular stimulus, resulting in an increase in its intracellular or extracellular levels (Nievergelt et al., 2019).” In support of the theorized clustering of genetic effects on PTSD into TNF modulating pathways, one of the top significant genes within that data set was TNF alfa induced protein 3 (Nievergelt et al., 2019). Perhaps more importantly, both

TNF and Interleukin 1 beta are proinflammatory cytokines, suggesting an aggregation of genetic effects into biological systems which regulate pro-inflammatory cytokine expression (Nievergelt et al., 2019). Additionally, a number of studies have found evidence to support a role of interleukin 1 beta in PTSD (A. K. Smith et al., 2011). Further, one pre-clinical study found that stress-enhanced fear learning, which is an animal model of PTSD, led to a region-specific increase in interleukin 1 beta mRNA and protein expression dorsal hippocampus (Jones,

Lebonville, Barrus, & Lysle, 2015). Intriguingly, post-stress administration of an interleukin 1 beta antagonist prevented the stress-enhanced fear learning, suggesting a possible causal role of increased interleukin 1 beta expression and activity in at least one aspect of PTSD (Jones et al.,

2015). Much like the enriched TNF gene-set, the enriched interleukin 1 beta gene-set is comprised of numerous individual genes which exert effects on many diverse biological processes (G. S. E. A. Database, 2020), implying that they are peripheral genes (Boyle et al.,

2017). Based on this evidence, both of the enriched gene-sets identified in “European and

European Americans” samples have identified groups of peripheral genes which might contribute to PTSD risk largely through their aggregated effects on stress-induced alterations in Genetic Risk Factors for PTSD….. 30 proinflammatory cytokine expression. Results from the gene-set analysis performed in the

“African and African-Americans” combined-sex sample are somewhat more difficult to integrate into this theory but may ultimately converge into a common causal system.

In contrast with “European and European Americans” enriched gene-sets, the two enriched gene-sets found in the “African and African-Americans” combined sex sample are not directly related to the impact of pro-inflammatory cytokines on PTSD (Nievergelt et al., 2019).

The top enriched gene-set in that sample is defined by a role in metabolism (“interacting selectively and non-covalently with any enzyme”) (Nievergelt et al., 2019). Though this gene-set is not primarily defined based on direct convergence into pro-inflammatory or immunological pathways, a recent synthesis of the literature demonstrates that inflammatory and metabolic aspects of PTSD may represent different facets of a common systemic pathology

(Mellon, Gautam, Hammamieh, Jett, & Wolkowitz, 2018). In support of this view, two of the five statistically significant genes within the thioesterase enzyme gene-set were TNF related

(Nievergelt et al., 2019). Additionally, it is worth noting that this gene-set included several additional inflammation related genes and was defined by interaction with an enzyme for lipid metabolism (Nievergelt et al., 2019), perhaps supporting the theorized link between metabolism and immunology in PTSD. The other enriched gene-set in that sample was related to a pleiotropic cytokine (i.e. a cytokine that has both pro-inflammatory and anti-inflammatory properties), but only included genes which are up-regulated by that cytokine in pituitary cancer cells (“genes up-regulated in GH3 cells after treatment with leukemia inhibitory factor”)

(Banner, Patterson, Allchorne, Poole, & Woolf, 1998; Nievergelt et al., 2019). In spite of its name, leukemia inhibitory factor has a wide array of biological functions that are not specific to leukemia, many of which impact brain cells and are relevant to behavior (Banner et al., 1998). Of Genetic Risk Factors for PTSD….. 31 particular relevance to PTSD, pre-clinical models show that leukemia inhibitory factor regulates glucocorticoid receptor expression in the HPA axis (Kariagina, Zonis, Afkhami, Romanenko, &

Chesnokova, 2005). Though it is curious that enrichment was detected in a gene-set specifically tailored to pituitary cancer cells rather than brain cells, this may have been due to methodology.

The Psychiatric Genomics Consortium study used hypothesis blind gene-set analysis and the database which provided the gene-sets did not include any brain-specific leukemia inhibitory factor gene-sets (M. Database; Nievergelt et al., 2019). Taken as a whole, though no gene-set was enriched in more than one sample, all enriched gene-sets may have been related to a common systemic dysfunction.

The hypothesis-blind gene-set analysis performed in the most recent Psychiatric

Genomics Consortium GWAS of PTSD identified four enriched gene-sets and highlighted the possible concentration of genetic risk into genes which interact with inflammatory cytokines.

Though the findings are highly pertinent to PTSD, a hypothesis-driven approach might yield additional insights, specifically with respect to putative targets for pharmacotherapy.

1.6 Neurotransmitter Receptors and Curated Gene-sets: The Case For a Hypothesis-Driven Gene-set Analysis Though one gene-set analysis has already been performed on the most recent Psychiatric

Genomics Consortium GWAS, gene-set analysis describes a diverse set of flexible techniques which can be tailored to fit specific research questions (De Leeuw et al., 2016). Thus, by tailoring the approach to fit a pertinent topic, this study was designed to address a gap in the literature. One particular topic with clear clinical implications is enrichment of PTSD risk in neurotransmitter receptor gene-sets (Krystal et al., 2017). Genetic Risk Factors for PTSD….. 32

It is broadly accepted that neurotransmitters influence emotions, cognition, behavior, and the pathogenesis of psychiatric disorders such as PTSD (A Kato et al., 2013). Importantly, once they are released into the synaptic cleft, the effects of neurotransmitters largely depend upon the receptors to which they bind (A Kato et al., 2013; Albert & Vahid-Ansari, 2018; Bankson &

Cunningham, 2001; Pillai et al., 2017). Thus, genetic mutations which contribute to individual differences in receptor availability or function are of interest to psychiatric researchers (Day &

Tuite, 1998; Jacobsen, Vanderluit, Slack, & Albert, 2008). Glutamatergic, GABAergic, cholinergic, dopaminergic, serotonergic, adrenergic, neuropeptide Y, opioid, and groups are of particular interest in PTSD research (Krystal et al., 2017). In pharmacotherapy of PTSD and other psychiatric disorders, a common strategy is to increase synaptic levels of a given neurotransmitter and thus increase activation of all receptors to which it binds (A Kato et al., 2013; Krystal et al., 2017). Thus, gene-sets which include all receptors for a given neurotransmitter may capture concentration of genetic effects into groups of molecular targets that are acted upon by drugs such as reuptake inhibitors or releasing agents (A Kato et al.,

2013). However, the gene-sets used for gene-set analysis in the recent Psychiatric Genomics

Consortium study were attained from MsigDB, which does not include complete gene-sets for all receptors in those aforementioned systems (M. Database; Nievergelt et al., 2019). Thus, curated gene-sets may be needed to fully explore genetic enrichment among these functionally related groups of genes.

Though the combined effects of groups of receptors is a topic of interest, the specific effects of influential receptor subtypes is equally pertinent (Artigas, 2016; Ossowska et al.,

2001). For example, though many common antidepressants work by increasing synaptic neurotransmitter levels many current and candidate drugs target one specific receptor or a few Genetic Risk Factors for PTSD….. 33 specific receptors (A Kato et al., 2013; Albert & Vahid-Ansari, 2018; Bankson & Cunningham,

2001; Pillai et al., 2017). Thus, in order to give a more detailed analysis of this topic, it is important to consider not only the gene or genes which encode the receptor itself but also the gene or genes which encode for the intracellular signaling mechanisms through which a particular receptor exerts its effects. In the KEGG pathways database, genes encoding the proteins involved in synaptic intracellular signaling pathways are well described for glutamate,

GABA, acetylcholine, dopamine, and serotonin ("KEGG Pathway "). Thus, there are pre-curated and easily applied gene-sets which can be adapted in order to examine the topic of neurotransmitter receptors from two important angles. However, MsigDB does not include any gene-sets which isolate any of the specific receptors in any of these systems (G. S. E. A.

Database, 2020). By examining both groups of receptors and specific receptor subtype systems, this proposed study aims to provide both breadth and depth.

Glutamate is the most abundant excitatory neurotransmitter in the human nervous system

(Artigas, Celada, & Bortolozzi, 2018). As such, it affects a wide array of functions through excitatory receptors (Jooyeon & N, 2011). The four classes of glutamate receptors are encoded by a set of 24 total genes (Brown, Sweatt, & Kaas, 2019). A number of reviews have suggested that glutamate is among the most influential neurotransmitters in PTSD pathogenesis (Averill et al., 2017). In one positron emission tomography (PET), mGluR5 availability was correlated with

PTSD (Holmes et al., 2017). Additionally, one magnetic resonance spectroscopy (SPECT) study found a correlation between glutamate- complex blood metabolites and PTSD

(Jooyeon & N, 2011). Further, an expert group selected by the PTSD Psychopharmacology

Working Group included NMDA antagonists among the most highly recommended possible drug targets (Krystal et al., 2017). However, an analysis of a postmortem sample found increased Genetic Risk Factors for PTSD….. 34 expression of SHANK1 but no difference in mGluR5 expression, suggesting that the observed increases in mGluR5 likely results from SHANK1’s role in anchoring mGluR5 to the cell surface

(Holmes et al., 2017). Glutamatergic receptor polymorphisms have been implicated in anxiety disorders and transporter in PTSD (Jooyeon & N, 2011). However, it is unclear whether those findings are robust (Duncan, Cooper, et al., 2018). Based on the cumulative evidence, glutamatergic receptors are among the most promising molecular targets for PTSD. However, the degree to which variation in glutamatergic genes plays a role is unclear.

In contrast with glutamate, GABA is the most abundant inhibitory neurotransmitter in the human nervous system (Tokarski, Kusek, & Hess, 2011). As such, it affects a wide array of functions through inhibitory receptors (Chebib & Johnston, 1999). The three main types of

GABA receptors are encoded by a set of 21 total genes (E. M. Jorgensen, 2005). A collection of studies suggests involvement of GABA in PTSD. For example, one in vivo PET study found a correlation between PTSD and GABAA (Jooyeon & N, 2011). Further, a number of studies find altered response to GABAergic drugs in PTSD (Krystal et al., 2017) and one SPECT study found a correlation between GABA metabolites in blood and PTSD (Jooyeon & N, 2011).

Additionally, heterozygosity (i.e. the presence of one major allele and one minor allele) of one

SNP within the GABRB3 gene (encodes the GABAA receptor β3 subunit) was correlated with

PTSD in one candidate gene study (Feusner et al., 2001). Overall, though GABA and glutamate have opposite effects on neural excitability, evidence suggests that GABA and glutamate may each play a role PTSD pathogenesis and, perhaps, treatment. However, more evidence is needed to assess the impact of genetic variants related to the receptors of those neurotransmitters.

While not as abundant as glutamate or GABA, acetylcholine has important functions for parasympathetic nervous system activity, attention, memory, and motivation (Picciotto, Higley, Genetic Risk Factors for PTSD….. 35

& Mineur, 2012). The two classes of cholinergic receptors are encoded by a set of 14 genes

(Johnston, Staines, Klein, & Marshall-Gradisnik, 2016). One study each has found correlations of peripheral levels of cholinergic receptor A and nicotinic with PTSD

(Jooyeon & N, 2011). Additionally, pre-clinical studies suggest that cholinergic drugs may improve behavioral sequalae of PTSD (Rao Sun et al., 2017). Importantly,

Inhibitors have improved PTSD-like symptoms in a case study and a small pilot trial (Krystal et al., 2017). Though the role of acetylcholine in PTSD is not as well studied as some other neurotransmitters, preliminary evidence suggests that it might be a worthwhile target to include in this study.

Dopamine plays multiple important biological roles, most notably reward system mediation (Rasheed et al., 2010). Dopamine receptors are grouped into two receptor families, encompassing 5 total subtypes each of which is encoded by a single gene (Doll, Bath, Daw, &

Frank, 2016). Dopamine is often associated with PTSD and major depressive disorder, among other mood disorders (Li et al., 2016). Of note, a number of candidate gene studies have found correlations of dopamine receptors and dopamine transporter with PTSD (Li et al., 2016; A. K.

Smith et al., 2011). Further, the aforementioned expert group suggested the dopamine receptor

D2 as a promising target (Krystal et al., 2017). One candidate gene study found a correlation with the A1 allele of DRD2 (gene which encodes the ) and replicated that finding (Li et al., 2016). However, a later study reported no correlation between any allele of D2 gene and PTSD, demonstrating the need to use caution in interpreting candidate gene results

(Blum et al., 2019). Due to dopamine’s role in reward-motivated behavior, which may influence depressive symptoms in PTSD, this gene-set is highly worth investigating. Genetic Risk Factors for PTSD….. 36

Like dopamine, serotonin is a monoamine neurotransmitter and plays multiple important biological roles (G. M. Sullivan et al., 2005). In contrast with dopamine, it is most notably implicated in mood regulation (Van Praag & De Haan, 1980). Serotonin exerts its effects through seven families of receptors, which are encoded by a total of 18 genes (Żmudzka, Sałaciak, Sapa,

& Pytka, 2018). Though the only two FDA approved pharmacotherapies for PTSD are both selective serotonin reuptake inhibitors (SSRIs), it is unclear whether serotonin is the ideal molecular target for PTSD (Krystal et al., 2017). However, a number of candidate gene studies have linked polymorphisms of the serotonin transporter with PTSD (Jooyeon & N, 2011). In comparison to SERT, specific serotonin receptors have received relatively little attention in

PTSD research. However, one PET study found that availability of the serotonin-1A receptor and expression of the minor allele of a SNP within the locus of the HTR1A gene (encodes serotonin-

1A receptor) were both positively correlated with PTSD (G. M. Sullivan et al., 2013). Further, several pre-clinical studies suggest that dysregulated expression of 5-HT1A might be a consequence of trauma exposure and might lead to common trauma sequalae seen in PTSD (Lin,

Liu, & Sun, 2017). In order to explore an underexamined aspect of the serotonergic system, an investigation of a receptor gene-set may be informative.

Adrenergic receptors are activated by several catecholamines, most notably epinephrine and norepinephrine (Ciccarelli, Sorriento, Coscioni, Iaccarino, & Santulli, 2017). Like serotonin and dopamine, noradrenaline is another important monoamine neurotransmitter with multiple roles (Ciccarelli et al., 2017). In contrast with acetylcholine, it is the primary neurotransmitter of the sympathetic nervous system (Jooyeon & N, 2011). As such, it is implicated in the fight or flight response and likely highly relevant to fear-related PTSD symptoms (Liberzon et al., 2014).

There are 9 subtypes of adrenergic receptors, each encoded by a single gene (Ciccarelli et al., Genetic Risk Factors for PTSD….. 37

2017). One candidate gene study found correlation of PTSD and beta-2 adrenergic receptor

(Liberzon et al., 2014). Importantly, alpha 1 receptor antagonists were noted as a promising target by the aforementioned group of experts chosen by the PTSD Psychopharmacology

Working Group (Krystal et al., 2017). Due to noradrenaline’s role in fear response and fear learning (Inoue, 1993), the adrenergic receptor gene-set is among the most theoretically likely to contain core genes for PTSD.

Neuropeptide-Y is the most abundant peptide in the human central nervous system, though it also is involved in the peripheral nervous system (Sah & Geracioti, 2005).

Neuropeptide-Y is believed to be involved in anxiety and stress reduction, reduced pain perception, among other functions (Sah & Geracioti, 2005). Each of its four known receptor subtypes are metabotropic and encoded by a single gene; a fifth receptor is suggested by receptor-binding profiles but still preliminary (Reichmann & Holzer, 2016). Thus, there are five known genes in the gene-set. Recently, neuropeptide-Y has become increasingly of interest to

PTSD researchers and several recent reviews have detailed its potential role in PTSD etiology

(Sah & Geracioti, 2005). Further, the PTSD Working Group panel noted it as a promising molecular drug target (Krystal et al., 2017). Given that relatively little research exists exploring the relationship between neuropeptide-Y and PTSD in humans, neuropeptide-Y related gene-sets may be worth exploring with respect to PTSD.

Finally, cannabinoid receptors and opioid receptors are two groups of receptors worth exploring through gene-set analysis of PTSD. Opioid receptors respond to any of three endogenous opioid peptides and play a role in pain modulation and mood; the five types are each encoded by a single gene (C. Stein, 2016). Cannabinoid receptors respond to endocannabinoids and the system includes just 2 receptors with one gene each Genetic Risk Factors for PTSD….. 38

(Moreira, Grieb, & Lutz, 2009). Kappa opioid receptors and CB1 receptors have both been found to be correlated with PTSD in one PET study each (Jooyeon & N, 2011). Further, the expert group listed cannabinoid receptors second on its list of promising molecular targets (Krystal et al., 2017). Thus, it is worth creating a gene-set for cannabinoid receptors and a gene-set for opioid receptors.

While most of this section has focused on neurotransmitter receptors, receptors exert their effects through complex intracellular cascades which involve numerous additional proteins

(Réus, Generoso, Rodrigues, & Quevedo, 2019). Recently, these pathways have gained increased attention in psychiatric research, especially with respect to the potential for biased agonist drugs which target specific intracellular pathways of g-protein coupled receptors via preferential activation of g-proteins (Komatsu, Fukuchi, & Habata, 2019; Seyedabadi, Ghahremani, &

Albert, 2019). In order to understand neurotransmitter receptors and their psychological import, it might be critical to examine their intracellular signaling cascades (Albert & Vahid-Ansari,

2018; Rojas & Fiedler, 2016).

When viewed as a whole, the evidence from this section demonstrates that glutamatergic,

GABAergic, cholinergic, dopaminergic, serotonergic, adrenergic, neuropeptide Y, opioid, and cannabinoid receptor groups are all promising targets for future research in the pathogenesis and treatment of PTSD. In spite of recent advances, much remains unknown about the degree to which genetic differences related to these receptor groups impact PTSD risk (Krystal et al.,

2017). Currently, numerous receptors might be key to understanding and treating PTSD, but the relative importance of these receptors compared to each other and to other biological mechanisms of PTSD is unknown (Akiki & Abdallah, 2018; Zoladz & Diamond, 2013). Further, to the extent that neurotransmitter receptors are important in the development and maintenance Genetic Risk Factors for PTSD….. 39 of PTSD, it is unclear whether observed differences in these receptors are due to the genes which encode for them and / or for the intracellular signaling proteins through which they exert their effects (Banerjee et al., 2017; Duncan, Cooper, et al., 2018). In order to examine these topics, the current study harnesses the statistical flexibility of gene-set analysis in order to examine numerous gene-sets that may help shed light om these issues. Though the current study is preliminary, findings may inform next steps in research.

Genetic Risk Factors for PTSD….. 40

2.0 Hypotheses Due to practical considerations, this work used summary statistics derived from a GWAS with far fewer participants (2,424 cases; 7,113 controls) than initially proposed (23,212 cases;

151,447 controls). The primary objective of the initially proposed study was to explore whether specific groups of neurotransmitter receptor genes are enriched for PTSD. Though the current data lacks sufficient statistical power to provide a robust test of initial hypotheses, the initial hypotheses remain of interest with respect to next steps for this line of research. Preliminary results from this study will be interpreted both respect to next steps and (tentatively) with respect to the initial hypotheses. Thus, initial hypotheses are presented below.

Previously reviewed evidence suggests that a number of neurotransmitters are influential in

PTSD, with molecular imaging, candidate gene findings, and clinical pharmacology experts specifically highlighting glutamate (Artigas et al., 2018; Krystal et al., 2017). Though less is known about neuropeptide Y, its inclusion as a target that was listed by an expert panel of PTSD pharmacotherapy researchers demonstrates that top minds in the field view it as a promising candidate (Krystal et al., 2017). Thus, the following working hypotheses are offered:

1) First, enriched gene-sets will include glutamate. 2) Second, enriched gene-sets will include neuropeptide Y. 3) Third, due to the exploratory nature of this study, it is hypothesized that this test will identify at least one enriched neurotransmitter receptor gene-set that was not included in hypothesis one or two.

An additional objective of this study is to explore whether any particular receptors for glutamate, GABA, acetylcholine, dopamine, or serotonin are enriched for PTSD. Previously reviewed evidence from neuroimaging studies suggests that metabotropic glutamate receptor 5, serotonin 1-A, serotonin 1-B, GABA(A), CB1, and kappa-opioid receptors may have a role in the etiology of PTSD (G. M. Sullivan et al., 2013). Further, candidate gene studies have found Genetic Risk Factors for PTSD….. 41 evidence for dopamine D2 and serotonin 1-A (Li et al., 2016; G. M. Sullivan et al., 2013). Thus, the following working hypothesis are offered:

4) Fourth, enriched gene-sets will include D2 5) Fifth, enriched gene-sets will include serotonin 1-A 6) Sixth, due to the exploratory nature of this study, it is hypothesized that this test will identify at least one enriched receptor subtype gene-set that was not included in hypothesis one or two. Of note, while the proposed study is hypothesis driven and tailored to a specific research question, it is also designed to strike a balance between cultivating a tailored approach while also exploring a wide range of neurotransmitter receptor gene-sets. As such, included hypotheses are best described as working hypotheses.

Genetic Risk Factors for PTSD….. 42

3.0 Method 3.1 PGC Methodology This study used summary statistics from a Psychiatric Genomics Consortium study as input data. Thus, in order to define certain fundamental aspects of the data for this study (e.g. definitions of cases and controls, indices of critical variables, inclusion / exclusion criteria, data cleaning), a description of relevant aspects of contributing studies is included.

3.1a Data Overview This study used summary statistics from a Psychiatric Genomics Consortium study as input data. This dissertation was initially proposed to use summary statistics from the most recent Psychiatric Genomics Consortium study. However, due to the COVID-19 shelter-in-place orders, it was not possible to access a computer with the requisite computing power. In order to still provide analyses, the present study used data from the second most recent Psychiatric

Genomics Consortium study (Duncan, Ratanatharathorn, et al., 2018). Input data were summary statistics derived from a GWAS with 2,424 cases and 7,113 controls (Duncan, Ratanatharathorn, et al., 2018), rather than the initially planned 23,212 cases; 151,447 controls (Nievergelt et al.,

2019). Input data included SNPs, labeled using standardized naming conventions (rs followed by a number) and p-values for each SNP. Due to the use of summary statistics, no subject recruitment was necessary, and no individual-level data were collected or used for this study.

Due to the use of data collected from the Psychiatric Genomics Consortium, many of the methodology relevant to data collection are a function of past studies.

3.1b PGC Contributing Studies: Sample Characteristics and Methodology Genetic Risk Factors for PTSD….. 43

The Psychiatric Genomics Consortium sample from which GWAS summary statistics used in this study were derived was comprised of participants who initially participated in any one of nine studies which contributed data for pooled analysis (Duncan, Ratanatharathorn, et al.,

2018). All contributing studies were conducted in the United States of American, all assessed

PTSD using DSM-IV criteria, and all divided participants into cases and controls (Duncan,

Ratanatharathorn, et al., 2018). However, methods for assessing PTSD case status varied between studies (Duncan, Ratanatharathorn, et al., 2018).

Of the nine studies which contributed data used in this study, three used self-repot measures and seven used in-person interviews delivered by trained interviewers (Duncan,

Ratanatharathorn, et al., 2018). Of the four studies using self-repot measures of PTSD, two used the PTSD Symptom Scale (PSS) and one used the PTSD Checklist (PCL) (Duncan,

Ratanatharathorn, et al., 2018). Briefly, the PSS and PCL are both 17 item measures that have been shown to have good psychometric properties in clinical samples (Foa, Johnson, Feeny, &

Treadwell, 2001; Weathers, Litz, Herman, Huska, & Keane, 1993). However, these measures differ in several noteworthy ways. First, while the PSS queries symptom frequency over the past two weeks, the PCL queries symptom severity over the past month (Foa et al., 2001; Weathers et al., 1993). Second, the scaling of these measures is somewhat different. While the PSS uses a four-point scale of frequency measures ranging from 0 (not at all) to 3 (3 to 5 or more times per week / very much / almost always), the PCL uses a five-point scale of severity measures ranging from 1 (not at all) to 5 (extremely) (Foa et al., 2001; Weathers et al., 1993). Third, the definitions of PTSD or probable PTSD are different. For the PSS, there is no established cutoff for PTSD

(Coffey, Gudmundsdottir, Beck, Palyo, & Miller, 2006). However, a symptom-level cutoff of 1

(once per week or less / a little bit / once in a while) has been used (Coffey et al., 2006). While Genetic Risk Factors for PTSD….. 44 the Psychiatric Genomics Consortium study did not specify which cutoff they used (and the constituent studies performed dimensional analyses), they did state in the supplemental material that they used a cutoff for each symptom (Duncan, Ratanatharathorn, et al., 2018). Conversely, the PCL includes an established cutoff score based on overall symptom score (30 or higher indicates probable PTSD)(Weathers et al., 1993). Fourth, while the PSS queries symptoms over the past two weeks, the PCL queries symptoms over the past month (Foa et al., 2001; Weathers et al., 1993). Fifth, there are some small differences in wording for some of the symptom questions (Foa et al., 2001; Weathers et al., 1993). For example, for symptom number 16, the

PSS queries how often participants have been “overly alert” while the PCL queries how much participants have been bothered by “being ‘super alert’ or watchful on guard” (Foa et al., 2001;

Weathers et al., 1993).

Of the four studies using in-person interviews, two used the Clinician Administered

PTSD Scale (CAPS), one used the Structured Clinical Interview for DSM-IV Disorders (SCID), one used the Semi-Structured Assessment for Drug Dependence and Alcoholism (SSADDA), two used the Diagnostic Interview Schedule for DSM-IV (DIS-IV), and one used a study- specific interview that mirrored the PCL (Duncan, Ratanatharathorn, et al., 2018). Briefly, the

CAPs is a 30-item structured interview and is considered the gold standard assessment for PTSD; the severity of each symptom is measured on a 5 point Likert Scale (Bremner, Steinberg,

Southwick, Johnson, & Charney, 1993). The SCID is a semi-structured interview guide that is often used in both clinical and research settings and has been shown to have good psychometric properties in clinical samples (Bremner et al., 1993). The SSADDA is a semi-structured interview that was primarily designed for the assessment of substance use disorders and their relationship to other DSM disorders (Bremner et al., 1993). The SSADDA has been shown to Genetic Risk Factors for PTSD….. 45 have moderate interrater reliability and substantial test-retest reliability for PTSD (Pierucci-

Lagha et al., 2005). The DIS-IV is a structured interview that has been shown to have good psychometric properties for PTSD (Breslau, Kessler, & Peterson, 1998).

Perhaps exacerbating the differences in PTSD diagnostic tools, some studies also differed with respect to the index trauma for PTSD. More specifically, five studies queried symptoms with respect to a self-selected most severe trauma, one queried with respect to two traumas (the most severe and a randomly chosen trauma; PTSD cases were defined as any who met criteria for either trauma), one separately assessed based on adult deployment-related trauma and also based on non-deployment trauma (PTSD cases met for either), and one queried with respect to

“up to three traumatic events” (Duncan, Ratanatharathorn, et al., 2018). Studies also varied with respect to methodology for assessing trauma exposure. One study assessed 25 types of trauma using the brief trauma questionnaire, one assessed 13 types of trauma using the traumatic events inventory, one assessed 19 types of trauma as part of the PCL-C interview, one assessed 15 trauma types using a study-specific list, on assessed 12 specific traumas using the SSADDA, two assessed 16 types of trauma using a combination of a modified childhood trauma questionnaire and life events checklist, and two assessed a combination of 16 categories of deployment-related

7 nondeployment related adult traumatic events and four adverse childhood experiences

(Duncan, Ratanatharathorn, et al., 2018).

While differences in methodology and incomplete reporting within the Psychiatric

Genomics Consortium study make it difficult to examine exact differences in exposure between- studies (i.e. these summary statistics are drawn from a study-specific genetic ancestry group and no detailed exposure statistics are provided by Psychiatric Genomics Consortium), one can infer from available information that these samples differed with respect to exposure. For example, Genetic Risk Factors for PTSD….. 46 two of the studies were conducted in military samples, one was conducted in a national guard samples, three were conducted in subsamples of large-scale studies of substance abuse, one was conducted in a study of nurses health, and two were conducted in samples designed to be representative of PTSD in the general United States population (Duncan, Ratanatharathorn, et al.,

2018). Further, while two studies allowed for the inclusion of trauma-exposed controls, seven did not (Duncan, Ratanatharathorn, et al., 2018). The Psychiatric Genomics Consortium study did not provide detailed reporting of the distribution of trauma-exposure in controls within the specific subset of the GWAS data which contributed to the summary statistics used in this work, but did report that 87.7% of controls were trauma-exposed (Duncan, Ratanatharathorn, et al.,

2018). Based on available evidence, the present work is best characterized as consisting of a mixed-trauma sample (Duncan, Ratanatharathorn, et al., 2018).

In addition to differences in sample characteristics and phenotyping methodology, the nine studies which contributed to the GWAS summary statistics used in this work differed with respect to the exact SNPs included in analysis. Though a comprehensive list of exactly which

SNPs are genotyped directly, imputed, or missing from each study is not available (likely due to the fact that direct genotyping includes tens or hundreds of thousands of SNPs while imputation includes millions), it is worth noting that these differences are study-specific and thus covary systematically with other methodological differences (Duncan, Ratanatharathorn, et al., 2018).

Underlying reasons for differences in which SNPs are sequenced and imputed in any given

GWAS include different methodologies for tagging and sequencing DNA, differences in sample size, and random chance (Bergen & Petryshen, 2012).

Though detailed descriptions of demographics, phenotypes, and genotypes were not readily available for the initially planned sample that would have used summary statistics from a Genetic Risk Factors for PTSD….. 47 more recent and larger GWAS, it is worth noting that the planned sample was even more heterogeneous. The larger sample included data from a combined 60 contributing studies

(Nievergelt et al., 2019). Further, that sample included studies which used PTSD definitions from DSM-III, DSM-IV, and DSM-5 (Nievergelt et al., 2019). Additionally, that sample included studies from several different nations and included an even wider range of PTSD diagnostic and screening tools, trauma types, and other between-study discrepancies (Nievergelt et al., 2019). Much like the sample used here, that study included controls who were mostly trauma-exposed but also included some who were not (exact percentage not reported)

(Nievergelt et al., 2019). In spite of differences in methodology between studies, the very high replication rate of large GWAS performed in other complex phenotypes (including analyses across geographically and ancestrally diverse populations) have led to a general consensus that the increased statistical power makes up for other limitations (Bergen & Petryshen, 2012;

Marigorta & Navarro, 2013; Marigorta, Rodríguez, Gibson, & Navarro, 2018).

3.1c PGC Study Design Overview The Psychiatric Genomics Consortium analyses, which were performed on pooled data collected from the above-detailed contributing studies, provided summary statistics which were analyzed in the current study. Of note, the current study only used the summary statistics from one of the four genetic ancestry groups included within the overall GWAS, titled “European

American.” Though the Psychiatric Genomics Consortium has also made summary statistics available from the GWAS meta-analyses of genetic ancestry groups titled “transethnic” and

“African American,” both of those files had higher SNP counts (Duncan, Ratanatharathorn, et al., 2018), which led to larger file sizes and required increased computing power that was not available at the time of this analysis. Genetic Risk Factors for PTSD….. 48

As an overview, the GWAS analysis which created the summary statistics used in this work as input data can be divided into six steps. Each of the six steps are described in brief here and are described in more detail in the sections following this paragraph. First, ancestry assignment was performed on individual-level SNP data in each of the 11 total contributing studies (nine of which were included in this work, two of which did not include any participants assigned to the “European American” genetic ancestry group) (Duncan, Ratanatharathorn, et al.,

2018). Based on ancestry assignment, the 11 total contributing studies were divided into 19 distinct data sub-sets (Duncan, Ratanatharathorn, et al., 2018). For each data sub-set, all individuals were participants of the same original contributing study prior and were assigned to the same genetic ancestry group based on Psychiatric Genomic Consortium ancestry assignment

(Duncan, Ratanatharathorn, et al., 2018). Second, after ancestry assignment, quality control procedures were performed on individual-level SNP data in each of the 19 distinct data sub-sets

(Duncan, Ratanatharathorn, et al., 2018). Third, SNP imputation was performed within each of the 19 data sub-sets (Duncan, Ratanatharathorn, et al., 2018). Fourth, relatedness testing was performed and individuals showing familial relation were removed (Duncan, Ratanatharathorn, et al., 2018). Fifth, GWAS was performed within each of the 19 data sub-sets (Duncan,

Ratanatharathorn, et al., 2018). Sixth, various stratified and combined meta-analyses were performed using the 19 data-subset GWAS analyses as input data (Duncan, Ratanatharathorn, et al., 2018). The combined-sex meta-analysis provided the data for the current study.

In general, the GWAS study from which the larger, initially proposed dataset were to be drawn used the same methodological steps (though it had 60 contributing studies rather than 11)

(Nievergelt et al., 2019). At the end of each section below, any differences in design are briefly stated. Genetic Risk Factors for PTSD….. 49

3.1d PGC Ancestry Assignment Assignment into genetic ancestry groups, termed ancestry assignment, was performed for each individual based on genetically inferred ancestry (Duncan, Ratanatharathorn, et al., 2018).

Ancestry of each participant was inferred based on analyses performed within each of the 11 contributing studies by comparing raw data within each study to data from a large reference panel using SNPweights software (Chen et al., 2013; Duncan, Ratanatharathorn, et al., 2018).

SNPweights software utilizes external reference panels derived from HapMap, which is a publicly available database containing millions of directly sequenced genetic variants from diverse populations (Chen et al., 2013; I. H. Consortium, 2003). Within SNPweights, Principal

Components Analysis is performed on HapMap data to determine SNP weights, which are an index of the relative predictive capacity of each SNP in terms of geographic ancestry (Chen et al., 2013). The SNP weights derived from the reference panel were separately applied to each of the 11 contributing studies in order to estimate the percentage ancestry of each individual in each study (Duncan, Ratanatharathorn, et al., 2018). The ancestry panels used were derived from

HapMap phase 3 (Pemberton, Wang, Li, & Rosenberg, 2010). Included ancestry panels were titled “African Ancestry (YRI)”, “European Ancestry (CEU)”, and “Asian (ASI),” and “Native

American (NAT)”(Chen et al., 2013). The SNP weights were applied to a Principal Components

Analysis of the entire Psychiatric Genomic Consortium study sample and cutoff scores for ancestry assignment were determined based on visual inspection of principal components plots

(Duncan, Ratanatharathorn, et al., 2018).

Of the 11 total studies, 10 were conducted in the United States; participants from each of those studies were grouped into geographic ancestry groups titled “European American,”

“African American,” and “Latino American” based on cutoffs determined from examining the Genetic Risk Factors for PTSD….. 50 principal components plots (Duncan, Ratanatharathorn, et al., 2018). The cutoff for “European

American” was ancestry estimation ⩾90% European (Duncan, Ratanatharathorn, et al., 2018).

The cutoffs for the “African American” group were twofold. The first cutoff was ⩾90% total of ancestry estimation from the combined scores of African and European; the second was <3% total combined from Asian plus Native American (Duncan, Ratanatharathorn, et al., 2018). The geographic ancestry group titled “Latino American” was defined based on a threefold cutoff of

⩾85% combined European plus Native American, <10% African, and also <3% Asian (Duncan,

Ratanatharathorn, et al., 2018). Each United States data subset was comprised of participants who were placed in the same geographic ancestry group using cutoff scores and HapMap3 data and whose data was collected during the same initial study (Duncan, Ratanatharathorn, et al.,

2018). Due to demographic differences within the original studies, some geographic ancestry groups were larger than others. Overall, there were a total of 7 “African American” data subsets,

9 “European American” data subsets, and 1 “Latino American” data subset (Duncan,

Ratanatharathorn, et al., 2018).

In addition to the 10 studies from United States samples, one additional study was conducted in South Africa (Duncan, Ratanatharathorn, et al., 2018). Due to a lack of any available robust external reference panels for South African populations, the South African study participant data was based entirely on principal components analysis and visual identification of two distinct groups (Duncan, Ratanatharathorn, et al., 2018). The participants whose data were collected from a single South African study were split into two separate subsets (Duncan,

Ratanatharathorn, et al., 2018). Overall, ancestry assignment procedures led to the creation of 19 total data subsets comprised of data from the 11 initial studies (Duncan, Ratanatharathorn, et al.,

2018). Genetic Risk Factors for PTSD….. 51

The current work uses summary statistics from a meta-analysis of the results of the nine total GWAS analyses performed on “European American” data subsets (one per contributing study) (Duncan, Ratanatharathorn, et al., 2018). Of note, the summary statistics for the prior planned analysis used identical procedures with respect to ancestry assignment and had an identical cutoff score for the sample titled “European and European Americans” as was used in this study (Nievergelt et al., 2019). Though the cutoff score was the same, the demographics of that group were more diverse due to the use of studies conducted in both the United States and internationally (Nievergelt et al., 2019).

3.1e PGC Data Reduction After ancestry assignment, steps two through four (as defined in section 3.1c) can be grouped together as data reduction procedures. Data reduction comprised of quality control procedures on each of the 19 data subsets, SNP imputation on each of the 19 data sub-sets, and relatedness testing.

Quality control procedures consisted of removing SNPs and individuals from analysis based on possible systemic errors and were performed in second-generation PLINK (Chang et al., 2015; Duncan, Ratanatharathorn, et al., 2018). In each of the 19 data sub-sets, SNPs were removed from a particular sub-set (but not the others) if they were missing in > 5% of individuals or if they failed Hardy-Weinberg equilibrium, which is a test to see if a fundamental assumption of GWAS called the Hardy-Weinberg principle is upheld within a given dataset

(Duncan, Ratanatharathorn, et al., 2018; Wittke-Thompson, Pluzhnikov, & Cox, 2005). The

Hardy-Weinberg principle states that allele frequencies should not vary between different generations of a given population unless an evolutionary force causes a drastic change and non Genetic Risk Factors for PTSD….. 52

(Wittke-Thompson et al., 2005). Hardy-Weinberg equilibrium was tested within each data subset at a level of P<1 x 10-6 in controls and at P<1 x 10-10 in cases and any individual SNP which was statistically significant at that level (indicating a between-generation difference) was removed from that data subset (Duncan, Ratanatharathorn, et al., 2018). After SNPs were removed based on missingness and Hardy-Weinberg equilibrium, individual participants were removed if they had a SNP missingness of > 2%, heterozygosity >|20|, or if they failed a genetic sex check

(Duncan, Ratanatharathorn, et al., 2018). Finally, after removing individuals, SNPs were again examined and were removed if missingness was >2% or if the difference in missingness between controls and cases was > 2% (Duncan, Ratanatharathorn, et al., 2018). No data was provided regarding the total number of SNPs or participants omitted (Duncan, Ratanatharathorn, et al.,

2018). Finally, principal components-based outlier removal was performed in order to ensure case-control matching within and across data subsets. Principal Components Analysis was performed using FastPCA on high-quality SNPs (i.e. those that were not removed due to quality control, had a minor allele frequency above 5%, and were not likely to be misidentified due to nucleotide properties or linkage disequilibrium) in each of the 19 sub-sets and outliers were removed based on visual examination of scatterplots until cases and controls appears to be matched with respect to principal components (Duncan, Ratanatharathorn, et al., 2018). Finally, the principal components analysis-based outlier removal procedures were repeated across all data sub-sets (Duncan, Ratanatharathorn, et al., 2018).

After FastPCA, SNP imputation was performed within each of the 19 data sub-sets using

PGC Pipeline (Duncan, Ratanatharathorn, et al., 2018; Ripke et al., 2014). Imputation relies upon the existence of haplotypes, which are groups of SNPs within the same which are positioned near one another within a particular segment of the chromosome and are inherited Genetic Risk Factors for PTSD….. 53 together as a group (Gabriel et al., 2002). Using Hidden Markov Modeling, PGC Pipeline analyzes raw genotype data in order to estimate haplotypes within the study sample (Ripke et al.,

2014). Then, using an external reference panel of haplotypes maps provided by phase one of the

1000 genomes project, PGC Pipeline imputes missing SNPs within haplotypes identified using raw data (G. P. Consortium, 2010; Ripke et al., 2014; Siva, 2008).

After imputation, relatedness testing was performed on combined samples within geographic ancestry groups (Duncan, Ratanatharathorn, et al., 2018). For each pair of related individuals (p̂ > 20%), one of the pair was dropped; whenever possible, cases were retained and controls were dropped (Duncan, Ratanatharathorn, et al., 2018).

Of note, the more recent Psychiatric Genomics Consortium study which was initially proposed as input data used the exact same procedures as described in this section, with the lone exception being that they used the more recent phase 3 reference data from the 1000 genomes project for imputation (G. P. Consortium, 2015; Nievergelt et al., 2019).

3.1f PGC Analyses The analytic design of the study consisted of GWAS analyses conducted within data subsets followed by stratified meta-analysis (Duncan, Ratanatharathorn, et al., 2018). First, sex- stratified and combined-sex GWAS analysis was performed in each of the 19 data subsets using

PLINK; three GWAS were performed within each subset (females only, males only, combined- sex) for a total of 57 GWAS (Chang et al., 2015; Duncan, Ratanatharathorn, et al., 2018). GWAS analysis was performed using additive linear modeling with the first 10 principal components

(derived from FastPCA as described in the previous section) as covariates (Duncan,

Ratanatharathorn, et al., 2018). The test statistic was the Cochran-Armitage trend test, which is a Genetic Risk Factors for PTSD….. 54 modified Pearson chi-squared test that accounts for expected ordering within categorical variables is commonly used in GWAS to account for ordering within SNPs (homozygous no minor allele, heterozygous, homozygous two minor alleles) (Duncan, Ratanatharathorn, et al.,

2018; C. R. Mehta, Patel, & Senchaudhuri, 1998). Results from the subsample GWAS tests were not reported but rather were used for subsequent stratified meta-analyses (Duncan,

Ratanatharathorn, et al., 2018).

Individual GWAS results were used as input for meta-analysis, which was conducted using METAL with fixed effects and inverse variance weighting (Duncan, Ratanatharathorn, et al., 2018; Willer, Li, & Abecasis, 2010). Meta-analysis was performed within the two geographic ancestry groups that were large enough for adequate power (“African Americans” and “European

Americans”) as well as for a “transethnic” group which included all 19 data subsets (Duncan,

Ratanatharathorn, et al., 2018). Additionally, meta-analysis was performed on each of the sex- stratified groups from each of those same three geographic ancestry groups (Duncan,

Ratanatharathorn, et al., 2018). For the purposes of this study, results from the meta-analysis of the combined sex “European Americans” were used as summary statistics.

The study from which the originally planned summary statistics would have been drawn used almost all of the same procedures (Duncan, Ratanatharathorn, et al., 2018; Nievergelt et al.,

2019). The only exception to this was that GWAS on data subsets drawn from family and twin studies were analyzed separately in that study using appropriate analytical methods whereas the current work does not include any data obtained from twin or family studies (Duncan,

Ratanatharathorn, et al., 2018; Nievergelt et al., 2019).

3.2 Data Acquisition Genetic Risk Factors for PTSD….. 55

Psychiatric Genomics Consortium summary statistics do not include any individual data.

As such, they are available for anyone to download for scholarly purposes (Duncan,

Ratanatharathorn, et al., 2018). Data for this study were acquired through the Psychiatric

Genomics Consortium website at the following link https://www.med.unc.edu/Psychiatric

Genomics Consortium/results-and-downloads/.

Initially, data were acquired for analysis of the publicly available “Freeze 2 EUR

Overall” data, which included summary statistics from a “European and European American” meta-analysis which yielded genome-wide significant SNPs and had a very large sample size

(23,212 cases; 151,447 controls) (Nievergelt et al., 2019). However, due to the COVID-19 pandemic and the associated shelter in place orders, the investigator of this study was unable to access a computer with sufficient RAM to process a file of that size. Thus, the publicly available

“European American” data, which included summary statistics from a methodologically similar but smaller meta-analysis (2,424 cases; 7,113 controls) were acquired for the current work

(Duncan, Ratanatharathorn, et al., 2018). Of note, the feasibility of analyzing the “African

American” and “Transethnic” summary statistics was also explored but those datasets each contained higher SNP counts causing files to be larger than currently available computing power would allow (Duncan, Ratanatharathorn, et al., 2018).

3.3 Data Reduction Data reduction was performed in several steps. First, the European American meta- analysis summary statistics data were downloaded as a .txt file. The original .txt file contained

11,991,833 rows and six columns per row. Each row in the original file represented a single SNP and the rows were as follows: rs number, Chr number, odds ratio, p-value, minor allele, minor Genetic Risk Factors for PTSD….. 56 allele frequency (Duncan, Ratanatharathorn, et al., 2018). Next, in order to be reformatted into the appropriate two-column format, the original summary statistics file was transferred into Java and reformatted into a .txt file with 11,991,833 rows and two columns per row (only the first and fourth column from the original file were kept) (Yoon et al., 2018a). Aside from the deletion of columns, no other changes were made. In the final file, each row represents a single SNP, with the rs number located in the left column and the p-value of that SNP’s correlation with PTSD in the right column (Yoon, 2018).

Second, a manually curated gene-set file was created in order to specify gene-sets of interest. Gene-set files are used in GSA-SNP2 to map genes to gene-sets (Yoon et al., 2018a).

The gene-set file is a .txt file and was created using Notepad software (Keely Jr, Young, &

Palay, 2002). The gene-set files are tab delimited and include in each row the name of the gene- set followed by the name of each gene in the gene-set. Within the gene-set file was all gene-sets created for this study. Gene-sets were manually created based on the previously stated goals of exploring genetic risk aggregation into genes which encode for (1) groups of receptors that are activated by a common neurotransmitter and (2) specific receptor subtypes and their intracellular signaling proteins, two types of gene-sets were created.

“Neurotransmitter receptor group” gene-sets were created in order to explore genetic risk aggregation into genes which encode for groups of receptors that are activated by a common neurotransmitter (or group of neurotransmitters such as opioids). As such, they are comprised of all protein encoding genes which encode receptors of the neurotransmitter of interest.

Neurotransmitter receptor gene-sets were created for each of the following neurotransmitters: glutamate, GABA, acetylcholine, dopamine, serotonin, adrenergic, NPY, opiod, cannabinoid.

“Neurotransmitter receptor group” gene-sets were created via a review of the literature. First, Genetic Risk Factors for PTSD….. 57 reviews and/or studies with extensive discussion of genes encoding for each of these neurotransmitters were identified and an initial list of genes was generated (Brown et al., 2019;

Ciccarelli et al., 2017; Doll et al., 2016; Johnston et al., 2016; E. M. Jorgensen, 2005; Moreira et al., 2009; Reichmann & Holzer, 2016; Żmudzka et al., 2018). Second, the Genecards dataset was used to confirm the initial list (https://www.genecards.org/, 2020).

“Receptor subtype” gene-sets were created in order to explore genetic risk aggregation into genes which encode for specific receptor subtypes and their intracellular signaling proteins.

As such, they are comprised of genes which encode the receptor subtype of interest in addition to genes which encode the intracellular signaling mechanisms activated by the receptor subtype of interest. “Receptor subtype” gene-sets include one gene-set for each known subtype of glutamate receptors, dopamine receptors, GABA receptors, serotonin receptors, and Acetylcholine receptors. The genes included in each gene-set were derived from synaptic signaling pathways found on the publicly available Kyoto Encyclopedia of Genes and Genomes (KEGG) database

(Aoki & Kanehisa, 2005). The KEGG database is developed by Kanehisa Laboratories for the purpose of providing multilevel maps of biological systems which integrate information from genome sequencing with higher-level biological systems and is considered to be a “gold standard” resource for gene-set analysis (Aoki & Kanehisa, 2005; Mooney & Wilmot, 2015).

Receptor subtypes for analysis were chosen from among the “neurotransmitter receptor group” neurotransmitters based on whether or not a relevant synaptic pathway within KEGG was available. In order to create gene-sets which capture genetic variants that may alter the signaling of specific receptor subtypes, each of the five available KEGG synaptic signaling pathways were manually subdivided based on receptor subtypes. The relevant molecules for each pathway and the corresponding genes were already specified in KEGG (Aoki & Kanehisa, 2005). While the Genetic Risk Factors for PTSD….. 58 original KEGG pathways included all genes relevant to the synaptic signaling of the neurotransmitter, each “receptor subtype” gene-set used in this analysis contained only the genes specifically involved in encoding the receptor subtype of interest and its intracellular signaling proteins.

3.4 Data Analysis Data analysis was performed using GSA-SNP2 software (Yoon et al., 2018a). GSA-

SNP2 is a graphical user interface software designed specifically for efficient GSA-SNP2 analysis (Yoon et al., 2018a). In order to analyze data in GSA-SNP2, five data files are needed

(Yoon, 2018). In order to map SNPs to genes, a SNP-gene mapping file and SNP position file are both required. In order to calculate gene-based p-values, an input file is needed (Yoon, 2018). In order to map genes to gene-sets, a gene-set file is needed (Yoon, 2018). In order to correct for

LD during the calculation of gene-set values, an adjacent gene correlation file is needed (Yoon,

2018).

The first step in using GSA-SNP2 is to specify an input file (Yoon, 2018). In this case, the input file was a .txt file comprised of Psychiatric Genomics Consortium PTSD summary statistics (edited in Java as previously described) (Duncan, Ratanatharathorn, et al., 2018). The next step is to specify the gene padding, which influences the mapping of SNPs to genes (Yoon,

2018). Gene padding refers to the amount of space past the from the 5′ and 3′-end of a gene to be included within the gene locus (Yoon et al., 2018a). SNPs were mapped to genes based on loci.

Padding sizes of 5kb, 15kb, 20 kb, and 50 kb have all been recommended in the literature and no consensus has been reached (Yoon et al., 2018a). Based on recommendations from the creators of GSA-SNP2 to choose either 10 kb or 20 kb as well as research which suggests that more than Genetic Risk Factors for PTSD….. 59

90% of SNPs affecting expression are located within 15 kb from the 5′ and 3′-end of the gene, padding was set at 20 kb (Syvänen, 2001; Yoon, 2018). The use of 20 kb padding ensures that most SNPs directly involved in determining expression of a given gene are mapped to that gene while also limiting the inclusion of functionally unrelated SNPs within the gene locus (Yoon et al., 2018a). Next, a reference SNP-gene mapping file was specified (Yoon, 2018). In keeping with recommendations from the literature and in order to account for the use of 20 kb padding size, a reference genome titled “db19_20k” was from downloaded from the GSA-SNP2 website and used in analysis; the file contained gene mapping data from the 1000 genomes project in a format that is compatible with GSA-SNP2 (Yoon, 2018). In order to ensure appropriate removal of adjacent genes, the next step was to specify an adjacent gene correlation file which identified adjacent genes within pathways of interest and alternatively removed them only if they were found to have high positive genotype correlations (>0.5) (Yoon, 2018). Based on recommendations from the literature, the “EUR_Adjacent_correlation” file from the 1000 genomes project was used as reference data for adjacent gene removal (Yoon et al., 2018a). On average, this correction removes <1% of all SNPs (Yoon et al., 2018a). Then, for each analysis, the manually curated gene-set file (described earlier) was specified in order to map genes to gene-sets (Yoon, 2018). Finally, based on recommendations from the literature, a filter was set to remove pathways with a very low or a very high number of genes (Yoon, 2018). Based on recommendations from the literature, analysis was first run with the minimum set at 10 and the maximum was set at 200 (Yoon et al., 2018a). Next, in order to examine the numerous very small gene-sets included in this study, alternative analysis was run with the minimum set at seven, five, and one (Yoon et al., 2018a). Genetic Risk Factors for PTSD….. 60

Once all files were specified, GSA-SNP2 was used to test for gene-set enrichment. Like all gene-set analysis methods, GSA-SNP2 involves a two-tier structure (Yoon et al., 2018a). First, the SNP associations were used to calculate gene associations (Yoon et al., 2018a; Yoon et al.,

2018b). Second, gene associations are used to perform a set of bivariate hypothesis tests (Yoon et al., 2018a). Though all gene-set analysis fundamentally uses this structure, the specific calculations involved vary (De Leeuw et al., 2016). Though a review of all of the ways in which

GSA-SNP2 varies from other methods is beyond the scope of this work, a brief review of the calculations used is informative.

First, GSA-SNP2 derives gene scores from GWAS summary statistic p-values (Yoon et al.,

2018a). After assigned SNPs to genes based on steps described above, SNA-SNP2 uses the p- values of SNPs within a given gene locus to estimate the p-value of that gene (Yoon et al.,

2018a). One limitation of the use of summary statistics is that it limits the ability to use certain advanced methods for estimating gene values (De Leeuw et al., 2016; Yoon et al., 2018a). For example, principal components analysis regression based on individual data is used in MAGMA to calculate gene scores (De Leeuw et al., 2016; Yoon et al., 2018a). In gene-set analysis, genes are often represented by the SNP with the best p-value (De Leeuw et al., 2016; Yoon et al.,

2018a). However, this method may introduce bias, especially bias due to overestimation of genes with more SNPs (Yoon et al., 2018a). In order to reduce bias, GSA-SNP2, gene values estimates are calculated using an adjusted gene score for the SNP with the best p-value (Yoon et al.,

2018a). GSA-SNP2 controls for gene length (number of SNPs per gene) by fitting a monotone cubic spline trend (Yoon et al., 2018a). In other words, GSA-SNP2 quantifies the likelihood of a gene having a significant p-value as a function of number of SNPs by fitting dual monotonic curves, which is shown to substantially lower type-I error rate and increase power (Yoon et al., Genetic Risk Factors for PTSD….. 61

2018a). The dual curves are averaged, yielding an upward trending curve which is incorporated into the calculation of each gene’s p-value (Yoon et al., 2018a). The adjusted gene score is calculated by subtracting the estimated gene score (based on number of SNPs) on the upward trending curve from the log of the best p-value among SNPs from the gene (Yoon et al., 2018a).

For a given gene, the log score of the predicted best p value (y axis) for a gene with a given number of SNPs is subtracted from the log score of the actual highest p-value for that gene

(Yoon et al., 2018a). Adjusted gene scores for a gene are calculated using the formula

where is the best p-value in the gene locus and is the predicted gene score derived from the trend curve based on the number of SNPs within the gene locus (Yoon et al., 2018a).

After creating adjusted gene scores, GSA-SNP2 computes a competitive bivariate hypothesis test for each specified gene-set (Yoon et al., 2018a). Each hypothesis test compares the correlation of the genes within the gene-set to all of the genes from the GWAS that are not included in the gene-set (Yoon et al., 2018a). The null hypothesis is that the genes in the gene-set are no more strongly associated with PTSD than are a random group of genes with the same number of genes(Yoon et al., 2018a). In GSA-SNP2, the bivariate hypothesis test is mean-based and is measured using a modified one-way Z-test with a Benjamini-Hochberg adjustment for multiple testing; the one-tailed z-test is shown to better control type-II error (Yoon et al., 2018a).

A general problem seen in many variations of gene-set analysis is that models may be biased towards larger gene-sets (De Leeuw et al., 2016; Mooney & Wilmot, 2015). In GSA-SNP2, the

Z-test is modified in order to account for gene-set size. The unadjusted base formula is given as

with PJ-bar representing the mean of all adjusted gene scores in the gene-set PJ, m and σ representing the mean and standard deviation of all adjusted gene scores, and NJ Genetic Risk Factors for PTSD….. 62 representing the number of genes in PJ. In order to better account for the impact of gene-set size, the above formula is adjusted by replacing σ with , where is the total number of genes analyzed. The modified Z-statistic is shown to have slightly increased statistical power(Yoon et al., 2018a).

GSA-SNP2 efficiently performs all of the above calculations and provides a gene-set name, size of gene-set, count of detected genes, adjusted z-score, adjusted p-value, q-value and list of member genes (Yoon et al., 2018a). The p-value identifies nominally significant gene-sets that are not corrected for multiple comparisons, while the q-value is used as the test of gene-set enrichment and is a measure of the false discovery rate that is adjusted for multiple comparisons

(Yoon et al., 2018a). Genes with a q-value <.05 are interpreted as enriched (i.e. are statistically significant after controlling for multiple comparisons) (Yoon et al., 2018a). Q value uses the

Benjamini-Hotchberg correction for multiple testing (Yoon et al., 2018a).

In order to interpret results, the following steps were taken. First, manual calculations were performed to note which gene-sets were filtered out. GSA-SNP2 did not provide error messages during any analyses run during this work, but gene-set filtering was checked by counting the gene-sets in the output as well as the gene-sets in the gene-set file and comparing.

Gene-sets may be filtered out due to having fewer genes than the specified gene-set minimum filter (Yoon, 2018). Second, the gene-set size was compared to the detected genes count. This allows one to see if any genes are missing from within gene-sets; missing genes occur when no

SNPs assigned to that gene are present in the input data (Yoon, 2018). Missing SNPs are common in GWAS studies, which can only impute SNPs for which adequate haplotype-relevant data is available in the raw data (Bergen & Petryshen, 2012). Third, the adjusted z-score provides an estimate of each gene-set’s degree of correlation with the phenotype of interest (relative to all Genetic Risk Factors for PTSD….. 63 other genes) (Yoon, 2018). It is considered an adjusted z-score because of the previously mentioned adjustment for number of genes (Yoon et al., 2018a). Positive z-scores indicate higher than normal correlation while negative z-scores indicate lower than normal correlation (Yoon et al., 2018a). Fourth, the p-value provides a measure of nominal significance (i.e. statistical significance at the p < .05 level prior to correction for multiple comparisons) (Yoon et al.,

2018a). Fifth, and most importantly, the Q-value provides a measure of statistical significance based on the false discover rate, which includes a correction for multiple comparisons (aka gene- set enrichment) (Yoon et al., 2018a). The Q-value is the most important statistic for interpreting results because a gene-set is only said to be enriched if it is below the .05 Q-value threshold

(Yoon, 2018; Yoon et al., 2018a). Sixth, for each gene-set, gene scores are provided for all genes in the gene-set (Yoon, 2018). The gene score provides an estimate of each gene’s relative degree of correlation with the phenotype of interest (relative to the expected degree of correlation for a randomly chosen gene with the same amount of SNPs) (Yoon et al., 2018a).

3.5 Post-Hoc Analysis One recommended use of preliminary analysis is to refine study methodology (P. G.

Smith, Morrow, & Ross, 2015). In addition to the planned analyses described above, a post-hoc analysis was run using an alternate pathway file. The post-hoc pathway file included two alterations in the organization and composition of gene-sets which were designed to explore and contextualize observations from the results of planned analysis (results are described in detail in section four).

First, in order to explore and contextualize missing gene-sets and provide an ability to indirectly examine research questions related to those gene-sets, two composite missing data Genetic Risk Factors for PTSD….. 64 gene-sets were created, titled “missing receptors” and “missing intracellular”. “Missing receptors” was a neurotransmitter receptor composite group comprised of all neurotransmitter receptor gene-sets which were excluded from planned analysis due to unanticipated limitations in the ability of GSA-SNP2 to handle very small gene-sets. “Missing intracellular” was an intracellular signaling composite group of all intracellular signaling molecules that were excluded from the planned analysis, also due to difficulties with very small gene-sets. Though these groups were not designed to be interpreted with regards to any one theory-driven unitary function, they were intended to allow some insights into the possible relevance of missing very small gene-set data to the fundamental research questions. Second, based on observations covered in the results section as well as theory and parsimony, intracellular signaling pathway gene-sets were created. These differed from the initially tested “receptor subtype” gene-sets in two important ways. First, they did not include any receptor coding genes but rather were only focused on signaling molecules. Second, they were organized based on pathways, rather than receptors.

One single analysis was run which included the newly created “missing receptors” gene- set, the newly created “intracellular signaling” gene-sets, and each of the “neurotransmitter receptor” gene-sets from the prior analysis. All methods for analysis followed those of the main analysis. Additionally, analyses were attempted using both the “African American” and “Trans- ethnic” geographic ancestry groups in order to increase generalizability and to compare results.

However, those analyses were not able to be computed using GSA-SNP2 with currently available computing power, due to having file sizes which were larger than the “European

American” dataset.

Genetic Risk Factors for PTSD….. 65

4.0 Results 4.1 Planned Analysis Manual comparison of the number of gene-sets included in analysis to the number of gene-sets included in the pathway file that GSA-SNP2 filtered out all gene-sets with fewer than

10 genes. Unexpectedly, this occurred even when setting the minimum gene-set size below 10

(via the graphical user interface), including analysis run with the minimum set at seven, five, and one. In order to further examine a possible cause for the error, a model was run that only included gene-sets of fewer than 10 genes. When running those models, GSA-SNP2 indicated that it loaded and analyzed the data and an output file was created. However, the output file did not include any results (only column headers were included). This appears to have been a software constraint related to GSA-SNP2. Of note, GSA-SNP2 is not a commercial product, is not widely used in research, and was primarily developed in order to examine large gene-sets

(Yoon et al., 2018a). A list of all of the original 35 gene-sets included in the pathway file is shown in figure 1.

Overall, 35 of the original 51 gene-sets were analyzed while 16 were excluded. Of the nine “neurotransmitter receptor group” gene-sets, five were excluded and four were included.

The five excluded “neurotransmitter receptor group” gene-sets were dopamine (five genes), adrenergic (nine genes), cannabinoid (two genes), opioid (five genes), and neuropeptide-Y (four genes). The four included “neurotransmitter receptor group” gene-sets were GABA (21 genes), glutamate (24 genes), acetylcholine (14 genes), and serotonin (17 genes). Of the 42 “receptor subtype” gene-sets, 11 were excluded and 31 were included. Of the 11 excluded

“neurotransmitter receptor group” gene-sets, five belonged to the serotonin 3 family of receptors

(5-HT3A, 5-HT3B, 5-HT3C, 5-HT3D, 5-HT3E), three were muscarinic acetylcholine receptors (M1, Genetic Risk Factors for PTSD….. 66

M3, M5), two were ionotropic glutamatergic receptors (KA, AMPA), and the other one was a

GABA receptor (GABAC). The 31 included “receptor subtype” gene-sets are included in table 1.

Next, in order to examine the extent of missing genes within the included gene-sets, gene size count was compared to detected gene count. Of the 35 gene-sets included in analysis, just 13 included all of the genes which were included within their pathway file. Of the 23 gene-sets which were missing at least one gene, 13 were missing one gene, eight were missing two, and two were missing three. Detailed information regarding gene-set size and detected gene count for specific analyzed gene-sets is included in table 1. Gene-sets with missing genes were investigated manually in order to determine the cause of missing genes. Based on manual examination, missing genes were attributable to a lack of relevant SNPs included within the

GWAS summary statistics. For example, the gene HTR1C, which is the protein encoding gene for the 5-HT1C receptor was one of two missing genes in the “receptor subtype” pathway for 5-

HT1C. An inspection of the “db19_20k” SNP-gene mapping file (Yoon, 2018) showed that all known variants for this gene were correctly mapped to the gene; the missingness was not a result of file misspecification. For HTR1C, there are six known common variants (rs3813929, rs518147, rs6318, rs17260600, rs6644093, and rs1801412) (Baou et al., 2016). An inspection of the summary statistics from the Psychiatric Genomics Consortium GWAS that were used as input data in this study showed that none of those SNPs were included in the analysis (Duncan,

Ratanatharathorn, et al., 2018). Similarly, other missing genes were found to be lacking SNPs in the summary statistics (Duncan, Ratanatharathorn, et al., 2018). Of note, genes for which some but not all SNPs were missing were included in the gene-set analysis.

After ascertaining an understanding of missing data, z-scores were examined in order to provide an overview of direction of gene-set scores. Specific z-scores are provided in table 1 and Genetic Risk Factors for PTSD….. 67 an overview of positive and negative z-scores is provided here. In the context of gene-set analysis with a competitive hypothesis test, the z-score for each gene-set represents an estimate of the comparative strength of correlation of that gene-set with the phenotype of interest relative to the average strength of correlation for all other genes in the gene-set (De Leeuw et al., 2016).

Thus, in this study, negative z-scores suggest that the average of genes within the gene-set is estimated to be less correlated with PTSD than the average gene; positive z-scores suggest the opposite. An examination of z-scores shows that the average of genes within 17 total gene-sets were estimated to be less strongly correlated with PSD than the average of all other genes. Due to the use of a right-sided significance test, the statistical significance of negative z-scores cannot be examined using GSA-SNP2 (Yoon et al., 2018a). In contrast, for the 18 gene-sets with positive z-scores, p-values and Q-value were examined in order to determine nominal significance and gene-set enrichment, respectively.

Due to the importance of controlling for multiple comparisons, the p-value does not indicate gene-set enrichment (De Leeuw et al., 2016). However, the p-values in these data do provide information about the nominal significance for positive z-scores (De Leeuw et al., 2016).

The p-value indicates that seven of the 18 gene-sets with positive z-scores were nominally significant at the p ≤. 05 level. In order of lowest p-value to highest, the nominally significant gene-sets were 5-HT2B receptor subtype, 5-HT2C receptor subtype, mGluR5 receptor subtype, mGluR1 receptor subtype, 5-HT2A receptor subtype, GABA neurotransmitter receptors, and glutamate neurotransmitter receptors. Of note, serotonin and glutamate related gene-sets showed great variation; the most negative z-score belonged to the “neurotransmitter receptors group” gene-set for serotonin and the 13 most negative z-scores were all “receptor subtype” gene-sets of either serotonin or glutamate receptor subtypes. In contrast, zero of the nine analyzed gene-sets Genetic Risk Factors for PTSD….. 68 from acetylcholine and dopamine combined were in either the top eight or the bottom 13 z- scores. None of the nominally significant gene-sets were approached the cutoff for gene-set enrichment based on the Q-value. P-values and Q-values for all analyzed gene-sets are shown in table 1.

Finally, gene scores were examined for the genes within the analyzed gene-sets. All gene scores for each gene-set are listed in supplementary figure s2, in order from highest to lowest gene score. An overview of relevant patterns is also provided here. Of note, each of the top five gene-sets share the same top three most significant genes (PLA2G4F gene score = 2.10831,

PLA2G4D gene score = 2.00666, and PLA2G4E gene score = 1.94461). Each of those three genes encode intracellular signaling belonging to the A2 group (Kudo &

Murakami, 2002). No other gene-sets tested included any genes. In addition to the commonality of the top three genes across all of the top five gene-sets, the top five gene-sets were all “receptor subtype” gene-sets and each represented a receptor subtype which activated one of two intracellular signaling pathways. Due to the intricate nature of these shared intracellular pathways, the majority of genes included in the gene-sets of receptors which activate those pathways were derived from those pathways and therefore overlapped. For example, three of the top five were part of the same serotonin 2 family (5-HT2B, 5-HT2C, 5-HT2A) and activate a common postsynaptic intracellular signaling pathway as defined by KEGG (Aoki

& Kanehisa, 2005). The gene-set for 5-HT2B only activated that one major intracellular pathway in KEGG and therefore the only gene in that gene-set that did not appear in the other serotonin-2 gene-sets was the receptor encoding gene (HTR2B), which was estimated to have a positive gene score and was tenth highest among the 24 genes detected in that gene-set (0.244506). The 5-

HT2C gene-set has no unique genes; the HTR2C gene which encodes for 5-HT2C receptors was Genetic Risk Factors for PTSD….. 69

missing due to the GWAS not having any corresponding p-values. The gene-set for 5-HT2A had just two unique genes (HTR2A gene score = -0.19768 and astrocyte signaling enzyme encoding gene TRPC1 gene score = -0.474553). Similarly, the mGluR5 and mGluR1 gene-sets differed only with respect to their receptor coding genes (GRM5 gene score = 1.21134 and GRM1 gene score = .304609). Further, many of the receptor subtype gene-sets that did not reach nominal significance were highly overlapping based on a shared major intracellular signaling pathway; those pathways also had similar rank within the analyzed gene-sets with respect to z-score and p- value. Based on the observation that the gene-set scores for “receptor subtype” gene-sets were primarily determined by intracellular signaling genes as well as the clear pharmacological importance of intracellular pathways (Réus et al., 2019), post-hoc analysis was conducted to further explore these pathways

4.2 Post-Hoc Analyses

Post-hoc analyses were performed in order to explore general characteristics of missing data and to test a more parsimonious organization of gene-sets. The theory-driven methodology for post-hoc gene-set creation is detailed in the methods section. In addition to the post-hoc gene- sets, the “neurotransmitter receptor groups” gene-sets that were able to be analyzed in the planned analysis were included in post-hoc analysis.

Based on manual examination of the gene-sets, the “missing intracellular” gene-set was removed due to insufficient genes. Further, intracellular pathway gene-sets with fewer than 10 genes were removed. The pathway file used for post-hoc analysis contained a total of 21 gene- sets. A list of post-hoc gene-sets is included in figure 2. Gene size, gene count, and z-score statistics are detailed in table 2 and are comparable to what was seen in the main analysis. Post- Genetic Risk Factors for PTSD….. 70 hoc z-score trends were in line with planned analysis z-score trends; of the 21 pathways, 11 had positive z-scores and 10 had negative z-scores. The 11 positive z-scores were examined for nominal significance and for gene-set enrichment.

In line with expectations, nominally significant gene-sets were highly concordant with the nominally significant gene-sets from the initial analysis. In order from lowest p-value to highest, the four post-hoc gene-sets that were nominally significant at the p ≤. 05 level were as follows: serotonin-2 family receptor postsynaptic intracellular signaling, glutamate mGluR1/mGluR5 postsynaptic intracellular signaling, GABA receptors group, and glutamate receptors group. Additionally, z-score ranks of the gene-sets that were not nominally significant were mostly in line with the planned analysis z-scores. For example, the middle ranked post-hoc gene-set represented a postsynaptic intracellular signaling pathway activated by serotonin-4 family, serotonin-6 family, and serotonin-7 family receptors. In the planned analysis, the z-score rankings of the “receptor subtype” gene-sets for the receptor that activate that pathway were all in the 14 to 17 range (out of 35). Similarly, the “receptor subtype” gene-sets that corresponded with the lower ranked “intracellular signaling” gene-sets generally clustered toward the bottom of the z-scores. P-values, Q-values, and z-scores for all post hoc gene-sets are shown in table 2.

The missing “missing receptors” gene-set had the lowest z-score and highest p-value of any gene-sets included. Examination of the gene scores output suggested that each of the individual gene-sets excluded from the planned analysis would have likely had negative z-scores.

The highest individual gene score for this gene-set was lower than the highest gene from any other gene-set (ADR1D gene score = 0.890306). Further, for each included neurotransmitter, detected genes with negative gene scores outnumbered detected genes with positive gene scores. Genetic Risk Factors for PTSD….. 71

Gene score data for the “missing receptors” gene-set is included in supplementary figure s2 along with the gene score data for the other post-hoc gene-sets.

Genetic Risk Factors for PTSD….. 72

5.0 Discussion Based on the current results, the author failed to support any of the six hypotheses tested.

However, due to technical constraints, the current study used a substantially smaller sample size than initially planned and some hypotheses were untestable. The lack of statistical significance based on the Q-score is in line with expectations in this reduced data set; GWAS studies which do not yield genome-wide significant SNPs are not expected to have sufficient power to yield gene-set enrichment (De Leeuw et al., 2016; Duncan, Ratanatharathorn, et al., 2018). Due to low statistical power, the results from this study should be seen as preliminary. Though results should be interpreted with caution, the current study can be viewed as a preliminary study and may inform next steps in this program of research.

5.1 Effect of Gene-set Size Minimum on Hypothesis Testability Due to limitations in the ability for GSA-SNP2 to conduct analysis in small gene-sets, some hypotheses are either unable to be tested or are only partially able to be tested using GSA-

SNP2. For example, the neuropeptide-Y gene-set includes four genes. Given the small number of genes in the gene-set, hypothesis two “enriched gene-sets will include neuropeptide-Y” is untestable using GSA-SNP2. This software constraint is not listed in the software manual nor in the publication explaining the methodology (Yoon, 2018; Yoon et al., 2018a) and therefore represents a useful observation for next steps. Future studies will have to either utilize different software or exclude these hypotheses.

In contrast, three planned hypotheses are fully testable using GSA-SNP2. Due to the fact that 25 genes were included in the “glutamate receptor subtypes” gene-set, hypothesis one

“enriched gene-sets will include glutamate” is testable in gsa-snp2. Further, the two primary Genetic Risk Factors for PTSD….. 73

“receptor subtype” hypotheses are both testable with the 10 gene limitation. Both dopamine D2 receptor and serotonin 5-HT1A receptor gene-sets include far more than ten genes, allowing hypothesis four “enriched gene-sets will include D2” and hypothesis five “enriched gene-sets will include serotonin-1A” to be tested. Future studies could use GSA-SNP2 to examine these hypotheses in larger samples.

The two exploratory hypotheses are partially testable using GSA-SNP2. Hypothesis three

“that this test will identify at least one enriched neurotransmitter receptor gene-set that was not included in hypothesis one or two” is testable for less than half of the initially planned data. Of the seven “neurotransmitter receptor” gene-sets that are included within hypothesis four, four are unable to be tested due to having fewer than 10 genes. Dopamine, adrenergic, opioid, and cannabinoid receptor group gene-sets are excluded by GSA-SNP2. The three receptor group gene-sets for hypothesis three that GSA-SNP2 can test are GABA, serotonin, and acetylcholine.

Hypothesis six, “it is hypothesized that this test will identify at least one enriched receptor subtype gene-set that was not included in hypothesis four or five” is testable in GSA-SNP2 for a majority of initially planned gene-sets, though a substantial minority cannot be examined in

GSA-SNP2. Eleven of the originally planned 42 subtypes for exploration in hypothesis six are unable to be included in GSA-SNP2 analysis. Of those 11, five belong to the serotonin 3 family of receptors, three are muscarinic acetylcholine receptor, two are glutamatergic ionotropic receptors, and the other one is GABAc. Overall, these observations inform the author’s evaluation of whether or not GSA-SNP2 is an appropriate software for future investigations.

5.2 Preliminary Results: Lessons Learned and Next Steps Genetic Risk Factors for PTSD….. 74

Due to the preliminary nature of this work, results are first discussed in relation to their bearings on next steps for the current study. One clear next step is to test the current hypotheses using summary statistics from larger studies as input data. The preliminary findings from this study highlight the importance of very high sample sizes for gene-set analysis. The lack of detected gene-set enrichment may be due to insufficient statistical power. Thus, while the results do not support any of the tested hypotheses, they do not disconfirm them either. Based on insights gained from this study, the use of a larger sample may not be the only adjustment that should be made in the next steps of analysis.

One important lesson learned in this work is that several of this study’s hypotheses may be untestable in GSA-SNP2, regardless of sample size. In gene-set analysis, the exclusion of gene-sets of less than ten is a common strategy due to concerns regarding statistical power (De

Leeuw et al., 2016). In spite of the statistical justifications for ignoring gene-sets of nine genes or fewer, very small gene-sets may contain critical information (Ryan Sun, Hui, Bader, Lin, &

Kraft, 2019). Recently, a simulation study tested the statistical power of several gene-set analysis methods when applied to small gene-sets and found the Generalized Berk-Jones statistic, which can be implemented using a package in R, to have the best performance (Ryan Sun et al., 2019).

Thus, in order to test hypothesis two and to test exploratory hypotheses three and six with respect to the gene-sets initially proposed, running analyses in R using the Generalized Berk-Jones statistic package would be preferable.

Another useful insight provided by this preliminary study is that hypotheses and gene- sets related to specific neurotransmitter receptor subtypes should be refined and should focus on intracellular signaling pathways. The planned analysis revealed that “receptor subtype” gene- sets, which were conceptualized in order to capture the combined impact of all protein-encoding Genetic Risk Factors for PTSD….. 75 genes relevant to synaptic signaling of specific receptor subtypes, clustered together based on shared intracellular mechanisms. The planned analysis was designed to reflect the common research approach of focusing on specific receptor subtypes, which is reflected in the high volume of candidate gene studies which focus on specific receptor subtypes (Duncan, Cooper, et al., 2018). “Receptor subtype” gene-sets included genes encoding for the receptor subtype of interest as well as genes encoding for each of the intracellular signaling proteins involved in each of the intracellular pathways activated by a given subtype. Though KEGG synaptic signaling pathways indicate that several receptors activate multiple intracellular signaling pathways, many involve one major pathway (i.e. one pathway accounted for a majority of the genes in the gene- set for most receptor subtypes) (Aoki & Kanehisa, 2005). In the initially planned analysis, z- scores and p-values clustered according to shared major pathways. Thus, refining “receptor subtype” gene-sets by creating gene-sets based on intracellular signaling pathways may reduce dimensionality and increase interpretability. Further, a focus on intracellular signaling aligns with the literature; there has been a recent increase in focus on intracellular signaling in relation to disease etiology and pharmacotherapy design (Komatsu et al., 2019; Réus et al., 2019;

Sniecikowska, Newman-Tancredi, & Kolaczkowski, 2019). Further, the use of “intracellular pathways” gene-sets, along with “neurotransmitter receptor group” gene-sets may increase theoretical parsimony and pharmacological applicability by providing a clear distinction between gene-sets examining receptor characteristics versus gene-sets examining intracellular pathways

(Berg et al., 1998). Based on these findings, next steps in this study should investigate use the refined “intracellular pathways” gene-sets rather than “receptor subtype” gene-sets.

In contrast with the “receptor subtype” gene-sets, preliminary results increase confidence that the “neurotransmitter receptor groups” gene-sets were appropriately defined for the purposes Genetic Risk Factors for PTSD….. 76 of this investigation. One difficult methodological decision was whether or not to include genes which encode proteins that impact expression in that gene-set (e.g. transcription factors, transfer

RNA, etc.). On one hand, a comprehensive examination of genes which impact the levels and characteristics of receptor proteins necessarily must include genes which encode for modifying proteins such as transcription factors (Albert, 2012). However, much like intracellular signaling proteins, modifying proteins are not distinct to any one specific receptor subtype

(Albert, 2012). Additionally, like intracellular signaling proteins, the number of gene expression modifying proteins which act upon any given receptor subtype (and therefore the number of genes needed to encode all of them) far exceed the number of genes which encode that receptor protein (Albert, 2012). Thus, had “neurotransmitter receptor group” gene-sets included the genes encoding for their expression-modification proteins, results for those gene-sets would have primarily reflected the impact of expression-modification proteins and would not have been particularly informative with respect to genetic variation in receptor protein encoding genes.

Based on these observations, next steps in this study should use the current “neurotransmitter receptor group” gene-sets. Overall, a gene-set analysis using the Generalized Berk-Jones R package to analyze the “intracellular signaling” and “neurotransmitter receptor group” gene-sets might be the ideal next step for this study.

5.3 Limitations The primary limitation is a large reduction in sample size and statistical power relative to planned analysis, which causes this work to be preliminary. The originally proposed dataset had

23,212 cases and 151,447 controls and was the only available PTSD GWAS which identified genome-wide significant SNPs (Nievergelt et al., 2019). Genome-wide significant SNPs are generally viewed as a sign that a gene-set analysis has adequate power, while a lack thereof Genetic Risk Factors for PTSD….. 77 indicates the opposite (De Leeuw et al., 2016). However, due to a nationwide state of emergency, the researcher was unable to access Virginia Tech computers with sufficient RAM for data processing. Thus, the dataset included in this work included 2,424 cases and 7,113 controls

(Duncan, Ratanatharathorn, et al., 2018). Though the study was likely underpowered to detect effects, analyses were run in a smaller sample, and preliminary results were presented. In addition to limitations in statistical power, the use of a smaller sample size led to the inclusion of fewer total SNPs (Duncan, Ratanatharathorn, et al., 2018; Nievergelt et al., 2019). While the decrease in number of SNPs led to a smaller file size, it also contributed to the relatively high number of missing gene scores in the gene-set analysis and may have also biased gene scores due to missing SNPs (De Leeuw et al., 2016).

Beyond the limited sample size, the summary statistics used for this analysis had several additional limitations. The GWAS from which summary statistics were attained entailed the meta-analysis of nine smaller studies (Duncan, Ratanatharathorn, et al., 2018). Contributing studies varied in terms of sample composition and methodology. For example, studies examined a wide array of populations, including military populations, nurses, and the general population

(Duncan, Ratanatharathorn, et al., 2018). As a result, samples varied with respect to trauma type

(Duncan, Ratanatharathorn, et al., 2018). Additionally, some studies included controls who were not exposed to trauma, while some did not (Duncan, Ratanatharathorn, et al., 2018). Due to the covariance of heritability with environment and the covariance of specific genetic effects with trauma type, this may have led to between-study differences in the degree to which SNPs were correlated with PTSD. Further, methodological differences in those nine studies led to discrepancies in both genotyping and phenotyping. For example, studies used various methods of measuring PTSD and trauma exposure, likely leading to between-study differences in underlying Genetic Risk Factors for PTSD….. 78 phenotypes among both cases and controls (e.g. severity threshold for case status, differences in accuracy, qualitative differences in sensitivity to specific presentations of symptoms) (Duncan,

Ratanatharathorn, et al., 2018). Additionally, the use of different sample sizes and different methodologies for sequencing SNPs led to inconsistencies in which SNPs were genotyped or able to be imputed in the various contributing studies (Duncan, Ratanatharathorn, et al., 2018).

Given that each of these differences varied by study, between-study differences in methodological biases most likely varied systematically. While the use of study-specific analysis followed by meta-analysis partially addressed those concerns, this approach is not guaranteed to fully account for methodological differences and can also decrease power and increase false- discovery rate (Faye et al., 2011). Of note, inconsistency between contributing studies is just as substantial of an issue (if not more so) in the initially planned data set, which included participants evaluated for PTSD using a wider variety of measures and using any one of DSM-

III, DSM-IV, or DSM-5 criteria for PTSD (Nievergelt et al., 2019). However, in spite of the considerable limitations introduced by these inconsistencies, large GWAS results of complex disorders have had relatively high replicability, suggesting that the benefits of increased statistical power likely outweigh the limitations introduced by inconsistent methodology of contributing studies (Bergen & Petryshen, 2012; Marigorta et al., 2018). Thus, while these limitations must be taken into consideration, they do not invalidate the usefulness of the GWAS approach.

In addition to sample characteristics, this study was not able to test interactive effects, which is a noteworthy limitation (Duncan, Ratanatharathorn, et al., 2018; Lencz & Malhotra,

2015). A major aim of gene-set analysis is to detect patterns with respect to the grouping of those main effects into biologically meaningful groups (De Leeuw et al., 2016). However, PTSD is a Genetic Risk Factors for PTSD….. 79 complex genetic disorder that is defined with respect to a reaction to environment (Duncan,

Cooper, et al., 2018). Further, it is highly possible that a high degree of risk for PTSD is driven by allosteric effects (i.e. gene x gene interactions) (Duncan, Cooper, et al., 2018). Though the

Psychiatric Genomics Consortium includes the examination of interactive effects among its stated goals, no currently available GWAS of PTSD includes interactions (C. M. Nievergelt et al., 2018). Thus, the inability to incorporate gene x gene interactions or gene x environment interactions is a limitation of the current work.

One final limitation of GWAS is that the focus on common SNPs does not account for the effects of rare variants (i.e. genetic mutations found in less than 1% of the population) and copy number variants (i.e. variation in the number of repeats in a segment of the genome)

(Manolio et al., 2009). Several studies have indicated that these types of variants may be critical in understanding disease risk (Manolio et al., 2009). Thus, while GWAS is an important aspect of defining risk loci, it should be noted that this technique does not incorporate all important forms of genetic variation.

In addition to limitations derived from the input data, the analytic approach of this study has several noteworthy limitations. First, though methods which use summary statistics as input data are considered valid and meaningful, simulation studies suggest that they have lower statistical power and higher type-I error rate than methods which use individual level data (De

Leeuw et al., 2016). Second, though GSA-SNP2 has many laudable properties, it also has several limitations. For one, the formula used to estimate gene-score is based on the gene-size adjusted best p-value from within the gene (Yoon et al., 2018a). However, underlying assumption that the best p-value from within the gene, adjusted for the expected best p-value of a gene with the same number of SNPs, represents the gene as a whole may not always be accurate (De Leeuw et al., Genetic Risk Factors for PTSD….. 80

2016). Additionally, hypothesis-driven gene-sets tend to be limited by available knowledge as well as by the statistical limitations of the reliance of aggregated correlations used in gene-set analysis (De Leeuw et al., 2016). In this particular study, the “neurotransmitter receptor group” gene-sets were designed to focus on protein encoding genes in order to provide parsimonious results. Though this choice allows clear interpretation regarding the function of the included genes, it does not account for the importance of genetic variation in genes which encode transcription factors and other regulators of gene expression (Albert, Le François, & Vahid-

Ansari, 2019). Further, though the consolidation and reclassification of “receptor subtype” gene- sets into “intracellular signaling” gene-sets may increase parsimony, the “intracellular signaling” gene-set definition still has noteworthy limitations. For example, it does not account for the cell- type specificity of intracellular signaling, which causes some but not all of the genes included in the gene-sets to be influential in any given cell (Rojas & Fiedler, 2016). Due to the high overlap of those cell-specific pathways (Rojas & Fiedler, 2016), it was determined that the limitation of not accounting for cell-specificity may be preferable to the inclusion of many overlapping pathways. However, this conundrum illustrates the general principle that most decisions regarding gene-set definitions entail choosing one set of limitations over another, especially when investigating complex disorders for which biological etiology is not well known (De

Leeuw et al., 2016). In addition to the aforementioned limitations, this analysis used a homogeneous sample of individuals who all live in the same country (USA) and who were all classified as belonging to the same geographic ancestry group (“European American”), which limits generalizability. The use of a homogeneous sample is problematic. First, in order to assure that vulnerable populations can be given the best available treatment, it is critical to directly study those populations (Benjet et al., 2016). Second, the use of more diverse samples is Genetic Risk Factors for PTSD….. 81 sometimes critical in interpreting the possible causal mechanism through which genes and gene- sets exert their effects on behavior (Logue et al., 2015). For example, due to group-average differences in the presence of inverted segments of a particular inversion region (i.e. differences in how frequently a segment of a chromosome is inherited in such a way that SNP locations along the chromosome are flipped from front to back), one GWAS study was only able to identify the true causal SNP underlying a particular GWAS association initially found in a

“European American” sample when including members of other geographic ancestry groups in the analysis (Levey et al., 2019). Thus, it will be critical for future studies to examine diverse populations, especially using samples that are diverse enough for meaningful examination across geographic ancestry groups.

5.4 Future Directions In section 5.2, next steps were discussed with respect to methodological modifications to the current study. In this section, the current study and overall program of research to which it belongs are discussed within the context of the broader literature, including several future directions which extend beyond the hypotheses explored in the current work.

A clear future direction for this line of research is to continue increasing statistical power in GWAS studies of PTSD (De Leeuw et al., 2016). The Psychiatric Genomics Consortium and

Million Veteran Program are both continuing to collect larger samples for future studies (Levey et al., 2019; Nievergelt et al., 2019). As statistical power increases, it is expected that additional genome-wide significant SNPs will be identified and GWAS findings will become more robust

(Levey et al., 2019; Nievergelt et al., 2019). As summary statistics from robust PTSD GWAS become increasingly available, hypothesis-driven gene-set analysis may be most immediately Genetic Risk Factors for PTSD….. 82 useful in examining big-picture questions. Several relevant big-picture questions can be explored by applying the current analytical approach (or a modified version of it). The analysis explored in the current study, if applied to a bigger dataset, may help with answering two important big- picture questions. First, are genetic risk effects concentrated within gene-sets which contribute to neurotransmitter receptors and/or their intracellular signaling pathways? Second, if so, into which neurotransmitter systems and/or intracellular pathways are they most concentrated?

Answering questions such as these may be particularly important, due to the tendency for most pharmacological approaches to focus on neurotransmitter receptors (Cavanagh & Mathias,

2008). Depending on the answers, such research could have major implications for the understanding of genetic aspects of etiology and may inform future research questions.

A major challenge in genetic research of complex disorders is reconciling the current evidence supporting the omnigenic hypothesis with the need to identify druggable disease mechanisms for pharmacological advancement (Boyle et al., 2017). In the introduction section, I reviewed evidence to suggest that gene-set analysis of PTSD GWAS results might be helpful in linking diffusely acting genes with higher-level biological pathways within which genetic risk is concentrated (Duncan, Ratanatharathorn, et al., 2018; Nievergelt et al., 2019). One aim of the current study was to examine gene-sets which focus on neurotransmitter receptors that are currently identified as promising pharmacotherapy targets (Krystal et al., 2017). However, if genetic risk for PTSD is not concentrated within neurotransmitter receptors or their intracellular pathways, it will be important to examine whether the lack of enrichment (relative to other types of molecules) is found at other levels of analysis (e.g. transcriptomic, peripheral blood)

(Horrobin, 2001). However, if genetic risk effects are concentrated within neurotransmitter receptors and/or their pathways, further examination of the specific receptors and pathways, both Genetic Risk Factors for PTSD….. 83 at the genetic level and other levels analysis (e.g. transcriptomic, peripheral blood, neuroimaging, post-morbid, pharmacological, etc.), may help inform pharmacological studies.

Regardless of findings for neurotransmitter receptor and intra-cellular gene-sets, it will be important to examine other aspects of neurotransmitter systems (e.g. neurotransmitter metabolizing genes, neurotransmitter transporter genes, etc.).

In addition to increasing sample size, the Psychiatric Genomics Consortium has already communicated its plans to address limitations of this study by conducting analyses which account for trauma type and gene x environment interactions (Logue et al., 2015). Further, the

Psychiatric Genomics Consortium has expressed a commitment to deep phenotyping (i.e. extensive characterization of distinct disease presentations yielding highly precise and computationally accessible information which can be linked with genomic data and other biological data), which is a critical aspect of converting robust correlations into etiological insights (Logue et al., 2015). As a compliment to these future GWAS studies, gene-set analysis may be helpful in integrating information about SNP-level correlation with higher-level biological pathways. For example, gene-set analyses may be helpful in exploring the degree to which enriched gene-sets for PTSD covary with trauma-type as well as the degree to which these trauma-type specific gene-sets converge onto similar biological pathways. A proof-of-concept for such an approach already exists in the transcription-based gene-set analysis literature. A previous transcriptome study applied gene-set analysis to peripheral gene expression data and found that while a majority of enriched gene-sets were trauma-specific, a substantial minority were enriched across all trauma types (Breen et al., 2018). Further, many of the trauma-specific gene-sets converged into similar immunological pathways (Breen et al., 2018). A similar study applied to GWAS data would be informative as to whether or not findings are similar at the level Genetic Risk Factors for PTSD….. 84 of the genome. Additionally, study of gene x environment interaction gene-set enrichment might help to further isolate specific pathways for which the effects of genetic variants may interact with specific aspects of trauma (Logue et al., 2015). Pairing these investigations with deep phenotyping may be informative in specifying clinically relevant genetic subtypes, which may be differentially expressed depending on trauma type (T. P. Sullivan et al., 2006). In order to maximize the potential for genetic association and deep phenotyping to generate insights into clinical subtypes, studies which use a genotype-first may be particularly informative (Mefford,

2009).

In contrast with GWAS, the use of a genotype-first approach entails starting with a gene of interest and investigating which phenotypes and clinical presentations are associated with variants in that gene (Mefford, 2009; Stessman, Bernier, & Eichler, 2014). While genotype-first studies are typically conducted at the gene level with the goal of identifying “genetic subtypes” of disorders, gene-set analysis using this approach can help to define “molecular subtypes” based on the pathways impacted by combinations of genes (Stessman et al., 2014). The promise of a genotype-first approach has been demonstrated in autism; two different genes within the “beta- catenin/Wnt-signaling pathway” have been found to lead to two different “genetic subtypes” of autism that may have phenotypically identifiable and clinically relevant features (Stessman et al.,

2014). Of note, while these two “genetic subtypes” are defined by genes which are found within a common pathway, they have opposite effects on the pathway; one is associated with excess cell proliferation while the other is associated with excess cell death (Stessman et al., 2014). Using a genotype-first approach, future studies should build on findings of enriched genotypes and gene- sets in PTSD by investigating specific “genetic subtypes” and “molecular subtypes” with respect to cross-diagnosis endophenotypes of trauma-related psychopathology. Ideally, such approaches Genetic Risk Factors for PTSD….. 85 should be conducted along with “deep phenotyping” (i.e. extensive characterization of distinct disease presentations yielding highly precise and computationally accessible information which can be linked with genomic data and other biological data) using a combination of extensive clinical data and biological data at multiple levels (e.g. neuroimaging, psychophysiology, etc.).

In all future directions, analyses should be conducted in diverse geographic ancestry groups with the ultimate goal of describing genetic causes of PTSD and other trauma-related psychopathology across geographic ancestry. Though current “trans-ethnic” GWAS has not yielded genome-wide significant SNPs for PTSD, gene-set analysis in current data should still investigate the “trans-ethnic” sample in addition to the “European Ancestry” and “African

Ancestry” samples in order to provide a comprehensive and generalizable view of current data.

5.5 Conclusion This analysis was initially designed to investigate two important research questions. First, are gene-sets comprised of neurotransmitter receptor groups and/or neurotransmitter receptor subtypes and their intracellular signaling pathways genetically enriched. Second, if so, which specific neurotransmitter receptor groups and/or subtype-intracellular pathways are enriched.

Due to a limited sample size, the study was not sufficiently powered to provide a robust analysis.

However, preliminary results yielded two useful methodological insights. First, GSA-SNP2 may not be the ideal statistical software with which to test these hypotheses. Second, by reconceptualization and consolidation of “receptor subtypes” gene-sets into “intracellular signaling” gene-sets may lead to increased parsimony and interpretability with respect to pharmacology. In order to maximize the advantages of increased statistical power, analysis in Genetic Risk Factors for PTSD….. 86 larger samples might be best conducted using the Generalized Berk-Jones R package to analyze the “intracellular signaling” and “neurotransmitter receptor group” gene-sets.

Genetic Risk Factors for PTSD….. 87

Glossary Active Chromatin - chromatin that can be transcribed into RNA Chromatin - DNA-protein complex which provides the structure for , prevents DNA damage, and plays a role in regulating DNA expression and replication by packaging DNA molecules into condensed structures (more compact forms = DNA less able to be activated) Combined Annotation-Dependent Depletion score - a machine-learning based estimation of deleteriousness based on simulated estimates of selective pressure) Core Genes – are thought to exert concentrated effects on a particular biological mechanism by either encoding or directly regulating expression of molecular targets Deep Phenotyping - extensive characterization of distinct disease presentations yielding highly precise and computationally accessible information which can be linked with genomic data and other biological data Enriched Gene Set - a gene-set that has been identified as being comprised of genes that are especially strongly correlated with the disorder of interest (i.e. the correlation is stronger than that of the average genes to an extent that exceeds what one would expect to find by chance) Expression quantitative trait loci analysis (eQTL analysis) - analysis of which loci in the genome explain variation in levels of a given mRNA Genotype-first approach – A bottom-up approach which seeks to identify the behavioral and psychiatric effects of specific genotype variants Genome-wide significant SNP - a SNP that has reached the threshold for statistical significance in a GWAS study after correcting for multiple comparisons Geographic Ancestry Group - a group of participants in a specific study that has been defined using genetic cutoff criteria specific to that study and is comprised of individuals who have been estimated to meet cutoff criteria for that group based on the statistical methods used to estimate genealogical history in that study GSA-SNP2 - the name of the analysis I performed as well as the name of the software I used to perform it GWAS - genome-wide association study; enables the concurrent testing of genetic variants across the entire genome in relation to PTSD Haplotype - a group of SNPs within the same chromosome which are positioned near one another within a particular segment of the chromosome and are inherited together as a group Inactive Chromatin - genetically inactive chromatin (i.e. chromatin that cannot be transcribed into RNA) (e.g. constitutive heterochromatin) Intergenic DNA / Intergenic SNP - DNA sequences located between genes which are always non-coding DNA Genetic Risk Factors for PTSD….. 88

Inversion region - segment of a chromosome which has a heritable variant with an end to end reversal such that SNP locations are flipped from front to back Linkage Disequilibrium - the tendency for SNPs which are located near one another in the genome to have nonrandom correlations with one another Locus - particular position on a chromosome MAGMA - a particular commonly used form of gene set analysis as well as the software developed to perform MAGMA analysis Magnetic resonance spectroscopy (SPECT) - provides an index of level of certain chemicals (e.g. neurotransmitters) in living brain tissue by detecting electromagnetic signals attained from atomic nuclei within molecules Non-coding DNA / Non-coding SNP - sequences of DNA that do not encode protein / a SNP located within noncoding DNA; sometimes regulate coding DNA expression Non-coding RNA - RNA molecule that is not translated into protein; many non-coding RNA have important functions including regulating gene expression Omnigenic hypothesis – states that a majority of genetic risk for complex genetic disorders is derived from genes that have small effects on many biological processes but are not specific to any one core biological process Peripheral Genes - are thought to exert nontrivial effects on a broad group of biological processes through indirect effects Pleiotropy - the influence of one gene on multiple theoretically nonoverlapping phenotypes Population Stratification = group-level differences in average allelic frequencies between ancestral groups, especially those which may confound GWAS findings positron emission tomography (PET) - provides an index of receptor availability (or other metabolic processes depending upon the radiotracer used) by detecting a radiotracer which can bind to a ligand of interest (i.e. can bind to a receptor of interest) or which is involved in the metabolic process of interest in some measurable way Race - a socially constructed concept which has historically been used to place individuals into vaguely defined hierarchical categories based on a preconceived notion that cosmetic phenotypic traits, geographical ancestry, and individual characteristics covary SNP - single-nucleotide polymorphism; Each SNP represents one locus (specific location on the genome) for which one of two possible nucleotide variants, known as alleles, may appear

Genetic Risk Factors for PTSD….. 89

References A Kato, T., Yamauchi, Y., Horikawa, H., Monji, A., Mizoguchi, Y., Seki, Y., . . . Kanba, S.

(2013). Neurotransmitters, psychotropic drugs and microglia: clinical implications for

psychiatry. Current medicinal chemistry, 20(3), 331-344.

Afifi, T. O., Asmundson, G. J., Taylor, S., & Jang, K. L. J. C. p. r. (2010). The role of genes and

environment on trauma exposure and posttraumatic stress disorder symptoms: a review of

twin studies. 30(1), 101-112.

Akiki, T. J., & Abdallah, C. G. (2018). Are There Effective Psychopharmacologic Treatments

for PTSD? The Journal of clinical psychiatry, 80(3), 0-0.

Albert, P. R. (2012). Transcriptional regulation of the 5-HT1A receptor: implications for mental

illness. Philosophical Transactions of the Royal Society B: Biological Sciences,

367(1601), 2402-2415.

Albert, P. R., Le François, B., & Vahid-Ansari, F. (2019). Genetic, epigenetic and

posttranscriptional mechanisms for treatment of major depression: the 5-HT1A receptor

gene as a paradigm. Journal of psychiatry and neuroscience, 44(3), 164.

Albert, P. R., & Vahid-Ansari, F. (2018). The 5-HT1A receptor: signaling to behavior.

Biochimie.

Aoki, K. F., & Kanehisa, M. (2005). Using the KEGG database resource. Current protocols in

bioinformatics, 11(1), 1.12. 11-11.12. 54.

Artigas, F. (2016). The role of brain 5-HT1A receptors in the actions of antidepressant drugs.

30th CINP World Congress.

Artigas, F., Celada, P., & Bortolozzi, A. (2018). Can we increase the speed and efficacy of

antidepressant treatments? Part II. Glutamatergic and RNA interference strategies.

European Neuropsychopharmacology, 28(4), 457-482. Genetic Risk Factors for PTSD….. 90

Ashley-Koch, A. E., Garrett, M. E., Gibson, J., Liu, Y., Dennis, M. F., Kimbrel, N. A., . . .

Hauser, M. A. (2015). Genome-wide association study of posttraumatic stress disorder in

a cohort of Iraq–Afghanistan era veterans. Journal of affective disorders, 184, 225-234.

Association, A. P. (2017). Clinical practice guideline for the treatment of PTSD. In.

Atwoli, L., Stein, D. J., Koenen, K. C., & McLaughlin, K. A. (2015). Epidemiology of

posttraumatic stress disorder: prevalence, correlates and consequences. Current opinion

in psychiatry, 28(4), 307.

Averill, L. A., Purohit, P., Averill, C. L., Boesl, M. A., Krystal, J. H., & Abdallah, C. G. (2017).

Glutamate dysregulation and glutamatergic therapeutics for PTSD: Evidence from human

studies. Neuroscience letters, 649, 147-155.

Banerjee, S. B., Morrison, F. G., & Ressler, K. J. (2017). Genetic approaches for the study of

PTSD: advances and challenges. Neuroscience letters, 649, 139-146.

Bankson, M. G., & Cunningham, K. A. (2001). 3, 4-Methylenedioxymethamphetamine

(MDMA) as a unique model of serotonin receptor function and serotonin-dopamine

interactions. Journal of Pharmacology and Experimental Therapeutics, 297(3), 846-852.

Banner, L. R., Patterson, P. H., Allchorne, A., Poole, S., & Woolf, C. J. (1998). Leukemia

inhibitory factor is an anti-inflammatory and analgesic cytokine. Journal of

Neuroscience, 18(14), 5456-5462.

Baou, M., Boumba, V. A., Petrikis, P., Rallis, G., Vougiouklakis, T., & Mavreas, V. (2016). A

review of genetic alterations in the serotonin pathway and their correlation with psychotic

diseases and response to atypical antipsychotics. Schizoprenia research, 170(1), 18-29. Genetic Risk Factors for PTSD….. 91

Beauchaine, T. P., & Constantino, J. N. (2017). Redefining the endophenotype concept to

accommodate transdiagnostic vulnerabilities and etiological complexity. Biomarkers in

medicine, 11(9), 769-780.

Belmont, J. W., & Leal, S. M. J. C. a. r. (2005). Complex phenotypes and complex genetics: an

introduction to genetic studies of complex traits. 7(3), 180-187.

Benjet, C., Bromet, E., Karam, E., Kessler, R., McLaughlin, K., Ruscio, A., . . . Hill, E. (2016).

The epidemiology of traumatic event exposure worldwide: results from the World Mental

Health Survey Consortium. Psychological medicine, 46(2), 327-343.

Berg, K. A., Maayani, S., Goldfarb, J., Scaramellini, C., Leff, P., & Clarke, W. P. (1998).

Effector pathway-dependent relative efficacy at serotonin type 2A and 2C receptors:

evidence for agonist-directed trafficking of receptor stimulus. Molecular pharmacology,

54(1), 94-104.

Bergen, S. E., & Petryshen, T. L. (2012). Genome-wide association studies (GWAS) of

schizophrenia: does bigger lead to better results? Current opinion in psychiatry, 25(2),

76.

Blum, K., Gondré-Lewis, M., Modestino, E., Lott, L., Baron, D., Siwicki, D., . . . Oscar-Berman,

M. (2019). Understanding the Scientific Basis of Post-traumatic Stress Disorder (PTSD):

Precision Behavioral Management Overrides Stigmatization. Molecular neurobiology,

56(11), 7836-7850.

Bosker, F., Hartman, C., Nolte, I., Prins, B., Terpstra, P., Posthuma, D., . . . De Geus, E. J. M. p.

(2011). Poor replication of candidate genes for major depressive disorder using genome-

wide association data. 16(5), 516. Genetic Risk Factors for PTSD….. 92

Bouaziz, M., Ambroise, C., & Guedj, M. (2011). Accounting for population stratification in

practice: a comparison of the main strategies dedicated to genome-wide association

studies. PloS one, 6(12).

Boyle, E. A., Li, Y. I., & Pritchard, J. K. J. C. (2017). An expanded view of complex traits: from

polygenic to omnigenic. 169(7), 1177-1186.

Breen, M. S., Tylee, D. S., Maihofer, A. X., Neylan, T. C., Mehta, D., Binder, E. B., . . .

Risbrough, V. B. (2018). PTSD blood transcriptome mega-analysis: shared inflammatory

pathways across biological sex and modes of trauma. Neuropsychopharmacology, 43(3),

469.

Bremner, J. D., Steinberg, M., Southwick, S. M., Johnson, D. R., & Charney, D. S. (1993). Use

of the Structured Clinical Interview for DSM-IV Dissociative Disorders for systematic

assessment of dissociative symptoms in posttraumatic stress disorder. The American

Journal of Psychiatry.

Breslau, N., Kessler, R., & Peterson, E. L. (1998). Post‐traumatic stress disorder assessment with

a structured interview: reliability and concordance with a standardized clinical interview.

International Journal of Methods in Psychiatric Research, 7(3), 121-127.

Briggs‐Gowan, M. J., Carter, A. S., Clark, R., Augustyn, M., McCarthy, K. J., & Ford, J. D.

(2010). Exposure to potentially traumatic events in early childhood: differential links to

emergent psychopathology. Journal of Child Psychology and Psychiatry, 51(10), 1132-

1140.

Brown, J. A., Sweatt, J. D., & Kaas, G. A. (2019). Locus-specific DNA methylation assays to

study glutamate receptor regulation. In Glutamate Receptors (pp. 167-188): Springer. Genetic Risk Factors for PTSD….. 93

Cassiers, L. L. M., Sabbe, B. G., Schmaal, L., Veltman, D. J., Penninx, B. W., & Van Den Eede,

F. (2018). Structural and functional brain abnormalities associated with exposure to

different childhood trauma subtypes: A systematic review of neuroimaging findings.

Frontiers in psychiatry, 9, 329.

Caulfield, T., Fullerton, S. M., Ali-Khan, S. E., Arbour, L., Burchard, E. G., Cooper, R. S., . . .

Kahn, J. (2009). Race and ancestry in biomedical research: exploring the challenges.

Genome medicine, 1(1), 8.

Cavanagh, J., & Mathias, C. (2008). Inflammation and its relevance to psychiatry. Advances in

Psychiatric Treatment, 14(4), 248-255.

Chang, C. C., Chow, C. C., Tellier, L. C., Vattikuti, S., Purcell, S. M., & Lee, J. J. (2015).

Second-generation PLINK: rising to the challenge of larger and richer datasets.

Gigascience, 4(1), s13742-13015-10047-13748.

Chantarujikapong, S. I., Scherrer, J. F., Xian, H., Eisen, S. A., Lyons, M. J., Goldberg, J., . . .

True, W. R. (2001). A twin study of generalized anxiety disorder symptoms, panic

disorder symptoms and post-traumatic stress disorder in men. Psychiatry research,

103(2-3), 133-145.

Chebib, M., & Johnston, G. A. (1999). The ‘ABC’of GABA receptors: a brief review. Clinical

experimental pharmacology physiology, 26(11), 937-940.

Chen, C.-Y., Pollack, S., Hunter, D. J., Hirschhorn, J. N., Kraft, P., & Price, A. L. (2013).

Improved ancestry inference using weights from external reference panels.

Bioinformatics, 29(11), 1399-1406.

Ciccarelli, M., Sorriento, D., Coscioni, E., Iaccarino, G., & Santulli, G. (2017). Adrenergic

receptors. In Endocrinology of the Heart in Health and Disease (pp. 285-315): Elsevier. Genetic Risk Factors for PTSD….. 94

Coffey, S. F., Gudmundsdottir, B., Beck, J. G., Palyo, S. A., & Miller, L. (2006). Screening for

PTSD in motor vehicle accident survivors using the PSS‐SR and IES. Journal of

traumatic stress, 19(1), 119-128.

Coleman, J. R., Gaspar, H. A., Bryois, J., & Breen, G. J. B. P. (2019). The genetics of the mood

disorder spectrum: genome-wide association analyses of over 185,000 cases and 439,000

controls.

Consortium, G. P. (2010). A map of variation from population-scale sequencing.

Nature, 467(7319), 1061.

Consortium, G. P. (2015). A global reference for human genetic variation. Nature, 526(7571),

68-74.

Consortium, I. H. (2003). The international HapMap project. Nature, 426(6968), 789.

Database, G. S. E. A. (2020).

GO_POSITIVE_REGULATION_OF_TUMOR_NECROSIS_FACTOR_SUPERFAMIL

Y_CYTOKINE_PRODUCTION. https://www.gsea-

msigdb.org/gsea/msigdb/cards/GO_POSITIVE_REGULATION_OF_TUMOR_NECROSI

S_FACTOR_SUPERFAMILY_CYTOKINE_PRODUCTION.

Database, M. https://www.gsea-msigdb.org/gsea/msigdb/collections.jsp.

David, S. P., Murthy, N. V., Rabiner, E. A., Munafó, M. R., Johnstone, E. C., Jacob, R., . . .

Grasby, P. M. (2005). A functional genetic variation of the serotonin (5-HT) transporter

affects 5-HT1A receptor binding in humans. Journal of Neuroscience, 25(10), 2586-

2590.

Day, D., & Tuite, M. F. (1998). Post-transcriptional gene regulatory mechanisms in eukaryotes:

an overview. The Journal of Endocrinology, 157(3), 361-371. Genetic Risk Factors for PTSD….. 95

De Leeuw, C. A., Neale, B. M., Heskes, T., & Posthuma, D. J. N. R. G. (2016). The statistical

properties of gene-set analysis. 17(6), 353.

Devlin, B., & Roeder, K. (1999). Genomic control for association studies. Biometrics, 55(4),

997-1004.

Doll, B. B., Bath, K. G., Daw, N. D., & Frank, M. J. (2016). Variability in dopamine genes

dissociates model-based and model-free reinforcement learning. Journal of Neuroscience,

36(4), 1211-1222.

Duncan, L. E., Cooper, B. N., & Shen, H. (2018). Robust Findings From 25 Years of PTSD

Genetics Research. Current Psychiatry Reports, 20(12), 115.

Duncan, L. E., Ratanatharathorn, A., Aiello, A. E., Almli, L. M., Amstadter, A. B., Ashley-Koch,

A. E., . . . Bisson, J. (2018). Largest GWAS of PTSD (N= 20 070) yields genetic overlap

with schizophrenia and sex differences in heritability. Molecular psychiatry, 23(3), 666.

Ekblad, S., Jaranson, J. M. J. J. P. W., Boris Drožđek. Broken Spirits. The treatment of

traumatized asylum seekers, r., war, & Pp, t. v. (2004). Psychosocial rehabilitation. 609-

636.

Faye, L. L., Sun, L., Dimitromanolakis, A., & Bull, S. B. J. S. i. m. (2011). A flexible genome‐

wide bootstrap method that accounts for rankingand threshold‐selection bias in GWAS

interpretation and replication study design. 30(15), 1898-1912.

Feusner, J., Ritchie, T., Lawford, B., Young, R. M., Kann, B., & Noble, E. P. (2001). GABAA

receptor β3 subunit gene and psychiatric morbidity in a post-traumatic stress disorder

population. Psychiatry research, 104(2), 109-117. Genetic Risk Factors for PTSD….. 96

Foa, E. B., Johnson, K. M., Feeny, N. C., & Treadwell, K. R. (2001). The Child PTSD Symptom

Scale: A preliminary examination of its psychometric properties. Journal of clinical child

psychology, 30(3), 376-384.

Gabriel, S. B., Schaffner, S. F., Nguyen, H., Moore, J. M., Roy, J., Blumenstiel, B., . . . Faggart,

M. (2002). The structure of haplotype blocks in the human genome. Science, 296(5576),

2225-2229.

Galatzer-Levy, I. R., & Bryant, R. A. (2013). 636,120 ways to have posttraumatic stress disorder.

Perspectives on Psychological Science, 8(6), 651-662.

García-Campos, M. A., Espinal-Enríquez, J., & Hernández-Lemus, E. J. F. i. p. (2015). Pathway

analysis: state of the art. 6, 383.

Gelernter, J., Sun, N., Polimanti, R., Pietrzak, R., Levey, D. F., Bryois, J., . . . Radhakrishnan, K.

J. N. n. (2019). Genome-wide association study of post-traumatic stress disorder

reexperiencing symptoms in> 165,000 US veterans. 22(9), 1394-1401.

Gentes, E. L., Dennis, P. A., Kimbrel, N. A., Rissling, M. B., Beckham, J. C., Workgroup, V.

M.-A. M., & Calhoun, P. S. (2014). DSM-5 posttraumatic stress disorder: Factor

structure and rates of diagnosis. Journal of psychiatric research, 59, 60-67.

Gillespie, C. F., Phifer, J., Bradley, B., & Ressler, K. J. (2009). Risk and resilience: genetic and

environmental influences on development of the stress response. Depression and Anxiety,

26(11), 984-992.

Gómez-Lázaro, E., Arregi, A., Beitia, G., Vegas, O., Azpiroz, A., & Garmendia, L. (2011).

Individual differences in chronically defeated male mice: behavioral, endocrine, immune,

and neurotrophic changes as markers of vulnerability to the effects of stress. Stress,

14(5), 537-548. Genetic Risk Factors for PTSD….. 97

Grant, D. M., Beck, J. G., Marques, L., Palyo, S. A., & Clapp, J. D. (2008). The structure of

distress following trauma: Posttraumatic stress disorder, major depressive disorder, and

generalized anxiety disorder. Journal of Abnormal Psychology, 117(3), 662.

Greenland, S., Senn, S. J., Rothman, K. J., Carlin, J. B., Poole, C., Goodman, S. N., & Altman,

D. G. (2016). Statistical tests, P values, confidence intervals, and power: a guide to

misinterpretations. European journal of epidemiology, 31(4), 337-350.

Guffanti, G., Galea, S., Yan, L., Roberts, A. L., Solovieff, N., Aiello, A. E., . . . Uddin, M.

(2013). Genome-wide association study implicates a novel RNA gene, the lincRNA

AC068718. 1, as a risk factor for post-traumatic stress disorder in women.

Psychoneuroendocrinology, 38(12), 3029-3038.

Hellwege, J. N., Keaton, J. M., Giri, A., Gao, X., Velez Edwards, D. R., & Edwards, T. L.

(2017). Population stratification in genetic association studies. Current protocols in

human genetics, 95(1), 1.22. 21-21.22. 23.

Holmans, P., Green, E. K., Pahwa, J. S., Ferreira, M. A., Purcell, S. M., Sklar, P., . . .

Consortium, W. T. C.-C. (2009). analysis of GWA study data sets

provides insights into the biology of bipolar disorder. The American Journal of Human

Genetics, 85(1), 13-24.

Holmes, S., Girgenti, M., Davis, M., Pietrzak, R., & DellaGioia, N. (2017). Altered metabotropic

glutamate receptor 5 markers in PTSD: In vivo and postmortem evidence. PNAS, 135(1-

3), 200.

Horrobin, D. F. (2001). Phospholipid metabolism and depression: the possible roles of

phospholipase A2 and coenzyme A‐independent transacylase. Human

Psychopharmacology: Clinical and Experimental, 16(1), 45-52. Genetic Risk Factors for PTSD….. 98

Hoskins, M., Pearce, J., Bethell, A., Dankova, L., Barbui, C., Tol, W. A., . . . Chen, H. (2015).

Pharmacotherapy for post-traumatic stress disorder: systematic review and meta-analysis.

The British Journal of Psychiatry, 206(2), 93-100. https://www.genecards.org/. (2020).

Huckins, L., Breen, M. S., Chatzinakos, C., Hartmann, J., Klengel, T., da Silva Almeida, A. C., .

. . Klengel, C. (2019). Analysis of Genetically Regulated Gene Expression identifies a

trauma type specific PTSD gene, SNRNP35. Cell Press Sneak Peek.

Inoue, T. (1993). Effects of conditioned fear stress on monoaminergic systems in the rat brain.

The Hokkaido journal of medical science, 68(3), 377-390.

Jacobsen, K. X., Vanderluit, J. L., Slack, R. S., & Albert, P. R. (2008). HES1 regulates 5-HT1A

receptor gene transcription at a functional polymorphism: essential role in developmental

expression. Molecular Cellular Neuroscience, 38(3), 349-358.

Johnson, W., Turkheimer, E., Gottesman, I. I., & Bouchard Jr, T. J. (2009). Beyond heritability:

Twin studies in behavioral research. Current directions in psychological science, 18(4),

217-220.

Johnston, S., Staines, D., Klein, A., & Marshall-Gradisnik, S. (2016). A targeted genome

association study examining transient receptor potential ion channels, acetylcholine

receptors, and adrenergic receptors in Chronic Fatigue Syndrome/Myalgic

Encephalomyelitis. BMC Medical Genetics, 17(1), 79.

Jones, M. E., Lebonville, C. L., Barrus, D., & Lysle, D. T. (2015). The role of brain interleukin-1

in stress-enhanced fear learning. Neuropsychopharmacology, 40(5), 1289-1296.

Jooyeon, J., & N, E. (2011). Molecular Neuroimaging in Posttraumatic Stress Disorder. Exp

Neurobiol, 24(1), 3-24. Genetic Risk Factors for PTSD….. 99

Jorgensen, E. M. (2005). Gaba. In WormBook: The Online Review of C. elegans Biology

[Internet]: WormBook.

Jorgensen, T. J., Ruczinski, I., Kessing, B., Smith, M. W., Shugart, Y. Y., & Alberg, A. J.

(2009). Hypothesis-driven candidate gene association studies: practical design and

analytical considerations. American journal of epidemiology, 170(8), 986-993.

Kariagina, A., Zonis, S., Afkhami, M., Romanenko, D., & Chesnokova, V. (2005). Leukemia

inhibitory factor regulates glucocorticoid receptor expression in the hypothalamic-

pituitary-adrenal axis. American Journal of physiology-endocrinology metabolism,

289(5), E857-E863.

Keely Jr, L. B., Young, D. A., & Palay, A. J. (2002). Pen-based interface for a notepad computer.

In: Google Patents.

KEGG Pathway https://www.genome.jp/kegg-bin/show_pathway?hsa04727, 10.

Kenakin, T., & Christopoulos, A. (2013). Signalling bias in new drug discovery: detection,

quantification and therapeutic impact. Nature reviews drug discovery, 12(3), 205.

Kilpatrick, D. G., Resnick, H. S., Milanak, M. E., Miller, M. W., Keyes, K. M., & Friedman, M.

(2013). National estimates of exposure to traumatic events and PTSD prevalence using

DSM‐IV and DSM‐5 criteria. Journal of traumatic stress, 26(5), 537-547.

Koenen, K. C., Hitsman, B., Lyons, M. J., Niaura, R., McCaffery, J., Goldberg, J., . . . Tsuang,

M. (2005). A twin registry study of the relationship between posttraumatic stress disorder

and nicotine dependence in men. Archives of general psychiatry, 62(11), 1258-1265.

Koenen, K. C., Lyons, M. J., Goldberg, J., Simpson, J., Williams, W. M., Toomey, R., . . .

Wolfe, J. (2003). A high risk twin study of combat-related PTSD comorbidity. Twin

Research Human Genetics, 6(3), 218-226. Genetic Risk Factors for PTSD….. 100

Komatsu, H., Fukuchi, M., & Habata, Y. (2019). Potential utility of biased GPCR signaling for

treatment of psychiatric disorders. International journal of molecular sciences, 20(13),

3207.

Krystal, J. H., Davis, L. L., Neylan, T. C., Raskind, M. A., Schnurr, P. P., Stein, M. B., . . .

Huang, G. D. (2017). It is time to address the crisis in the pharmacotherapy of

posttraumatic stress disorder: a consensus statement of the PTSD Psychopharmacology

Working Group. Biological psychiatry, 82(7), e51-e59.

Kudo, I., & Murakami, M. (2002). Phospholipase A2 enzymes. Prostaglandins and other lipid mediators, 68, 3-58.

Lee, E. C., Whitehead, A. L., Jacques, R. M., & Julious, S. A. (2014). The statistical

interpretation of pilot trials: should significance thresholds be reconsidered? BMC

Medical Research Methodology, 14(1), 41.

Lencz, T., & Malhotra, A. J. M. p. (2015). Targeting the schizophrenia genome: a fast track

strategy from GWAS to clinic. 20(7), 820.

Levey, D. F., Gelernter, J., Polimanti, R., Zhou, H., Cheng, Z., Aslan, M., . . . Bryois, J. J. b.

(2019). Reproducible Risk Loci and Psychiatric Comorbidities in Anxiety: Results from~

200,000 Million Veteran Program Participants. 540245.

Li, L., Bao, Y., He, S., Wang, G., Guan, Y., Ma, D., . . . Zhang, D. (2016). The association

between genetic variants in the dopaminergic system and posttraumatic stress disorder: a

meta-analysis. Medicine, 95(11).

Liberzon, I., King, A. P., Ressler, K. J., Almli, L. M., Zhang, P., Ma, S. T., . . . Galea, S. (2014).

Interaction of the ADRB2 gene polymorphism with childhood trauma in predicting adult

symptoms of posttraumatic stress disorder. JAMA psychiatry, 71(10), 1174-1182. Genetic Risk Factors for PTSD….. 101

Lin, L., Liu, G., & Sun, M. (2017). Effect of 5-HT1 A receptor in hippocampal CA1 region on

spatial memory of PTSD rats. Chinese Journal of Pathophysiology, 33(1), 98-103.

Logue, M. W., Amstadter, A. B., Baker, D. G., Duncan, L., Koenen, K. C., Liberzon, I., . . .

Ressler, K. J. J. N. (2015). The Psychiatric Genomics Consortium Posttraumatic Stress

Disorder Workgroup: posttraumatic stress disorder enters the age of large-scale genomic

collaboration. 40(10), 2287.

Manolio, T. A., Collins, F. S., Cox, N. J., Goldstein, D. B., Hindorff, L. A., Hunter, D. J., . . .

Chakravarti, A. (2009). Finding the missing heritability of complex diseases. Nature,

461(7265), 747-753.

Marchini, J., Cardon, L. R., Phillips, M. S., & Donnelly, P. (2004). The effects of human

population structure on large genetic association studies. Nature Genetics, 36(5), 512-

517.

Marian, A. J. J. T. R. (2012). Molecular genetic studies of complex phenotypes. 159(2), 64-79.

Marigorta, U. M., & Navarro, A. (2013). High trans-ethnic replicability of GWAS results implies

common causal variants. PLoS genetics, 9(6).

Marigorta, U. M., Rodríguez, J. A., Gibson, G., & Navarro, A. (2018). Replicability and

prediction: lessons and challenges from GWAS. Trends in Genetics, 34(7), 504-517.

Mateu, E., Calafell, F., Ramos, M. D., Casals, T., & Bertranpetit, J. (2002). Can a place of origin

of the main cystic fibrosis mutations be identified? The American Journal of Human

Genetics, 70(1), 257-264.

McFarlane, A. C., & Bookless, C. (2001). The effect of PTSD on interpersonal relationships:

Issues for emergency service workers. Sexual Relationship Therapy, 16(3), 261-267. Genetic Risk Factors for PTSD….. 102

Mefford, H. C. (2009). Genotype to phenotype—discovery and characterization of novel

genomic disorders in a “genotype-first” era. Genetics in Medicine, 11(12), 836-842.

Mehta, C. R., Patel, N. R., & Senchaudhuri, P. (1998). Exact power and sample-size

computations for the Cochran-Armitage trend test. Biometrics, 1615-1621.

Mehta, D., & Binder, E. B. (2012). Gene× environment vulnerability factors for PTSD: the HPA-

axis. Neuropharmacology, 62(2), 654-662.

Mellon, S. H., Gautam, A., Hammamieh, R., Jett, M., & Wolkowitz, O. M. (2018). Metabolism,

metabolomics, and inflammation in posttraumatic stress disorder. Biological psychiatry,

83(10), 866-875.

Middeldorp, C., Cath, D., Van Dyck, R., & Boomsma, D. (2005). The co-morbidity of anxiety

and depression in the perspective of genetic epidemiology. A review of twin and family

studies. Psychological medicine, 35(5), 611-624.

Miller, M., Maniates, H., Wolf, E., Logue, M., Schichman, S., Stone, A., . . . McGlinchey, R.

(2018). CRP polymorphisms and DNA methylation of the AIM2 gene influence

associations between trauma exposure, PTSD, and C-reactive protein. Brain, behavior,

and immunity, 67, 194-202.

Mooney, M. A., & Wilmot, B. J. A. J. o. M. G. P. B. N. G. (2015). Gene set analysis: A step‐by‐

step guide. 168(7), 517-527.

Moreira, F. A., Grieb, M., & Lutz, B. (2009). Central side-effects of therapies based on CB1

cannabinoid receptor agonists and antagonists: focus on anxiety and depression. Best

practice research Clinical endocrinology metabolism

23(1), 133-144. Genetic Risk Factors for PTSD….. 103

Nichols, D. E., & Nichols, C. D. (2008). Serotonin receptors. Chemical reviews, 108(5), 1614-

1641.

Nievergelt, C., Maihofer, A., Dalvie, S., Duncan, L., Ratanatharathorn, A., Ressler, K., . . .

Koenen, K. (2018). 157. Large-scale genetic characterization of PTSD: addressing

heterogeneity across ancestry, sex, and trauma. Biological psychiatry, 83(9), S64.

Nievergelt, C. M., Ashley-Koch, A. E., Dalvie, S., Hauser, M. A., Morey, R. A., Smith, A. K., &

Uddin, M. (2018). Genomic approaches to posttraumatic stress disorder: the psychiatric

genomic consortium initiative. Biological psychiatry, 83(10), 831-839.

Nievergelt, C. M., Maihofer, A. X., Klengel, T., Atkinson, E. G., Chen, C.-Y., Choi, K. W., . . .

Gelernter, J. (2019). International meta-analysis of PTSD genome-wide association

studies identifies sex-and ancestry-specific genetic risk loci. Nature Communications,

10(1), 1-16.

Nievergelt, C. M., Maihofer, A. X., Mustapic, M., Yurgil, K. A., Schork, N. J., Miller, M. W., . .

. O’Connor, D. T. (2015). Genomic predictors of combat stress vulnerability and

resilience in US Marines: a genome-wide association study across multiple ancestries

implicates PRTFDC1 as a potential PTSD gene. Psychoneuroendocrinology, 51, 459-

471.

NIMH. (2018). What Are Single Nucleotide Polymorphisms (SNPs)?

https://ghr.nlm.nih.gov/primer/genomicresearch/snp, 46(10), e60-e60.

Ohgidani, M., Kato, T. A., Sagata, N., Hayakawa, K., Shimokawa, N., Sato-Kasai, M., & Kanba,

S. (2016). TNF-α from hippocampal microglia induces working memory deficits by acute

stress in mice. Brain, behavior, and immunity, 55, 17-24. Genetic Risk Factors for PTSD….. 104

Ossowska, G., Nowak, G., Kata, R., Klenk-Majewska, B., Danilczuk, Z., & Żebrowska-Łupina,

I. (2001). Brain monoamine receptors in a chronic unpredictable stress model in rats.

Journal of neural transmission, 108(3), 311-319.

Pemberton, T. J., Wang, C., Li, J. Z., & Rosenberg, N. A. (2010). Inference of unexpected

genetic relatedness among individuals in HapMap Phase III. The American Journal of

Human Genetics, 87(4), 457-464.

Picciotto, M. R., Higley, M. J., & Mineur, Y. S. (2012). Acetylcholine as a neuromodulator:

cholinergic signaling shapes nervous system function and behavior. Neuron, 76(1), 116-

129.

Pickrell, J. K., Marioni, J. C., Pai, A. A., Degner, J. F., Engelhardt, B. E., Nkadori, E., . . .

Pritchard, J. K. J. N. (2010). Understanding mechanisms underlying human gene

expression variation with RNA sequencing. 464(7289), 768.

Pierucci-Lagha, A., Gelernter, J., Feinn, R., Cubells, J. F., Pearson, D., Pollastri, A., . . .

Kranzler, H. R. (2005). Diagnostic reliability of the Semi-structured Assessment for Drug

Dependence and Alcoholism (SSADDA). Drug alchohol dependence, 80(3), 303-312.

Pietrzak, R. H., Goldstein, R. B., Southwick, S. M., & Grant, B. F. (2011). Prevalence and Axis I

comorbidity of full and partial posttraumatic stress disorder in the United States: results

from Wave 2 of the National Epidemiologic Survey on Alcohol and Related Conditions.

Journal of anxiety disorders, 25(3), 456-465.

Pillai, R., Zhang, M., Yang, J., Mann, J. J., Oquendo, M., Parsey, R., & DeLorenzo, C. (2017).

602. PET Imaging of Individual Raphe Nuclei in Major Depressive Disorder: Physiologic

Insight and Diagnostic Utility. J Biological Psychiatry, 81(10), S243-S244. Genetic Risk Factors for PTSD….. 105

Plomin, R., DeFries, J. C., Knopik, V. S., & Neiderhiser, J. M. (2016). Top 10 replicated

findings from behavioral genetics. Perspectives on Psychological Science, 11(1), 3-23.

Price, A. L., Patterson, N. J., Plenge, R. M., Weinblatt, M. E., Shadick, N. A., & Reich, D.

(2006). Principal components analysis corrects for stratification in genome-wide

association studies. Nature Genetics, 38(8), 904-909.

Quittner, A. L., Schechter, M. S., Rasouliyan, L., Haselkorn, T., Pasta, D. J., & Wagener, J. S.

(2010). Impact of socioeconomic status, race, and ethnicity on quality of life in patients

with cystic fibrosis in the United States. Chest, 137(3), 642-650.

Rasheed, N., Ahmad, A., Pandey, C. P., Chaturvedi, R. K., Lohani, M., & Palit, G. (2010).

Differential response of central dopaminergic system in acute and chronic unpredictable

stress models in rats. Neurochemical Research, 35(1), 22-32.

Reichmann, F., & Holzer, P. (2016). Neuropeptide Y: a stressful review. Neuropeptides, 55, 99-

109.

Réus, G. Z., Generoso, J. S., Rodrigues, A. L. S., & Quevedo, J. (2019). Intracellular Signaling

Pathways Implicated in the Pathophysiology of Depression. In Neurobiology of

Depression (pp. 97-109): Elsevier.

Richter-Levin, G., Stork, O., & Schmidt, M. V. (2018). Animal models of PTSD: a challenge to

be met. Molecular psychiatry, 1.

Ripke, S., Neale, B. M., Corvin, A., Walters, J. T., Farh, K.-H., Holmans, P. A., . . . Huang, H.

(2014). Biological insights from 108 schizophrenia-associated genetic loci. Nature,

511(7510), 421-427.

Rojas, P. S., & Fiedler, J. L. (2016). What do we really know about 5-HT1A receptor signaling

in neuronal cells? Frontiers in cellular neuroscience, 10, 272. Genetic Risk Factors for PTSD….. 106

Rosenberg, N. A., Pritchard, J. K., Weber, J. L., Cann, H. M., Kidd, K. K., Zhivotovsky, L. A., &

Feldman, M. W. (2002). Genetic structure of human populations. Science, 298(5602),

2381-2385.

Safran, M., Dalah, I., Alexander, J., Rosen, N., Iny Stein, T., Shmoish, M., . . . Krug, H. J. D.

(2010). GeneCards Version 3: the human gene integrator. Database,

https://www.genecards.org/cgi-bin/carddisp.pl?gene=SH3RF3.

Sah, R., & Geracioti, T. (2005). Neuropeptide Y and posttraumatic stress disorder. Molecular

psychiatry, 52(5), 546-553.

Sartor, C. E., Grant, J. D., Lynskey, M. T., McCutcheon, V. V., Waldron, M., Statham, D. J., . . .

Martin, N. G. (2012). Common heritable contributions to low-risk trauma, high-risk

trauma, posttraumatic stress disorder, and major depression. Archives of general

psychiatry, 69(3), 293-299.

Sartor, C. E., McCutcheon, V., Pommer, N., Nelson, E., Grant, J., Duncan, A., . . . Heath, A.

(2011). Common genetic and environmental contributions to post-traumatic stress

disorder and alcohol dependence in young women. Psychological medicine, 41(7), 1497-

1505.

Seyedabadi, M., Ghahremani, M. H., & Albert, P. R. (2019). Biased signaling of

coupled receptors (GPCRs): Molecular determinants of GPCR/transducer selectivity and

therapeutic potential. Pharmacology Therapeutics.

Shishkina, G. T., Kalinina, T. S., Berezova, I. V., & Dygalo, N. N. (2012). Stress-induced

activation of the brainstem Bcl-xL gene expression in rats treated with fluoxetine:

correlations with serotonin metabolism and depressive-like behavior.

Neuropharmacology, 62(1), 177-183. Genetic Risk Factors for PTSD….. 107

Siva, N. (2008). 1000 Genomes project. In: Nature Publishing Group.

Skelton, K., Ressler, K. J., Norrholm, S. D., Jovanovic, T., & Bradley-Davino, B. (2012). PTSD

and gene variants: new pathways and new thinking. Neuropharmacology, 62(2), 628-637.

Smith, A. K., Conneely, K. N., Kilaru, V., Mercer, K. B., Weiss, T. E., Bradley, B., . . . Ressler,

K. J. (2011). Differential immune system DNA methylation and cytokine regulation in

post‐traumatic stress disorder. American Journal of medical genetics part B:

neuropsychiatric genetics, 156(6), 700-708.

Smith, P. G., Morrow, R. H., & Ross, D. A. (2015). Field trials of health interventions: a

toolbox: OUP Oxford.

Smoller, J. W. (2016). The genetics of stress-related disorders: PTSD, depression, and anxiety

disorders. Neuropsychopharmacology, 41(1), 297.

Sniecikowska, J., Newman-Tancredi, A., & Kolaczkowski, M. (2019). From Receptor

Selectivity to Functional Selectivity: The Rise of Biased Agonism in 5-HT1A Receptor

Drug Discovery. Current topics in medicinal chemistry, 19(26), 2393-2420.

Spencer, D., Venkataraman, M., Higgins, S., Stevenson, K., & Weller, P. (1994). Cystic fibrosis

in children from ethnic minorities in the West Midlands. Respiratory medicine, 88(9),

671-675.

Stein, C. (2016). Opioid receptors. Annual review of medicine, 67, 433-451.

Stein, M. B., Jang, K. L., Taylor, S., Vernon, P. A., & Livesley, W. J. (2002). Genetic and

environmental influences on trauma exposure and posttraumatic stress disorder

symptoms: a twin study. American Journal of Psychiatry, 159(10), 1675-1681.

Stessman, H. A., Bernier, R., & Eichler, E. E. (2014). A genotype-first approach to defining the

subtypes of a complex disease. Cell, 156(5), 872-877. Genetic Risk Factors for PTSD….. 108

Sullivan, C., Jones, R. T., Hauenstein, N., & White, B. (2017). Development of the Trauma-

Related Anger Scale. Assessment, 1073191117711021.

Sullivan, G. M., Ogden, R. T., Huang, Y. y., Oquendo, M. A., Mann, J. J., & Parsey, R. V.

(2013). Higher in vivo serotonin‐1a binding in posttraumatic stress disorder: A PET study

with [11C] WAY‐100635. Depression and Anxiety, 30(3), 197-206.

Sullivan, G. M., Oquendo, M. A., Simpson, N., Van Heertum, R. L., Mann, J. J., & Parsey, R. V.

(2005). Brain serotonin1A receptor binding in major depression is related to psychic and

somatic anxiety. Biological psychiatry, 58(12), 947-954.

Sullivan, P. F. (2007). Spurious genetic associations. Biological psychiatry, 61(10), 1121-1126.

Sullivan, T. P., Fehon, D. C., Andres‐Hyman, R. C., Lipschitz, D. S., & Grilo, C. M. (2006).

Differential relationships of childhood abuse and neglect subtypes to PTSD symptom

clusters among adolescent inpatients. Journal of Traumatic Stress: Official Publication of

The International Society for Traumatic Stress Studies, 19(2), 229-239.

Sun, R., Hui, S., Bader, G. D., Lin, X., & Kraft, P. (2019). Powerful gene set analysis in GWAS

with the Generalized Berk-Jones statistic. PLoS genetics, 15(3), e1007530.

Sun, R., Zhang, W., Bo, J., Zhang, Z., Lei, Y., Huo, W., . . . Gu, X. (2017). Spinal activation of

alpha7-nicotinic acetylcholine receptor attenuates posttraumatic stress disorder-related

chronic pain via suppression of glial activation. Neuroscience, 344, 243-254.

Syvänen, A.-C. J. N. R. G. (2001). Accessing genetic variation: genotyping single nucleotide

polymorphisms. 2(12), 930.

Tokarski, K., Kusek, M., & Hess, G. (2011). 5HT-receptors modulate GABAergic transmission

in rat hippocampal CA1 area. Journal of Physiology and Pharmacology, 62(5), 535. Genetic Risk Factors for PTSD….. 109

Tol, W. A., Barbui, C., Bisson, J., Cohen, J., Hijazi, Z., Jones, L., . . . Seedat, S. (2014). World

Health Organization guidelines for management of acute stress, PTSD, and bereavement:

key challenges on the road ahead. PLoS medicine, 11(12), e1001769.

True, W. R., Rice, J., Eisen, S. A., Heath, A. C., Goldberg, J., Lyons, M. J., & Nowak, J. (1993).

A twin study of genetic and environmental contributions to liability for posttraumatic

stress symptoms. Archives of general psychiatry, 50(4), 257-264.

Van Praag, H., & De Haan, S. (1980). Central serotonin deficiency—a factor which increases

depression vulnerability? Journal Acta Psychiatrica Scandinavica, 61(S280), 89-96.

Wang, K., Li, M., & Hakonarson, H. (2010). Analysing biological pathways in genome-wide

association studies. Nature Reviews Genetics, 11(12), 843.

Weathers, F. W., Litz, B. T., Herman, D. S., Huska, J. A., & Keane, T. M. (1993). The PTSD

Checklist (PCL): Reliability, validity, and diagnostic utility. Paper presented at the annual

convention of the international society for traumatic stress studies, San Antonio, TX.

Wei, P., Tang, H., & Li, D. (2012). Insights into pancreatic cancer etiology from pathway

analysis of genome-wide association study data. PloS one, 7(10), e46887.

Wen, H.-J., Walsh, M. P., Yan, I. K., Takahashi, K., Fields, A., & Patel, T. (2018). Functional

Modulation of Gene Expression by Ultraconserved Long Non-coding RNA TUC338

during Growth of Human Hepatocellular Carcinoma. iScience, 2, 210-220.

Willer, C. J., Li, Y., & Abecasis, G. R. (2010). METAL: fast and efficient meta-analysis of

genomewide association scans. Bioinformatics, 26(17), 2190-2191.

Wittke-Thompson, J. K., Pluzhnikov, A., & Cox, N. J. (2005). Rational inferences about

departures from Hardy-Weinberg equilibrium. The American Journal of Human

Genetics, 76(6), 967-986. Genetic Risk Factors for PTSD….. 110

Xian, H., Chantarujikapong, S. I., Scherrer, J. F., Eisen, S. A., Lyons, M. J., Goldberg, J., . . .

True, W. R. (2000). Genetic and environmental influences on posttraumatic stress

disorder, alcohol and drug dependence in twin pairs. Drug alchohol dependence, 61(1),

95-102.

Xie, P., Kranzler, H. R., Yang, C., Zhao, H., Farrer, L. A., & Gelernter, J. (2013). Genome-wide

association study identifies new susceptibility loci for posttraumatic stress disorder.

Biological psychiatry, 74(9), 656-663.

Yamamoto, S., Morinobu, S., Takei, S., Fuchikami, M., Matsuki, A., Yamawaki, S., & Liberzon,

I. (2009). Single prolonged stress: toward an animal model of posttraumatic stress

disorder. Depression and Anxiety, 26(12), 1110-1117.

Yoon, S. (2018). GSA-SNP2 User's Manual.

Yoon, S., Nguyen, H. C. T., Yoo, Y. J., Kim, J., Baik, B., Kim, S., . . . Nam, D. (2018a).

Efficient pathway enrichment and network analysis of GWAS summary data using GSA-

SNP2. Nucleic Acids Research, 46(10), e60-e60.

Yoon, S., Nguyen, H. C. T., Yoo, Y. J., Kim, J., Baik, B., Kim, S., . . . Nam, D. J. N. a. r.

(2018b). Efficient pathway enrichment and network analysis of GWAS summary data

using GSA-SNP2. 46(10), e60-e60.

Yudell, M., Roberts, D., DeSalle, R., & Tishkoff, S. (2016). Taking race out of human genetics.

Science, 351(6273), 564-565.

Zhao, H., Mitra, N., Kanetsky, P. A., Nathanson, K. L., & Rebbeck, T. R. (2018). A practical

approach to adjusting for population stratification in genome-wide association studies:

principal components and propensity scores (PCAPS). Statistical applications in genetics

molecular biology, 17(6). Genetic Risk Factors for PTSD….. 111

Zhu, M., & Zhao, S. J. I. j. o. b. s. (2007). Candidate gene identification approach: progress and

challenges. 3(7), 420.

Żmudzka, E., Sałaciak, K., Sapa, J., & Pytka, K. (2018). Serotonin receptors in depression and

anxiety: Insights from animal studies. Life Sciences.

Zoellner, L. A., Pruitt, L. D., Farach, F. J., Jun, J. J. J. D., & anxiety. (2014). Understanding

heterogeneity in PTSD: fear, dysphoria, and distress. Depression and Anxiety, 31(2), 97-

106.

Zoladz, P. R., & Diamond, D. (2016). Psychosocial predator stress model of PTSD based on

clinically relevant risk factors for trauma-induced psychopathology. Posttraumatic Stress

Disorder: From Neurobiology to Treatment, 125, 125-143.

Zoladz, P. R., & Diamond, D. M. (2013). Current status on behavioral and biological markers of

PTSD: a search for clarity in a conflicting literature. Neuroscience and biobehavioral

reviews, 37(5), 860-895.

Genetic Risk Factors for PTSD….. 112

Table 1 – Results

Detected Gene-Set Type Size Genes Z-score p-value Q-value Serotonin_5ht2B R 25 24 2.3601 0.0091351 0.319729 Serotonin_5ht2C R 25 23 2.29544 0.010854 0.319729 Glutamate_mGluR5 R 39 39 2.22041 0.0131954 0.319729 Glutamate_mGluR1 R 39 39 1.97689 0.0240272 0.319729 Serotonin_5ht2A R 26 25 1.92692 0.0269947 0.319729 GABA NT 21 18 1.73509 0.0413628 0.319729 Glutamate NT 24 24 1.70639 0.0439675 0.319729 Glutamate_NMDA R 20 20 1.3551 0.0876935 0.383659 Acetylcholine_nAChR R 47 45 1.1689 0.121222 0.471417 Dopamine_D1 R 57 55 0.858285 0.195368 0.683787 GABA_GABAB R 39 38 0.81539 0.207425 0.683787 Dopamine_D5 R 32 32 0.78713 0.215603 0.683787 GABA_GABAA R 17 14 0.535156 0.296271 0.797653 Serotonin_5ht7 R 13 13 0.491124 0.311669 0.797653 Acetylcholine NT 14 13 0.478441 0.316168 0.797653 Serotonin_5ht4 R 13 13 0.410958 0.340552 0.797653 Serotonin_5ht6 R 13 13 0.400976 0.344219 0.797653 Dopamine_D2 R 57 56 0.038178 0.484773 0.942614 Acetylcholine_M4 R 54 54 -0.213056 0.584358 1 Acetylcholine_M2 R 54 54 -0.431854 0.667076 1 Dopamine_D3 R 26 26 -0.441106 0.670432 1 Dopamine_D4 R 26 26 -0.453774 0.675004 1 Serotonin_5ht1F R 54 53 -0.508179 0.694336 1 Glutamate_mGluR6 R 17 17 -0.58003 0.719053 1 Glutamate_mGluR4 R 40 38 -0.620253 0.732454 1 Serotonin_5ht1A R 54 53 -0.761094 0.776699 1 Serotonin_5ht5A R 54 53 -0.773389 0.780354 1 Serotonin_5ht1E R 54 53 -0.804708 0.789506 1 Serotonin_5ht1B R 54 53 -0.809067 0.790762 1 Glutamate_mGluR8 R 39 38 -0.867762 0.807238 1 Glutamate_mGluR3 R 41 38 -0.920916 0.821453 1 Glutamate_mGluR7 R 39 38 -0.954109 0.829986 1 Serotonin_5ht1D R 54 53 -0.957307 0.830794 1 Glutamate_mGluR2 R 41 38 -1.06011 0.855453 1 Serotonin NT 17 16 -1.53436 0.937529 1

Genetic Risk Factors for PTSD….. 113

Table 2 - Post-Hoc Results

Detected Gene-Set Type Size Genes Z-score p-value Q-value Serotonin2_Intra Intra(Post) 24 23 2.29544 0.010854 0.227933 Glutamate_mGluR1_5_Intra Intra(Post) 38 38 1.93987 0.026198 0.275078 GABA NT 21 18 1.73509 0.041363 0.28954 Glutamate NT 25 24 1.70639 0.043968 0.28954 Glutamate_NMDA_Intra Intra(Post) 13 13 1.4602 0.072117 0.302891 Acetylcholine_nAChR_Intra Intra(Post) 41 39 1.116 0.132212 0.462742 Dopamine_D1_Intra Intra(Post) 56 54 0.906016 0.182464 0.547391 GABA_GABAB_Intra Intra(Post) 39 38 0.81539 0.207425 0.547391 Dopamine_D5_Intra Intra(Post) 31 31 0.746897 0.227563 0.547391 Acetylcholine NT 14 13 0.478441 0.316168 0.663953 Serotonin4_6_7_Intra Intra(Post) 12 12 0.471537 0.318629 0.663953 Dopamine_D2_Intra Intra(Post) 55 54 -0.00667 0.502662 0.879659 Acetylcholine_M2_M4_Presynaptic Intra(Pre) 23 23 -0.15705 0.562398 0.90849 Dopamine_D3_D4_Intra Intra(Post) 25 25 -0.29354 0.615445 0.923167 Acetylcholine_M2_M4_Intra Intra(Post) 51 51 -0.44755 0.672762 0.941867 Serotonin1_5a_Intra Intra(Post) 53 52 -0.71321 0.762141 1 Glutamate_mGluR7_8_PreSynaptic Intra(Pre) 38 37 -0.92619 0.822825 1 Glutamate_mGluR2_3_4_PreSynaptic Intra(Pre) 40 37 -0.92619 0.822825 1 Glutamate_mGluR2_3_4_6_7_8 Intra(Post) 17 16 -1.0666 0.856924 1 Serotonin NT 17 16 -1.53436 0.937529 1 Missing_Receptors Missing 25 20 -1.96954 0.975554 1

Genetic Risk Factors for PTSD….. 114

Figure 1 - Planned Gene-sets

Glutamate GRIK1 GRIK2 GRIK3 GRIK4 GRIK5 GRIA1 GRIA2 GRIA3 GRIA4 GRIN3A GRIN3B GRIN1 GRIN2A GRIN2B GRIN2C GRIN2D GRM1 GRM5 GRM2 GRM3 GRM4 GRM6 GRM7 GRM8

GABA GABRA1 GABRA2 GABRA3 GABRA4 GABRA5 GABRA6 GABRB1 GABRB2 GABRB3 GABRD GABRE GABRG1 GABRG2 GABRG3 GABRP GABRQ GABRR1 GABRR2 GABRR3 GABBR1 GABBR2

Acetylcholine CHRM5 CHRM3 CHRM1 CHRM2 CHRM4 CHRNA3 CHRNA4 CHRNA7 CHRNB2 CHRNB4 CHRNA6 CHRM1 CHMR3 CHRM5

Dopamine DRD1 DRD2 DRD3 DRD4 DRD5

Serotonin HTR2A HTR2B HTR2C HTR3A HTR3B HTR3C HTR3D HTR3E HTR4 HTR6 HTR7 HTR1A HTR1B HTR1D HTR1E HTR1F HTR5A

Adrenergic ADRA1A ADRA1B ADRA1D ADRA2A ADRA2B ADRA2C ADRB1 ADRB2 ADRB3

NPY NPY1R NPY2R PPYR1 NPY5R

Opioid OPRD1 OPRK1 OPRM1 OPRL1 OGFR

Cannabinoid CNR1 CNR2

Glutamate_KA GRIK1 GRIK2 GRIK3 GRIK4 GRIK5

Glutamate_AMPA GRIA1 GRIA2 GRIA3 GRIA4

Glutamate_NMDA GRIN3A GRIN3B GRIN1 GRIN2A GRIN2B GRIN2C GRIN2D PPP3CA PPP3CB PPP3CC PPP3R1 PPP3R2 DLG4 DLGAP1 SHANK1 SHANK2 SHANK3 HOMER1 HOMER2 HOMER3

Glutamate_mGluR1 GRM1 GNAS ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNAQ PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG PLA2G4A PLA2G4B PLA2G4C PLA2G4D PLA2G4E PLA2G4F JMJD7-PLA2G4B PLD1 PLD2 MAPK1 MAPK3 HOMER1 HOMER2 HOMER3 ITPR1 ITPR2 ITPR3

Glutamate_mGluR5 GRM5 GNAS ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNAQ PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG PLA2G4A PLA2G4B PLA2G4C PLA2G4D PLA2G4E PLA2G4F JMJD7-PLA2G4B PLD1 PLD2 MAPK1 MAPK3 HOMER1 HOMER2 HOMER3 ITPR1 ITPR2 ITPR3

Glutamate_mGluR2 GRM2 GNAI1 GNAI2 GNAI3 GNAO1 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNB1 GNB2 GNB3 GNB4 Genetic Risk Factors for PTSD….. 115

GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 CACNA1A GRK2 GRK3

Glutamate_mGluR3 GRM3 GNAI1 GNAI2 GNAI3 GNAO1 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 CACNA1A GRK2 GRK3

Glutamate_mGluR4 GRM4 GNAI1 GNAI2 GNAI3 GNAO1 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 CACNA1A

Glutamate_mGluR6 GRM6 GNAI1 GNAI2 GNAI3 GNAO1 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG

Glutamate_mGluR7 GRM7 GNAI1 GNAI2 GNAI3 GNAO1 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 CACNA1A

Glutamate_mGluR8 GRM8 GNAI1 GNAI2 GNAI3 GNAO1 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 CACNA1A

Glutamate_KA GRIK1 GRIK2 GRIK3 GRIK4 GRIK5 CACNA1A

GABA_GABAA GABRA1 GABRA2 GABRA3 GABRA4 GABRA5 GABRA6 GABRB1 GABRB2 GABRB3 GABRD GABRE GABRG1 GABRG2 GABRG3 GABRP GABRQ GPHN

GABA_GABAB GABBR1 GABBR2 GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 CACNA1A CACNA1B CACNA1C CACNA1D CACNA1F CACNA1S KCNJ6

GABA_GABAC GABRR1 GABRR2 GABRR3

Acetylcholine_M5 CHRM5 GNA11 GNAQ PLCB1 PLCB2 PLCB3 PLCB4

Acetylcholine_M3 CHRM3 GNA11 GNAQ PLCB1 PLCB2 PLCB3 PLCB4

Acetylcholine_M1 CHRM1 GNA11 GNAQ PLCB1 PLCB2 PLCB3 PLCB4

Acetylcholine_M2 CHRM2 GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PIK3R6 PIK3R5 PIK3CG HRAS KRAS NRAS MAP2K1 MAPK1 MAPK3 FOS ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG Genetic Risk Factors for PTSD….. 116

CREB3 CREB1 CREB3L4 ATF4 CREB3L2 CREB3L3 CREB3L1 CREB5 CACNA1A CACNA1B

Acetylcholine_M4 CHRM4 GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PIK3R6 PIK3R5 PIK3CG HRAS KRAS NRAS MAP2K1 MAPK1 MAPK3 FOS ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG CREB3 CREB1 CREB3L4 ATF4 CREB3L2 CREB3L3 CREB3L1 CREB5 CACNA1A CACNA1B

Acetylcholine_nAChR CHRNA3 CHRNA4 CHRNA7 CHRNB2 CHRNB4 CHRNA6 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG CREB3 CREB1 CREB3L4 ATF4 CREB3L2 CREB3L3 CREB3L1 CREB5 CAMK4 CAMK2A CAMK2B CAMK2D CAMK2G FYN JAK2 PIK3CA PIK3CB PIK3CD PIK3R1 PIK3R2 PIKR3 AKT3 AKT1 AKT2 BCL2 CACNA1C CACNA1D CACNA1F CACNA1S

Dopamine_D1 DRD1 CALY GNAQ PLCB1 PLCB2 PLCB3 PLCB4 ITPR1 ITPR2 ITPR3 CALML1 CALML2 CALML3 CALML4 CALML5 CALML6 CAMK2A CAMK2B CAMK2D CAMK2G PPP3CA PPP3CB PPP3CC PRKCA PRKCB PRKCG GNAL GNAS ADCY5 PRKACA PRKACB PRKACG CREB3 CREB1 ATF2 ATF6B CREB3L4 ATF4 CREB3L2 CREB3L3 CREB3L1 CREB5 MAPK14 MAPK8 MAPK11 MAPK9 MAPK10 MAPK13 MAPK12 FOS PPP1R1B PPP1CA PPP1CB PPP1CC SCN1A CACNA1C CACNA1D

Dopamine_D5 DRD5 GNAL GNAS ADCY5 PRKACA PRKACB PRKACG CREB3 CREB1 ATF2 ATF6B CREB3L4 ATF4 CREB3L2 CREB3L3 CREB3L1 CREB5 MAPK14 MAPK8 MAPK11 MAPK9 MAPK10 MAPK13 MAPK12 FOS PPP1R1B PPP1CA PPP1CB PPP1CC SCN1A CACNA1C CACNA1D

Dopamine_D3 DRD3 GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 KCNJ5 KCNJ6 KCNJ9

Dopamine_D4 DRD4 GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 KCNJ5 KCNJ6 KCNJ9

Dopamine_D2 DRD2 GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 KCNJ5 KCNJ6 KCNJ9 ARRB1 ARRB2 PPP2R3B PPP2R3C PPP2CA PPP2CB PPP2R1A PPP2R1B PPP2R2A PPP2R2B PPP2R2C PPP2R3A PPP2R5A PPP2R5B PPP2R5C PPP2R5D PPP2R5E PPP2R2D AKT3 AKT1 AKT2 GSK3A GSK3B ARNTL CLOCK KIF5A KIF5B KIF5C GRIN2A GRIN2B TH

Serotonin_5ht2A HTR2A CACNA1C CACNA1D CACNA1F CACNA1S GNAQ PLCB1 PLCB2 PLCB3 PLCB4 ITPR1 ITPR2 ITPR3 PRKCA PRKCB PRKCG MAPK1 MAPK3 Genetic Risk Factors for PTSD….. 117

PLA2G4B PLA2G4E PLA2G4F PLA2G4D PLA2G4A PLA2G4C JMJD7-PLA2G4B TRPC1

Serotonin_5ht2B HTR2B CACNA1C CACNA1D CACNA1F CACNA1S GNAQ PLCB1 PLCB2 PLCB3 PLCB4 ITPR1 ITPR2 ITPR3 PRKCA PRKCB PRKCG MAPK1 MAPK3 PLA2G4B PLA2G4E PLA2G4F PLA2G4D PLA2G4A PLA2G4C JMJD7-PLA2G4B

Serotonin_5ht2C HTR2C CACNA1C CACNA1D CACNA1F CACNA1S GNAQ PLCB1 PLCB2 PLCB3 PLCB4 ITPR1 ITPR2 ITPR3 PRKCA PRKCB PRKCG MAPK1 MAPK3 PLA2G4B PLA2G4E PLA2G4F PLA2G4D PLA2G4A PLA2G4C JMJD7-PLA2G4B

Serotonin_5ht3A HTR3A

Serotonin_5ht3B HTR3B

Serotonin_5ht3C HTR3C

Serotonin_5ht3D HTR3D

Serotonin_5ht3E HTR3E

Serotonin_5ht4 HTR4 GNAS ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Serotonin_5ht6 HTR6 GNAS ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Serotonin_5ht7 HTR7 GNAS ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Serotonin_5ht1A HTR1A GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG CASP3 ITPR1 ITPR2 ITPR3 DUSP1 MAPK1 MAPK3 HRAS KRAS NRAS ARAF RAF1 BRAF MAP2K1 ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Serotonin_5ht1B HTR1B GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG CASP3 ITPR1 ITPR2 ITPR3 DUSP1 MAPK1 MAPK3 HRAS KRAS NRAS ARAF RAF1 BRAF MAP2K1 ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Serotonin_5ht1D HTR1D GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG CASP3 ITPR1 ITPR2 ITPR3 DUSP1 MAPK1 MAPK3 HRAS KRAS NRAS ARAF RAF1 BRAF MAP2K1 ADCY5 PRKACA Genetic Risk Factors for PTSD….. 118

PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Serotonin_5ht1E HTR1E GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG CASP3 ITPR1 ITPR2 ITPR3 DUSP1 MAPK1 MAPK3 HRAS KRAS NRAS ARAF RAF1 BRAF MAP2K1 ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Serotonin_5ht1F HTR1F GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG CASP3 ITPR1 ITPR2 ITPR3 DUSP1 MAPK1 MAPK3 HRAS KRAS NRAS ARAF RAF1 BRAF MAP2K1 ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Serotonin_5ht5A HTR5A GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG CASP3 ITPR1 ITPR2 ITPR3 DUSP1 MAPK1 MAPK3 HRAS KRAS NRAS ARAF RAF1 BRAF MAP2K1 ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Genetic Risk Factors for PTSD….. 119

Figure 2 - Post-hoc Gene-sets

Glutamate GRIK1 GRIK2 GRIK3 GRIK4 GRIK5 GRIA1 GRIA2 GRIA3 GRIA4 GRIN3A GRIN3B GRIN1 GRIN2A GRIN2B GRIN2C GRIN2D TRPC1 GRM1 GRM5 GRM2 GRM3 GRM4 GRM6 GRM7 GRM8

GABA GABRA1 GABRA2 GABRA3 GABRA4 GABRA5 GABRA6 GABRB1 GABRB2 GABRB3 GABRD GABRE GABRG1 GABRG2 GABRG3 GABRP GABRQ GABRR1 GABRR2 GABRR3 GABBR1 GABBR2

Acetylcholine CHRM5 CHRM3 CHRM1 CHRM2 CHRM4 CHRNA3 CHRNA4 CHRNA7 CHRNB2 CHRNB4 CHRNA6 CHRM1 CHMR3 CHRM5

Serotonin HTR2A HTR2B HTR2C HTR3A HTR3B HTR3C HTR3D HTR3E HTR4 HTR6 HTR7 HTR1A HTR1B HTR1D HTR1E HTR1F HTR5A

Missing_Receptors ADRA1A ADRA1B ADRA1D ADRA2A ADRA2B ADRA2C ADRB1 ADRB2 ADRB3 NPY1R NPY2R PPYR1 NPY5R OPRD1 OPRK1 OPRM1 OPRL1 OGFR CNR1 CNR2

Serotonin2_Intra CACNA1C CACNA1D CACNA1F CACNA1S GNAQ PLCB1 PLCB2 PLCB3 PLCB4 ITPR1 ITPR2 ITPR3 PRKCA PRKCB PRKCG MAPK1 MAPK3 PLA2G4B PLA2G4E PLA2G4F PLA2G4D PLA2G4A PLA2G4C JMJD7-PLA2G4B

Serotonin4_6_7_Intra GNAS ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Serotonin1_5a_Intra GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG CASP3 ITPR1 ITPR2 ITPR3 DUSP1 MAPK1 MAPK3 HRAS KRAS NRAS ARAF RAF1 BRAF MAP2K1 ADCY5 PRKACA PRKACB PRKACG KCND2 KCNN2 GABRB1 GABRB2 GABRB3 RAPGEF3 APP

Glutamate_NMDA_Intra PPP3CA PPP3CB PPP3CC PPP3R1 PPP3R2 DLG4 DLGAP1 SHANK1 SHANK2 SHANK3 HOMER1 HOMER2 HOMER3

Glutamate_mGluR1_5_Intra GNAS ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNAQ PLCB1 PLCB2 PLCB3 PLCB4 PRKCA PRKCB PRKCG PLA2G4A PLA2G4B PLA2G4C PLA2G4D PLA2G4E PLA2G4F JMJD7-PLA2G4B PLD1 PLD2 MAPK1 MAPK3 HOMER1 HOMER2 HOMER3 ITPR1 ITPR2 ITPR3

Glutamate_mGluR2_3_4_6_7_8_PostSynaptic GNAI1 GNAI2 GNAI3 GNAO1 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG Genetic Risk Factors for PTSD….. 120

Glutamate_mGluR2_3_4_PreSynaptic GNAI1 GNAI2 GNAI3 GNAO1 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 CACNA1A GRK2 GRK3

Glutamate_mGluR7_8_PreSynaptic GNAI1 GNAI2 GNAI3 GNAO1 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 CACNA1A

GABA_GABAB_Intra GABBR1 GABBR2 GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 CACNA1A CACNA1B CACNA1C CACNA1D CACNA1F CACNA1S KCNJ6

Acetylcholine_M2_M4_Intra GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 PIK3R6 PIK3R5 PIK3CG HRAS KRAS NRAS MAP2K1 MAPK1 MAPK3 FOS ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG CREB3 CREB1 CREB3L4 ATF4 CREB3L2 CREB3L3 CREB3L1 CREB5

Acetylcholine_M2_M4_Auto GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 CACNA1A CACNA1B

Acetylcholine_nAChR_Intra ADCY1 ADCY2 ADCY3 ADCY4 ADCY5 ADCY6 ADCY7 ADCY8 ADCY9 PRKACA PRKACB PRKACG CREB3 CREB1 CREB3L4 ATF4 CREB3L2 CREB3L3 CREB3L1 CREB5 CAMK4 CAMK2A CAMK2B CAMK2D CAMK2G FYN JAK2 PIK3CA PIK3CB PIK3CD PIK3R1 PIK3R2 PIKR3 AKT3 AKT1 AKT2 BCL2 CACNA1C CACNA1D CACNA1F CACNA1S

Dopamine_D1_Intra CALY GNAQ PLCB1 PLCB2 PLCB3 PLCB4 ITPR1 ITPR2 ITPR3 CALML1 CALML2 CALML3 CALML4 CALML5 CALML6 CAMK2A CAMK2B CAMK2D CAMK2G PPP3CA PPP3CB PPP3CC PRKCA PRKCB PRKCG GNAL GNAS ADCY5 PRKACA PRKACB PRKACG CREB3 CREB1 ATF2 ATF6B CREB3L4 ATF4 CREB3L2 CREB3L3 CREB3L1 CREB5 MAPK14 MAPK8 MAPK11 MAPK9 MAPK10 MAPK13 MAPK12 FOS PPP1R1B PPP1CA PPP1CB PPP1CC SCN1A CACNA1C CACNA1D

Dopamine_D5_Intra GNAL GNAS ADCY5 PRKACA PRKACB PRKACG CREB3 CREB1 ATF2 ATF6B CREB3L4 ATF4 CREB3L2 CREB3L3 CREB3L1 CREB5 MAPK14 MAPK8 MAPK11 MAPK9 MAPK10 MAPK13 MAPK12 FOS PPP1R1B PPP1CA PPP1CB PPP1CC SCN1A CACNA1C CACNA1D

Dopamine_D3_D4_Intra GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 KCNJ5 KCNJ6 KCNJ9 Genetic Risk Factors for PTSD….. 121

Dopamine_D2_Intra GNB1 GNB2 GNB3 GNB4 GNB5 GNAI1 GNAI2 GNAI3 GNAO1 GNG3 GNG4 GNG5 GNG7 GNG10 GNG11 GNGT1 GNGT2 GNG13 GNG2 GNG12 GNG8 KCNJ3 KCNJ5 KCNJ6 KCNJ9 ARRB1 ARRB2 PPP2R3B PPP2R3C PPP2CA PPP2CB PPP2R1A PPP2R1B PPP2R2A PPP2R2B PPP2R2C PPP2R3A PPP2R5A PPP2R5B PPP2R5C PPP2R5D PPP2R5E PPP2R2D AKT3 AKT1 AKT2 GSK3A GSK3B ARNTL CLOCK KIF5A KIF5B KIF5C GRIN2A GRIN2B

Genetic Risk Factors for PTSD….. 122

Supplementary Figure S1 - Gene Scores for Planned Gene-sets

Serotonin_5ht2B PLA2G4F(2.10831); PLA2G4D(2.00666); PLA2G4E(1.94461); PRKCB(1.12043); GNAQ(0.856944); PLCB1(0.657282); PRKCA(0.530621); PLA2G4B(0.483477); JMJD7-PLA2G4B(0.464105); PLA2G4C(0.440997); HTR2B(0.403711); CACNA1C(0.356703); CACNA1D(0.296217); PRKCG(0.0786838); PLCB4(-0.0532556); ITPR2(-0.0756829); ITPR1(-0.10489); PLA2G4A(-0.136318); MAPK3(-0.182594); ITPR3(-0.239205); CACNA1S(-0.357408); MAPK1(-0.515604); PLCB3(-0.622332); PLCB2(-0.796177);

Serotonin_5ht2C PLA2G4F(2.10831); PLA2G4D(2.00666); PLA2G4E(1.94461); PRKCB(1.12043); GNAQ(0.856944); PLCB1(0.657282); PRKCA(0.530621); PLA2G4B(0.483477); JMJD7-PLA2G4B(0.464105); PLA2G4C(0.440997); CACNA1C(0.356703); CACNA1D(0.296217); PRKCG(0.0786838); PLCB4(-0.0532556); ITPR2(-0.0756829); ITPR1(-0.10489); PLA2G4A(-0.136318); MAPK3(-0.182594); ITPR3(-0.239205); CACNA1S(-0.357408); MAPK1(-0.515604); PLCB3(-0.622332); PLCB2(-0.796177);

Glutamate_mGluR5 PLA2G4F(2.10831); PLA2G4D(2.00666); PLA2G4E(1.94461); ADCY7(1.458); GRM5(1.21134); PLD2(1.14245); PRKCB(1.12043); ADCY1(0.954473); GNAQ(0.856944); GNAS(0.844842); PLCB1(0.657282); PRKCA(0.530621); PLA2G4B(0.483477); JMJD7-PLA2G4B(0.464105); PLA2G4C(0.440997); ADCY3(0.225875); HOMER3(0.187757); ADCY9(0.132965); ADCY2(0.107415); PRKCG(0.0786838); ADCY6(-0.0204362); PLCB4(-0.0532556); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); PLD1(-0.135981); PLA2G4A(-0.136318); MAPK3(-0.182594); HOMER2(-0.207691); PRKACA(-0.225821); ITPR3(-0.239205); ADCY4(-0.302726); HOMER1(-0.343684); ADCY8(-0.460334); MAPK1(-0.515604); PRKACB(-0.593236); PLCB3(-0.622332); PRKACG(-0.669971); PLCB2(-0.796177);

Glutamate_mGluR1 PLA2G4F(2.10831); PLA2G4D(2.00666); PLA2G4E(1.94461); ADCY7(1.458); PLD2(1.14245); PRKCB(1.12043); ADCY1(0.954473); GNAQ(0.856944); GNAS(0.844842); PLCB1(0.657282); PRKCA(0.530621); PLA2G4B(0.483477); JMJD7-PLA2G4B(0.464105); PLA2G4C(0.440997); GRM1(0.304609); ADCY3(0.225875); HOMER3(0.187757); ADCY9(0.132965); ADCY2(0.107415); PRKCG(0.0786838); ADCY6(-0.0204362); PLCB4(-0.0532556); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); PLD1(-0.135981); PLA2G4A(-0.136318); MAPK3(-0.182594); HOMER2(-0.207691); PRKACA(-0.225821); ITPR3(-0.239205); ADCY4(-0.302726); HOMER1(-0.343684); ADCY8(-0.460334); MAPK1(-0.515604); PRKACB(-0.593236); PLCB3(-0.622332); PRKACG(-0.669971); PLCB2(-0.796177);

Serotonin_5ht2A PLA2G4F(2.10831); PLA2G4D(2.00666); PLA2G4E(1.94461); PRKCB(1.12043); GNAQ(0.856944); PLCB1(0.657282); PRKCA(0.530621); PLA2G4B(0.483477); JMJD7-PLA2G4B(0.464105); PLA2G4C(0.440997); CACNA1C(0.356703); CACNA1D(0.296217); PRKCG(0.0786838); PLCB4(-0.0532556); ITPR2(-0.0756829); ITPR1(-0.10489); PLA2G4A(-0.136318); MAPK3(-0.182594); HTR2A(-0.19768); ITPR3(- 0.239205); CACNA1S(-0.357408); TRPC1(-0.474553); MAPK1(-0.515604); PLCB3(-0.622332); PLCB2(- 0.796177);

GABA GABBR1(1.02476); GABRR2(0.984362); GABBR2(0.890302); GABRB2(0.776778); GABRA4(0.764706); GABRR3(0.59692); GABRA2(0.536947); GABRG2(0.460323); GABRG3(0.35292); GABRG1(0.270733); GABRB3(0.185329); GABRA5(0.163624); GABRB1(0.0429372); GABRA1(-0.0266599); GABRP(-0.0508056); GABRA6(-0.231377); GABRR1(-0.401135); GABRD(-0.622532);

Glutamate GRM5(1.21134); GRM6(1.19216); GRIN1(1.1748); GRM4(1.15314); GRIK5(0.915416); GRIA4(0.660596); GRIK1(0.624287); GRIN3B(0.623253); GRIA1(0.383217); GRIN2D(0.331225); Genetic Risk Factors for PTSD….. 123

GRM1(0.304609); GRIK4(0.244594); GRM8(0.243446); GRIK3(0.135137); GRM3(0.0480824); GRIN2B(0.0397144); GRIK2(0.0396233); GRIN2C(-0.0732666); GRM7(-0.0739135); GRIA2(-0.378383); GRM2(-0.463511); TRPC1(-0.474553); GRIN2A(-0.496191); GRIN3A(-0.609511);

Glutamate_NMDA PPP3CB(2.13408); GRIN1(1.1748); PPP3R1(0.786684); GRIN3B(0.623253); SHANK2(0.530936); DLG4(0.419713); SHANK3(0.336771); GRIN2D(0.331225); SHANK1(0.300812); DLGAP1(0.216453); HOMER3(0.187757); GRIN2B(0.0397144); PPP3CA(0.026846); PPP3CC(-0.0304729); GRIN2C(-0.0732666); HOMER2(-0.207691); PPP3R2(-0.258932); HOMER1(-0.343684); GRIN2A(- 0.496191); GRIN3A(-0.609511);

Acetylcholine_nAChR ADCY7(1.458); AKT3(1.367); CAMK2D(1.21548); AKT2(1.1076); ADCY1(0.954473); CAMK2B(0.672307); CREB3L1(0.663382); CHRNB4(0.644963); CHRNA3(0.618276); PIK3R2(0.590356); CHRNA4(0.568833); CREB5(0.481566); AKT1(0.469481); CACNA1C(0.356703); BCL2(0.301821); CACNA1D(0.296217); CREB3L4(0.292783); CAMK2G(0.268395); CAMK2A(0.231646); ADCY3(0.225875); CREB3L3(0.177996); ATF4(0.144112); CREB1(0.139283); ADCY9(0.132965); CREB3(0.108695); ADCY2(0.107415); FYN(0.0513743); CHRNA6(0.0447898); ADCY6(-0.0204362); PIK3CB(-0.0278242); PIK3CD(-0.0545999); CAMK4(-0.0607568); ADCY5(-0.128395); PRKACA(-0.225821); CHRNB2(-0.277029); ADCY4(-0.302726); JAK2(-0.345736); CACNA1S(-0.357408); CREB3L2(-0.392186); PIK3CA(-0.402222); ADCY8(-0.460334); PRKACB(-0.593236); CHRNA7(-0.638258); PRKACG(-0.669971); PIK3R1(-0.742498);

Dopamine_D1 PPP3CB(2.13408); CAMK2D(1.21548); PRKCB(1.12043); MAPK13(0.927615); GNAQ(0.856944); MAPK8(0.847746); GNAS(0.844842); MAPK14(0.753329); FOS(0.744517); CAMK2B(0.672307); CREB3L1(0.663382); PLCB1(0.657282); ATF2(0.556153); PRKCA(0.530621); MAPK9(0.521339); CREB5(0.481566); MAPK11(0.417754); MAPK12(0.396692); CACNA1C(0.356703); CACNA1D(0.296217); CREB3L4(0.292783); CAMK2G(0.268395); CAMK2A(0.231646); SCN1A(0.228832); ATF6B(0.184584); CREB3L3(0.177996); GNAL(0.176089); ATF4(0.144112); CREB1(0.139283); CREB3(0.108695); PRKCG(0.0786838); PPP3CA(0.026846); PPP3CC(-0.0304729); PLCB4(-0.0532556); ITPR2(-0.0756829); DRD1(-0.100758); ITPR1(-0.10489); ADCY5(-0.128395); PRKACA(-0.225821); CALML4(-0.228977); ITPR3(-0.239205); MAPK10(-0.261223); PPP1CB(-0.288708); CALY(-0.301501); CREB3L2(-0.392186); PPP1R1B(-0.476736); PRKACB(-0.593236); PLCB3(-0.622332); PRKACG(-0.669971); PPP1CC(-0.672875); CALML3(-0.698534); CALML6(-0.720422); CALML5(-0.737076); PPP1CA(-0.785583); PLCB2(-0.796177);

GABA_GABAB GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); GABBR1(1.02476); ADCY1(0.954473); GABBR2(0.890302); KCNJ6(0.374662); GNGT1(0.37302); CACNA1C(0.356703); CACNA1A(0.332833); CACNA1D(0.296217); GNG10(0.269162); ADCY3(0.225875); CACNA1B(0.179502); GNG11(0.168354); ADCY9(0.132965); GNB5(0.128179); ADCY2(0.107415); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); GNB1(-0.146567); GNGT2(-0.190679); ADCY4(-0.302726); CACNA1S(-0.357408); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); GNAI3(-0.684467); GNAI2(-0.870932);

Dopamine_D5 MAPK13(0.927615); MAPK8(0.847746); GNAS(0.844842); MAPK14(0.753329); FOS(0.744517); CREB3L1(0.663382); ATF2(0.556153); MAPK9(0.521339); CREB5(0.481566); MAPK11(0.417754); MAPK12(0.396692); CACNA1C(0.356703); CACNA1D(0.296217); CREB3L4(0.292783); DRD5(0.24907); SCN1A(0.228832); ATF6B(0.184584); CREB3L3(0.177996); Genetic Risk Factors for PTSD….. 124

GNAL(0.176089); ATF4(0.144112); CREB1(0.139283); CREB3(0.108695); ADCY5(-0.128395); PRKACA(- 0.225821); MAPK10(-0.261223); PPP1CB(-0.288708); CREB3L2(-0.392186); PPP1R1B(-0.476736); PRKACB(-0.593236); PRKACG(-0.669971); PPP1CC(-0.672875); PPP1CA(-0.785583);

GABA_GABAA GABRB2(0.776778); GABRA4(0.764706); GABRA2(0.536947); GABRG2(0.460323); GABRG3(0.35292); GABRG1(0.270733); GABRB3(0.185329); GABRA5(0.163624); GABRB1(0.0429372); GABRA1(-0.0266599); GABRP(-0.0508056); GABRA6(-0.231377); GPHN(-0.396183); GABRD(-0.622532);

Serotonin_5ht7 GNAS(0.844842); APP(0.799182); GABRB2(0.776778); KCND2(0.438621); RAPGEF3(0.427181); GABRB3(0.185329); HTR7(0.15563); GABRB1(0.0429372); KCNN2(-0.0382039); ADCY5(-0.128395); PRKACA(-0.225821); PRKACB(-0.593236); PRKACG(-0.669971);

Acetylcholine CHRM4(0.680809); CHRNB4(0.644963); CHRNA3(0.618276); CHRNA4(0.568833); CHRM1(0.308285); CHRM1(0.308285); CHRM3(0.185959); CHRNA6(0.0447898); CHRM5(-0.0899186); CHRM5(-0.0899186); CHRNB2(-0.277029); CHRM2(-0.277482); CHRNA7(-0.638258);

Serotonin_5ht4 GNAS(0.844842); APP(0.799182); GABRB2(0.776778); KCND2(0.438621); RAPGEF3(0.427181); GABRB3(0.185329); GABRB1(0.0429372); HTR4(-0.0168005); KCNN2(-0.0382039); ADCY5(-0.128395); PRKACA(-0.225821); PRKACB(-0.593236); PRKACG(-0.669971);

Serotonin_5ht6 GNAS(0.844842); APP(0.799182); GABRB2(0.776778); KCND2(0.438621); RAPGEF3(0.427181); GABRB3(0.185329); GABRB1(0.0429372); KCNN2(-0.0382039); HTR6(-0.0382694); ADCY5(-0.128395); PRKACA(-0.225821); PRKACB(-0.593236); PRKACG(-0.669971);

Dopamine_D2 GNG8(2.10561); ARNTL(1.79205); AKT3(1.367); GNB3(1.18685); AKT2(1.1076); PPP2R2D(0.623372); TH(0.622009); PPP2R1B(0.605917); AKT1(0.469481); GSK3A(0.387131); PPP2R5D(0.386355); KCNJ6(0.374662); GNGT1(0.37302); PPP2R2A(0.332969); PPP2R2C(0.269586); GNG10(0.269162); GNG11(0.168354); GNB5(0.128179); ARRB1(0.127419); KCNJ5(0.11836); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GRIN2B(0.0397144); PPP2R5C(0.0319423); PPP2R2B(0.0226157); PPP2CB(0.00482329); KIF5C(0.00349316); GNG13(0.00264592); GNG12(- 0.00591415); GNG3(-0.0134605); GNG7(-0.0666816); PPP2CA(-0.077368); PPP2R5E(-0.104118); KCNJ9(- 0.108818); GNB1(-0.146567); KCNJ3(-0.150505); GNGT2(-0.190679); KIF5A(-0.226009); PPP2R5A(- 0.26817); DRD2(-0.275043); PPP2R5B(-0.299161); PPP2R3A(-0.313783); CLOCK(-0.337872); GSK3B(- 0.366811); GNAO1(-0.367373); GNG4(-0.373484); KIF5B(-0.453313); GNG5(-0.457399); PPP2R1A(- 0.465772); GRIN2A(-0.496191); GNG2(-0.500408); PPP2R3C(-0.558653); ARRB2(-0.619929); GNAI3(- 0.684467); GNAI2(-0.870932);

Acetylcholine_M4 GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); ADCY1(0.954473); FOS(0.744517); CHRM4(0.680809); CREB3L1(0.663382); CREB5(0.481566); GNGT1(0.37302); CACNA1A(0.332833); CREB3L4(0.292783); GNG10(0.269162); ADCY3(0.225875); CACNA1B(0.179502); CREB3L3(0.177996); PIK3CG(0.169337); GNG11(0.168354); NRAS(0.167908); ATF4(0.144112); CREB1(0.139283); ADCY9(0.132965); GNB5(0.128179); CREB3(0.108695); ADCY2(0.107415); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(-0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); PIK3R6(-0.129567); GNB1(-0.146567); MAPK3(-0.182594); GNGT2(-0.190679); PRKACA(-0.225821); PIK3R5(-0.233324); ADCY4(-0.302726); KRAS(-0.357043); GNAO1(-0.367373); GNG4(-0.373484); CREB3L2(-0.392186); Genetic Risk Factors for PTSD….. 125

GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); MAPK1(-0.515604); PRKACB(-0.593236); HRAS(-0.624); PRKACG(-0.669971); GNAI3(-0.684467); GNAI2(-0.870932);

Acetylcholine_M2 GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); ADCY1(0.954473); FOS(0.744517); CREB3L1(0.663382); CREB5(0.481566); GNGT1(0.37302); CACNA1A(0.332833); CREB3L4(0.292783); GNG10(0.269162); ADCY3(0.225875); CACNA1B(0.179502); CREB3L3(0.177996); PIK3CG(0.169337); GNG11(0.168354); NRAS(0.167908); ATF4(0.144112); CREB1(0.139283); ADCY9(0.132965); GNB5(0.128179); CREB3(0.108695); ADCY2(0.107415); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(- 0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); PIK3R6(-0.129567); GNB1(- 0.146567); MAPK3(-0.182594); GNGT2(-0.190679); PRKACA(-0.225821); PIK3R5(-0.233324); CHRM2(- 0.277482); ADCY4(-0.302726); KRAS(-0.357043); GNAO1(-0.367373); GNG4(-0.373484); CREB3L2(- 0.392186); GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); MAPK1(-0.515604); PRKACB(- 0.593236); HRAS(-0.624); PRKACG(-0.669971); GNAI3(-0.684467); GNAI2(-0.870932);

Dopamine_D3 GNG8(2.10561); GNB3(1.18685); KCNJ6(0.374662); GNGT1(0.37302); GNG10(0.269162); GNG11(0.168354); GNB5(0.128179); KCNJ5(0.11836); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); GNG7(-0.0666816); KCNJ9(-0.108818); GNB1(-0.146567); KCNJ3(-0.150505); GNGT2(-0.190679); GNAO1(-0.367373); GNG4(- 0.373484); DRD3(-0.392331); GNG5(-0.457399); GNG2(-0.500408); GNAI3(-0.684467); GNAI2(- 0.870932);

Dopamine_D4 GNG8(2.10561); GNB3(1.18685); KCNJ6(0.374662); GNGT1(0.37302); GNG10(0.269162); GNG11(0.168354); GNB5(0.128179); KCNJ5(0.11836); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); GNG7(-0.0666816); KCNJ9(-0.108818); GNB1(-0.146567); KCNJ3(-0.150505); GNGT2(-0.190679); GNAO1(-0.367373); GNG4(- 0.373484); DRD4(-0.430856); GNG5(-0.457399); GNG2(-0.500408); GNAI3(-0.684467); GNAI2(- 0.870932);

Serotonin_5ht1F GNG8(2.10561); GNB3(1.18685); PRKCB(1.12043); HTR1F(0.934115); APP(0.799182); GABRB2(0.776778); CASP3(0.698174); PLCB1(0.657282); PRKCA(0.530621); KCND2(0.438621); RAPGEF3(0.427181); GNGT1(0.37302); GNG10(0.269162); GABRB3(0.185329); GNG11(0.168354); NRAS(0.167908); GNB5(0.128179); RAF1(0.0997595); GNAI1(0.0796953); PRKCG(0.0786838); GNB2(0.0537978); GNB4(0.0443189); GABRB1(0.0429372); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(-0.0134605); KCNN2(-0.0382039); PLCB4(- 0.0532556); GNG7(-0.0666816); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); GNB1(- 0.146567); MAPK3(-0.182594); GNGT2(-0.190679); PRKACA(-0.225821); ITPR3(-0.239205); BRAF(- 0.258152); KRAS(-0.357043); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); DUSP1(- 0.491877); GNG2(-0.500408); MAPK1(-0.515604); PRKACB(-0.593236); PLCB3(-0.622332); HRAS(-0.624); PRKACG(-0.669971); GNAI3(-0.684467); PLCB2(-0.796177); GNAI2(-0.870932);

Glutamate_mGluR6 ADCY7(1.458); GRM6(1.19216); ADCY1(0.954473); ADCY3(0.225875); ADCY9(0.132965); ADCY2(0.107415); GNAI1(0.0796953); ADCY6(-0.0204362); ADCY5(-0.128395); PRKACA(-0.225821); ADCY4(-0.302726); GNAO1(-0.367373); ADCY8(-0.460334); PRKACB(-0.593236); PRKACG(-0.669971); GNAI3(-0.684467); GNAI2(-0.870932); Genetic Risk Factors for PTSD….. 126

Glutamate_mGluR4 GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); GRM4(1.15314); ADCY1(0.954473); GNGT1(0.37302); GNG10(0.269162); ADCY3(0.225875); GNG11(0.168354); ADCY9(0.132965); GNB5(0.128179); ADCY2(0.107415); GNAI1(0.0796954); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); GNB1(-0.146567); GNGT2(-0.190679); PRKACA(-0.225821); ADCY4(-0.302726); GNAO1(-0.367373); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); PRKACB(-0.593236); PRKACG(-0.669971); GNAI3(-0.684467); GNAI3(-0.684467); GNAI2(-0.870932); GNAI2(-0.870932);

Serotonin_5ht1A GNG8(2.10561); GNB3(1.18685); PRKCB(1.12043); APP(0.799182); GABRB2(0.776778); CASP3(0.698174); PLCB1(0.657282); PRKCA(0.530621); KCND2(0.438621); RAPGEF3(0.427181); GNGT1(0.37302); GNG10(0.269162); GABRB3(0.185329); GNG11(0.168354); NRAS(0.167908); GNB5(0.128179); RAF1(0.0997595); GNAI1(0.0796953); PRKCG(0.0786838); GNB2(0.0537978); GNB4(0.0443189); GABRB1(0.0429372); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(-0.0134605); KCNN2(-0.0382039); PLCB4(-0.0532556); GNG7(-0.0666816); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); GNB1(-0.146567); HTR1A(-0.163322); MAPK3(- 0.182594); GNGT2(-0.190679); PRKACA(-0.225821); ITPR3(-0.239205); BRAF(-0.258152); KRAS(- 0.357043); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); DUSP1(-0.491877); GNG2(- 0.500408); MAPK1(-0.515604); PRKACB(-0.593236); PLCB3(-0.622332); HRAS(-0.624); PRKACG(- 0.669971); GNAI3(-0.684467); PLCB2(-0.796177); GNAI2(-0.870932);

Serotonin_5ht5A GNG8(2.10561); GNB3(1.18685); PRKCB(1.12043); APP(0.799182); GABRB2(0.776778); CASP3(0.698174); PLCB1(0.657282); PRKCA(0.530621); KCND2(0.438621); RAPGEF3(0.427181); GNGT1(0.37302); GNG10(0.269162); GABRB3(0.185329); GNG11(0.168354); NRAS(0.167908); GNB5(0.128179); RAF1(0.0997595); GNAI1(0.0796953); PRKCG(0.0786838); GNB2(0.0537978); GNB4(0.0443189); GABRB1(0.0429372); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(-0.0134605); KCNN2(-0.0382039); PLCB4(-0.0532556); GNG7(-0.0666816); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); GNB1(-0.146567); MAPK3(-0.182594); GNGT2(- 0.190679); HTR5A(-0.216674); PRKACA(-0.225821); ITPR3(-0.239205); BRAF(-0.258152); KRAS(- 0.357043); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); DUSP1(-0.491877); GNG2(- 0.500408); MAPK1(-0.515604); PRKACB(-0.593236); PLCB3(-0.622332); HRAS(-0.624); PRKACG(- 0.669971); GNAI3(-0.684467); PLCB2(-0.796177); GNAI2(-0.870932);

Serotonin_5ht1E GNG8(2.10561); GNB3(1.18685); PRKCB(1.12043); APP(0.799182); GABRB2(0.776778); CASP3(0.698174); PLCB1(0.657282); PRKCA(0.530621); KCND2(0.438621); RAPGEF3(0.427181); GNGT1(0.37302); GNG10(0.269162); GABRB3(0.185329); GNG11(0.168354); NRAS(0.167908); GNB5(0.128179); RAF1(0.0997595); GNAI1(0.0796953); PRKCG(0.0786838); GNB2(0.0537978); GNB4(0.0443189); GABRB1(0.0429372); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(-0.0134605); KCNN2(-0.0382039); PLCB4(-0.0532556); GNG7(-0.0666816); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); GNB1(-0.146567); MAPK3(-0.182594); GNGT2(- 0.190679); PRKACA(-0.225821); ITPR3(-0.239205); BRAF(-0.258152); HTR1E(-0.352573); KRAS(- 0.357043); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); DUSP1(-0.491877); GNG2(- 0.500408); MAPK1(-0.515604); PRKACB(-0.593236); PLCB3(-0.622332); HRAS(-0.624); PRKACG(- 0.669971); GNAI3(-0.684467); PLCB2(-0.796177); GNAI2(-0.870932); Genetic Risk Factors for PTSD….. 127

Serotonin_5ht1B GNG8(2.10561); GNB3(1.18685); PRKCB(1.12043); APP(0.799182); GABRB2(0.776778); CASP3(0.698174); PLCB1(0.657282); PRKCA(0.530621); KCND2(0.438621); RAPGEF3(0.427181); GNGT1(0.37302); GNG10(0.269162); GABRB3(0.185329); GNG11(0.168354); NRAS(0.167908); GNB5(0.128179); RAF1(0.0997595); GNAI1(0.0796953); PRKCG(0.0786838); GNB2(0.0537978); GNB4(0.0443189); GABRB1(0.0429372); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(-0.0134605); KCNN2(-0.0382039); PLCB4(-0.0532556); GNG7(-0.0666816); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); GNB1(-0.146567); MAPK3(-0.182594); GNGT2(- 0.190679); PRKACA(-0.225821); ITPR3(-0.239205); BRAF(-0.258152); KRAS(-0.357043); GNAO1(- 0.367373); HTR1B(-0.371485); GNG4(-0.373484); GNG5(-0.457399); DUSP1(-0.491877); GNG2(- 0.500408); MAPK1(-0.515604); PRKACB(-0.593236); PLCB3(-0.622332); HRAS(-0.624); PRKACG(- 0.669971); GNAI3(-0.684467); PLCB2(-0.796177); GNAI2(-0.870932);

Glutamate_mGluR8 GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); ADCY1(0.954473); GNGT1(0.37302); GNG10(0.269162); GRM8(0.243446); ADCY3(0.225875); GNG11(0.168354); ADCY9(0.132965); GNB5(0.128179); ADCY2(0.107415); GNAI1(0.0796954); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); GNB1(-0.146567); GNGT2(-0.190679); PRKACA(-0.225821); ADCY4(-0.302726); GNAO1(-0.367373); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); PRKACB(-0.593236); PRKACG(-0.669971); GNAI3(-0.684467); GNAI3(-0.684467); GNAI2(-0.870932); GNAI2(-0.870932);

Glutamate_mGluR3 GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); ADCY1(0.954473); GNGT1(0.37302); GNG10(0.269162); ADCY3(0.225875); GNG11(0.168354); ADCY9(0.132965); GNB5(0.128179); ADCY2(0.107415); GNAI1(0.0796954); GNAI1(0.0796953); GNB2(0.0537978); GRM3(0.0480824); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); GNB1(-0.146567); GNGT2(-0.190679); PRKACA(-0.225821); ADCY4(-0.302726); GNAO1(-0.367373); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); PRKACB(-0.593236); PRKACG(-0.669971); GNAI3(-0.684467); GNAI3(-0.684467); GNAI2(-0.870932); GNAI2(-0.870932);

Glutamate_mGluR7 GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); ADCY1(0.954473); GNGT1(0.37302); GNG10(0.269162); ADCY3(0.225875); GNG11(0.168354); ADCY9(0.132965); GNB5(0.128179); ADCY2(0.107415); GNAI1(0.0796954); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); GRM7(-0.0739135); ADCY5(-0.128395); GNB1(-0.146567); GNGT2(-0.190679); PRKACA(-0.225821); ADCY4(-0.302726); GNAO1(-0.367373); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); PRKACB(-0.593236); PRKACG(-0.669971); GNAI3(-0.684467); GNAI3(-0.684467); GNAI2(-0.870932); GNAI2(-0.870932);

Serotonin_5ht1D GNG8(2.10561); GNB3(1.18685); PRKCB(1.12043); APP(0.799182); GABRB2(0.776778); CASP3(0.698174); PLCB1(0.657282); PRKCA(0.530621); KCND2(0.438621); RAPGEF3(0.427181); GNGT1(0.37302); GNG10(0.269162); GABRB3(0.185329); GNG11(0.168354); NRAS(0.167908); GNB5(0.128179); RAF1(0.0997595); GNAI1(0.0796953); PRKCG(0.0786838); GNB2(0.0537978); GNB4(0.0443189); GABRB1(0.0429372); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(-0.0134605); KCNN2(-0.0382039); PLCB4(-0.0532556); GNG7(-0.0666816); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); GNB1(-0.146567); MAPK3(-0.182594); GNGT2(- Genetic Risk Factors for PTSD….. 128

0.190679); PRKACA(-0.225821); ITPR3(-0.239205); BRAF(-0.258152); KRAS(-0.357043); GNAO1(- 0.367373); GNG4(-0.373484); GNG5(-0.457399); DUSP1(-0.491877); GNG2(-0.500408); MAPK1(- 0.515604); PRKACB(-0.593236); PLCB3(-0.622332); HRAS(-0.624); PRKACG(-0.669971); GNAI3(- 0.684467); PLCB2(-0.796177); GNAI2(-0.870932); HTR1D(-1.01472);

Glutamate_mGluR2 GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); ADCY1(0.954473); GNGT1(0.37302); GNG10(0.269162); ADCY3(0.225875); GNG11(0.168354); ADCY9(0.132965); GNB5(0.128179); ADCY2(0.107415); GNAI1(0.0796954); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); GNB1(-0.146567); GNGT2(-0.190679); PRKACA(-0.225821); ADCY4(-0.302726); GNAO1(-0.367373); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); ADCY8(-0.460334); GRM2(-0.463511); GNG2(-0.500408); PRKACB(-0.593236); PRKACG(-0.669971); GNAI3(-0.684467); GNAI3(-0.684467); GNAI2(-0.870932); GNAI2(-0.870932);

Serotonin HTR1F(0.934115); HTR2B(0.403711); HTR3D(0.244506); HTR7(0.15563); HTR4(-0.0168005); HTR6(-0.0382694); HTR1A(-0.163322); HTR2A(-0.19768); HTR3C(-0.211874); HTR5A(-0.216674); HTR3A(- 0.274707); HTR1E(-0.352573); HTR1B(-0.371485); HTR3E(-0.641259); HTR3B(-0.719955); HTR1D(- 1.01472);

Genetic Risk Factors for PTSD….. 129

Supplementary Figure S2 - Gene Scores for Each Post-hoc Gene-set

Serotonin2_Intra PLA2G4F(2.10831); PLA2G4D(2.00666); PLA2G4E(1.94461); PRKCB(1.12043); GNAQ(0.856944); PLCB1(0.657282); PRKCA(0.530621); PLA2G4B(0.483477); JMJD7-PLA2G4B(0.464105); PLA2G4C(0.440997); CACNA1C(0.356703); CACNA1D(0.296217); PRKCG(0.0786838); PLCB4(-0.0532556); ITPR2(-0.0756829); ITPR1(-0.10489); PLA2G4A(-0.136318); MAPK3(-0.182594); ITPR3(-0.239205); CACNA1S(-0.357408); MAPK1(-0.515604); PLCB3(-0.622332); PLCB2(-0.796177);

Glutamate_mGluR1_5_Intra PLA2G4F(2.10831); PLA2G4D(2.00666); PLA2G4E(1.94461); ADCY7(1.458); PLD2(1.14245); PRKCB(1.12043); ADCY1(0.954473); GNAQ(0.856944); GNAS(0.844842); PLCB1(0.657282); PRKCA(0.530621); PLA2G4B(0.483477); JMJD7-PLA2G4B(0.464105); PLA2G4C(0.440997); ADCY3(0.225875); HOMER3(0.187757); ADCY9(0.132965); ADCY2(0.107415); PRKCG(0.0786838); ADCY6(-0.0204362); PLCB4(-0.0532556); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); PLD1(-0.135981); PLA2G4A(-0.136318); MAPK3(-0.182594); HOMER2(-0.207691); PRKACA(-0.225821); ITPR3(-0.239205); ADCY4(-0.302726); HOMER1(-0.343684); ADCY8(-0.460334); MAPK1(-0.515604); PRKACB(-0.593236); PLCB3(-0.622332); PRKACG(-0.669971); PLCB2(-0.796177);

GABA GABBR1(1.02476); GABRR2(0.984362); GABBR2(0.890302); GABRB2(0.776778); GABRA4(0.764706); GABRR3(0.59692); GABRA2(0.536947); GABRG2(0.460323); GABRG3(0.35292); GABRG1(0.270733); GABRB3(0.185329); GABRA5(0.163624); GABRB1(0.0429372); GABRA1(-0.0266599); GABRP(-0.0508056); GABRA6(-0.231377); GABRR1(-0.401135); GABRD(-0.622532);

Glutamate GRM5(1.21134); GRM6(1.19216); GRIN1(1.1748); GRM4(1.15314); GRIK5(0.915416); GRIA4(0.660596); GRIK1(0.624287); GRIN3B(0.623253); GRIA1(0.383217); GRIN2D(0.331225); GRM1(0.304609); GRIK4(0.244594); GRM8(0.243446); GRIK3(0.135137); GRM3(0.0480824); GRIN2B(0.0397144); GRIK2(0.0396233); GRIN2C(-0.0732666); GRM7(-0.0739135); GRIA2(-0.378383); GRM2(-0.463511); TRPC1(-0.474553); GRIN2A(-0.496191); GRIN3A(-0.609511);

Glutamate_NMDA_Intra PPP3CB(2.13408); PPP3R1(0.786684); SHANK2(0.530936); DLG4(0.419713); SHANK3(0.336771); SHANK1(0.300812); DLGAP1(0.216453); HOMER3(0.187757); PPP3CA(0.026846); PPP3CC(-0.0304729); HOMER2(-0.207691); PPP3R2(-0.258932); HOMER1(-0.343684);

Acetylcholine_nAChR_Intra ADCY7(1.458); AKT3(1.367); CAMK2D(1.21548); AKT2(1.1076); ADCY1(0.954473); CAMK2B(0.672307); CREB3L1(0.663382); PIK3R2(0.590356); CREB5(0.481566); AKT1(0.469481); CACNA1C(0.356703); BCL2(0.301821); CACNA1D(0.296217); CREB3L4(0.292783); CAMK2G(0.268395); CAMK2A(0.231646); ADCY3(0.225875); CREB3L3(0.177996); ATF4(0.144112); CREB1(0.139283); ADCY9(0.132965); CREB3(0.108695); ADCY2(0.107415); FYN(0.0513743); ADCY6(- 0.0204362); PIK3CB(-0.0278242); PIK3CD(-0.0545999); CAMK4(-0.0607568); ADCY5(-0.128395); PRKACA(-0.225821); ADCY4(-0.302726); JAK2(-0.345736); CACNA1S(-0.357408); CREB3L2(-0.392186); PIK3CA(-0.402222); ADCY8(-0.460334); PRKACB(-0.593236); PRKACG(-0.669971); PIK3R1(-0.742498);

Dopamine_D1_Intra PPP3CB(2.13408); CAMK2D(1.21548); PRKCB(1.12043); MAPK13(0.927615); GNAQ(0.856944); MAPK8(0.847746); GNAS(0.844842); MAPK14(0.753329); FOS(0.744517); CAMK2B(0.672307); CREB3L1(0.663382); PLCB1(0.657282); ATF2(0.556153); PRKCA(0.530621); MAPK9(0.521339); CREB5(0.481566); MAPK11(0.417754); MAPK12(0.396692); CACNA1C(0.356703); CACNA1D(0.296217); CREB3L4(0.292783); CAMK2G(0.268395); CAMK2A(0.231646); SCN1A(0.228832); ATF6B(0.184584); CREB3L3(0.177996); GNAL(0.176089); ATF4(0.144112); CREB1(0.139283); Genetic Risk Factors for PTSD….. 130

CREB3(0.108695); PRKCG(0.0786838); PPP3CA(0.026846); PPP3CC(-0.0304729); PLCB4(-0.0532556); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); PRKACA(-0.225821); CALML4(-0.228977); ITPR3(-0.239205); MAPK10(-0.261223); PPP1CB(-0.288708); CALY(-0.301501); CREB3L2(-0.392186); PPP1R1B(-0.476736); PRKACB(-0.593236); PLCB3(-0.622332); PRKACG(-0.669971); PPP1CC(-0.672875); CALML3(-0.698534); CALML6(-0.720422); CALML5(-0.737076); PPP1CA(-0.785583); PLCB2(-0.796177);

GABA_GABAB_Intra GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); GABBR1(1.02476); ADCY1(0.954473); GABBR2(0.890302); KCNJ6(0.374662); GNGT1(0.37302); CACNA1C(0.356703); CACNA1A(0.332833); CACNA1D(0.296217); GNG10(0.269162); ADCY3(0.225875); CACNA1B(0.179502); GNG11(0.168354); ADCY9(0.132965); GNB5(0.128179); ADCY2(0.107415); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); GNB1(-0.146567); GNGT2(-0.190679); ADCY4(-0.302726); CACNA1S(-0.357408); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); GNAI3(-0.684467); GNAI2(-0.870932);

Dopamine_D5_Intra MAPK13(0.927615); MAPK8(0.847746); GNAS(0.844842); MAPK14(0.753329); FOS(0.744517); CREB3L1(0.663382); ATF2(0.556153); MAPK9(0.521339); CREB5(0.481566); MAPK11(0.417754); MAPK12(0.396692); CACNA1C(0.356703); CACNA1D(0.296217); CREB3L4(0.292783); SCN1A(0.228832); ATF6B(0.184584); CREB3L3(0.177996); GNAL(0.176089); ATF4(0.144112); CREB1(0.139283); CREB3(0.108695); ADCY5(-0.128395); PRKACA(-0.225821); MAPK10(- 0.261223); PPP1CB(-0.288708); CREB3L2(-0.392186); PPP1R1B(-0.476736); PRKACB(-0.593236); PRKACG(-0.669971); PPP1CC(-0.672875); PPP1CA(-0.785583);

Acetylcholine CHRM4(0.680809); CHRNB4(0.644963); CHRNA3(0.618276); CHRNA4(0.568833); CHRM1(0.308285); CHRM1(0.308285); CHRM3(0.185959); CHRNA6(0.0447898); CHRM5(-0.0899186); CHRM5(-0.0899186); CHRNB2(-0.277029); CHRM2(-0.277482); CHRNA7(-0.638258);

Serotonin4_6_7_Intra GNAS(0.844842); APP(0.799182); GABRB2(0.776778); KCND2(0.438621); RAPGEF3(0.427181); GABRB3(0.185329); GABRB1(0.0429372); KCNN2(-0.0382039); ADCY5(-0.128395); PRKACA(-0.225821); PRKACB(-0.593236); PRKACG(-0.669971);

Dopamine_D2_Intra GNG8(2.10561); ARNTL(1.79205); AKT3(1.367); GNB3(1.18685); AKT2(1.1076); PPP2R2D(0.623372); PPP2R1B(0.605917); AKT1(0.469481); GSK3A(0.387131); PPP2R5D(0.386355); KCNJ6(0.374662); GNGT1(0.37302); PPP2R2A(0.332969); PPP2R2C(0.269586); GNG10(0.269162); GNG11(0.168354); GNB5(0.128179); ARRB1(0.127419); KCNJ5(0.11836); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GRIN2B(0.0397144); PPP2R5C(0.0319423); PPP2R2B(0.0226157); PPP2CB(0.00482329); KIF5C(0.00349316); GNG13(0.00264592); GNG12(-0.00591415); GNG3(- 0.0134605); GNG7(-0.0666816); PPP2CA(-0.077368); PPP2R5E(-0.104118); KCNJ9(-0.108818); GNB1(- 0.146567); KCNJ3(-0.150505); GNGT2(-0.190679); KIF5A(-0.226009); PPP2R5A(-0.26817); PPP2R5B(- 0.299161); PPP2R3A(-0.313783); CLOCK(-0.337872); GSK3B(-0.366811); GNAO1(-0.367373); GNG4(- 0.373484); KIF5B(-0.453313); GNG5(-0.457399); PPP2R1A(-0.465772); GRIN2A(-0.496191); GNG2(- 0.500408); PPP2R3C(-0.558653); ARRB2(-0.619929); GNAI3(-0.684467); GNAI2(-0.870932);

Acetylcholine_M2_M4_Auto GNG8(2.10561); GNB3(1.18685); GNGT1(0.37302); CACNA1A(0.332833); GNG10(0.269162); CACNA1B(0.179502); GNG11(0.168354); GNB5(0.128179); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); Genetic Risk Factors for PTSD….. 131

GNG7(-0.0666816); GNB1(-0.146567); GNGT2(-0.190679); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); GNG2(-0.500408); GNAI3(-0.684467); GNAI2(-0.870932);

Dopamine_D3_D4_Intra GNG8(2.10561); GNB3(1.18685); KCNJ6(0.374662); GNGT1(0.37302); GNG10(0.269162); GNG11(0.168354); GNB5(0.128179); KCNJ5(0.11836); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); GNG7(-0.0666816); KCNJ9(-0.108818); GNB1(-0.146567); KCNJ3(-0.150505); GNGT2(-0.190679); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); GNG2(-0.500408); GNAI3(-0.684467); GNAI2(- 0.870932);

Acetylcholine_M2_M4_Intra GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); ADCY1(0.954473); FOS(0.744517); CREB3L1(0.663382); CREB5(0.481566); GNGT1(0.37302); CREB3L4(0.292783); GNG10(0.269162); ADCY3(0.225875); CREB3L3(0.177996); PIK3CG(0.169337); GNG11(0.168354); NRAS(0.167908); ATF4(0.144112); CREB1(0.139283); ADCY9(0.132965); GNB5(0.128179); CREB3(0.108695); ADCY2(0.107415); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(-0.0134605); ADCY6(- 0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); PIK3R6(-0.129567); GNB1(-0.146567); MAPK3(- 0.182594); GNGT2(-0.190679); PRKACA(-0.225821); PIK3R5(-0.233324); ADCY4(-0.302726); KRAS(- 0.357043); GNAO1(-0.367373); GNG4(-0.373484); CREB3L2(-0.392186); GNG5(-0.457399); ADCY8(- 0.460334); GNG2(-0.500408); MAPK1(-0.515604); PRKACB(-0.593236); HRAS(-0.624); PRKACG(- 0.669971); GNAI3(-0.684467); GNAI2(-0.870932);

Serotonin1_5a_Intra GNG8(2.10561); GNB3(1.18685); PRKCB(1.12043); APP(0.799182); GABRB2(0.776778); CASP3(0.698174); PLCB1(0.657282); PRKCA(0.530621); KCND2(0.438621); RAPGEF3(0.427181); GNGT1(0.37302); GNG10(0.269162); GABRB3(0.185329); GNG11(0.168354); NRAS(0.167908); GNB5(0.128179); RAF1(0.0997595); GNAI1(0.0796953); PRKCG(0.0786838); GNB2(0.0537978); GNB4(0.0443189); GABRB1(0.0429372); GNG13(0.00264592); GNG12(-0.00591415); MAP2K1(-0.0121188); GNG3(-0.0134605); KCNN2(-0.0382039); PLCB4(-0.0532556); GNG7(-0.0666816); ITPR2(-0.0756829); ITPR1(-0.10489); ADCY5(-0.128395); GNB1(-0.146567); MAPK3(-0.182594); GNGT2(- 0.190679); PRKACA(-0.225821); ITPR3(-0.239205); BRAF(-0.258152); KRAS(-0.357043); GNAO1(- 0.367373); GNG4(-0.373484); GNG5(-0.457399); DUSP1(-0.491877); GNG2(-0.500408); MAPK1(- 0.515604); PRKACB(-0.593236); PLCB3(-0.622332); HRAS(-0.624); PRKACG(-0.669971); GNAI3(- 0.684467); PLCB2(-0.796177); GNAI2(-0.870932);

Glutamate_mGluR7_8_PreSynaptic GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); ADCY1(0.954473); GNGT1(0.37302); GNG10(0.269162); ADCY3(0.225875); GNG11(0.168354); ADCY9(0.132965); GNB5(0.128179); ADCY2(0.107415); GNAI1(0.0796954); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); GNB1(-0.146567); GNGT2(-0.190679); PRKACA(-0.225821); ADCY4(-0.302726); GNAO1(-0.367373); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); PRKACB(-0.593236); PRKACG(-0.669971); GNAI3(-0.684467); GNAI3(-0.684467); GNAI2(-0.870932); GNAI2(-0.870932);

Glutamate_mGluR2_3_4_PreSynaptic GNG8(2.10561); ADCY7(1.458); GNB3(1.18685); ADCY1(0.954473); GNGT1(0.37302); GNG10(0.269162); ADCY3(0.225875); GNG11(0.168354); ADCY9(0.132965); GNB5(0.128179); ADCY2(0.107415); GNAI1(0.0796954); GNAI1(0.0796953); GNB2(0.0537978); GNB4(0.0443189); GNG13(0.00264592); GNG12(-0.00591415); GNG3(-0.0134605); Genetic Risk Factors for PTSD….. 132

ADCY6(-0.0204362); GNG7(-0.0666816); ADCY5(-0.128395); GNB1(-0.146567); GNGT2(-0.190679); PRKACA(-0.225821); ADCY4(-0.302726); GNAO1(-0.367373); GNAO1(-0.367373); GNG4(-0.373484); GNG5(-0.457399); ADCY8(-0.460334); GNG2(-0.500408); PRKACB(-0.593236); PRKACG(-0.669971); GNAI3(-0.684467); GNAI3(-0.684467); GNAI2(-0.870932); GNAI2(-0.870932);

Glutamate_mGluR2_3_4_6_7_8_PostSynaptic ADCY7(1.458); ADCY1(0.954473); ADCY3(0.225875); ADCY9(0.132965); ADCY2(0.107415); GNAI1(0.0796953); ADCY6(-0.0204362); ADCY5(-0.128395); PRKACA(-0.225821); ADCY4(-0.302726); GNAO1(-0.367373); ADCY8(-0.460334); PRKACB(-0.593236); PRKACG(-0.669971); GNAI3(-0.684467); GNAI2(-0.870932);

Serotonin HTR1F(0.934115); HTR2B(0.403711); HTR3D(0.244506); HTR7(0.15563); HTR4(-0.0168005); HTR6(-0.0382694); HTR1A(-0.163322); HTR2A(-0.19768); HTR3C(-0.211874); HTR5A(-0.216674); HTR3A(- 0.274707); HTR1E(-0.352573); HTR1B(-0.371485); HTR3E(-0.641259); HTR3B(-0.719955); HTR1D(- 1.01472);

Missing Receptors (Organized by receptor)

Dopamine: DRD1(0.134093); DRD4(-.00281582); DRD3(-0.469232)

Adrenergic: ADRA1D(0.890306); ADRA1B(0.742379); ADRA2C(0.215671); ADRB2(0.00124975); ADRB3(- 0.293838); ADRA2B(-0.425166); ADRB1(-0.526369); ADRA1A(-0.725393); ADRA2A(-0.728913);

Cannabinoid: CNR1(0.304161); CNR2(-0.321743);

Opioid: OGFR(0.237652); OPRD1(-0.563886); OPRL1(-0.716653);

NPY: NPY1R(-0.429355); NPY5R(-0.552257); NPY2R(-0.698366);