FROM MOUSE TO MAN: HOW THE HUMAN POST MORTEM MICROBIOME RELATES TO LOCAL AREA CRIME LEVEL By Christine Carole Kwiatkowski A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Criminal Justice – Doctor of Philosophy 2022 ABSTRACT Violence is a widespread public health and justice system problem with far-reaching consequences for victims, offenders, and their communities. Aggression, the cognitive and behavioral antecedent to violent action, is mainly understood in terms of the psychosocial risk factors that increase the likelihood of aggressive behavioral strategies. Neighborhood context is a principal risk factor for violent crime perpetration, but the mechanisms that mediate the effect of the environment on individual-level aggression behavior are poorly understood, especially the biological factors that may contribute to our understanding of violent behavior. In order to gain a better understanding of mechanisms that precipitate violence in specific geographic contexts, this dissertation explores the relationship between aggression behavior and the gut microbiome, a spatially determined physiological system that affects human health and behavior. In humans, post mortem microbiome studies show a loss of biodiversity in high crime census block groups that can be leveraged, alongside indicators of socioeconomic status, to predict block group crime level, supplying critical foundational support to explore how differences in gut microbiome composition relate to human aggression behavior. The overall goal of this research is to connect basic science findings in mice to correlational evidence of a relationship between the human gut microbiome and violent crime exposures, thus revealing how community health affects individuals and supplying a potential target for future intervention. In honor of my grandfathers, Władysław Kwiatkowski and Anthony Ambroziak, who collectively fought a war, crossed an ocean, and overcame abject poverty to give me this opportunity. Your sacrifices will never be forgotten. iii ACKNOWLEDGEMENTS The research presented in this dissertation was supported through the Avielle Foundation’s Basic Neuroscience Research Grant Program and Michigan State University’s College of Natural Sciences Dissertation Completion Fellowship. The Human Post Mortem Microbiome (HPMM) research was funded by a grant from the National Institute of Justice, Office of Justice Programs, U.S. Department of Justice, awarded (2014- DN-BX-K008) to Jennifer L. Pechal, Carl J. Schmidt, Heather R. Jordan, and M. Eric Benbow. Points of view in this document are those of the authors and do not necessarily represent the official position or policies of the U.S. Department of Justice or Department of Defense. This work was completed with significant contributions from Sierra Kaszubinski who processed all of the raw sequencing data from both animal and human microbiome study samples and dedicated innumerable training hours instructing on DNA extraction and methods of microbiome data analysis. To this end, I would also like to thank Dr. Jennifer Pechal who offered consistent guidance and feedback for the work presented in Chapter 5 as well as Drs. Andrew Eagle, Andrew Bender, Hope Akaeze, and Sam Golden who collaborated on the work presented in Chapter 3. I am extremely fortunate to have worked alongside two highly collaborative scientists at Michigan State University who were willing to bridge disciplines for an unconventional project. Drs. April Zeoli and A.J. Robison have imparted a tremendous amount of knowledge and encouragement throughout my graduate school experience. Without their support, this interdisciplinary project would not be possible. I would also like to thank my committee, Drs. Michelle Mazei-Robison, Adam Zwickle, Chris Melde, Adam Moeser, Nara Parameswaran, and v iv Eric Benbow for offering instructive and thoughtful feedback on my project. I truly stand on the shoulders of giants. I have also had the distinct pleasure of working alongside amazing scientists who are also incredible people. Foremost, I would like to thank Ken Moon for his friendship, mentorship, and countless hours of feeding mice probiotics and antibiotics. I would also like to thank Drs. Claire Manning, Andrew Eagle, Liz Williams, Marie Doyle, and Natalia Duque-Wilckens for their mentorship, friendship, and scientific discussions. I am also immeasurably grateful for my new collaborations with Chiho Sugimoto and Abigal Barrett who have readily offered their friendship as well as assistance and feedback on my project. Finally, I would like to thank Dr. Jean-Pierre Nshimyimana for being a willing sounding board and source of feedback. Most importantly, I would like to recognize the unconditional love and support of my family who have been the foundation on which I stand to take risks in this world. Most of all, I would like to thank my best friend and partner, Nate, who is a constant source of joy and support, not only on this journey but always. His love has made me a better person and scientist. vi v TABLE OF CONTENTS CHAPTER 1: INTRODUCTION ................................................................................................... 1 Introduction ......................................................................................................................... 1 Background ......................................................................................................................... 7 The Current Study ............................................................................................................... 9 CHAPTER 2: METHODOLOGY ................................................................................................ 10 Site Selection .................................................................................................................... 10 Aggregate Data ................................................................................................................. 15 Analytic Strategy .............................................................................................................. 20 Dependent and Independent Variables ............................................................................. 21 Analysis ............................................................................................................................ 23 CHAPTER 3: RESULTS .............................................................................................................. 27 Overview ........................................................................................................................... 27 CHAPTER 4: DISCUSSION ........................................................................................................ 46 Conclusion ........................................................................................................................ 51 CHAPTER 5: CONCLUSIONS AND FUTURE DIRECTIONS ................................................ 53 Summary ........................................................................................................................... 53 Implications and Future Directions................................................................................... 54 Ethical Considerations ...................................................................................................... 58 WORKS CITED ........................................................................................................................... 59 APPENDIX ................................................................................................................................... 77 vii vi CHAPTER 1: INTRODUCTION Introduction Aggression predicates acts of violence that serve to control, traumatize, maim, or kill victims. According to the Federal Bureau of Investigation (2017), however, a conservative estimate of violent crime (e.g., murder, rape, robbery, and aggravated assault) in Michigan is 450 crimes per 100,000 persons annually, with aggravated assault offenses accounting for approximately 69% of all violent crime. Within clinical populations, the risk of aggression is elevated among those with paranoid psychoses, substance abuse disorder, and antisocial personality disorder (Eronen, Angermeyer, & Schulze, 1998), posing a challenge to patient care within the justice and healthcare systems. Importantly, survivors of violence often suffer long-term health consequences of abuse, such as increased risk for post-traumatic stress disorder, chronic pain, and generally higher health care utilization (Campbell, 2002; Rivara et al., 2019). Thus, aggression and violence represent serious public health problems with negative effects on quality of life, health, and vitality for both perpetrators and victims. Spatial location is a critical risk factor for crime perpetration, underscoring the importance of contextual, or environmental, factors in driving offender behavior. Crime mapping has thus become a tradition of criminological study in an effort to understand broader patterns of behavior and the local area factors that increase risk of crime. Historically, criminologists of the 19th century first demonstrated the coupling of crime and place (e.g., Guerry, 1833), but it was the work of Shaw and McKay (1942) that first explained the link in terms of social and environmental factors that disrupt community through social disorganization. In particular, residential instability, population heterogeneity, and low socioeconomic status emerged as predictors of neighborhood delinquency that persisted in defining risk across three different time periods irrespective of 1 shifting neighborhood demographics (Shaw & McKay, 1942), suggesting that factors within a disorganized neighborhood context rather than individual-level traits drive human health and behavioral patterns. This foundational research led to the identification of a suite of social and environmental risk factors for criminal behavior that are spatially clustered in disadvantaged neighborhoods (Bursik & Grasmick, 1993; Sampson & Groves, 1989; Sampson, Raudenbush, & Earls, 1997; Shaw & McKay, 1942). This has led many to speculate that there is a neighborhood effect wherein mechanisms of social control breakdown (in response to disorganization) to create a social (i.e., low cohesion) and physical environment where crime is more readily perpetrated. However, most individuals who live in neighborhoods that have a high risk of crime do not commit crime or otherwise engage in deviant behavior (Bursik & Grasmick, 1993). Therefore, there is important variation in this so-called “neighborhood effect” on crime that is not accounted for by current explanations of criminal behavior, highlighting a need for additional research to explain individual variation in response to the psychosocial risk factors associated with deviance. Indeed, research to date is limited in its ability to provide (1) a mechanism through which social and environmental exposures influence individual action and (2) why individuals exposed to similar social and environmental factors vary in their engagement in deviant behavior. Understanding how these factors affect the brain, the final common mediator of risk, is essential for understanding the etiology of deviant/aggressive behavior. Previous research demonstrates that there are significant differences in brain activity between individuals with a history of violent behavior and non-violent controls (e.g., Leutgeb et al., 2015). Specifically, increased risk for human aggression is associated with pathophysiological changes in brain regions underlying emotional regulation and the body’s response to stress (e.g., threat). Impairments in the prefrontal 2 cortex (PFC), the seat of executive function, and the amygdala, the center of fear and aggression (Rosell & Siever, 2015), limit the ability to process and regulate emotion, enabling the expression of violating behaviors (Sapolsky, 2017). Importantly, similar pathophysiological changes are observed in the brains of individuals who grow up living in impoverished socioeconomic conditions (Liberzon et al., 2015). This raises the question as to how exposures in disadvantaged neighborhood contexts affect the brain. Critically, because the brain does not operate in isolation from the rest of the body, consideration of other physiological systems that could mediate the relationship between the physical environment and the brain must be accounted for. Emerging research about the human microbiome, the communities of microbes that inhabit the body, suggests that it has a role as a physiological regulator in health and behavior (Knight, 2015), an effect that depends on the environment and its spatial location. Moreover, recent evidence suggests a role for the gut microbiome in human psychological disposition and mental health. For example, Borre and colleagues (2014) review how perturbations in the normal developmental trajectory of the human microbiome due to both lifestyle choices (e.g., formula versus breast feeding) and environmental insults (e.g., pathogen exposure) can disrupt physical growth, permanently alter critical brain signaling pathways, and predispose an individual to neurodevelopmental psychiatric disorders. Research also shows that differential microbial community structures in the gut are associated with different temperaments in children (Christian et al., 2015) as well as major depression and anxiety in adults (Jiang et al., 2015; Jiang et al., 2018). Together, these findings suggest that gut dysbiosis (i.e., imbalance) is associated with various expressions of pathological behavior and aberrations in brain health. Evidence of the specific effect of the gut microbiome on human aggression is limited and does not directly relate to behavior. In one study, changes in inmate diet were associated with 3 reductions in aggression and prison rule-breaking (Gesch, Hammond, Hampson, Eves, & Crowder, 2002; Zaalberg, Nijman, Bulten, Stroosma, & van der Staak, 2010). This effect was attributed to changes in metabolic function (Gesch et al., 2002), but given recent findings regarding the effect of diet on the gut microbiome (Sonnenburg & Sonnenburg, 2015), the underlying mechanism for this behavioral change among the prisoners remains unclear. More specifically, other work has shown that changes in the gut microbiome are associated with changes in the neurocognitive antecedents of aggressive behavior. For example, a recent human imaging study showed reductions in brain activity related to vigilance (e.g., threat tracking) following probiotic treatment (Tillisch et al., 2013), and other research demonstrated reductions in self-reported aggression in response to probiotics (Steenbergen et al., 2015). Taken together, this research suggests a relationship between the microbiome of the gut and sensitivity to emotional cues that could be important for the manifestation of aggressive behavior in humans. Importantly, the microbiome of the gut has a known role in sensory and neurochemical pathways that communicate with the brain to modulate behavior (Mayer et al., 2014). Its composition is also highly dependent on lifestyle choices and environmental exposures (Clarke, O'Mahony, Dinan, & Cryan, 2014; von Mutius & Vercelli, 2010). Thus, perturbations in the gut microbiome caused by environmental input, such as those observed in disadvantaged neighborhoods (e.g., low access to nutritional foods, toxin exposure), could have a profound effect on behavior, including deviant behavior. However, research is needed to investigate the potential link between the human microbiome, brain health, and aggression behavior to uncover potential mechanisms that could explain why certain contexts are so risky for both health and behavior. Though there is increasing interest in studying the biological factors associated with aggressive behavior from a criminal justice perspective (Rafter et al., 2016), only limited work has 4 been done. Neuroscientific research shows the importance of pathophysiological changes in the brain for the expression of aggression behavior (Bannon, Salis, & O'Leary, 2015), which involves a dynamic interplay of biological, psychological, and social or environmental factors. Given the established contributions of both environmental and social neighborhood factors to individual- level risk for aggressive behavior, the pressing question becomes how context interacts with physiology, and to what extent there is an effect of physiology beyond long-established risk factors like residential instability, population heterogeneity, and low socioeconomic status. One potential mechanism is through the gut-brain axis, which is both shaped by the environment and has a role in human behavior, including pathological behavior. However, the role of the gut-brain axis in aggression and aggressive behavior remains largely unexplored. The Biopsychosocial Model In the 1970s, the medical community experienced a paradigm shift with the introduction of George L. Engel’s biopsychosocial model (Engel, 1977). Engel (1977) criticized the biomedical model for its cold reductionism, asserting that biology alone could not explain the complexity of human health and behavioral phenomena. As a solution, Engel (1977) proposed the biopsychosocial model as a paradigm to explain human health and behavior with respect to the combination of biological, psychological, and social or environmental risk factors that work together and interact to affect human health and behavior. Naturally, the main tenet of the biopsychosocial model is that complex health and behavioral phenomena cannot be explained by narrowly defined mechanisms. For example, disease may be characterized by pathophysiological changes in the body, but the severity of the illness and its prognosis worsen in the presence of psychosocial risk factors, such as psychological distress or poor living conditions (Engel, 1977). In this view, human health and behavioral phenomena may have a classic presentation, but vary 5 according to other biological (e.g., comorbidities), psychological, and social or environmental risk factors. This perspective expands the influences of context, or the social and environmental exposures that nuance both human experience and physiology, and supplies a critical framework for integrating natural and social science to advance the study of aggression behavior. At the onset of an acute social interaction, the actors and their biology arrive conditioned from past experience. They import experience gained through development, socialization, and social learning (i.e. culture) as reference for appropriate behavior. Together, the actors' histories, social abilities, and other factors, such as current emotional and psychological state, frame each actor's perception of environmental and social cues, and shape the behavioral response (Crick & Dodge, 1994). As such, the role of biology is to simply enable an established propensity towards aggression and deviance (Sapolsky, 2017). This comprehensive explanation of aggression can be accounted for within a biopsychosocial framework, creating a critical need to integrate biology in criminological research to understand the full set of factors known to influence aggressive and violent behavior (Fishbein, 1990). The human microbiome is a primary candidate for integration into a biopsychosocial model of aggression behavior in humans. The microbiome, which varies according to both physical (Afshinnekoo et al., 2015; Yatsunenko et al., 2012) and social (Song et al., 2013) exposures, is an environmental factor that is known to affect host physiology, including the structure and function of the brain (Sherwin, Bordenstein, Quinn, Dinan, & Cryan, 2019). Importantly, the human microbiome is specifically known to vary by socioeconomic conditions with a significantly higher abundance of pathogens and lower biodiversity associated with urban blight and unhealthy living conditions (Pearson et al., 2019), forging another potential link between the human microbiome and disadvantaged neighborhood contexts where crime occurs. Together, these features position 6 the gut microbiome as a potential novel mechanism through which the environment can affect individual behavior in a spatially determined way. Background In an ordinary context, aggression is a behavioral tool used to compete for resources. When aggression is expressed outside the context of competition, it is pathological (Nelson & Trainor, 2007). When it becomes harmful, it is deviant. When a law prohibits the behavior and the individual is caught, it is criminal. Interpersonal aggression is conceptualized in many ways in the literature but can generally be subdivided into two types. Broadly, reactive aggression is considered emotional and impulsive while instrumental aggression is considered calculated and goal-oriented, either for gain or pleasure (J. Haller & Kruk, 2006; Nelson & Trainor, 2007; Sapolsky, 2017). Both reactive and instrumental aggression are associated with pathophysiological changes in several key brain regions in the clinical literature. In particular, areas implicated in emotion and executive control function differently or less optimally preceding deviant behavior. Though many regions are known to participate in processing emotional stimuli, the cortico- amygdala loop is among the most studied. The basic function of this circuitry is to (1) detect emotional stimuli, and (2) properly interpret the stimuli to select an appropriate behavioral response. Deficits in the structure and function of the amygdala and prefrontal cortex (PFC), critical players in circuits underlying emotional regulation and social behavior, result from changes in the molecular environment of the body. Indeed, brain activity depends on the coordinated signaling of chemicals messengers for normal function. Adverse events (i.e., stressors) can excessively activate biological processes that suppress neural growth factors, limit social learning, and alter developmental trajectory. When activity changes, there is an underlying change in chemical 7 messaging, or signaling cascades. It is the inimitable set of life experiences garnered through unique social and environmental exposures in combination with unique genetic factors that creates variation in human behavior and adds complexity to understanding behavioral phenomena, like aggression. Acknowledging that the brain does not operate in isolation, it is critical to examine the relationship between other physiological systems and the brain dysfunction driving behavioral aggression. The human microbiome is one such system, both known for its profound effects on health and behavior as well as its contextually specific (microbial) community structure, positioning it as a likely conduit between environmental exposures, the brain, and spatially determined behaviors (e.g., aggression). The microbiome is an environmental factor that affects host physiology and behavior, potentially via the gut-brain axis (Knight, 2015; Sherwin et al., 2019), a collection of signaling pathways that describes mechanisms through which microbiota can engage in physiological processes throughout the body. For example, the vagus nerve, known for its role in visceral sensation (e.g., cramping), directly innervates the gut where afferent projections interact with microbial metabolites and intermediary enteroendocrine cells that are regulated, in part, by gut bacteria. Moreover, gut bacteria produce neurotransmitters, short-chain fatty acid (SCFA), and other signaling molecules as well as inflammatory agents (e.g., lipopolysaccharide) that induce an immune response. The gut microbiota therefore interact with vital body systems, directly modulating host physiology, and by extension health and even behavior. Critically, although in its infancy, there is evidence to suggest that the gut microbiome plays a role in regulating the function of key aggression-related brain regions. 8 The Current Study The current study is an interdisciplinary exploration of the role of the gut microbiome in aggressive behavior, underscoring the translational relevance of my preclinical work with the Michigan State University’s Neuroscience Program. Human microbiome data were obtained through collaboration with the Human Post Mortem Microbiome (HPMM) research team who supplied secondary sequencing data from their study of the post mortem human microbiome across Wayne County, Michigan (Pechal, Schmidt, Jordan, & Benbow, 2018). Secondary sequencing data from samples representing the digestive track (i.e., mouth and rectum) collected from decedents during routine autopsy in 2014-2015 were used as a proxy measure for the ante mortem gut microbiome to investigate the relationship between microbial community structure and local area crime. These data supplied descriptive information, including the diversity and distribution of bacteria, as well as geospatial information for both the (1) death event and (2) home locations of each decedent, placing each subject in the local area where they had known social and environmental exposures. Violent crime data from the Detroit Police Department (DPD) as well as socioeconomic data from the United States Census Bureau (USCB) were paired with microbiome data to examine the relationship between disadvantage, the human gut microbiome, and patterns of local area aggression behavior. 9 CHAPTER 2: METHODOLOGY Site Selection Detroit is a recovering city, trying to regain itself after years of political corruption, rapid population decline, and the mass exodus of industry (LeDuff, 2013). Once considered America’s arsenal of democracy for its manufacturing efforts during World War II (Baime, 2014), it is now plagued with healthcare disparities, poverty, and crime. In 2015, Detroit’s population was estimated to be approximately 690,074 individuals, accounting for only 7% of the state of Michigan’s total population (U.S. Census Bureau, 2015) but a disproportionate amount of its disadvantage. For example, Detroit has a higher mortality rate for seven of the ten leading causes of death in Michigan, including heart disease, cancer, stroke, unintentional injuries, diabetes mellitus, pneumonia/influenza, and kidney disease compared to both Michigan and the United States at large (Michigan Department of Community Health, 2014). The 2015 American Community Survey estimates that 30% of Detroit’s 365,528 housing units are vacant. Among the homes that are occupied, only 49% are owner-occupied and a staggering 25% do not have access to a car in a city where there is limited public transportation. The median household income is $25,764, but 49% of households earn less than $25,000 per year. The unemployment rate is high with 25% of Detroit’s eligible workforce unemployed (U.S. Census Bureau, 2015). In addition to its poor health and poverty, Detroit has an infamous reputation. Popular media consistently characterizes Detroit as America’s most dangerous city, and even rank it among one of the most dangerous places in the world (Fisher, 2015; "The world’s most dangerous cities," 2017). According to the Federal Bureau of Investigation, the City of Detroit accounted for 36% of Michigan’s violent crime in 2015 and ranked in the top three major cities in violent crime and homicide (Federal Bureau of Investigation, 2017). Detroit is thus characterized by disadvantage, 10 instability, and crime. Given this unfortunate profile, Detroit is the ideal place to study how patterns of violent crime relate to local area social and environmental risk factors. Aggregate Data: The United States Census Bureau American Community Survey In an attempt to reconstruct contextual features at the time of HPMM data collection, 5- year socioeconomic variable estimates for Michigan in 2015 were obtained from the 2015-2019 American Community Survey (ACS). Unlike single-year estimates, five-year survey estimates are drawn from larger sample sizes with sampling at smaller census geographies and have the benefit of years of subsequent data collection to inform estimates, rendering five-year variable estimates the most reliable (Gaquin & Ryan, 2019). The 2015-2019 ACS data are the most recent five-year product release from the USCB and were selected because the majority of the HPMM samples were collected during 2015 (61%). Data from the 2010-2014 ACS were not included to reduce the chance of introducing variation in estimates between ACS surveys related to changing census geographies or question modification. Importantly, 2010 decennial census data were excluded from the current study in an effort to minimize threats to internal validity driven by the profound impact of the 2008 recession on economic security nationwide. The 2010 data reflect a distinct period of history that may inaccurately represent the socioeconomic conditions during 2014 and 2015. Therefore, the 5-year estimates for 2015 from the 2015-2019 ACS were collected for the State of Michigan. 11 Table 1 | 2015 American Community Survey Variables Study Variable Operationalization Unemployment Estimate of the total count of work eligible adults not in the labor force Household Income Estimate of the median household income during the past 12 months Population of Black Residents Estimate of the total count of Black residents Vacancy Estimate of the total count of vacant housing units Tenure Estimate of the total count of occupied housing units Population of Ages ≥ 16 Estimate of the total count of population aged 16 years or older Population Estimate of the total count of the population Unemployment Rate = Unemployment / Population of Ages ≥ 16 Years Percent of Black Residents = Population of Black Residents / Population Vacancy Rate = Vacancy / ( Vacancy + Tenure) Sociodemographic variables estimating the total population and population aged 16 and older (i.e., eligible workforce), unemployment, median household income, the representation of Black citizens, vacancy status, and total tenure were collected at the census block group level (Table 1), the smallest geographic unit of analysis available for the ACS, representing between 600 and 3,000 people clustered in a contiguous area (USCB, 2021a). All demographic measures were recorded by the Census per survey respondent self-identification and report. Census variables were collected in accordance with Office of Management and Budget’s Federal Interagency Work Group for Research on Race and Ethnicity. According to the most recent ACS estimates (USCB, 2019), the City of Detroit includes a diverse population that is 77.9% Black or African American, 14.7 White, 1.7% Asian, 0.7% American Indian or Alaskan Native, 3.3% of another race, and 1.7% multi-racial. While 77.9% of Detroit residents self-identify as Black or African American, this ranges from 83.1% to 100% per block group. Due to the current and historical practice of societal underinvestment in residential areas with high Black representation, the percentage Black could be a signal for under-investment not measured by other covariates and is therefore included 12 in these models. Altogether, these ACS data provide basic social and environmental factors that characterize Detroit block groups and serve as socioeconomic covariates in statistical modeling. Aggregate Data: Detroit Police Department Open-Source Crime Data Open-source violent crime data for the City of Detroit were obtained from the city’s open data portal. Specifically, victim-based data enumerating the number of aggravated assaults, nonfatal shootings, and homicides reported to the Detroit Police Department (DPD) during 2014- 2015 were collected. Aggravated assaults reported to DPD were collected from a larger dataset documenting major crimes reported to DPD during 2011-2014 that were compiled for submission to public safety surveillance programs. To extract aggravated assaults from these data, a crime category and date variable query was used to filter the dataset for aggravated assault crimes that occurred only during 2014. Likewise, data for nonfatal shootings reported to DPD in 2014-2017 were downloaded and filtered for incidents that occurred during 2014 and 2015. Finally, homicide data for incidents reported to DPD in 2014-2017 was downloaded and subset by the incident date variable to obtain years 2014 and 2015. All three datasets included latitude and longitude data, providing the location of where each incident occurred. These data were thus transformed into three separate spatial point datasets. These spatial point data were assigned to census block group polygons and summed to calculate a raw count of incidents within the block group. Totals were divided by the block group population and multiplied by 1,000 to generate per capita aggravated assault, nonfatal shootings, and homicide rates. Each block group was then categorized as having high, medium, or low levels of per capita assault, shootings, and homicide using quantile thresholds from variable summary statistics. These categorizations were subsequently used to code block groups as having high, medium, or low overall crime. A block group was considered to have high crime if it was also 13 coded as having high levels of assaults, shootings, and homicides. Likewise, a block group was considered to have low crime if it was coded as having low levels of assaults, shootings, and homicides. Notably, a designation of “medium” crime level can be interpreted as a variable crime level. Importantly, overall crime rates were not calculated given significant variation reporting across crime type (see Appendix). Together with USCB socioeconomic data, the combined data supplied critical contextual information for the human post mortem microbiome data. Aggregate Data: The Human Post Mortem Microbiome To examine the relationships between the human gut microbiome and local area crime and sociodemographic factors, secondary 16S ribosomal RNA (rRNA) amplicon sequencing data from the Human Post Mortem Microbiome (HPMM) project were obtained through collaboration with HPMM colleagues. In their original work, Pechal and colleagues (2018) collected post mortem microbiome samples from the ears, eyes, nose, mouth, and rectum from 188 decedents during routine autopsy with the Wayne County, Michigan, Medical Examiner (ME) between 2014 and 20151. Samples were subsequently characterized via a common genetic marker, allowing for the identification of constituent microbiota. In addition, ME reports provided basic demographic, health, and law enforcement information, detailing circumstances surrounding the individual’s death. Accompanying sample metadata therefore included variables for age, race, sex, body mass index (BMI), manner of death (e.g., natural), cause of death (e.g., cardiovascular disease), and the estimated postmortem interval, which leverages milestones in decomposition to determine the estimated time between death and discovery. Central to the current study, decedent death event and home location geospatial information were also collected. Together, these data provide covariates which have a role in shaping the human gut microbiome. 1 Some 2015 cases were processed in 2016. 14 The HPMM dataset has broad forensic applications, but principal among them is its ability to represent the ante mortem human microbiome. Indeed, the majority of the subjects were sampled within 24-48 hours of death, limiting the effects of decomposition on microbial community composition. Moreover, variation in the manner in which individuals died (i.e., natural, accidental, homicide, and suicide) led to the inclusion of decedents of varying age and health, creating opportunity to evaluate the post mortem microbiome as a tool for exploring living human health. As a first step, initial analyses showed stability in microbial community composition during the first 24-48 hours after death, especially among rectal samples, suggesting that decomposition processes do not dramatically shift microbial communities immediately after death. In a further proof of concept analysis, the initial HPMM investigation demonstrated that features of the post mortem microbiome of the mouth, such as decreased phylogenetic diversity, significantly predicted the presence of ante mortem heart disease among decedents found less than 24 hours after death. Taken together, the overall stability of the post mortem microbiome and predictive power of these data suggest that they may serve as an appropriate proxy measure of ante mortem microbial communities, rendering the human post mortem microbiome a viable tool for examining human health in living populations (Pechal et al., 2018). As such, 16S rRNA amplicon sequencing data from mouth and rectal samples (i.e., digestive track samples) from this study were used as proxy measures to characterize the human gut microbiome among decedents who lived and died across the City of Detroit. Aggregate Data United States Census Bureau To compare the effects of local area socioeconomic factors on the composition of the human post mortem microbiome, basic demographic information for the State of Michigan was 15 obtained from the United States Census Bureau (USCB) using a Census application programming interface (API) key2 and the tidycensus package for R and RStudio (Walker & Herman, 2021). The USCB is the United States federal repository for both the decennial census, a population enumeration effort in accordance with Article I, Section 2 of the United States Constitution, and the American Community Survey (ACS), an annual estimation of the nation’s demographic and economic characteristics (USCB, 2021b) for policy decision-making. The tidycensus package facilitates digital data collection from both of these surveys. a b c Figure 1 | Maps of Michigan Census Block Groups. a-c) Schematic showing the 2015 ACS census block groups (BG) represented in study geographies. a) Michigan, 8,205 BG; b) Wayne County, MI, 1,822 BG; c) Detroit, MI, 879 BG. Geospatial information for the State of Michigan was downloaded from USCB using a combination of the tidycensus and tigris packages for R and RStudio (Walker, 2020; Walker & Herman, 2021). The tidycensus package facilitates USCB survey data download with accompanying attribute information, including polygon geometry for the census geography to which the data are assigned (e.g., block group). The tigris package, however, facilitates the download of shapefiles for USCB geographies, such as metropolitan areas in specific states, providing critical cartographic information. Shapefiles store geospatial vector data that describe physical space as points (e.g., locations), lines (e.g., streets), and polygons (e.g., neighborhoods), 2 Census API requests can be made at https://api.census.gov/data/key_signup.html. 16 defining geographies for data selection and mapping operations. In particular, USCB observes legally bounded small geographies, such as counties and incorporated places, but also creates statistical entities for other small geographies, like census core-based statistical areas (i.e., metropolitan areas), to examine local data trends (USCB, 2012). To access the geographical boundaries observed during the 2015 ACS, geospatial information for counties, core-based statistical areas, and incorporated places in Michigan was collected using the tigris package. These data were subsequently integrated with census demographic and economic data to pare down, or clip, census survey data to study-relevant geographies. A series of data processing steps were conducted to isolate Wayne County and City of Detroit geospatial and ACS survey data (Fig 1a-c). Shapefiles for the legal boundaries observed during the 2015 ACS for Michigan counties and incorporated places were downloaded and queried using logical operators searching the “NAME” variable for “Wayne” and “Detroit”, respectively. Shapefiles for nationwide core-based statistical areas drawn for the 2015 ACS were also downloaded and similarly queried for “Detroit” to identify its metro area. Each geography was then joined with 2015 ACS demographic and economic data using the sf package for R and RStudio (Pebesma, 2018). Importantly, all Michigan ACS survey data that were assigned to block groups that did not fall within the shapefiles were eliminated, thereby “clipping” the census data to areas under study. The resulting geospatial dataset was projected to the World Geodetic System 1984 (WGS84) coordinate reference system for downstream processing and analysis. Detroit Police Department Open Source Crime Data The City of Detroit has an open data portal from which three datasets were downloaded (https://data.detroitmi.gov/). Specifically, datasets enumerating the number of aggravated assaults, nonfatal shootings, and homicides reported to the Detroit Police Department (DPD) during 2014- 17 2015 were collected. Aggravated assaults reported to DPD were collected from a larger dataset documenting major crimes reported to DPD during 2011-2014 that were compiled for submission to state and federal public safety surveillance programs, such as the Michigan Incident Crime Reporting (MICR)3 and National Incident Based Reporting System (NIBRS) data repositories. To extract aggravated assaults from these data, a crime category and date variable query was used to filter the dataset for aggravated assault crimes that occurred only during 2014. Likewise, a victim- based dataset documenting nonfatal shootings reported to DPD during 2014-2017 calendar years was downloaded and queried for incidents that occurred during 2014 and 2015 using the incident date variable. Finally, a third, victim-based dataset detailing homicides reported to DPD during 2014-2017 calendar years was downloaded and queried for incidents that occurred during 2014 and 2015 using the incident date variable. These data are victim-based, spatial point datasets with latitude and longitude coordinates projected to the WGS84 coordinate reference system, providing the precise location of where each incident occurred for assignment to census block group polygons. Human Post Mortem Microbiome Data Secondary 16S rRNA amplicon sequencing data from the Human Post Mortem Microbiome (HPMM) project (www.ebi.ac.uk/ena; PRJEB22642) were obtained through collaboration with HPMM colleagues. For the original project, Pechal and colleagues (2018) collected post mortem microbiome samples from five different body locations, including the ears, eyes, nose, mouth, and rectum, for 188 decedents during routine autopsy examinations conducted by the Wayne County, Michigan, Medical Examiner (ME) between 2014 and 20154. The ME 3 The MICR data was extensively evaluated for this project and deemed insufficient due to inadequate address data entry. 4 Some 2015 cases were processed during 2016. 18 report also provided basic demographic and health information as well as the circumstances surrounding the individual’s death. Sample metadata therefore included variables for age, race, sex, body mass index (BMI), manner of death (e.g., natural), cause of death (e.g., cardiovascular disease), and the estimated postmortem interval, denoting the estimated time between death and discovery. Critically, the metadata include geospatial information for the decedent’s home and death event locations. Together, these data constitute the basic factors which could drive the microbial profile reflected in the sample data. The HPMM dataset provides geospatial information for decedent home and death event locations, both of which were geocoded to examine the distribution of HPMM sample characteristics in physical space. Specifically, location information was provided via geographical coordinates and/or postal address, depending on the nature of the circumstances surrounding the decedent and his death. For example, some decedents were homeless and therefore had no known home location. To avoid excluding these individuals from analysis, their death event locations were duplicated as an approximation of their sleep location, and by extension, home area (Pearson et al., 2019). Additionally, some deaths occurred at locations without postal addresses, like intersections or open fields. These death events were manually assigned the next nearest postal address to the given coordinates or intersection using Google Maps Street View. Together, these strategies facilitated the collection of standard postal addresses for all home and death event locations in the dataset. These standard postal addresses were subsequently formatted for automated geocoding using the ggmap package for R and RStudio (Kahle & Wickham, 2013), which leverages Google Maps to generate geographical coordinates for known postal addresses. Together, these strategies yielded geocoordinates for a complete spatial point dataset that was 19 filtered for locations within Detroit and subsequent block group assignment (see below), thereby rendering characteristics of decedent gut microbiome spatial attributes. Analytic Strategy Microbiome Data Analysis (performed with S. Kaszubinski) To obtain meaningful sample characterizations from raw DNA sequences, a series of computationally-intensive data processing steps were conducted. First, the study team’s collaborating HPMM forensic data analyst pulled raw sequences through a standardized data processing pipeline to classify taxa present in the samples using the Quantitative Insights Into Microbial Ecology 2 (QIIME 2) program v2018.11 (Bolyen et al., 2019). Briefly, this entailed assembling paired-end sequence reads from the raw fastq files provided by MSU Genomics Core, removing poor quality sequences, and taxonomic assignment, the details of which are outlined by the HPMM team elsewhere (Kaszubinski, Pechal, Schmidt, et al., 2020). For classification, sequences were binned into operational taxonomic units (OTUs), or working groups of sequences with 99% similarity. From these OTUs, representative sequences were aligned to the SILVA small subunit database v132 (Quast et al., 2013). Non-bacterial sequences were removed from the dataset and final taxonomy tables were exported to CSV files to be used as input data for downstream analysis. Next, the taxonomy tables were imported into R and RStudio for further data processing. In these final steps, a long-form dataset was created such that each OTU represented a row and every column was one of the decedents’ five samples. The OTUs were labeled according to their taxonomic classification (down to the genus level), empty rows were removed, and the reformatted table was merged with combined study metadata using the phyloseq package for R and RStudio (McMurdie & Holmes, 2013). In an effort to standardize sample microbial 20 communities, taxa with less than 0.01% of the mean library size, or the number of sequences, were dropped from the dataset. Sequence libraries were subsequently normalized via rarefaction, whereby sequence libraries were randomly subsampled to a specified minimum library size to prevent the effects of sample size bias on microbial community composition. Guided by rarefaction curves, human and mouse data were rarefied to standardize library size. Using normalized sequence libraries, measures of relative abundance and community diversity metrics were calculated to characterize the decedent’s gut microbiome. Statistics and Model Selection Statistical analysis was conducted using a combination of R and RStudio version 4.0.5 (R Core Team, 2021; RStudio Team, 2020). Data distributions were examined with Shapiro-Wilks tests of normality and Bartlett’s tests of homogeneity of variance to select the appropriate parametric or nonparametric tests. Dependent and Independent Variables For this analysis, the primary dependent variable was an ordinal, pseudo-composite crime variable with three levels describing census block groups with high, medium (i.e., fluctuating), and low crime relative to levels of assault, nonfatal shootings, and homicide. To construct this categorical measure of overall crime, quantile thresholds were used to designate each block group as having low, medium, or high levels of per capita assault, shootings, and homicide. These designations within each crime type were subsequently used to code block groups as having high, medium, or low overall crime. Importantly, overall crime rates (as opposed to the categorical variable described here) were not calculated given variation in reporting across crime type. This variable was selected from the suite of described crime variables (i.e., per capita rates) during the 21 initial stages of data analysis with exploratory analyses that established the potential microbiome- crime exposure link. Independent variable collection was informed by criminological research and theory that emphasizes the role of contextual factors in disrupting social processes in human communities. As such, USCB socioeconomic variables, including household income, vacancy rate, unemployment rate, and the percentage of residents who identify as Black in census block groups were used alongside HPMM variables that characterize the human post mortem gut microbiome. These included measures of relative abundance as well as alpha and beta diversity metrics. Relative abundance was calculated as a value between 0 and 1, indicating the percentage of the community that each operational taxonomic units (OTUs), or working groups of sequences with 99% similarity, occupies (McGill et al., 2007). This measure was used to compare the distribution of microbiota between samples across different levels of local area crime. Relatedly, diversity metrics were used to describe the structure of microbial community membership. In particular, alpha diversity measures, including observed richness, Chao1 diversity, Shannon diversity, and Inverse Simpson diversity, were used as summary statistics to quantify the richness and evenness of taxa within samples. Conversely, beta was used as a summary statistic of (dis)similarity, or the degree of overlap, among samples within each crime level condition (Morgan & Huttenhower, 2012). The diversity of each sample was compared to the diversity of every other sample, generating a dissimilarity matrix among the sample microbial communities. Together, measures of relative abundance and community diversity provide basic descriptive statistics for microbiome community structure that were used to determine the extent to which features of the human post mortem gut microbiome could predict what type of living crime exposure decedents experienced, accounting for all other socioeconomic factors. 22 Research Questions In this cross-sectional, non-experimental investigation of violent crime, characteristics of the human post mortem gut microbiome were compared between census block groups5 with varying levels of violent crime. These comparisons were made with respect to both (1) death event and (2) home locations. In doing so, three primary research questions were answered: 1. Is there a relationship between the human gut microbiome and local area violent crime levels? 2. Does the human gut microbiome vary according to (i) local area crime level and (ii) other socioeconomic factors? 3. Which taxa, if any, drive the relationship between the human gut microbiome and local area socioeconomic factors? Analysis Given the exploratory nature of this project, data analysis was conducted in a series of data- driven steps. Sample data were subdivided by body site and rectal and mouth samples were separately analyzed. Model building proceeded with bivariate comparisons between microbiome diversity metrics and structural features and crime variables, including block group crime level as well as per capita assault, nonfatal shooting, and homicide rates. The distribution of microbiome sequencing data is unknown but non-normal, zero-inflated, and overdispersed, requiring the use of nonparametric tests. Alpha diversity metrics (i.e., observed richness, Chao1, Shannon diversity, and Inverse Simpson diversity) were calculated for both mouth and rectal samples. Shapiro-Wilks 5 This analysis was repeated using census tracts and DPD police precincts as the unit of analysis. These analyses returned negative findings, likely due to introducing variation in study variables aggregated to larger geographic areas. 23 normality tests confirmed the use of nonparametric correlations with Kendall’s tau and Kruskal- Wallis tests. Here, a Bonferroni correction was used to correct for multiple comparisons. For beta diversity, permutational multivariate analysis of variance (PERMANOVA) was used with 999 permutations and concomitant tests for differences in beta dispersion, or homogeneity of variance, between conditions. Permutational analysis of variance is a nonparametric model that tests group differences, similar to parametric analysis of variance (ANOVA). Instead of means testing, however, PERMANOVA evaluates group differences based on user-specified distances that reflect similarities between sample clusters (M. J. Anderson, 2001, 2017). As such, PERMANOVA evaluates differences across centroids of clusters (e.g., block group crime level). Here, dissimilarity matrices were calculated using Unifrac distances, thereby taking phylogenetic relationships into account in exploring differences between samples. In accordance with other HPMM research (Kaszubinski, Pechal, Smiles, et al., 2020), multinomial logistic regression was selected to model the probability of membership in one of the three block group crime level categories using maximum likelihood estimation. Multinomial logistic regression is an ideal modeling technique because it does not require normally distributed data, linearity, or homoscedasticity. In this case, the model was used to predict block group placement in one of the three crime level categories using a combination of socioeconomic variables and microbiome measures collected from decedents assigned to the block groups as either their death event or home location. Model building was conducted in a stepwise fashion, beginning with diversity metrics and proceeding with the subsequent addition of socioeconomic covariates. Models were evaluated for goodness-of-fit with the Akaike information criterion (AIC) and log likelihood indices relative to a null model. Model performance was assessed via model 24 rate of correct block group (BG) classification, but models were compared to the null model using a Chi-squared difference test. McFadden’s pseudo R2 was also reported. To investigate whether specific taxa or socioeconomic predictors drive the relationship between gut microbiome composition and local area crime, a random forest classification model with 500 trees was built (Kaszubinski, Pechal, Schmidt, et al., 2020; Y. Zhang et al., 2019). Random forest modeling is a machine learning strategy that utilizes a supervised learning approach to classify data within large and complex datasets. Briefly, labeled training data that pairs input values (e.g., socioeconomic and OTU data) with their class label (e.g., block group crime level) is used to derive a function to classify novel, unlabeled inputs, or test data. This cross-validation step assigns prediction error to the model, which can be leveraged to generate an importance score for specific taxa (or other metadata) by estimating the increase in error associated with removing them as a predictor. Random forest classification was piloted to identify OTUs and socioeconomic predictors that differentiate the composition of the gut microbiome between high, medium, and low violent crime levels. Given that random forest modeling is sensitive to unbalanced groups, however, alternative modeling strategies were implemented. In particular, negative binomial mixed models (NBMM) with maximum likelihood estimation were used as a follow-up to random forest modeling to identify specific taxa associated with block group crime level and per capita homicide. This modeling strategy was selected to account for features of microbiome data that are poorly dealt with by other analytic strategies (Zhang et al., 2017; Zhang & Yi, 2020). Specifically, microbiome data are count data with natural dependencies created by an inherent structure related to taxonomic levels (i.e., phylum, genus), phylogenetic relationships, gene function, etc. Moreover, the relative abundance of an OTU necessarily affects the abundance of all others in the sample. Together, these features violate the 25 classic assumption of independence that most statistical tests require. In addition, sample OTU data are nested within each decedent, incorporating unknown host factors that thus become random effects. Models thus included OTUs as the modeled count variable, crime and socioeconomic variables as fixed effects, and sample ID, a decedent indicator, as the random effect. To incorporate the crime level variable, low, medium, and high crime conditions were dummy coded with the low condition as the reference category. Overall, NBMM was utilized to identify microbiota that are differentially abundant, depending on contextual factors. 26 CHAPTER 3: RESULTS Overview The current study demonstrates a potential link between human gut microbiome composition and local area crime level. Secondary sequencing data from both mouth and rectum samples were used to calculate diversity metrics that describe the human post mortem microbiome as a proxy measure for the ante mortem gut microbiome. These metrics were subsequently assigned to both the decedents’ death event and home locations to determine if the structure of the gut microbiome was unique in different physical contexts such that it could be used to predict the local area crime level, and by extension, the types of social and environmental exposures in the block group. Demographics The final sample included 98 decedents with a death event or home location in the City of Detroit, 85 of whom both lived and died in Detroit. The full sample (n = 98) was comprised of decedents with an average age of 43.9 years and an average body mass index (BMI) of 27.4. Fifty- nine percent were male and 41% female. There were only two racial identities represented in the sample, 76% of whom were Black and 24% of whom were White. Most cases occurred in the Spring (52%) with the majority of decedents found less than 48 hours post mortem (89%). The manner in which decedents died included accident (34%), homicide (33%), suicide (6%), and natural (28%) events, categories that included drug-related (29%), gunshot (26%), cardiovascular (22%), and other (23%) causes. From the full sample, 93 had complete case information on their death event location and 82 had complete case information related to their home location. Table 2 presents basic demographic information on the sample subdivided by block group, or local area, crime level. 27 Samples The original HPMM contained 855 samples representing five body areas and 30,906 taxa. From this dataset, 162 rectum and 173 mouth samples were extracted. Libraries with less than 0.01% the mean number of sequences were trimmed, and libraries were rarefied to 5,000 reads. After these quality control and normalization measures, Detroit samples were selected. There were a total of 146 rectum samples, representing 1,191 taxa, and 170 mouth samples, representing 913 taxa, for death event and home locations in the final Detroit HPMM dataset (Table 3). 28 Table 2 | Wayne County Medical Examiner Case Report Data Death Event Location Home Location Low Medium High Low Medium High Crime Crime Crime Crime Crime Crime (n = 6) (n = 75) (n = 12) (n = 8) (n = 72) (n = 2) Age Mean 41.3 44.2 42.2 47.1 43.8 46.0 SD (11.3) (14.5) (15.9) (16.0) (13.9) (26.9) BMI Mean 26.2 27.6 26.7 26.9 27.5 30.4 SD (4.01) (7.15) (5.21) (3.43) (7.04) (3.96) Sex Male 1 31 5 4 44 2 Female 5 44 7 4 28 0 Race White 3 16 5 2 15 1 Black 3 59 7 6 57 1 MoD Accident 1 22 9 1 26 2 Homicide 2 28 1 3 24 0 Natural 3 21 1 4 19 0 Suicide 0 4 1 0 3 0 Event Hospital 0 16 8 1 19 0 Location Indoors 5 42 2 4 41 1 Outdoors 1 13 1 2 8 1 Vehicle 0 4 1 1 4 0 PMI < 48 HRS 5 65 11 7 62 2 > 48 HRS 1 10 1 1 10 0 Table 3 | Human Post Mortem Microbiome Sample Count by Body Site Death Event Home Location (n = 164) (n = 152) Low Medium High Total Low Medium High Total Crime Crime Crime Crime Crime Crime Rectum 5 60 11 76 6 60 4 70 Mouth 6 69 13 88 7 70 5 82 29 Alpha Diversity | Death Event Location Crime Level Figure 2 | Alpha Diversity by Death Event Location. Schematic showing the distribution of alpha diversity values for observed richness, Chao1, Shannon diversity, and Inverse Simpson diversity in mouth and rectal samples across low, medium, and high levels of census block group crime at the decedents’ death even location. There were significant differences across levels of crime (p < 0.05) for each alpha diversity metric.in rectal samples. There were no significant findings in mouth samples. Alpha diversity is associated with block group crime level at decedent death event location To compare richness and evenness by local area crime level at the decedent's death event location, Kruskal-Wallis tests were conducted with crime level serving as the grouping variable. Though observed richness in rectum samples and Shannon diversity in mouth samples were normally distributed, sample size within conditions varied markedly between low, medium, and high crime areas, rendering the use of one-way ANOVA tests a less conservative choice. As such, 30 Alpha Diversity | Home Location Crime Level Figure 3 | Alpha Diversity by Home Location. Schematic showing the distribution of alpha diversity values for observed richness, Chao1, Shannon diversity, and Inverse Simpson diversity in mouth and rectal samples across low, medium, and high levels of census block group crime at the decedents’ home location. There were no significant differences in alpha diversity (p > 0.05) in mouth or rectum samples. Kruskal-Wallis tests with post hoc Nemenyi tests were globally employed to compare measures of alpha diversity across local area crime level at the death event locations. Results show a significant difference in the alpha diversity of rectum samples (n = 76) between low (n = 6), medium (n = 59), and high (n = 11) crime areas when comparing the decedents’ death event locations. Results indicate that there are significant differences in observed richness (H(2) = 8.1103, p = 0.017), Chao1 (H(2) = 7.842, p = 0.020), Shannon diversity (H(2) = 9.2975, p = 0.010), and Inverse Simpson diversity (H(2) = 7.3407, p = 0.025) indices among rectum samples across local area crime levels (Fig 2; Table 4). Specifically, observed richness among decedents found in areas with low (p = 0.036) and medium (p = 0.035) crime was significantly higher compared to high crime areas, but there was no difference between low and 31 medium crime areas (p = 0.505). Likewise, the Chao1 index, an indicator of sample richness, was significantly higher in low and medium crime areas compared to high crime areas; though, there was no difference in Chao1 index between medium and low crime areas (p = 0.515). Findings for Shannon diversity, an alpha diversity measure that accounts for both the abundance and distribution of taxa, follow the same pattern with significantly higher alpha diversity in low (0.013) and medium (0.038) crime areas compared to high crime areas but no difference between medium and low crime areas (p = 0.275). Finally, results show significantly higher Inverse Simpson diversity, another alpha diversity measure that accounts for richness and evenness, among low crime areas compared to high crime areas (p = 0.022), but no differences between medium and low (0.190) or medium and high crime areas (0.147) were found. There were no significant differences between low (n = 6), medium (n = 69), and high (n = 13) crime areas for any of the alpha diversity metrics among decedent mouth samples (n = 88). In addition, alpha diversity was further investigated by local area crime in the decedents’ home location. Repeat analysis according to the decedents’ home location showed no significant differences in alpha diversity in either mouth or rectal samples (Fig 3; Table 5). In sum, multiple metrics demonstrate significant differences in the alpha diversity of rectum samples between block group crime level assigned to the decedent death event, but not home, locations. Table 4 | Alpha Diversity by Death Event Location Crime Level Rectum Mouth M (SD) M (SD) Low Med High Low Med High Observed 140.600 104.550 65.455 80.500 82.652 78.385 Richness (72.189) (43.339) (34.992) (39.399) (49.226) (36.864) Chao1 Diversity 147.723 108.791 68.641 84.348 86.538 82.029 (77.443) (46.549) (36.913) (41.012) (51.691) (39.401) Shannon 3.805 3.320 2.715 2.520 2.783 2.827 Diversity (0.787) (0.714) (0.863) (0.833) (0.887) (0.706) Inverse Simpson 31.514 16.702 10.825 7.620 11.167 10.870 (17.476) (8.804) (6.475) (6.035) (9.095) (7.930) 32 Table 5 | Alpha Diversity by Home Location Crime Level Rectum Mouth M (SD) M (SD) Low Med High Low Med High Observed 112.667 102.850 71.500 70.714 82.029 95.800 Richness (64.920) (44.958) (19.416) (30.939) (47.265) (68.430) Chao1 119.311 107.227 72.857 74.121 85.938 97.887 Diversity (70.590) (48.121) (19.817) (31.429) (49.873) (70.579) Shannon 3.521 3.272 3.075 2.787 2.768 3.009 Diversity (0.681) (0.792) (0.556) (0.558) (0.880) (0.854) Inverse 22.204 16.399 14.113 10.275 10.706 13.428 Simpson (14.574) (8.699) (7.029) (7.857) (8.368) (13.116) Alpha diversity is not correlated with block group per capita assault, nonfatal shootings, or homicide at decedent death event or home locations To determine if alpha diversity of both rectum and mouth samples was associated with per capita rates of aggravated assault, nonfatal shootings, and homicide, nonparametric correlations were conducted using Kendall's tau with a Bonferroni correction to adjust for multiple comparisons. Findings revealed no significant associations between any of the continuous crime measures and observed richness, Chao1, Shannon diversity, or Inverse Simpson diversity in either the decedents’ death event (Fig 4-5) or home location (Fig 6-7), thereby confirming the choice of the categorical crime level variable as the primary dependent variable in subsequent analyses. In view of these findings, the rectum sample dataset for decedent death event location was selected for further analysis. To foretell any issues of multicollinearity while modeling, relationships between predictor variables were explored via Spearman rank correlation tests (Fig 8). Here, p-values were not adjusted to consider every source of potential model contamination. Unsurprisingly, household income was negatively associated with both block group 33 unemployment rate (r = -0.345, p < 0.0001) and vacancy rate (r = -0.247, p < 0.0001). Likewise, block group vacancy rate was also positively associated with unemployment rate (r = 0.146, p < 0.0001). In addition, there were small but significant correlations between the percentage of Black residents in a block group and socioeconomic and per capita crime variables. The percentage of Black residents was significantly associated with household income (r = -0.193, p < 0.0001), unemployment rate (0.091, p = 0.01), and vacancy rate (0.145, p = 0.01) as well as reductions in per capita nonfatal shootings (r = -0.222, p < 0.0001) and assault (r = -0.132, p < 0.0001), underscoring the potential of informal social control mechanisms and guardianship in driving down local area crime. Overall, these data demonstrate important covariation in the dataset that must be further examined during model building. 34 Alpha Diversity | Rectum Samples by Death Event Location Figure 4 | Rectum Sample Alpha Diversity and Per Capita Crime by Death Event Location. a-d) Correlograms depicting correlation results using Kendall’s tau to determine the relationship between block group per capita assault, nonfatal shootings, and homicide at decedent death event location and a) observed richness, b) Chao1, c) Shannon, and d) Inverse Simpson diversity in rectum samples. The color of the squares represents the strength and direction of the correlation coefficient, and the values represent p-values adjusted with a Bonferroni correction. 35 Alpha Diversity | Mouth Samples by Death Event Location Figure 5 | Mouth Sample Alpha Diversity and Per Capita Crime by Death Event Location. a-d) Correlograms depicting correlation results using Kendall’s tau to determine the relationship between block group per capita assault, nonfatal shootings, and homicide at decedent death event location and a) observed richness, b) Chao1, c) Shannon, and d) Inverse Simpson diversity in mouth samples. The color of the squares represents the strength and direction of the correlation coefficient, and the values represent p-values adjusted with a Bonferroni correction. 36 Alpha Diversity | Rectum Samples by Home Location Figure 6 | Rectum Sample Alpha Diversity and Per Capita Crime by Home Location. a-d) Correlograms depicting correlation results using Kendall’s tau to determine the relationship between block group per capita assault, nonfatal shootings, and homicide at decedent home location and a) observed richness, b) Chao1, c) Shannon, and d) Inverse Simpson diversity in rectum samples. The color of the squares represents the strength and direction of the correlation coefficient, and the values represent p-values adjusted with a Bonferroni correction. 37 Alpha Diversity | Mouth Samples by Home Location Figure 7 | Mouth Sample Alpha Diversity and Per Capita Crime by Home Location. a- d) Correlograms depicting correlation results using Kendall’s tau to determine the relationship between block group per capita assault, nonfatal shootings, and homicide at decedent home location and a) observed richness, b) Chao1, c) Shannon, and d) Inverse Simpson diversity in mouth samples. The color of the squares represents the strength and direction of the correlation coefficient, and the values represent p-values adjusted with a Bonferroni correction. 38 Figure 8 | Correlations for Model Covariates. Correlogram showing significant correlations among model covariates collected for decedent death event locations. Notably, there are significant relationships between household income and i) unemployment rate and ii) vacancy rate. There is also a significant association between unemployment rate and vacancy rate. The color of the squares represents the strength and direction of the correlation coefficient, and the values show unadjusted p-values. 39 Random forest classification fails with dramatically imbalanced groups In order to determine the OTUs driving the relationship between alpha diversity in rectum samples and block group crime level at decedent death locations, a random forest classification model was built. For this analysis, decedent demographic metadata and USCB socioeconomic covariates from their assigned block group were included with OTU data as predictors of block group crime level membership. Using 500 decision trees and 92 predictors at each decision tree split, the error estimate was 21.05%, suggesting that model predictors correctly predicted sample membership to its assigned block group 78.95% of the time. However, this low error rate was a result of the low number of cases in the low (n = 5) and high (n = 11) crime conditions being misclassified in the medium (n = 60) crime level group 100% of the time. As such, negative binomial mixed models were used as an alternative strategy to identify taxa that differ in abundance between low, medium, and high crime areas. Block group vacancy rate and alpha diversity drive block group classification To investigate the extent to which alpha diversity measures can predict block group placement in one of the three crime level categories beyond known risk factors for neighborhood dangerousness, sets of multinomial logistic regression models were fitted, corresponding to the four alpha diversity metrics presented (i.e., observed richness, Chao1, Shannon diversity, and Inverse Simpson diversity). Models were compared for goodness-of-fit relative to a null model. First, single variable models were fitted to assess standalone performance of model covariates (Models A1-7; Table 6). Findings indicate a significant effect of block group unemployment rate, vacancy rate, and Inverse Simpson diversity in predicting block group crime level; however, these models showed only a modest improvement in goodness-of-fit compared to the null model. Therefore, significant covariates from Models A1-7, including unemployment rate, vacancy rate, 40 and the Inverse Simpson diversity, were selected for further model building, accounting for population homogeneity. These predictors were used to construct Models B-G, comparing model fit to a null model. The addition of socioeconomic covariates improved model fit and block group classification in low, medium, and high crime conditions (Tables 7-8). Models B-G were all significant (Table 7), but the overall best performing models were Models D & G. Both Models D & G successfully predicted block group crime level 81.9% of the time. However, model fit was best for Model G. Model G shows a significant effect of block group vacancy rate such that for every one-unit increase in vacancy rate, the odds of classification in a high crime level block group increase (β = 24.3, p = 0.03). Likewise, for every one-unit decrease in decedent Inverse Simpson diversity, the odds of classification in a high crime block group increase (β = -0.17, p = 0.02). Importantly, unlike in random forest modeling, classification was not driven by the imbalanced groups. Block groups in every crime condition were both correctly and incorrectly classified, contributing to overall model performance. Taken together, these data demonstrate that there is a common gut microbiome structure within crime level conditions that drives block group classification. Table 6 | Single Variable Multinomial Logistic Regression Models A1-7 (n = 72) β SE p-value AIC % Black Residents 1.76 1.33 0.186 47.11 Unemployment Rate 10.85 4.57 0.018 42.19 Vacancy Rate 17.49 6.68 0.009 36.77 Observed -0.02 0.01 0.053 36.47 Chao1 -0.02 0.01 0.054 36.58 Shannon -2.08 1.11 0.062 35.52 Inverse Simpson -0.13 0.05 0.005 30.13 41 Table 7 | Multivariate Multinomial Logistic Regression with Inverse Simpson Diversity Inverse Simpson % Black Unemployment Vacancy Model B β -0.133 SE (0.047) p-value 0.005 Model C β -0.174 4.285 SE (0.064) 2.144 p-value 0.006 0.046 Model D β -0.113 -- 17.028 SE (0.054) -- 7.262 p-value 0.037 -- 0.019 Model E β -0.161 4.262 16.693 SE (0.077) 2.539 7.623 p-value 0.037 0.093 0.029 Model F β -0.193 2.853 -- 23.290 SE (0.083) 2.074 -- 11.516 p-value 0.020 0.169 -- 0.043 Model G β -0.171 -- -- 24.304 SE (0.071) -- -- 10.858 p-value 0.016 -- -- 0.025 Table 8 | Multivariate Multinomial Logistic Regression Model Fit Model Log-likelihood McFadden R2 X2 p BG % AIC Correct Correct Null -59.4 0.0 0.0 1.0 NA NA 46.8 Model B -41.1 0.14 13.9 0.0 58 80.6% 30.1 Model C -35.2 0.27 25.8 0.0 54 75.0% 27.3 Model D -37.0 0.23 22.1 0.0 59 81.9% 23.3 Model E -31.8 0.34 32.5 0.0 56 77.8% 20.8 Model F -30.3 0.37 35.5 0.0 59 81.9% 24.8 Model G -35.5 0.26 25.2 0.0 59 81.9% 21.0 Residential instability, population heterogeneity, and low socioeconomic status are important predictors of gut microbiome composition Microbiome OTUs, binned sequencing data used to describe bacterial abundance, from rectum samples assigned to decedent death locations were modeled using crime and socioeconomic variables to identify microbiota that have a significant relationship with block 42 group contextual factors. Given the significant relationships among socioeconomic covariates, four separate models were fitted using a dummy coded block group crime level variable alone (Fig 9; Model 1) as well as block group vacancy rate (Model 2), unemployment rate (Model 3), and percentage of Black residents (Model 4) together with block group crime level as fixed effects. In all four models, crime level and socioeconomic factors were significant predictors in modeling OTU data (see Appendix), with socioeconomic factors markedly increasing the number of OTUs predicted by block group contextual factors compared to the crime variable alone. Together, these four models collectively predicted the abundance of 52 different taxa (see Appendix), demonstrating local area contextual factors drive gut microbiome composition. Importantly, these Figure 9 | Microbial Families Associated with Crime Level. Schematic depicting 25 bacteria families significantly associated with block group crime level (Model 1). Each point corresponds to one taxon and its color corresponds the family. The error bars represent standard error of the estimate. 43 data supply critical evidence that gut microbiome composition varies according to local area crime exposures. Given that the primary dependent variable is categorical and there are limited cases in the low and high crime conditions, a follow-up NBMM analysis modeling OTU data with block group per capita homicide was conducted. Per capita homicide was selected for this supplemental analysis because it is considered a reliable indicator of local crime, and subsequently used in place of crime level as the model’s fixed effect. Results were similar to finding from Model 1, returning 18 taxa that were significantly associated with block group per capita homicide (Fig 10). Figure 10 | Microbial Families Associated with Per Capita Homicide. Schematic depicting 18 bacteria families significantly associated with block group per capita homicide. Each point corresponds to one taxon and its color corresponds the family. The error bars represent standard error of the estimate. 44 Interestingly, both crime variables have a significant relationship with Lachnospiraceae and the Clostridia class, in general, providing a direct connection to the project’s preclinical findings. Beta diversity did not differ between block group crime conditions To investigate the differences in beta diversity across levels of local area crime, a series of PERMANOVA tests were conducted using Unifrac distances. Results returned no significant difference in beta diversity among rectum samples across levels of block group crime at decedent death event locations (F(2) = 1.0534, R2 = 1.0534, p = 0.307). There were also no significant differences in beta dispersion among rectal samples between crime conditions (F(2) = 0.780, p = 0.45). Similarly, there were no significant differences detected for beta diversity of mouth samples by block group crime level at decedent death event locations (F(2) = 1.1527, R2 = 0.026, p = 0.166), nor any difference in beta dispersion across conditions (F(2) = 0.6643, p = 0549). A similar pattern emerged from analyses by decedent home location. PERMANOVA results indicated no difference in beta diversity of rectum samples between low, medium, and high crime block groups at decedent home locations (F(2) = 1.194, R2 = 0.034, p = 0.132). There was also no difference in home location beta dispersion of rectum samples (F(2) = 0.1468, p = 0.874). Finally, there were no significant differences in beta diversity (F(2) = 0.8684, R2 = 0.022, p = 0.77) or beta dispersion (F(2) = 1.1377, p = 0.323) of mouth samples across crime conditions at decedent home locations. Overall, this suggests that there is no significant deviation in microbiome community structure across block group crime level. 45 CHAPTER 4: DISCUSSION In this dissertation, findings demonstrate a potential link between human gut microbiome composition and local area crime level. Specifically, secondary sequencing data from mouth and rectum samples were used to calculate diversity metrics that describe the human post mortem microbiome as a proxy measure for the ante mortem gut microbiome. These metrics were subsequently assigned to both the decedents’ death event and home locations in an effort to determine if the structure of the gut microbiome was unique in different physical contexts such that it could be used to predict the local area crime level, and by extension, the types of social and environmental exposures in the block group. The data show a significant negative relationship between alpha diversity of rectum samples and block group crime level assigned to decedent death event locations. Negative binomial mixed models for rectum samples at the decedent death event location revealed 52 unique taxa that are associated with block group socioeconomic and crime variables. Importantly, subsets of taxa were identified that are associated with both block group crime level and per capita homicide. Further analysis (of rectum samples assigned to decedent death event locations) with multinomial logistic regression was used to determine the extent to which the structure of the human gut microbiome is a standalone marker of ante mortem social and environmental exposures. In sum, these data suggest that there is a link between human gut microbiome composition and local area violent crime level. The current study demonstrates an overall loss of gut microbiome diversity and change in distribution of certain taxa as local area crime increases, supplying a novel mechanism through which neighborhood context could affect health and behavior. This finding is in line with criminological research that has already established a link between environmental factors and local 46 crime patterns. For example, previous research documents that exposures to lead and other toxins in the environment are highest in impoverished areas (Mohai & Saha, 2006, 2007), and that those exposures increase the likelihood of neurocognitive deficits and externalizing behaviors (Denno, 1990; Narag, Pizarro, & Gibbs, 2009). This important criminological research provides well- documented evidence that bridges the first connection between physical context and human physiological states. Complementary neuroscientific research demonstrates that exposure to heavy metals, including lead, induces excitotoxicity and oxidative stress that causes cell death in the brain (Bondy, 2021; Li, Xia, Zorec, Parpura, & Verkhratsky, 2021), adding a molecular mechanism that explains the observed neurocognitive changes and behavioral deficits among those with high toxin exposure in their environment. Importantly, research shows that toxin exposure is positively associated with local violent crime patterns, especially in areas with the highest resource deprivation (Stretesky & Lynch, 2004), showcasing the interaction between human physiology and the environment. The current work shows a relationship between gut microbiome composition and local area crime that is modulated by socioeconomic factors, thereby expanding on this approach by combining criminological and basic neuroscientific research to provide another potential mechanism that may mediate the effect of a disadvantaged neighborhood context. The data presented here confirm with multiple metrics that the decedents with exposure to high crime areas have overall lower biodiversity in their gut microbiomes. This result is consistent with previous research that demonstrates a reduction in biodiversity associated with urban blighting (Pearson et al., 2019). Multiple studies demonstrate the relationship between crime and urban blight (Bogar & Beyer, 2015; Branas, Rubin, & Guo, 2012; Spelman, 1993), with some showing reductions in certain types of crime secondary to green remediation (Branas et al., 2016; Kondo et al., 2018). It is therefore unsurprising that, along with alpha diversity indicators, the best 47 model predicting block group crime level was its vacancy rate with high vacancy areas having higher crime rates. Though it is true that reductions in microbial diversity are associated with urban blight and blight is associated with increases in crime, evidence also suggests that environmental exposures drive gut microbiome composition (Afshinnekoo et al., 2015; Von Ehrenstein et al., 2000; von Mutius & Vercelli, 2010; Yatsunenko et al., 2012), raising important questions about other exposures in high crime areas that could affect gut microbiome composition. Previous work has demonstrated the forensic applications of the HPMM in its ability to predict forensic case attributes; however, the performance of rectal samples in predictions is relatively poor (Kaszubinski, Pechal, Smiles, et al., 2020; Pechal et al., 2018; Y. Zhang et al., 2019). Given that the post mortem microbiome of the rectum is especially stable (Pechal et al., 2018), the current study’s findings suggest that the microbial profile in post mortem rectal samples has potential non-forensic applications. To this end, the current study shows that rectal samples assigned to decedent death event location can be used in tandem with socioeconomic factors to predict local area crime level, establishing a potential gut-brain link related to differential exposures in low, medium, and high crime areas. Interestingly, the current study failed to show a relationship between measures of alpha diversity (i.e., observed richness, Chao1, Shannon diversity, and Inverse Simpson diversity) and any other measure of crime, suggesting that the relationship between the gut microbiome and local crime level exists in severe contexts where crime is very high or very low. Though the specific effect of crime exposure on gut microbiome composition is unknown, there are two principal mechanisms through which very high crime exposure could negatively affect the human gut microbiome. First, stress has a known role in shaping the gut microbiome both in humans and animal models (Foster, Rinaman, & Cryan, 2017; Moloney, Desbonnet, 48 Clarke, Dinan, & Cryan, 2014; C. Yang et al., 2017), playing a critical role in brain health. For example, one recent study in children reported that high stress was associated with lower alpha diversity (Michels et al., 2019). Interestingly, this relationship was found with Simpson, but not observed or Chao1 diversity, as shown here, suggesting that richness and evenness are both important in determining the effect of social and environmental exposures. Next, resource deprivation, a stressor in and of itself, may affect gut microbiome composition. While there is an overall dearth of research on this topic, researchers posit that both stress exposure and reduced access to varied sources of dietary fiber (e.g., food deserts) could explain, in part, gut microbiome differences related to inflammatory disease and health across geographic space (Harrison & Taren, 2018; Kau, Ahern, Griffin, Goodman, & Gordon, 2011). Given the connection between violence exposure and stress (Bingenheimer et al., 2005; Cisler et al., 2012), and the relationship of both with disadvantage, the current study supplies evidence that the gut microbiome is a potential candidate to consider as a mediator of the neighborhood effect. One important facet of the gut microbiome as a mediator of the neighborhood effect is its potential to explain, in part, individual-level variation in behavior within a biopsychosocial framework. Indeed, the structure of the gut microbiome depends on individual-level conditioning in response to psychosocial factors (e.g., stress, diet). In this view, unique social and environmental exposures shape the gut microbiome, changing host physiology and contributing to the variability in health and behavioral outcomes observed in disadvantaged communities. The gut microbiome is thus a biological factor that exerts its effects in accordance with other known psychological and social or environmental risk factors identified in the criminological literature. In addition, the current study identified specific taxa in relation to high crime contexts that were also identified in the project’s preclinical studies in relation to territorial, reactive aggression behavior in mice, 49 raising questions about neighborhood context and the role of the gut microbiome in human aggression that needs to be addressed in future research. The current study has a number of limitations. Detroit represents 139 square miles of land in Southeast Michigan, yet only 95 of the 185 HPMM subjects died within its city limits. Among those 95 decedents, only 72 had data from rectum samples, further paring down the sample size. Therefore, this study cannot be considered representative of all Detroit block groups. In addition, the current study is ecological research and cannot be used to make inferences about the individuals in the block groups (Zeoli, Paruk, Pizarro, & Goldstick, 2019), but rather, the types of exposures that could occur within a block group. Individual behaviors may have contributed to the observed findings in both decedent death event and home locations. Given the high number of deaths that occurred at hospitals6 and accidental deaths recorded in the death event location dataset for high crime areas, it is possible that other factors that affect gut microbiome composition, like drug use, could be driving the observed relationship between block group crime exposures and alpha diversity. Moreover, there may be variation in exposure to the decedent home locations among individuals who may have spent the majority of their time elsewhere (due to work or other lifestyle choices), contributing to the null findings. Because the current study has an ecological design, exploring the effect of individual-level factors on the gut microbiome was outside its scope. Additional observations are needed to improve future modeling that could be hampered by imbalanced groups and overall low number of individuals in the high crime condition. Another critical limitation of this study is endogeneity in modeling (Zohoori & Savitz, 1997). Environmental and social exposures that are known risk factors for crime, like family 6 Though there are many hospital deaths recorded in the high crime condition, they did not all occur at the same hospital. 50 structure (e.g., Bursik & Grasmick, 1993), also have an influence in driving microbial community composition (Song et al., 2013). In particular, the environment has a strong influence in shaping gut microbiome composition (Afshinnekoo et al., 2015; Pearson et al., 2019; Von Ehrenstein et al., 2000; von Mutius & Vercelli, 2010; Yatsunenko et al., 2012). Endogeneity thus becomes a problem when one model covariate (e.g., block group vacancy rate) has theoretical influence on both the outcome and other model covariates (e.g., alpha diversity). This is a particularly difficult limitation to overcome with microbiome and social and environmental research; however, future work should attempt more sophisticated modeling on larger datasets. Conclusion The current study shows that the structure of the human gut microbiome varies according to local area crime level and socioeconomic conditions. In particular, these data show that there is lower biodiversity in areas with high crime, an indicator of a less healthy microbiome. Thus, data presented here hypothesize a potential mechanism through which access to a particular diet, water supply, atmosphere, or physical space affects health and behavior. In the United States, socioeconomic deprivation is distributed across geographic space in a racially coded way. As such, the findings presented here demonstrate yet another way through which the persistent lack of investment and marginalization of Black communities has real health consequences that translates into a disproportionate health burden. Public health research has long shown how local food environments contribute to health disparities (Moore & Diez Roux, 2006; Wallace et al., 2021). Most immediately, the current study’s findings provide support for this public health research, providing a novel biomarker to demonstrate the impact of different physical contexts, with varying levels of crime exposure, socioeconomic resources, and healthy dietary choices, as well as how these different living 51 conditions affect health. This work could therefore provide additional rationale for bolstering greening efforts and urban food programs. Further research is needed to develop the human gut microbiome as a potential target for public health intervention. 52 CHAPTER 5: CONCLUSIONS AND FUTURE DIRECTIONS Summary The work presented in this dissertation contributes to the field of criminology in several ways. Overall, my preclinical data describe a novel connection between the gut microbiome and territorial, reactive aggression behavior in mice that holds translational relevance for human public health. In support of this clinical relevance, the human data show that local area crime level is associated with variation in the human post mortem gut microbiome, establishing a critical link in the field of criminology between environmental context and physiology. Together, these findings move toward a biopsychosocial model of aggression behavior, providing foundational information that begins to address a critical gap in criminological research. Though criminological theories discuss the ill effects of concentrated disadvantage, poor social attachment, and their negative psychological sequelae that increase risk for crime perpetration, current research has not elucidated the biological underpinnings of these dispositions or how they convert to behavioral action. That is to say, the mechanisms of how individuals in risky physical and social environments go from feeling to doing are poorly understood. By combining research in the social and natural sciences, the work of my dissertations demonstrates the relationship between the gut microbiome and aggression behavior that has important connections to human crime patterns, laying the foundation for further exploration of the gut microbiome in human aggression. Given the complexity of aggression behavior and the multitude of factors that increase risk for human aggression, the best solution is to pull aggression-related natural and social science work together into a biopsychosocial model, which this interdisciplinary work begins to do. To demonstrate the translational relevance of my preclinical work, the presented analysis explored the relationship between violent crime, socioeconomic conditions, and the human post 53 mortem gut microbiome. For these studies, secondary 16S ribosomal RNA (rRNA) amplicon sequencing data were obtained through collaboration with colleagues on the Human Post Mortem Microbiome (HPMM) research team, including data analyst Sierra Kaszubinski as well as Drs. Jennifer Pechal and Eric Benbow. These data provide correlational evidence that gut microbiome composition varies according to local area (i.e., census block group) crime level. Importantly, analysis of decedent rectum samples revealed that alpha diversity, the richness and evenness of gut bacteria, corresponding to decedent death event locations was significantly lower in high versus medium and low crime areas. Further investigation showed that, together with block group vacancy rate, alpha diversity could be used to predict block group category placement in low, medium, and high crime conditions. Further examination of bacteria abundance data with negative binomial mixed models identified specific taxa that varied by crime level, some of which belonged in the same Lachnospiracaeae family we identified as associated with aggression in male CD-1 aggressor mice described in my Aggression and the Gut-Brain Axis work. Taken together, these data provide support for a connection between crime exposure and changes in (gut) physiology that have implications for the future study of aggression behavior and violent crime in humans. Implications and Future Directions Biological changes in a disadvantaged neighborhood context There is a call in criminological research to incorporate biology into the framework of major criminological theories to supply information about the molecular underpinnings of criminal behavior in all its various presentations (Fishbein, 1990). Though not formally tested here, the work in this thesis provides foundational evidence in support of exploring how perturbations in the gut microbiome affect individual-level expressions of aggression behavior. While it is unlikely that all behaviors have a direct gut-brain link, there are a multitude of factors in disadvantaged 54 neighborhoods that could affect gut microbiome composition, thereby affecting individual physiology and behavior. Importantly, this line of reasoning is consistent with criminological research on neighborhoods and small places showing that structural features impair social and community processes such that psychosocial risk factors aggregate in geographic space to compound risk for negative health and behavioral outcomes. Living in an economically disadvantaged community, which in this study was associated with a higher block group percentage of Black residents, with high crime is a source of chronic stress, especially among children and adolescents. Among youth exposed to socioeconomic deprivation (i.e., low socioeconomic status; SES), research demonstrates an increase in both self- reported stress and cortisol reactivity (Brenner, Zimmerman, Bauermeister, & Caldwell, 2013; Hackman, Betancourt, Brodsky, Hurt, & Farah, 2012), revealing a direct connection between neighborhood context and a physiological mechanism that also affects (Xu, Lee, Zhang, & Frenette, 2020), and is affected by (Gao et al., 2018), the gut microbiome. Evidence also suggests that exposure to low socioeconomic status is associated with changes in the structure and function of key corticolimbic brain regions (Liberzon et al., 2015; Noble et al., 2015), an effect attributed, in part, to other psychosocial factors, such as maternal health, family relationship quality, and fewer community resources in low SES contexts (Hackman et al., 2010). Critically, such chronic stress exposure is also associated with pathophysiological changes in the structure and function of the corticolimbic brain regions, like the amygdala, shifting physiology towards threat sensitivity and reactivity (Kwiatkowski, Manning, Eagle, & Robison, 2020). In addition to socioeconomic deprivation, the majority of youth living in high crime areas are also routinely exposed to violence (Berman, Kurtines, Silverman, & Serafini, 1996; Finkelhor, Turner, Ormrod, & Hamby, 2009), which has been directly associated with an impaired stress 55 response among those affected (Theall, Shirtcliff, Dismukes, Wallace, & Drury, 2017). Altogether, there is overwhelming evidence that living within a crime hotspot has deleterious health consequences, both increasing risk for disease and mood disorders (Lowe et al., 2016; Robinson & Keithley, 2000; Weisburd & White, 2019). Moreover, violence exposure has an interactive effect with exposure to disadvantage, exacerbating the risk for major depression, anxiety, and even post-traumatic stress disorder (Stockdale et al., 2007), all of which are associated with perturbations in the gut microbiome (Foster et al., 2017). Importantly, exposure to socioeconomic deprivation and neighborhood violence is also a risk factor for crime perpetration. Early life adversity increases risk for criminal offending in a dose-dependent fashion (Appleyard et al., 2005; Hay et al., 2006). In a large cohort study examining adult socioeconomic burden, 81% of the subjects with criminal convictions experienced early life socioeconomic deprivation and other early life adversity (Caspi et al., 2016). In addition, youth who are exposed to violence show immediate deficits in attention and impulse control (Sharkey, Tirado-Strayer, Papachristos, & Raver, 2012), and are significantly more likely to perpetrate serious violence within two years of exposure (Bingenheimer et al., 2005). Though the mechanisms that facilitate this behavioral pattern are poorly understood, there is likely a complex interplay between stress, threat sensitization and reactivity, and other psychosocial factors that increase risk for aggression, raising the question as to how exposure to socioeconomic deprivation and violence in high crime areas changes physiology. In answer to this question, the data presented in this dissertation suggest that exposure to high crime areas affects the composition of the human gut microbiome, potentially mediating the epigenetic effects of social and environmental exposures. In recent consideration of the link between neighborhood context and aggression behavior, Lesham and Weisburd (2019) proposed that exposures in crime hotspots have epigenetic effects 56 whereby the chronically stressful social and environmental landscape interacts with individual- level genetic factors to change gene expression, the brain, and behavior. One canonical example comes from Caspi and colleagues (2002) who demonstrated a genetic variant of the monoamine oxidase A (MAOA) gene was associated with increased risk of a disposition favorable towards violence as well as antisocial and violent behavior, but only among those who had also experienced childhood maltreatment. Indeed, the environmental effects on gene expression are being more widely recognized in the study of aggression behavior (Tremblay, Vitaro, & Côté, 2018), with some studies estimating that a 50-50 split between genetic and environmental effects in explaining variation of aggression behavior in children (Lacourse et al., 2014; Tuvblad & Baker, 2011; Tuvblad & Beaver, 2013). Moreover, recent research shows multiple methylation sites associated with aggression behavior in humans, explained in part by maternal health behaviors and other chemical exposures (van Dongen et al., 2021), drawing attention to physiological processes that mediate such changes in cell signaling and gene expression, namely epigenetics. Future Work Further research is needed to increase the translational relevance of this dissertation with additional human data. Overall, additional HPMM samples are needed to cover a larger study geography and include more cases from distinctly low and high crime areas where the greatest differences in the HPMM analysis were observed. Current HPMM studies should also be supplemented with protein quantification of human brain tissue, potentially collected alongside post mortem microbiome samples during routine autopsy. In addition, sampling from living humans across varying levels of neighborhood crime and over repeated measurements could also be conducted. Moreover, self-reported measures of violent crime perpetration, psychosocial stress, socioeconomic status, and other lifestyle factors need to be collected directly from individuals 57 under study in order to make any direct inferences about human gut microbiome composition and aggression behavior. Together, these efforts could greatly advance a biopsychosocial model of human aggression. Ethical Considerations Though many criminologists are concerned about the life sciences bringing reductionism and determinism to the study of crime, this is highly unlikely with the biopsychosocial approach. The biopsychosocial model requires assessment of biological, psychological, and social or environmental factors that combine and interact to drive a behavioral response. This approach showcases several layers of explanation rather than a universal cause of aggression and deviance. Instead of being deterministic, the incorporation of biology in criminological research is a necessary result of the intertwining of these fields. As such, there will likely never be an absolute biomarker of aggression or deviance, or a single cure. To assume otherwise would be ethically and empirically incorrect. Though individual structures, such as the amygdala, PFC, and their circuitry, can change in function in a manner that is conducive to deviant behavior, they are part of a larger, more complicated system that may fail or compensate. There is never a guaranteed response based solely on physiology or solely on psychology or solely on environment. Instead, it is best to view biological processes as enablers of a propensity for aggression and deviance that is cast over time in the light of developing psychology in response to environmental influences. This is most appropriately understood in the context of a biopsychosocial approach where environmental cues and stressors shape psychological and biological function, and the work presented here offers a potential avenue forward for the biopsychosocial model. 58 WORKS CITED Afshinnekoo, E., Meydan, C., Chowdhury, S., Jaroudi, D., Boyer, C., Bernstein, N., . . . Mason, Christopher E. (2015). Geospatial Resolution of Human and Bacterial Diversity with City-Scale Metagenomics. Cell Systems, 1(1), 72-87. doi:10.1016/j.cels.2015.01.001 Anderson, D. J. (2012). Optogenetics, Sex, and Violence in the Brain: Implications for Psychiatry. Biological Psychiatry, 71(12), 1081-1089. doi:10.1016/j.biopsych.2011.11.012 Anderson, M. J. (2001). A new method for non-parametric multivariate analysis of variance. Austral Ecology, 26(1), 32-46. doi:https://doi.org/10.1111/j.1442-9993.2001.01070.pp.x Anderson, M. J. (2017). Permutational Multivariate Analysis of Variance (PERMANOVA) Wiley StatsRef (pp. 1-15). Appleyard, K., Egeland, B., van Dulmen, M. H. M., & Alan Sroufe, L. (2005). When more is not better: the role of cumulative risk in child behavior outcomes. Journal of Child Psychology and Psychiatry, 46(3), 235-245. doi:10.1111/j.1469-7610.2004.00351.x Baime, A. (2014). The Arsenal of Democracy: FDR, Detroit, and an Epic Quest to Arm an America at War: Houghton Mifflin Harcourt. Bannon, S. M., Salis, K. L., & O'Leary, K. D. (2015). Structural brain abnormalities in aggression and violent behavior. Aggression and Violent Behavior, 25, 323-331. doi:https://doi.org/10.1016/j.avb.2015.09.016 Bender, A. R., & Raz, N. (2015). Normal-appearing cerebral white matter in healthy adults: mean change over 2 years and individual differences in change. Neurobiology of Aging, 36(5), 1834-1848. doi:https://doi.org/10.1016/j.neurobiolaging.2015.02.001 Berman, S. L., Kurtines, W. M., Silverman, W. K., & Serafini, L. T. (1996). The Impact of Exposure to Crime and Violence on Urban Youth. American Journal of Orthopsychiatry, 66(3), 329-336. doi:https://doi.org/10.1037/h0080183 Berton, O., McClung, C. A., DiLeone, R. J., Krishnan, V., Renthal, W., Russo, S. J., . . . Nestler, E. J. (2006). Essential Role of BDNF in the Mesolimbic Dopamine Pathway in Social Defeat Stress. Science, 311(5762), 864. doi:10.1126/science.1120972 Bertsch, K., Grothe, M., Prehn, K., Vohs, K., Berger, C., Hauenstein, K., . . . Herpertz, S. C. (2013). Brain volumes differ between diagnostic groups of violent criminal offenders. European Archives of Psychiatry and Clinical Neuroscience, 263(7), 593-606. doi:10.1007/s00406-013-0391-6 Bingenheimer, J. B., Brennan, R. T., & Earls, F. J. (2005). Firearm Violence Exposure and Serious Violent Behavior. Science, 308(5726), 1323-1326. 59 Bogar, S., & Beyer, K. M. (2015). Green Space, Violence, and Crime: A Systematic Review. Trauma, Violence, & Abuse, 17(2), 160-171. doi:10.1177/1524838015576412 Bolyen, E., Rideout, J. R., Dillon, M. R., Bokulich, N. A., Abnet, C. C., Al-Ghalith, G. A., . . . Caporaso, J. G. (2019). Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nature Biotechnology, 37(8), 852-857. doi:10.1038/s41587- 019-0209-9 Bondy, S. C. (2021). Metal toxicity and neuroinflammation. Current Opinion in Toxicology, 26, 8-13. doi:https://doi.org/10.1016/j.cotox.2021.03.008 Braga, A. A., Papachristos, A. V., & Hureau, D. M. (2010). The Concentration and Stability of Gun Violence at Micro Places in Boston, 1980–2008. Journal of Quantitative Criminology, 26(1), 33-53. Branas, C. C., Kondo, M. C., Murphy, S. M., South, E. C., Polsky, D., & MacDonald, J. M. (2016). Urban Blight Remediation as a Cost-Beneficial Solution to Firearm Violence. American Journal of Public Health, 106(12), 2158-2164. doi:10.2105/AJPH.2016.303434 Branas, C. C., Rubin, D., & Guo, W. (2012). Vacant Properties and Violence in Neighborhoods. ISRN Public Health, 2012, 246142. doi:10.5402/2012/246142 Brenner, A. B., Zimmerman, M. A., Bauermeister, J. A., & Caldwell, C. H. (2013). Neighborhood Context and Perceptions of Stress Over Time: An Ecological Model of Neighborhood Stressors and Intrapersonal and Interpersonal Resources. American Journal of Community Psychology, 51(3-4), 544-556. doi:https://doi.org/10.1007/s10464- 013-9571-9 Brown, T. A. (2015). Confirmatory Factor Analysis for Applied Research (2 ed.). New York, NY, US: The Guilford Press. Bursik, R., Jr, & Grasmick, H. (1993). Neighborhoods and Crime: The Dimensions of Effective Community Control. Campbell, J. C. (2002). Health consequences of intimate partner violence. The Lancet, 359(9314), 1331-1336. doi:https://doi.org/10.1016/S0140-6736(02)08336-8 Caporaso, J. G., Lauber, C. L., Walters, W. A., Berg-Lyons, D., Huntley, J., Fierer, N., . . . Knight, R. (2012). Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. The ISME Journal, 6(8), 1621-1624. doi:10.1038/ismej.2012.8 60 Case, R. J., Boucher, Y., Dahllöf, I., Holmström, C., Doolittle, W. F., & Kjelleberg, S. (2007). Use of 16S rRNA and rpoB Genes as Molecular Markers for Microbial Ecology Studies. Applied and Environmental Microbiology, 73(1), 278-288. doi:10.1128/AEM.01177-06 Caspi, A., Houts, R. M., Belsky, D. W., Harrington, H., Hogan, S., Ramrakha, S., . . . Moffitt, T. E. (2016). Childhood forecasting of a small segment of the population with large economic burden. Nature Human Behaviour, 1, 1-10. Caspi, A., McClay, J., Moffitt, T. E., Mill, J., Martin, J., Craig, I. W., . . . Poulton, R. (2002). Role of Genotype in the Cycle of Violence in Maltreated Children. Science, 297(5582), 851-854. doi:10.1126/science.1072290 Chen, Y., Sprung, R., Tang, Y., Ball, H., Sangras, B., Kim, S. C., . . . Zhao, Y. (2007). Lysine Propionylation and Butyrylation are Novel Post-translational Modifications in Histones. Molecular & Cellular Proteomics, 6(5), 812-819. Choi, S.-W., & Friso, S. (2010). Epigenetics: A New Bridge between Nutrition and Health. Advances in Nutrition, 1(1), 8-16. doi:10.3945/an.110.1004 Christian, L. M., Galley, J. D., Hade, E. M., Schoppe-Sullivan, S., Kamp Dush, C., & Bailey, M. T. (2015). Gut microbiome composition is associated with temperament during early childhood. Brain, Behavior, and Immunity, 45, 118-127. doi:10.1016/j.bbi.2014.10.018 Cisler, J. M., Begle, A. M., Amstadter, A. B., Resnick, H. S., Danielson, C. K., Saunders, B. E., & Kilpatrick, D. G. (2012). Exposure to interpersonal violence and risk for PTSD, depression, delinquency, and binge drinking among adolescents: data from the NSA-R. Journal of Traumatic Stress, 25(1), 33-40. doi:10.1002/jts.21672 Clarke, G., O'Mahony, S. M., Dinan, T. G., & Cryan, J. F. (2014). Priming for health: gut microbiota acquired in early life regulates physiology, brain and behaviour. Acta Paediatrica, 103(8), 812-819. doi:10.1111/apa.12674 Coccaro, E. F., McCloskey, M. S., Fitzgerald, D. A., & Phan, K. L. (2007). Amygdala and Orbitofrontal Reactivity to Social Threat in Individuals with Impulsive Aggression. Biological Psychiatry, 62(2), 168-178. doi:http://dx.doi.org/10.1016/j.biopsych.2006.08.024 Contreras-Rodríguez, O., Pujol, J., Batalla, I., Harrison, B. J., Soriano-Mas, C., Deus, J., . . . Cardoner, N. (2015). Functional Connectivity Bias in the Prefrontal Cortex of Psychopaths. Biological Psychiatry, 78(9), 647-655. doi:http://dx.doi.org/10.1016/j.biopsych.2014.03.007 Crick, N. R., & Dodge, K. A. (1994). A Review and Reformulation of Social Information- Processing Mechanisms in Children's Social Adjustment. Psychological Bulletin, 115(1), 74-101. 61 Cryan, J. F., & Dinan, T. G. (2015). Gut microbiota: Microbiota and neuroimmune signalling- Metchnikoff to microglia. Nature Reviews: Gastroenterology & Hepatology, 12(9), 494- 496. doi:10.1038/nrgastro.2015.127 Datta, S. R., Anderson, D. J., Branson, K., Perona, P., & Leifer, A. (2019). Computational Neuroethology: A Call to Action. Neuron, 104(1), 11-24. doi:10.1016/j.neuron.2019.09.038 Davis, M., & Whalen, P. J. (2001). The amygdala: vigilance and emotion. Molecular Psychiatry, 6(1), 13-34. de Boer, S. F., van der Vegt, B., J., & Koolhaas, J. M. (2003). Individual Variation in Aggression of Feral Rodent Strains: A Standard for the Genetics of Aggression and Violence? Behavior Genetics, 33(5), 485-501. doi:http://dx.doi.org/10.1023/A:1025766415159 Denno, D. W. (1990). Biology and Violence: From Birth to Adulthood. Cambridge: Cambridge University Press. Dolan, M. C., & Fullam, R. S. (2009). Psychopathy and Functional Magnetic Resonance Imaging Blood Oxygenation Level-Dependent Responses to Emotional Faces in Violent Patients with Schizophrenia. Biological Psychiatry, 66(6), 570-577. doi:10.1016/j.biopsych.2009.03.019 Eagle, A. L., Manning, C. E., Williams, E. S., Bastle, R. M., Gajewski, P. A., Garrison, A., . . . Robison, A. J. (2020). Circuit-specific hippocampal ΔFosB underlies resilience to stress- induced social avoidance. Nat Commun, 11, 4484. doi:10.1038/s41467-020-17825-x Engel, G. L. (1977). The Need for a New Medical Model: A Challenge for Biomedicine. Science, 196(4286), 129-136. Erny, D., de Angelis, A. L. H., Jaitin, D., Wieghofer, P., Staszewski, O., David, E., . . . Buch, T. (2015). Host microbiota constantly control maturation and function of microglia in the CNS. Nature Neuroscience, 18(7), 965-977. Falkner, A. L., Grosenick, L., Davidson, T. J., Deisseroth, K., & Lin, D. (2016). Hypothalamic control of male aggression-seeking behavior. Nature Neuroscience, 19(4), 596-604. doi:10.1038/nn.4264 Federal Bureau of Investigation. (2017). Crime in the United States. Uniform Crime Reports, from U.S. Department of Justice https://ucr.fbi.gov/crime-in-the-u.s/2017/crime-in-the- u.s.-2017/home Finkelhor, D., Turner, H., Ormrod, R., & Hamby, S. L. (2009). Violence, Abuse, and Crime Exposure in a National Sample of Children and Youth. Pediatrics, 124(5), 1411. doi:10.1542/peds.2009-0467 62 Fishbein, D. H. (1990). Biological Perspectives in Criminology. Criminology, 28(1), 27-72. doi:https://doi.org/10.1111/j.1745-9125.1990.tb01317.x Fisher, D. (2015). America's Most Dangerous Cities: Detroit Can't Shake No. 1 Spot. Forbes. Retrieved from https://www.forbes.com/sites/danielfisher/2015/10/29/americas-most- dangerous-cities-detroit-cant-shake-no-1-spot/#11f4306c467d Foster, J. A., Rinaman, L., & Cryan, J. F. (2017). Stress & the gut-brain axis: Regulation by the microbiome. Neurobiology of Stress, 7, 124-136. doi:https://doi.org/10.1016/j.ynstr.2017.03.001 Fox, G. E., Pechman, K. R., & Woese, C. R. (1977). Comparative Cataloging of 16S Ribosomal Ribonucleic Acid: Molecular Approach to Procaryotic Systematics. International Journal of Systematic Bacteriology, 27(1), 44-57. doi:https://doi.org/10.1099/00207713-27-1-44 Gao, X., Cao, Q., Cheng, Y., Zhao, D., Wang, Z., Yang, H., . . . Yang, Y. (2018). Chronic stress promotes colitis by disturbing the gut microbiota and triggering immune system response. Proceedings of the National Academy of Sciences, 115(13), E2960. doi:10.1073/pnas.1720696115 Gaquin, D. A., & Ryan, M. M. (2019). The Who, What, and Where of America: Understanding the American Community Survey (7 ed.): Rowman & Littlefield. Gesch, C. B., Hammond, S. M., Hampson, S. E., Eves, A., & Crowder, M. J. (2002). Influence of supplementary vitamins, minerals and essential fatty acids on the antisocial behaviour of young adult prisoners: Randomised, placebo-controlled trial. British Journal of Psychiatry, 181(1), 22-28. doi:10.1192/bjp.181.1.22 Glenn, A. L., Raine, A., & Schug, R. A. (2009). The neural correlates of moral decision-making in psychopathy. Molecular Psychiatry, 14(1), 5-6. doi:http://www.nature.com/mp/journal/v14/n1/suppinfo/mp2008104s1.html Golden, S. A., Aleyasin, H., Heins, R., Flanigan, M., Heshmati, M., Takahashi, A., . . . Shaham, Y. (2017). Persistent conditioned place preference to aggression experience in adult male sexually-experienced CD-1 mice. Genes, Brain and Behavior, 16(1), 44-55. doi:10.1111/gbb.12310 Golden, S. A., Covington III, H. E., Berton, O., & Russo, S. J. (2011). A standardized protocol for repeated social defeat stress in mice. Nature Protocols, 6, 1183-1191. doi:10.1038/nprot.2011.361 Golden, S. A., Heshmati, M., Flanigan, M., Christoffel, D. J., Guise, K., Pfau, M. L., . . . Russo, S. J. (2016). Basal forebrain projections to the lateral habenula modulate aggression reward. Nature, 534, 688-692. doi:10.1038/nature18601 63 Golden, S. A., Jin, M., & Shaham, Y. (2019). Animal Models of (or for) Aggression Reward, Addiction, and Relapse: Behavior and Circuits. The Journal of Neuroscience, 39(21), 3996-4008. doi:10.1523/JNEUROSCI.0151-19.2019 Goodwin, N. L., Nilsson, S. R. O., & Golden, S. A. (2020). Rage Against the Machine: Advancing the study of aggression ethology via machine learning. Psychopharmacology, 237(9), 2569-2588. doi:10.1007/s00213-020-05577-x Gospic, K., Mohlin, E., Fransson, P., Petrovic, P., Johannesson, M., & Ingvar, M. (2011). Limbic Justice—Amygdala Involvement in Immediate Rejection in the Ultimatum Game. PLoS Biology, 9(5), 1-8. Guerry, A.-M. (1833). Essai sur La Statistique Morale de la France. Paris, France: Librairie Scientifique de Crochard. Hackman, D. A., Betancourt, L., Brodsky, N., Hurt, H., & Farah, M. (2012). Neighborhood disadvantage and adolescent stress reactivity. Frontiers in Human Neuroscience, 6(277). doi:10.3389/fnhum.2012.00277 Hackman, D. A., Farah, M. J., & Meaney, M. J. (2010). Socioeconomic status and the brain: mechanistic insights from human and animal research. Nature Reviews: Neuroscience, 11(9), 651-659. Haines, D. E. (Ed.) (2013). Fundamental Neuroscience for Basic and Clinical Applications (4 ed.): Elsevier/Saunders. Halász, J., Tóth, M., Kalló, I., Liposits, Z., & Haller, J. (2006). The activation of prefrontal cortical neurons in aggression—A double labeling study. Behavioural Brain Research, 175(1), 166-175. doi:https://doi.org/10.1016/j.bbr.2006.08.019 Haller, J. (2018). The role of central and medial amygdala in normal and abnormal aggression: A review of classical approaches. Neuroscience and Biobehavioral Reviews, 85, 34-43. doi:https://doi.org/10.1016/j.neubiorev.2017.09.017 Haller, J., & Kruk, M. R. (2006). Normal and abnormal aggression: human disorders and novel laboratory models. Neuroscience and Biobehavioral Reviews, 30(3), 292-303. doi:10.1016/j.neubiorev.2005.01.005 Hamady, M., & Knight, R. (2009). Microbial community profiling for human microbiome projects: Tools, techniques, and challenges. Genome Research, 19(7), 1141-1152. doi:10.1101/gr.085464.108 64 Hamm, J. A., & Hoffman, L. (2016). Working with Covariance: Using Higher-Order Factors in Structural Equation Modeling with Trust Constructs. In E. Shockley, T. M. S. Neal, L. M. PytlikZillig, & B. H. Bornstein (Eds.), Interdisciplinary Perspectives on Trust: Towards Theoretical and Methodological Integration (pp. 85-97). Cham: Springer International Publishing. Harrison, C. A., & Taren, D. (2018). How poverty affects diet to shape the microbiota and chronic disease. Nature Reviews Immunology, 18(4), 279-287. doi:10.1038/nri.2017.121 Harwell, M. R., Rubinstein, E. N., Hayes, W. S., & Olds, C. C. (1992). Summarizing Monte Carlo Results in Methodological Research: The One- and Two-Factor Fixed Effects ANOVA Cases. Journal of Educational Statistics, 17(4), 315-339. doi:10.3102/10769986017004315 Hay, C., Fortson, E. N., Hollist, D. R., Altheimer, I., & Schaible, L. M. (2006). The impact of community disadvantage on the relationship between the family and juvenile crime. Journal of Research in Crime and Delinquency, 43(4), 326-356. Heijtz, R. D., Wang, S., Anuar, F., Qian, Y., Björkholm, B., Samuelsson, A., . . . Pettersson, S. (2011). Normal gut microbiota modulates brain development and behavior. Proceedings of the National Academy of Sciences, 108(7), 3047-3052. doi:10.1073/pnas.1010529108 Hoban, A. E., Stilling, R. M., Moloney, G., Shanahan, F., Dinan, T. G., Clarke, G., & Cryan, J. F. (2017). The microbiome regulates amygdala-dependent fear recall. Molecular Psychiatry. doi:10.1038/mp.2017.100 Hoban, A. E., Stilling, R. M., Ryan, F. J., Shanahan, F., Dinan, T. G., Claesson, M. J., . . . Cryan, J. F. (2016). Regulation of prefrontal cortex myelination by the microbiota. Transl Psychiatry, 6(4), e774. doi:10.1038/tp.2016.42 Hsiao, E. Y., McBride, S. W., Hsien, S., Sharon, G., Hyde, E. R., McCue, T., . . . Mazmanian, S. K. (2013). Microbiota modulate behavioral and physiological abnormalities associated with neurodevelopmental disorders. Cell, 155(7), 1451-1463. doi:10.1016/j.cell.2013.11.024 Hu, L. t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55. doi:10.1080/10705519909540118 Jiang, H., Ling, Z., Zhang, Y., Mao, H., Ma, Z., Yin, Y., . . . Ruan, B. (2015). Altered fecal microbiota composition in patients with major depressive disorder. Brain, Behavior, and Immunity, 48, 186-194. doi:https://doi.org/10.1016/j.bbi.2015.03.016 Jiang, H., Zhang, X., Yu, Z., Zhang, Z., Deng, M., Zhao, J., & Ruan, B. (2018). Altered gut microbiota profile in patients with generalized anxiety disorder. Journal of Psychiatric Research, 104, 130-136. doi:https://doi.org/10.1016/j.jpsychires.2018.07.007 65 Jöreskog, K. G. (1969). A general approach to confirmatory maximum likelihood factor analysis. Psychometrika, 34(2), 183-202. doi:10.1007/BF02289343 Kahle, D., & Wickham, H. (2013). ggmap: Spatial Visualization with ggplot2. The R Journal, 5(1), 144-161. Kaszubinski, S. F., Pechal, J. L., Schmidt, C. J., Jordan, H. R., Benbow, M. E., & Meek, M. H. (2020). Evaluating Bioinformatic Pipeline Performance for Forensic Microbiome Analysis*,†,‡. Journal of Forensic Sciences, 65(2), 513-525. doi:https://doi.org/10.1111/1556-4029.14213 Kaszubinski, S. F., Pechal, J. L., Smiles, K., Schmidt, C. J., Jordan, H. R., Meek, M. H., & Benbow, M. E. (2020). Dysbiosis in the Dead: Human Postmortem Microbiome Beta- Dispersion as an Indicator of Manner and Cause of Death. Frontiers in Microbiology, 11(2212). doi:10.3389/fmicb.2020.555347 Kau, A. L., Ahern, P. P., Griffin, N. W., Goodman, A. L., & Gordon, J. I. (2011). Human nutrition, the gut microbiome and the immune system. Nature, 474(7351), 327-336. doi:10.1038/nature10213 Kiehl, K. A., Smith, A. M., Hare, R. D., Mendrek, A., Forster, B. B., Brink, J., & Liddle, P. F. (2001). Limbic abnormalities in affective processing by criminal psychopaths as revealed by functional magnetic resonance imaging. Biological Psychiatry, 50(9), 677-684. doi:http://dx.doi.org/10.1016/S0006-3223(01)01222-7 Kline, R. B. (2015). Principles and Practice of Structural Equation Modeling (4 ed.). New York, NY: Guilford Press. Knight, R. (2015). Follow Your Gut: The Enormous Impact of Tiny Microbes: Simon and Schuster. Kondo, M. C., Morrison, C., Jacoby, S. F., Elliott, L., Poche, A., Theall, K. P., & Branas, C. C. (2018). Blight Abatement of Vacant Land and Crime in New Orleans. Public Health Reports, 133(6), 650-657. doi:10.1177/0033354918798811 Koolhaas, J. M., Coppens, C. M., de Boer, S. F., Buwalda, B., Meerlo, P., & Timmermans, P. J. A. (2013). The Resident-intruder Paradigm: A Standardized Test for Aggression, Violence and Social Stress. Journal of Visualized Experiments(77), e4367. doi:doi:10.3791/4367 Kozich James, J., Westcott Sarah, L., Baxter Nielson, T., Highlander Sarah, K., & Schloss Patrick, D. (2013). Development of a Dual-Index Sequencing Strategy and Curation Pipeline for Analyzing Amplicon Sequence Data on the MiSeq Illumina Sequencing Platform. Applied and Environmental Microbiology, 79(17), 5112-5120. doi:10.1128/AEM.01043-13 66 Krishnan, V., Han, M.-H., Graham, D. L., Berton, O., Renthal, W., Russo, S. J., . . . Nestler, E. J. (2007). Molecular Adaptations Underlying Susceptibility and Resistance to Social Defeat in Brain Reward Regions. Cell, 131(2), 391-404. doi:https://doi.org/10.1016/j.cell.2007.09.018 Kudryavtseva, N. N., Bakshtanovskaya, I. V., & Koryakina, L. A. (1991). Social model of depression in mice of C57BL/6J strain. Pharmacology Biochemistry and Behavior, 38(2), 315-320. doi:https://doi.org/10.1016/0091-3057(91)90284-9 Kwiatkowski, C. C., Akaeze, H., Ndlebe, I., Goodwin, N., Eagle, A. L., Moon, K., . . . Robison, A. J. (2021). Quantitative standardization of resident mouse behavior for studies of aggression and social defeat. Neuropsychopharmacology, 46(9), 1584-1593. doi:10.1038/s41386-021-01018-1 Kwiatkowski, C. C., Manning, C. E., Eagle, A. L., & Robison, A. J. (2020). The neurobiology of police health, resilience, and wellness. In K. Papazoglou & D. M. Blumberg (Eds.), POWER (pp. 77-96): Academic Press. Kwiatkowski, C. C., Robison, A. J., & Zeoli, A. M. (2018). Brain Health and the Batterer. Violence and gender, 5(4), 199-201. doi:10.1089/vio.2018.0010 Lach, G., Fülling, C., Bastiaanssen, T. F. S., Fouhy, F., Donovan, A. N. O., Ventura-Silva, A. P., . . . Cryan, J. F. (2020). Enduring neurobehavioral effects induced by microbiota depletion during the adolescent period. Transl Psychiatry, 10(1), 382. doi:10.1038/s41398-020-01073-0 Lacourse, E., Boivin, M., Brendgen, M., Petitclerc, A., Girard, A., Vitaro, F., . . . Tremblay, R. E. (2014). A longitudinal twin study of physical aggression during early childhood: evidence for a developmentally dynamic genome. Psychological Medicine, 44(12), 2617- 2627. doi:10.1017/S0033291713003218 LaVeist, T., Pollack, K., Thorpe, R., Fesahazion, R., & Gaskin, D. (2011). Place, Not Race: Disparities Dissipate In Southwest Baltimore When Blacks and Whites Live Under Similar Conditions. Health Affairs, 30(10), 1880-1887. LeDuff, C. (2013). Detroit: An American Autopsy: Penguin. Leshem, R., & Weisburd, D. (2019). Epigenetics and Hot Spots of Crime: Rethinking the Relationship Between Genetics and Criminal Behavior. Journal of Contemporary Criminal Justice, 35(2), 186-204. doi:10.1177/1043986219828924 Leutgeb, V., Leitner, M., Wabnegger, A., Klug, D., Scharmüller, W., Zussner, T., & Schienle, A. (2015). Brain abnormalities in high-risk violent offenders and their association with psychopathic traits and criminal recidivism. Neuroscience, 308, 194-201. doi:http://dx.doi.org/10.1016/j.neuroscience.2015.09.011 67 Li, B., Xia, M., Zorec, R., Parpura, V., & Verkhratsky, A. (2021). Astrocytes in heavy metal neurotoxicity and neurodegeneration. Brain Research, 1752, 147234. doi:https://doi.org/10.1016/j.brainres.2020.147234 Liberzon, I., Ma, S. T., Okada, G., Shaun Ho, S., Swain, J. E., & Evans, G. W. (2015). Childhood poverty and recruitment of adult emotion regulatory neurocircuitry. Social Cognitive and Affective Neuroscience, 10(11), 1596-1606. Lingawi, N. W., Laurent, V., Westbrook, R. F., & Holmes, N. M. (2019). The role of the basolateral amygdala and infralimbic cortex in (re)learning extinction. Psychopharmacology, 236(1), 303-312. doi:10.1007/s00213-018-4957-x Lix, L. M., Keselman, J. C., & Keselman, H. J. (1996). Consequences of Assumption Violations Revisited: A Quantitative Review of Alternatives to the One-Way Analysis of Variance F Test. Review of Educational Research, 66(4), 579-619. doi:10.3102/00346543066004579 Lowe, S. R., Quinn, J. W., Richards, C. A., Pothen, J., Rundle, A., Galea, S., . . . Bradley, B. (2016). Childhood trauma and neighborhood-level crime interact in predicting adult posttraumatic stress and major depression symptoms. Child Abuse and Neglect, 51, 212- 222. doi:https://doi.org/10.1016/j.chiabu.2015.10.007 Lozier, L. M., Cardinale, E. M., VanMeter, J. W., & Marsh, A. A. (2014). Mediation of the Relationship Between Callous-Unemotional Traits and Proactive Aggression by Amygdala Response to Fear Among Children with Conduct Problems. JAMA Psychiatry, 71(6), 627-636. doi:10.1001/jamapsychiatry.2013.4540 Lueken, U., Hilbert, K., Stolyar, V., Maslowski, N. I., Beesdo-Baum, K., & Wittchen, H.-U. (2014). Neural substrates of defensive reactivity in two subtypes of specific phobia. Social Cognitive and Affective Neuroscience, 9(11), 1668-1675. doi:10.1093/scan/nst159 Mayer, E. A., Knight, R., Mazmanian, S. K., Cryan, J. F., & Tillisch, K. (2014). Gut microbes and the brain: paradigm shift in neuroscience. Journal of Neuroscience, 34(46), 15490- 15496. doi:10.1523/JNEUROSCI.3299-14.2014 McGill, B. J., Etienne, R. S., Gray, J. S., Alonso, D., Anderson, M. J., Benecha, H. K., . . . White, E. P. (2007). Species abundance distributions: moving beyond single prediction theories to integration within an ecological framework. Ecology Letters, 10(10), 995- 1015. doi:10.1111/j.1461-0248.2007.01094.x McMurdie, P. J., & Holmes, S. (2013). phyloseq: An R Package for Reproducible Interactive Analysis and Graphics of Microbiome Census Data. PloS One, 8(4), e61217. doi:10.1371/journal.pone.0061217 Meredith, W. (1993). Measurement invariance, factor analysis and factorial invariance. Psychometrika, 58(4), 525-543. doi:10.1007/BF02294825 68 Michels, N., Van de Wiele, T., Fouhy, F., O'Mahony, S., Clarke, G., & Keane, J. (2019). Gut microbiome patterns depending on children’s psychosocial stress: Reports versus biomarkers. Brain, Behavior, and Immunity, 80, 751-762. doi:https://doi.org/10.1016/j.bbi.2019.05.024 Michigan Department of Community Health. (2014). Number of Deaths and Age-adjusted Mortality Rates for the Ten Leading Causes of Death: Detroit City and Michigan and United States Residents. from Michigan Department of Health and Human Services http://www.mdch.state.mi.us/osr/Chi/Deaths/frame.html Miczek, K. A., de Boer, S. F., & Haller, J. (2013). Excessive aggression as model of violence: a critical evaluation of current preclinical methods. Psychopharmacology, 226(3), 445-458. doi:10.1007/s00213-013-3008-x Miczek, K. A., Faccidomo, S., de Almeida, R. M. M., Bannai, M., Fish, E. W., & Debold, J. F. (2004). Escalated Aggressive Behavior: New Pharmacotherapeutic Approaches and Opportunities. In J. Devine (Ed.), Youth Violence: Scientific Approaches to Prevention (Vol. 1036 of Annals of the New York Academy of Sciences, pp. 336-355). New York, NY, US: New York Academy of Sciences. Mobbs, D., Marchant, J. L., Hassabis, D., Seymour, B., Tan, G., Gray, M., . . . Frith, C. D. (2009). From Threat to Fear: The Neural Organization of Defensive Fear Systems in Humans. The Journal of Neuroscience, 29(39), 12236. Mobbs, D., Petrovic, P., Marchant, J. L., Hassabis, D., Weiskopf, N., Seymour, B., . . . Frith, C. D. (2007). When Fear Is Near: Threat Imminence Elicits Prefrontal-Periaqueductal Gray Shifts in Humans. Science (New York, N.Y.), 317(5841), 1079-1083. doi:10.1126/science.1144298 Mohai, P., & Saha, R. (2006). Reassessing racial and socioeconomic disparities in environmental justice research. Demography, 43(2), 383-399. doi:10.1353/dem.2006.0017 Mohai, P., & Saha, R. (2007). Racial Inequality in the Distribution of Hazardous Waste: A National-Level Reassessment. Social Problems, 54(3), 343-370. doi:10.1525/sp.2007.54.3.343 Moloney, R. D., Desbonnet, L., Clarke, G., Dinan, T. G., & Cryan, J. F. (2014). The microbiome: stress, health and disease. Mammalian Genome, 25(1), 49-74. doi:10.1007/s00335-013-9488-5 Moore, L. V., & Diez Roux, A. V. (2006). Associations of Neighborhood Characteristics With the Location and Type of Food Stores. American Journal of Public Health, 96(2), 325- 331. 69 Morgan, X. C., & Huttenhower, C. (2012). Chapter 12: Human Microbiome Analysis. PLoS Computational Biology, 8(12), e1002808. doi:10.1371/journal.pcbi.1002808 Mörkl, S., Butler, M. I., Holl, A., Cryan, J. F., & Dinan, T. G. (2020). Probiotics and the Microbiota-Gut-Brain Axis: Focus on Psychiatry. Current Nutrition Reports, 9(3), 171- 182. doi:10.1007/s13668-020-00313-5 Narag, R. E., Pizarro, J., & Gibbs, C. (2009). Lead Exposure and Its Implications for Criminological Theory. Criminal Justice and Behavior, 36(9), 954-973. doi:10.1177/0093854809339286 Nelson, R. J., & Trainor, B. C. (2007). Neural mechanisms of aggression. Nature Reviews Neuroscience, 8(7), 536-546. Neufeld, K. M., Kang, N., Bienenstock, J., & Foster, J. A. (2011). Reduced anxiety-like behavior and central neurochemical change in germ-free mice. Neurogastroenterology and Motility, 23(3), 255-e119. doi:10.1111/j.1365-2982.2010.01620.x Newman, E. L., Covington, H. E., Suh, J., Bicakci, M. B., Ressler, K. J., DeBold, J. F., & Miczek, K. A. (2019). Fighting Females: Neural and Behavioral Consequences of Social Defeat Stress in Female Mice. Biological Psychiatry, 86(9), 657-668. doi:https://doi.org/10.1016/j.biopsych.2019.05.005 Nilsson, S. R. O., Goodwin, N. L., Choong, J. J., Hwang, S., Wright, H. R., Norville, Z. C., . . . Golden, S. A. (2020). Simple Behavioral Analysis (SimBA) – an open source toolkit for computer classification of complex social behaviors in experimental animals. bioRxiv, 1- 29. doi:10.1101/2020.04.19.049452 Noble, K. G., Houston, S. M., Brito, N. H., Bartsch, H., Kan, E., Kuperman, J. M., . . . Libiger, O. (2015). Family Income, Parental Education and Brain Structure in Children and Adolescents. Nature Neuroscience, 18(5), 773-778. Olivier, B., & Young, L. J. (2002). Animal Models of Aggression. In K. L. Davis, D. Charney, J. T. Coyle, & C. Nemeroff (Eds.), Neuropsychopharmacology: The Fifth Generation of Progress (pp. 1699-1708). Philadelphia, PA: Lippincott Williams & Wilkins. Oyegbile, T. O., & Marler, C. A. (2005). Winning fights elevates testosterone levels in California mice and enhances future ability to win fights. Hormones and Behavior, 48(3), 259-267. doi:https://doi.org/10.1016/j.yhbeh.2005.04.007 Pearson, A. L., Rzotkiewicz, A., Pechal, J. L., Schmidt, C. J., Jordan, H. R., Zwickle, A., & Benbow, M. E. (2019). Initial Evidence of the Relationships between the Human Postmortem Microbiome and Neighborhood Blight and Greening Efforts. Annals of the American Association of Geographers, 109(3), 958-978. doi:10.1080/24694452.2018.1519407 70 Pebesma, E. (2018). Simple Features for R: Standardized Support for Spatial Vector Data. The R Journal, 10(1), 439-446. doi:10.32614/RJ-2018-009 Pechal, J. L., Schmidt, C. J., Jordan, H. R., & Benbow, M. E. (2018). A large-scale survey of the postmortem human microbiome, and its potential to provide insight into the living health condition. Scientific Reports, 8(1), 5724. doi:10.1038/s41598-018-23989-w Putnick, D. L., & Bornstein, M. H. (2016). Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Developmental Review, 41, 71-90. doi:https://doi.org/10.1016/j.dr.2016.06.004 Qin, Y., & Wade, P. A. (2018). Crosstalk between the microbiome and epigenome: messages from bugs. Journal of Biochemistry, 163(2), 105-112. doi:10.1093/jb/mvx080 Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T., Yarza, P., . . . Glöckner, F. O. (2013). The SILVA ribosomal RNA gene database project: improved data processing and web- based tools. Nucleic Acids Research, 41(D1), D590-D596. doi:10.1093/nar/gks1219 R Core Team. (2021). R: A language and environment for statistical computing. . Vienna, Austria: R Foundation for Statistical Computing. Retrieved from https://www.R- project.org/ Rafter, N., Posick, C., & Rocque, M. (2016). The Criminal Brain: Understanding Biological Theories of Crime (2 ed.). New York: New York University Press. Rivara, F., Adhia, A., Lyons, V., Massey, A., Mills, B., Morgan, E., . . . Rowhani-Rahbar, A. (2019). The Effects Of Violence On Health. Health Affairs, 38(10), 1622-1629. doi:10.1377/hlthaff.2019.00480 Robinson, F., & Keithley, J. (2000). The impacts of crime on health and health services: A literature review. Health, Risk & Society, 2(3), 253-266. doi:10.1080/713670168 Rosell, D. R., & Siever, L. J. (2015). The neurobiology of aggression and violence. CNS Spectr, 20(3), 254-279. doi:10.1017/s109285291500019x Rosseel, Y. (2012). lavaan: An R Package for Structural Equation Modeling. Journal of Statistical Software, 48(2), 1-36. doi:10.18637/jss.v048.i02 Rosshart, S. P., Herz, J., Vassallo, B. G., Hunter, A., Wall, M. K., Badger, J. H., . . . Rehermann, B. (2019). Laboratory mice born to wild mice have natural microbiota and model human immune responses. Science, 365(6452), eaaw4361. doi:10.1126/science.aaw4361 RStudio Team. (2020). RStudio: Integrated Development for R. Boston, MA: RStudio, Inc. Retrieved from http://www.rstudio.com/ 71 Sampson, R. (2012). Great American City: Chicago and the Enduring Neighborhood Effect. Chicago: University of Chicago Press. Sampson, R., & Groves, W. (1989). Community Structure and Crime: Testing Social- Disorganization Theory. American Journal of Sociology, 94(4), 774-802. Sampson, R., Raudenbush, S., & Earls, F. (1997). Neighborhoods and Violent Crime: A Multilevel Study of Collective Efficacy. Science, 277(5328), 918-924. Sapolsky, R. M. (2017). Behave: The Biology of Humans at Our Best and Worst. New York: Penguin Press. Satorra, A., & Bentler, P. M. (2010). Ensuring Positiveness of the Scaled Difference Chi-square Test Statistic. Psychometrika, 75(2), 243-248. doi:10.1007/s11336-009-9135-y Savidge, T. C. (2016). Epigenetic Regulation of Enteric Neurotransmission by Gut Bacteria. Frontiers in Cellular Neuroscience, 9(503). doi:10.3389/fncel.2015.00503 Schepper, J. D., Collins, F. L., Rios-Arce, N. D., Raehtz, S., Schaefer, L., Gardinier, J. D., . . . McCabe, L. R. (2019). Probiotic Lactobacillus reuteri Prevents Postantibiotic Bone Loss by Reducing Intestinal Dysbiosis and Preventing Barrier Disruption. Journal of Bone and Mineral Research, 34(4), 681-698. doi:https://doi.org/10.1002/jbmr.3635 Sharkey, P. T., Tirado-Strayer, N., Papachristos, A. V., & Raver, C. C. (2012). The Effect of Local Violence on Children’s Attention and Impulse Control. American Journal of Public Health, 102(12), 2287-2293. Shaw, C. R., & McKay, H. D. (1942). Juvenile Delinquency and Urban Areas. Chicago: University of Chicago Press. Sherwin, E., Bordenstein, S. R., Quinn, J. L., Dinan, T. G., & Cryan, J. F. (2019). Microbiota and the social brain. Science, 366(6465), eaar2016. doi:10.1126/science.aar2016 Shirtcliff, E. A., Vitacco, M. J., Graf, A. R., Gostisha, A. J., Merz, J. L., & Zahn-Waxler, C. (2009). Neurobiology of Empathy and Callousness: Implications for the Development of Antisocial Behavior. Behavioral Sciences and the Law, 27(2), 137-171. doi:10.1002/bsl.862 Song, S. J., Lauber, C., Costello, E. K., Lozupone, C. A., Humphrey, G., Berg-Lyons, D., . . . Nakielny, S. (2013). Cohabiting family members share microbiota with one another and with their dogs. Elife, 2, e00458. Sonnenburg, J., & Sonnenburg, E. (2015). The Good Gut: Taking Control of Your Weight, Your Mood, and Your Long-term Health. New York: Penguin Books. 72 Spelman, W. (1993). Abandoned buildings: Magnets for crime? Journal of Criminal Justice, 21(5), 481-495. doi:https://doi.org/10.1016/0047-2352(93)90033-J Steenbergen, L., Sellaro, R., van Hemert, S., Bosch, J. A., & Colzato, L. S. (2015). A randomized controlled trial to test the effect of multispecies probiotics on cognitive reactivity to sad mood. Brain, Behavior, and Immunity, 48, 258-264. doi:http://dx.doi.org/10.1016/j.bbi.2015.04.003 Stockdale, S. E., Wells, K. B., Tang, L., Belin, T. R., Zhang, L., & Sherbourne, C. D. (2007). The importance of social context: Neighborhood stressors, stress-buffering mechanisms, and alcohol, drug, and mental health disorders. Social Science and Medicine, 65(9), 1867-1881. doi:https://doi.org/10.1016/j.socscimed.2007.05.045 Stretesky, P. B., & Lynch, M. J. (2004). The Relationship between Lead and Crime. Journal of Health and Social Behavior, 45(2), 214-229. doi:10.1177/002214650404500207 Tan, J., McKenzie, C., Potamitis, M., Thorburn, A. N., Mackay, C. R., & Macia, L. (2014). The Role of Short-Chain Fatty Acids in Health and Disease. In F. W. Alt (Ed.), Advances in Immunology (Vol. 121, pp. 91-119): Academic Press. Theall, K. P., Shirtcliff, E. A., Dismukes, A. R., Wallace, M., & Drury, S. S. (2017). Association Between Neighborhood Violence and Biological Stress in Children. JAMA Pediatrics, 171(1), 53-60. doi:10.1001/jamapediatrics.2016.2321 Thurstone, L. L. (1935). The vectors of mind; multiple-factor analysis for the isolation of primary traits. Chicago, Ill.: The University of Chicago Press. Tillisch, K., Labus, J., Kilpatrick, L., Jiang, Z., Stains, J., Ebrat, B., . . . Naliboff, B. (2013). Consumption of Fermented Milk Product With Probiotic Modulates Brain Activity. Gastroenterology, 144(7), 1394-1401. doi:10.1053/j.gastro.2013.02.043 Tremblay, R. E., Vitaro, F., & Côté, S. M. (2018). Developmental Origins of Chronic Physical Aggression: A Bio-Psycho-Social Model for the Next Generation of Preventive Interventions. Annual Review of Psychology, 69, 383-407. doi:10.1146/annurev-psych- 010416-044030 Tuvblad, C., & Baker, L. A. (2011). Human Aggression Across the Lifespan: Genetic Propensities and Environmental Moderators. In R. Huber, D. L. Bannasch, & P. Brennan (Eds.), Advances in Genetics (Vol. 75, pp. 171-214): Academic Press. Tuvblad, C., & Beaver, K. M. (2013). Genetic and environmental influences on antisocial behavior. Journal of Criminal Justice, 41(5), 273-276. doi:10.1016/j.jcrimjus.2013.07.007 U.S. Census Bureau. (2015). Community Facts. American Community Survey, from U.S. Department of Commerce https://factfinder.census.gov/ 73 USCB. (2012). US Census Bureau Geographic Entities and Concepts. Retrieved from https://www.census.gov/content/dam/Census/data/developers/geoareaconcepts.pdf. USCB. (2021a). American Community Survey: Concepts & Definitions. Retrieved from https://www.census.gov/programs-surveys/acs/geography-acs/concepts-definitions.html USCB. (2021b). What We Do. Retrieved from https://www.census.gov/about/what.html Vacca, M., Celano, G., Calabrese, F. M., Portincasa, P., Gobbetti, M., & De Angelis, M. (2020). The Controversial Role of Human Gut Lachnospiraceae. Microorganisms, 8(4), 573-598. van Dongen, J., Hagenbeek, F. A., Suderman, M., Roetman, P. J., Sugden, K., Chiocchetti, A. G., . . . Boomsma, D. I. (2021). DNA methylation signatures of aggression and closely related constructs: A meta-analysis of epigenome-wide studies across the lifespan. Molecular Psychiatry. doi:10.1038/s41380-020-00987-x Von Ehrenstein, O. S., Von Mutius, E., Illi, S., Baumann, L., Böhm, O., & von Kries, R. (2000). Reduced risk of hay fever and asthma among children of farmers. Clinical and experimental allergy : journal of the British Society for Allergy and Clinical Immunology, 30(2), 187-193. von Mutius, E., & Vercelli, D. (2010). Farm living: effects on childhood asthma and allergy. Nature Reviews: Immunology, 10(12), 861-868. doi:10.1038/nri2871 Walker, K. (2020). tigris: Load Census TIGER/Line Shapefiles (Version version 1.0). Retrieved from https://CRAN.R-project.org/package=tigris Walker, K., & Herman, M. (2021). tidycensus: Load US Census Boundary and Attribute Data as 'tidyverse' and 'sf'-Ready Data Frames (Version R package version 0.11.4). Retrieved from https://CRAN.R-project.org/package=tidycensus Wallace, L. A., Paul, R., Gholizadeh, S., Zadrozny, W., Webster, C., Mayfield, M., & Racine, E. F. (2021). Neighborhood disadvantage and the sales of unhealthy products: alcohol, tobacco and unhealthy snack food. BMC Public Health, 21(1), 1361. doi:10.1186/s12889-021-11442-z Warren, B. L., Mazei-Robison, M. S., Robison, A. J., & Iñiguez, S. D. (2020). Can I Get a Witness? Using Vicarious Defeat Stress to Study Mood-Related Illnesses in Traditionally Understudied Populations. Biological Psychiatry, 88(5), 381-391. doi:https://doi.org/10.1016/j.biopsych.2020.02.004 Weisburd, D., Eck, J. E., Braga, A. A., Telep, C. W., Cave, B., Bowers, K., . . . Yang, S. M. (2016). Place Matters: Criminology for the Twenty-First Century. New York, New York: Cambridge University Press. 74 Weisburd, D., & White, C. (2019). Hot Spots of Crime Are Not Just Hot Spots of Crime: Examining Health Outcomes at Street Segments. Journal of Contemporary Criminal Justice, 35(2), 142-160. doi:10.1177/1043986219832132 Woese, C. R. (1987). Bacterial evolution. Microbiological Reviews, 51(2), 221-271. The world’s most dangerous cities. (2017). The Economist. Retrieved from http://www.economist.com/blogs/graphicdetail/2017/03/daily-chart-23 Xu, C., Lee, S. K., Zhang, D., & Frenette, P. S. (2020). The Gut Microbiome Regulates Psychological-Stress-Induced Inflammation. Immunity, 53(2), 417-428.e414. doi:https://doi.org/10.1016/j.immuni.2020.06.025 Yamawaki, Y., Yoshioka, N., Nozaki, K., Ito, H., Oda, K., Harada, K., . . . Akagi, H. (2018). Sodium butyrate abolishes lipopolysaccharide-induced depression-like behaviors and hippocampal microglial activation in mice. Brain Research, 1680, 13-38. doi:https://doi.org/10.1016/j.brainres.2017.12.004 Yang, C., Fujita, Y., Ren, Q., Ma, M., Dong, C., & Hashimoto, K. (2017). Bifidobacterium in the gut microbiota confer resilience to chronic social defeat stress in mice. Scientific Reports, 7, 45942. doi:10.1038/srep45942 Yang, Y., & Raine, A. (2009). Prefrontal structural and functional brain imaging findings in antisocial, violent, and psychopathic individuals: A meta-analysis. Psychiatry Research: Neuroimaging, 174(2), 81-88. doi:http://dx.doi.org/10.1016/j.pscychresns.2009.03.012 Yatsunenko, T., Rey, F. E., Manary, M. J., Trehan, I., Dominguez-Bello, M. G., Contreras, M., . . . Gordon, J. I. (2012). Human gut microbiome viewed across age and geography. Nature, 486(7402), 222-227. doi:10.1038/nature11053 Zaalberg, A., Nijman, H., Bulten, E., Stroosma, L., & van der Staak, C. (2010). Effects of Nutritional Supplements on Aggression, Rule‐Breaking, and Psychopathology Among Young Adult Prisoners. Aggressive Behavior, 36(2), 117-126. doi:doi:10.1002/ab.20335 Zeoli, A. M., Paruk, J. K., Pizarro, J. M., & Goldstick, J. (2019). Ecological Research for Studies of Violence: A Methodological Guide. Journal of interpersonal violence, 34(23-24), 4860-4880. doi:10.1177/0886260519871528 Zhang, L., Meng, J., Ban, Y., Jalodia, R., Chupikova, I., Fernandez, I., . . . Roy, S. (2019). Morphine tolerance is attenuated in germfree mice and reversed by probiotics, implicating the role of gut microbiome. Proceedings of the National Academy of Sciences, 116(27), 13523. doi:10.1073/pnas.1901182116 Zhang, X., Mallick, H., Tang, Z., Zhang, L., Cui, X., Benson, A. K., & Yi, N. (2017). Negative binomial mixed models for analyzing microbiome count data. BMC Bioinformatics, 18(1), 4. doi:10.1186/s12859-016-1441-7 75 Zhang, X., & Yi, N. (2020). NBZIMM: negative binomial and zero-inflated mixed models, with application to microbiome/metagenomics data analysis. BMC Bioinformatics, 21(1), 488. doi:10.1186/s12859-020-03803-z Zhang, Y., Pechal, J. L., Schmidt, C. J., Jordan, H. R., Wang, W. W., Benbow, M. E., . . . Tarone, A. M. (2019). Machine learning performance in a microbial molecular autopsy context: A cross-sectional postmortem human population study. PloS One, 14(4), e0213829. doi:10.1371/journal.pone.0213829 Zohoori, N., & Savitz, D. A. (1997). Econometric approaches to epidemiologic data: Relating endogeneity and unobserved heterogeneity to confounding. Annals of Epidemiology, 7(4), 251-257. doi:https://doi.org/10.1016/S1047-2797(97)00023-9 76 APPENDIX Table 9 | Negative Binomial Mixed Models of OTUs with Crime Level and Socioeconomic Predictors Family Genus Variable Estimate SE p p-adj Model Erysipelotrichaceae Clostridium:innocuum Crime -6.841 3.130 0.032 0.180 Model 1 Level Bifidobacteriaceae Bifidobacterium Crime -5.764 0.660 0.000 0.000 Model 1 Level Streptococcaceae Streptococcus Crime -2.963 0.416 0.000 0.000 Model 1 Level Bifidobacteriaceae Bifidobacterium Crime -2.655 0.554 0.000 0.000 Model 1 Level Desulfovibrionaceae Desulfovibrio Crime -2.584 0.420 0.000 0.000 Model 1 Level Burkholderiaceae Sutterella Crime -2.030 0.611 0.001 0.020 Model 1 Level FamilyXI W5053 Crime -1.977 0.598 0.001 0.020 Model 1 Level Coriobacteriaceae Collinsella Crime -1.936 0.683 0.006 0.057 Model 1 Level Corynebacteriaceae Corynebacterium1 Crime -1.756 0.337 0.000 0.000 Model 1 Level Veillonellaceae Unassigned Crime -1.709 0.811 0.038 0.190 Model 1 Level Lachnospiraceae Lachnoclostridium Crime -1.378 0.540 0.013 0.098 Model 1 Level Lachnospiraceae Ruminococcus:torques Crime -1.359 0.471 0.005 0.054 Model 1 Level FamilyXI Anaerococcus Crime -0.845 0.349 0.018 0.120 Model 1 Level Desulfovibrionaceae Bilophila Crime -0.773 0.367 0.038 0.190 Model 1 Level Burkholderiaceae Sutterella Crime -0.683 0.305 0.028 0.160 Model 1 Level Lachnospiraceae Roseburia Crime -0.651 0.249 0.011 0.088 Model 1 Level Ruminococcaceae RuminococcaceaeUCG-002 Crime -0.629 0.256 0.016 0.110 Model 1 Level Lachnospiraceae Blautia Crime -0.559 0.244 0.025 0.150 Model 1 Level Ruminococcaceae Subdoligranulum Crime -0.345 0.171 0.048 0.240 Model 1 Level Porphyromonadaceae Porphyromonas Crime 0.722 0.259 0.007 0.058 Model 1 Level Veillonellaceae Dialister Crime 0.820 0.291 0.006 0.057 Model 1 Level Ruminococcaceae Fastidiosipila Crime 1.063 0.366 0.005 0.054 Model 1 Level Porphyromonadaceae Porphyromonas Crime 1.495 0.484 0.003 0.036 Model 1 Level Rikenellaceae Alistipes Crime 2.118 0.850 0.015 0.110 Model 1 Level Porphyromonadaceae Porphyromonas Crime 3.573 0.780 0.000 0.000 Model 1 Level FamilyXI Anaerococcus Crime -8.415 3.090 0.008 0.094 Model 2 Level 77 Table 9 (cont’d) Erysipelotrichaceae Clostridium:innocuum Crime -6.429 3.218 0.049 0.300 Model 2 Level Streptococcaceae Streptococcus Crime -3.248 0.504 0.000 0.000 Model 2 Level Porphyromonadaceae Porphyromonas Crime -2.744 0.342 0.000 0.000 Model 2 Level Prevotellaceae Prevotella Crime -2.593 0.593 0.000 0.001 Model 2 Level Desulfovibrionaceae Desulfovibrio Crime -2.377 0.531 0.000 0.001 Model 2 Level Coriobacteriaceae Collinsella Crime -2.219 0.697 0.002 0.034 Model 2 Level Erysipelotrichaceae Erysipelatoclostridium Crime -1.902 0.774 0.016 0.160 Model 2 Level Lachnospiraceae Tyzzerella Crime -1.427 0.291 0.000 0.000 Model 2 Level Corynebacteriaceae Corynebacterium1 Crime -1.421 0.327 0.000 0.001 Model 2 Level Burkholderiaceae Sutterella Crime -1.194 0.227 0.000 0.000 Model 2 Level Lachnospiraceae Ruminococcus:torques Crime -1.017 0.372 0.008 0.094 Model 2 Level Desulfovibrionaceae Bilophila Crime -0.887 0.372 0.020 0.170 Model 2 Level FamilyXI Anaerococcus Crime -0.883 0.427 0.042 0.300 Model 2 Level Porphyromonadaceae Porphyromonas Crime -0.463 0.223 0.041 0.300 Model 2 Level Veillonellaceae Dialister Crime 0.595 0.290 0.044 0.300 Model 2 Level Porphyromonadaceae Porphyromonas Crime 0.613 0.265 0.023 0.180 Model 2 Level Veillonellaceae Negativicoccus Crime 0.708 0.277 0.012 0.130 Model 2 Level Porphyromonadaceae Porphyromonas Crime 1.205 0.502 0.019 0.170 Model 2 Level Rikenellaceae Alistipes Crime 2.286 0.731 0.003 0.036 Model 2 Level Streptococcaceae Streptococcus Vacancy -10.784 3.727 0.005 0.110 Model 2 Streptococcaceae Streptococcus Vacancy -3.692 1.314 0.006 0.120 Model 2 Peptostreptococcaceae Criibacteriumbergeronii Vacancy -3.629 0.951 0.000 0.036 Model 2 Lachnospiraceae Tyzzerella Vacancy -3.626 1.138 0.002 0.067 Model 2 Lachnospiraceae Ruminococcus:torques Vacancy -3.054 1.458 0.040 0.430 Model 2 FamilyXI Finegoldia Vacancy 3.244 1.241 0.011 0.180 Model 2 FamilyXI Peptoniphilus Vacancy 3.962 1.596 0.015 0.190 Model 2 Porphyromonadaceae Porphyromonas Vacancy 4.911 1.964 0.015 0.190 Model 2 FamilyXI W5053 Vacancy 7.078 2.160 0.002 0.067 Model 2 Ruminococcaceae Eubacterium:coprostanoligene Vacancy 7.331 2.467 0.004 0.100 Model 2 Corynebacteriaceae Corynebacterium1 Vacancy 27.633 12.757 0.034 0.400 Model 2 FamilyXI Anaerococcus Vacancy 40.051 12.100 0.002 0.067 Model 2 FamilyXI Anaerococcus % Black -4.304 2.079 0.042 0.260 Model 3 Corynebacteriaceae Corynebacterium1 % Black -3.108 0.970 0.002 0.032 Model 3 FamilyXI Peptoniphilus % Black -2.197 1.087 0.047 0.270 Model 3 78 Table 9 (cont’d) Actinomycetaceae Mobiluncus % Black -1.823 0.425 0.000 0.002 Model 3 Lachnospiraceae Blautia % Black 0.943 0.313 0.004 0.052 Model 3 Lachnospiraceae Lachnoclostridium % Black 1.172 0.472 0.015 0.150 Model 3 Bacteroidaceae Bacteroides % Black 1.217 0.451 0.009 0.100 Model 3 Enterobacteriaceae Escherichia-Shigella % Black 1.392 0.628 0.030 0.200 Model 3 Erysipelotrichaceae Erysipelatoclostridium % Black 1.494 0.569 0.011 0.120 Model 3 Bifidobacteriaceae Bifidobacterium % Black 1.497 0.379 0.000 0.005 Model 3 Burkholderiaceae Parasutterella % Black 1.986 0.716 0.007 0.090 Model 3 Burkholderiaceae Sutterella % Black 2.017 0.573 0.001 0.014 Model 3 Porphyromonadaceae Porphyromonas % Black 2.084 0.939 0.030 0.200 Model 3 Erysipelotrichaceae Holdemanella % Black 2.688 0.372 0.000 0.000 Model 3 Lachnospiraceae Blautia % Black 2.838 1.221 0.023 0.190 Model 3 Bacteroidaceae Bacteroides % Black 2.899 1.191 0.017 0.150 Model 3 Lachnospiraceae Agathobacter % Black 2.991 0.001 0.000 0.000 Model 3 Erysipelotrichaceae Clostridium:innocuum % Black 3.214 1.322 0.018 0.150 Model 3 Veillonellaceae Dialister % Black 4.077 1.058 0.000 0.005 Model 3 Lachnospiraceae Anaerostipes % Black 6.971 1.247 0.000 0.000 Model 3 Enterobacteriaceae Unassigned % Black 11.762 5.234 0.028 0.200 Model 3 Lachnospiraceae Fusicatenibacter % Black 12.311 5.845 0.039 0.250 Model 3 Lachnospiraceae Agathobacter Crime -6.982 0.001 0.000 0.000 Model 3 Level Porphyromonadaceae Porphyromonas Crime -4.951 0.586 0.000 0.000 Model 3 Level Acidaminococcaceae Phascolarctobacterium Crime -4.628 2.086 0.030 0.170 Model 3 Level Veillonellaceae Negativicoccus Crime -2.810 0.608 0.000 0.001 Model 3 Level Prevotellaceae Prevotella Crime -2.441 0.479 0.000 0.000 Model 3 Level Porphyromonadaceae Porphyromonas Crime -2.015 0.579 0.001 0.014 Model 3 Level Coriobacteriaceae Collinsella Crime -1.787 0.683 0.011 0.095 Model 3 Level Prevotellaceae Prevotella Crime -1.491 0.506 0.004 0.050 Model 3 Level Streptococcaceae Streptococcus Crime -1.476 0.529 0.007 0.066 Model 3 Level Desulfovibrionaceae Desulfovibrio Crime -1.474 0.649 0.026 0.160 Model 3 Level FamilyXI W5053 Crime -1.383 0.601 0.024 0.150 Model 3 Level Lachnospiraceae Tyzzerella Crime -1.331 0.376 0.001 0.013 Model 3 Level Lachnospiraceae Lachnoclostridium Crime -1.154 0.523 0.031 0.170 Model 3 Level Burkholderiaceae Sutterella Crime -1.080 0.358 0.004 0.045 Model 3 Level Lachnospiraceae Ruminococcus:torques Crime -1.051 0.267 0.000 0.005 Model 3 Level Ruminococcaceae RuminococcaceaeUCG-002 Crime -0.848 0.352 0.019 0.140 Model 3 Level 79 Table 9 (cont’d) Desulfovibrionaceae Bilophila Crime -0.779 0.373 0.040 0.190 Model 3 Level FamilyXI Anaerococcus Crime -0.771 0.288 0.009 0.085 Model 3 Level Erysipelotrichaceae Erysipelatoclostridium Crime -0.751 0.355 0.038 0.190 Model 3 Level Burkholderiaceae Sutterella Crime -0.719 0.311 0.024 0.150 Model 3 Level Erysipelotrichaceae Holdemanella Crime -0.704 0.232 0.003 0.045 Model 3 Level Bifidobacteriaceae Bifidobacterium Crime -0.676 0.237 0.006 0.060 Model 3 Level Actinomycetaceae Varibaculum Crime -0.640 0.296 0.034 0.180 Model 3 Level Bacteroidaceae Bacteroides Crime 0.565 0.281 0.048 0.210 Model 3 Level Actinomycetaceae Mobiluncus Crime 0.639 0.265 0.019 0.140 Model 3 Level Corynebacteriaceae Corynebacterium1 Crime 1.212 0.605 0.049 0.210 Model 3 Level FamilyXI Anaerococcus Crime 1.333 0.666 0.049 0.210 Model 3 Level Porphyromonadaceae Porphyromonas Crime 1.486 0.386 0.000 0.005 Model 3 Level Lachnospiraceae Anaerostipes Crime 1.759 0.778 0.027 0.160 Model 3 Level Rikenellaceae Alistipes Crime 2.389 0.987 0.018 0.140 Model 3 Level Erysipelotrichaceae Clostridium:innocuum Crime -7.207 3.182 0.026 0.170 Model 4 Level FamilyXI Anaerococcus Crime -6.592 3.192 0.043 0.220 Model 4 Level Bifidobacteriaceae Bifidobacterium Crime -5.462 0.699 0.000 0.000 Model 4 Level Veillonellaceae Negativicoccus Crime -5.054 0.523 0.000 0.000 Model 4 Level Porphyromonadaceae Porphyromonas Crime -4.178 1.864 0.028 0.170 Model 4 Level Desulfovibrionaceae Desulfovibrio Crime -3.358 0.469 0.000 0.000 Model 4 Level Erysipelotrichaceae Holdemanella Crime -2.828 1.131 0.015 0.110 Model 4 Level Bifidobacteriaceae Bifidobacterium Crime -2.731 0.468 0.000 0.000 Model 4 Level Erysipelotrichaceae Erysipelatoclostridium Crime -2.662 0.903 0.004 0.042 Model 4 Level Burkholderiaceae Sutterella Crime -2.091 0.626 0.001 0.019 Model 4 Level Coriobacteriaceae Collinsella Crime -1.848 0.688 0.009 0.077 Model 4 Level Veillonellaceae Unassigned Crime -1.690 0.810 0.040 0.220 Model 4 Level Peptostreptococcaceae Criibacteriumbergeronii Crime -1.614 0.513 0.002 0.031 Model 4 Level Prevotellaceae Prevotella Crime -1.609 0.534 0.004 0.042 Model 4 Level 80 Table 9 (cont’d) Lachnospiraceae Tyzzerella Crime -1.518 0.362 0.000 0.001 Model 4 Level Lachnospiraceae Dorea Crime -1.228 0.273 0.000 0.000 Model 4 Level Lachnospiraceae Ruminococcus:torquesgroup Crime -1.061 0.363 0.005 0.042 Model 4 Level Lachnospiraceae Lachnoclostridium Crime -0.759 0.323 0.021 0.150 Model 4 Level Desulfovibrionaceae Bilophila Crime -0.693 0.267 0.011 0.089 Model 4 Level Burkholderiaceae Sutterella Crime -0.660 0.310 0.036 0.210 Model 4 Level Actinomycetaceae Varibaculum Crime -0.649 0.318 0.045 0.220 Model 4 Level Ruminococcaceae UBA1819 Crime -0.552 0.263 0.040 0.220 Model 4 Level Prevotellaceae Prevotella6 Crime -0.450 0.196 0.025 0.170 Model 4 Level Veillonellaceae Dialister Crime 0.901 0.179 0.000 0.000 Model 4 Level Ruminococcaceae Faecalibacterium Crime 1.576 0.000 0.000 0.000 Model 4 Level Rikenellaceae Alistipes Crime 2.163 0.740 0.005 0.042 Model 4 Level Ruminococcaceae Faecalibacterium Unemploy. -26.970 0.000 0.000 0.000 Model 4 Ruminococcaceae Faecalibacterium Unemploy. -15.206 5.934 0.012 0.074 Model 4 Veillonellaceae Dialister Unemploy. -11.591 1.035 0.000 0.000 Model 4 Lachnospiraceae Agathobacter Unemploy. -11.515 3.291 0.001 0.009 Model 4 Lachnospiraceae Anaerostipes Unemploy. -11.456 2.847 0.000 0.002 Model 4 Ruminococcaceae RuminococcaceaeUCG-003 Unemploy. -10.481 2.435 0.000 0.002 Model 4 Streptococcaceae Streptococcus Unemploy. -9.397 2.298 0.000 0.002 Model 4 Bacteroidaceae Bacteroides Unemploy. -9.284 2.562 0.001 0.007 Model 4 Lachnospiraceae Ruminococcus:torquesgroup Unemploy. -3.403 1.553 0.032 0.130 Model 4 Desulfovibrionaceae Bilophila Unemploy. -2.797 1.141 0.017 0.084 Model 4 Corynebacteriaceae Lawsonella Unemploy. 2.079 0.925 0.028 0.120 Model 4 Peptococcaceae Peptococcus Unemploy. 2.922 0.912 0.002 0.018 Model 4 FamilyXI Peptoniphilus Unemploy. 3.014 1.101 0.008 0.059 Model 4 Porphyromonadaceae Porphyromonas Unemploy. 3.116 1.523 0.044 0.170 Model 4 FamilyXI Peptoniphilus Unemploy. 3.211 1.551 0.042 0.170 Model 4 Lachnospiraceae Tyzzerella Unemploy. 3.862 1.548 0.015 0.081 Model 4 FamilyXI Peptoniphilus Unemploy. 4.113 1.611 0.013 0.076 Model 4 Lachnospiraceae Dorea Unemploy. 4.437 1.166 0.000 0.004 Model 4 Veillonellaceae Dialister Unemploy. 4.654 0.766 0.000 0.000 Model 4 Rikenellaceae RikenellaceaeRC9gutgroup Unemploy. 4.973 1.972 0.014 0.079 Model 4 Peptostreptococcaceae Criibacteriumbergeronii Unemploy. 5.102 2.194 0.023 0.110 Model 4 FamilyXI Anaerococcus Unemploy. 5.565 2.328 0.019 0.091 Model 4 Veillonellaceae Dialister Unemploy. 5.854 1.413 0.000 0.002 Model 4 FamilyXI Peptoniphilus Unemploy. 5.918 1.834 0.002 0.018 Model 4 Porphyromonadaceae Porphyromonas Unemploy. 6.735 2.522 0.009 0.067 Model 4 Burkholderiaceae Parasutterella Unemploy. 7.196 2.592 0.007 0.056 Model 4 81 Table 9 (cont’d) Erysipelotrichaceae Holdemanella Unemploy. 7.253 3.264 0.029 0.120 Model 4 Ruminococcaceae Eubacterium:coprostanoligene Unemploy. 7.365 3.028 0.017 0.084 Model 4 Veillonellaceae Negativicoccus Unemploy. 9.263 2.240 0.000 0.002 Model 4 Bifidobacteriaceae Bifidobacterium Unemploy. 9.601 2.992 0.002 0.018 Model 4 Porphyromonadaceae Porphyromonas Unemploy. 10.049 3.207 0.003 0.022 Model 4 FamilyXI W5053 Unemploy. 17.847 6.910 0.012 0.074 Model 4 Porphyromonadaceae Porphyromonas Unemploy. 20.364 7.852 0.011 0.074 Model 4 82 Table 10 | 52 Unique Taxa Associated with Crime Level and Socioeconomic Predictors Phylum Class Order Family Genus Firmicutes Bacilli Lactobacillales Streptococcaceae Streptococcus Firmicutes Clostridia Clostridiales Lachnospiraceae Ruminococcus:torques Actinobacteria Actinobacteria Corynebacteriales Corynebacteriaceae Corynebacterium1 Bacteroidetes Bacteroidia Bacteroidales Porphyromonadaceae Porphyromonas Proteobacteria Gammaproteobacteria Betaproteobacteriales Burkholderiaceae Sutterella Proteobacteria Deltaproteobacteria Desulfovibrionales Desulfovibrionaceae Bilophila Actinobacteria Coriobacteriia Coriobacteriales Coriobacteriaceae Collinsella Actinobacteria Actinobacteria Bifidobacteriales Bifidobacteriaceae Bifidobacterium Bacteroidetes Bacteroidia Bacteroidales Prevotellaceae Prevotella Firmicutes Clostridia Clostridiales FamilyXI W5053 Firmicutes Erysipelotrichia Erysipelotrichales Erysipelotrichaceae Clostridium:innocuum Firmicutes Clostridia Clostridiales Ruminococcaceae RuminococcaceaeUCG-002 Bacteroidetes Bacteroidia Bacteroidales Rikenellaceae Alistipes Firmicutes Negativicutes Selenomonadales Veillonellaceae Unassigned Firmicutes Clostridia Clostridiales Ruminococcaceae Subdoligranulum Firmicutes Clostridia Clostridiales FamilyXI Anaerococcus Firmicutes Clostridia Clostridiales Lachnospiraceae Lachnoclostridium Firmicutes Clostridia Clostridiales Lachnospiraceae Blautia Firmicutes Clostridia Clostridiales Lachnospiraceae Roseburia Proteobacteria Deltaproteobacteria Desulfovibrionales Desulfovibrionaceae Desulfovibrio Firmicutes Clostridia Clostridiales Ruminococcaceae Fastidiosipila Firmicutes Negativicutes Selenomonadales Veillonellaceae Dialister Firmicutes Negativicutes Selenomonadales Veillonellaceae Negativicoccus Firmicutes Clostridia Clostridiales FamilyXI Finegoldia Firmicutes Clostridia Clostridiales Lachnospiraceae Tyzzerella Firmicutes Clostridia Clostridiales Ruminococcaceae Eubacterium:coprostanoligene Firmicutes Clostridia Clostridiales Peptostreptococcaceae Criibacteriumbergeronii Firmicutes Erysipelotrichia Erysipelotrichales Erysipelotrichaceae Erysipelatoclostridium Firmicutes Clostridia Clostridiales FamilyXI Peptoniphilus Actinobacteria Actinobacteria Actinomycetales Actinomycetaceae Varibaculum Firmicutes Clostridia Clostridiales Lachnospiraceae Agathobacter Firmicutes Clostridia Clostridiales Lachnospiraceae Anaerostipes Proteobacteria Gammaproteobacteria Betaproteobacteriales Burkholderiaceae Parasutterella Firmicutes Erysipelotrichia Erysipelotrichales Erysipelotrichaceae Holdemanella Bacteroidetes Bacteroidia Bacteroidales Bacteroidaceae Bacteroides Proteobacteria Gammaproteobacteria Enterobacteriales Enterobacteriaceae Unassigned Firmicutes Negativicutes Selenomonadales Acidaminococcaceae Phascolarctobacterium Actinobacteria Actinobacteria Actinomycetales Actinomycetaceae Mobiluncus Proteobacteria Gammaproteobacteria Enterobacteriales Enterobacteriaceae Escherichia-Shigella Firmicutes Clostridia Clostridiales Lachnospiraceae Fusicatenibacter Firmicutes Clostridia Clostridiales Peptococcaceae Peptococcus Firmicutes Clostridia Clostridiales Ruminococcaceae Faecalibacterium Firmicutes Clostridia Clostridiales Ruminococcaceae RuminococcaceaeUCG-003 Bacteroidetes Bacteroidia Bacteroidales Prevotellaceae Prevotella6 83 Table 10 (cont’d) Actinobacteria Actinobacteria Corynebacteriales Corynebacteriaceae Lawsonella Bacteroidetes Bacteroidia Bacteroidales Rikenellaceae RikenellaceaeRC9gut Firmicutes Clostridia Clostridiales Ruminococcaceae UBA1819 Firmicutes Clostridia Clostridiales Lachnospiraceae Dorea Actinobacteria Actinobacteria Actinomycetales Actinomycetaceae Actinomyces Firmicutes Clostridia Clostridiales Ruminococcaceae Ruminiclostridium5 Firmicutes Negativicutes Selenomonadales Veillonellaceae Megasphaera Bacteroidetes Bacteroidia Bacteroidales Marinifilaceae Odoribacter 84