The past decade has brought remarkable advances in the integration of social and biological models of health across the life course (see also Gruenewald, this volume). Research is beginning to specify multilevel connections between diverse social experiences—reflecting status, isolation, support, and stressors—and biological pathways, including neuroendocrine processes and intracellular mechanisms involving the genome (Miller, Chen, and Cole, 2009). Moreover, these multilevel complexities are now being studied with reference to life course models (principally, sensitive period, accumulation, and pathway; Shanahan and Hofer, 2011).
This chapter focuses on one promising subfield of this larger literature, social genomic studies of genetic transcription, and the opportunities and challenges that it presents for demographers and social epidemiologists who study aging and health. Social genomics was chosen as the focus for this chapter because this subfield attempts to interrelate social settings with gene expression by way of chains of mediating factors, and thus illustrates the promise and challenges of multilevel research in sharp relief.
The field of social genomics focuses on the mechanisms by which social experiences regulate genetic activity (Cole, 2009). Transcription regulation refers to processes that govern the rate at which DNA is transcribed into messenger RNA (mRNA), which in turn eventuates in proteins. Thus, social genomic studies of transcription focus on how social experiences influence mechanisms by which the information contained in the DNA is transcribed (or written out) into mRNA. This focus on transcription is important because mRNAs serve as the molecular building blocks for proteins, which are integral to virtually every biological process in the cell.
Social factors likely influence genetic activity by way of complex mediating chains involving many levels of analysis, possibly extending, for example, from political economies to people’s reactions to their immediate circumstance to intracellular mechanisms. By establishing these meditational links between social experiences and transcriptional activity, scientists can begin to understand how social experiences, like those associated with socioeconomic status (SES), affect physical and mental health.
The possibility that social genomics can articulate such mechanisms is highly significant, because most population research relies on nonexperimental survey data that make causal inference provisional. Human studies of transcription often suffer from the same limitations; however, these human studies can then inform nonhuman animal studies that use experimental designs, and even human experimental designs. When these diverse types of studies converge on a mechanistic model that links social experiences with expression profiles, causal inferences about social context and health are indeed strengthened. Thus, a major payoff of social genomics for demography and social epidemiology is the strengthening of causal claims by specifying the mechanisms that link social experience with behavior and health.
Viewed broadly, social genomic research to date suggests a “two-stage” historiography: (1) earlier research, conducted by social and behavioral scientists, identified putative social risk factors and (2) investigations, conducted largely by health and biological psychologists, then focused on how such risk factors “get under the skin,” eventuating in altered gene expression. As a result of this two-stage exchange, increasingly sophisticated models are emerging that aspire to mechanistically connect social risk factors, psychological and neuroendocrine mediators, and molecular processes that, in turn, explain the emergence and progression of diseases. These advances raise new questions for a “Stage Three,” the integration of population-based studies of social risks with gene expression mechanisms in the study of health and aging.
Drawing on early research in this area, which examined several acute and chronic stressors, the chapter begins by explaining the concept of transcription and, in general terms, research strategies that are used to study social factors and transcription. The chapter then reviews advances in social genomics with particular attention to studies of socioeconomic status, which are of central interest to population researchers. (The chapter could also have focused on social isolation as an illustrative example, and interested readers are directed to Cacioppo et al., 2011.) Finally, the chapter identifies a series of opportunities and challenges for Stage Three, population studies of health and aging that are informed by transcription studies. These opportunities and challenges include (1) the collection of multilevel data, including expression data, relevant biomarkers associated with stress
response, and neuroimaging data; (2) extensively longitudinal designs that can adjudicate among several life course models; (3) the refinement of measures of social context; (4) the use of large, diverse, population-based samples; and (5) the strategic use of diverse designs to strengthen causal claims and to examine the effects of socioeconomic status on gene expression in different policy settings, political economies, in societies characterized by different demographic compositions, and among migrants. In turn, such challenges and opportunities call for the innovative organization of interdisciplinary teams of scientists.
The overarching point of this chapter, then, is that population-based research originally inspired social genomic (transcription) studies, and the results of these investigations now suggest future directions for population-based studies of social location, health, and aging. Ideally, such studies will be designed to study multilevel, meditational processes as they extend over many decades of life.
SOCIOECONOMIC STATUS, GENE TRANSCRIPTION, AND INFLAMMATORY PROCESSES: STAGES ONE AND TWO
The Emergence of Transcription Studies of Social Experiences
Transcription depends on RNA polymerase (RNAP), an enzyme that attaches to the promoter region near a gene and that then synthesizes mRNA from the DNA template strand. In turn, the mRNA shuttles to the ribosomes, where encoded proteins are synthesized and then perform two broadly defined functions. First, the synthesized proteins maintain and regulate the basic metabolic processes necessary for life. The ENCODE Project—which draws on high-throughput sequencing to examine the functional elements of the human genome—reveals that transcription processes are quite complex, that the genome contains much more information than is expressed at any given time, and that cells are highly selective in which genes are expressed (ENCODE Project Consortium, 2011; Weinstock, 2007, and accompanying articles). This last point is evident when one considers that the diverse cells that make up the human body all develop and are regulated with instructions from identical copies of the individual’s DNA. Because each person’s DNA contains a massive amount of information, it must be selectively expressed to orchestrate coherent development.
Second, transcription is also an important (although not exclusive) mechanism of adaptation, allowing the cell to respond to changing circumstances. The process of adaptation depends on the regulation of RNAP and its attachment to the promoter region of a gene. The promoter region includes response elements, short sequences of DNA that can bind molecular flags known as transcription factors. Once a transcription factor attaches
to a target response element, it can promote or block the recruitment of RNAP. When the transcription factor attaches to the response element and recruits RNAP, the gene is said to be up-regulated, meaning that its rate of expression increases. Blocking (or repression) of RNAP leads to downregulation. Roughly 200 transcription factors operate in the mammalian genome, and each binds to a specific stretch of DNA (i.e., a response element), which is typically 10-20 but possibly up to 40 nucleotides in length. Research has identified the nucleotide sequence for many major transcription factors, and researchers can combine that information with data from the Human Genome Project to make inferences about patterns of transcription factor activity.
Transcription factors themselves are the product of intracellular signaling cascades and signals (or stimuli) originating from the environment. Heat shock factor was perhaps the first transcription factor identified that is responsive to environmental stimuli, discovered in 1974 in drosophila melongaster. This transcription factor is, not surprisingly, highly conserved across species and indeed observed in humans. However, that transcription was downstream from social experiences—that is, responsive to social circumstances by way of mediating chains—was only recently documented, beginning with studies of psychosocial factors and HIV-1 progression (Cole, 2008b).
Several streams of evidence (involving humans and nonhuman primate experiments) pointed to the activation of the sympathetic nervous system (SNS; typically associated with the body’s “flight or fight” response to stressors) in increasing the HIV’s transcription and replication and, hence, the progression of the disease. That is, psychosocial stressors and reactivity to stressors eventuate in faster progression of HIV-1 because of increased viral replication. Because the SNS is downstream from social stressors, however, this research suggested a broader possibility: that, beyond the specific example of HIV-1, social experiences could influence gene expression because of their effects on the nervous system.
Cole and his colleagues (2007) were among the first to examine this possibility, focusing on social isolation and the transcriptional profiles of white blood cells (leukocytes). Consistent with the two-stage historiography suggested earlier, this study began with social epidemiological and sociological research identifying social isolation as a risk factor for diverse forms of distress (e.g., Seeman, 1996), including physical diseases. In fact, Cole and his colleagues had earlier shown (1994) that social isolation was associated with increased activity in threat-related pathways of the SNS (for a recent review, see Irwin and Cole, 2011).
The study’s focus on white blood cells was highly strategic. Gene expression is likely tissue-specific, making the study of many of the body’s organs a highly invasive process. Peripheral blood cells, in contrast, may
be collected with minimal inconvenience to participants. Also, white blood cells are integral to host resistance to infection and inflammatory processes, which are implicated in cardiovascular disease, neurodegenerative disease, and likely depression. Thus, this study of social isolation joined two very different research traditions: a considerable body of population-based research on risk factors, and biological models of leukocytes, which offer a noninvasive window into the immune system and inflammatory processes associated with many common diseases of aging.
The basic issue, then, was whether genes were expressed differentially among socially isolated versus socially integrated adults. In turn, the answer to this question was examined in terms of three questions. First, are transcripts differentially expressed between socially isolated and integrated people? Fourteen healthy adults from the Chicago Health, Aging, and Social Relations Study were selected based on their relatively high or low scores on a standard measure of subjective loneliness, the University of California, Los Angeles, Loneliness Scale. Differential expression of mRNA was then examined using global gene expression profiling and was observed in 209 transcripts (of roughly 22,000 examined transcripts) correcting for a false discovery rate. Thus, the social isolates differed in their gene expression profiles from the socially integrated adults.
Second, what were the functions of these differentially expressed genes? The functions of these transcripts were then identified using the Gene Ontology (GO) catalog, a directory that lists the functions of gene products (see http://www.geneontology.org). The issue was whether the over- and underexpressed transcripts had biological functions that would indicate how social isolation could affect health. In fact, the GO categories indicated that lonely adults were, generally, showing greater activation of innate immune cells (e.g., pro-inflammatory signaling). And third, among the differentially expressed genes, could specific transcriptional pathways be identified that would help explain immune activation? This question was examined using TELiS, bioinformatics software that analyzes the prevalence of specific transcription factor binding motifs (TFBMs) in differentially expressed genes.
Several pathways were identified, but one—involving the glucocorticoid receptor (GR)—is particularly significant because of its replication by other studies, its biological plausibility, and its possible applicability beyond the specific case of social isolation. Glucocorticoids (Gs) are a class of steroid hormones that are integral to the immune system; when bound to a glucocorticoid receptor, the resulting complex migrates to the nucleus where it up-regulates genes that code for anti-inflammatory proteins and slows the expression of pro-inflammatory proteins by preventing other transcription factors from entering the nucleus. Thus, Gs and GRs are central players in inflammatory responses.
Cole and his colleagues found significantly less expression of GR-target genes in the leukocytes of socially isolated adults when compared to socially integrated ones. This pattern is consistent with the “glucocorticoid insensitivity hypothesis”: Chronically stressed individuals become insensitive to the anti-inflammatory actions of Gs (such as cortisol), reflecting the downregulation of GR transcription factors. That is, chronic stress lessens the ability of G-GR complex to work as a transcription factor that ultimately reduces inflammation (for experimental support, see Cole, Mendoza, and Capitanio, 2009; for evidence suggesting that even low levels of isolation could trigger G insensitivity, see Cole, 2008a).
Additional support for G resistance is found in Miller and his colleagues’ (2008) study of links between chronic stress and transcription, a genome-wide expression study (using microarray technology) of family care-givers of brain cancer patients and matched controls. They hypothesized that chronic stress elevates cortisol, which in turn leads to (1) an eventual, compensatory down-regulation of GRs in monocytes (a type of white blood cell that is integral to immune response) and (2) a consequent enhancement of pro-inflammatory signaling. Miller et al. (2009) examined the TFBMs in the promoter regions of differentially expressed genes. As expected, TFBMs occurred about 23 percent less frequently in genes associated with glucocorticoid receptors, and they occurred more frequently in genes associated with pro-inflammation in caregivers, compared to controls. Overall, the results suggested that social stress has a “transcriptional fingerprint” involving resistance to glucocorticoids and mild systemic inflammation (for an additional example focused on interpersonal stress, see Miller et al., 2009).
Transcriptional Studies of SES
In addition to studies of acute and chronic stressors, research has also examined SES, of central interest to demography and social epidemiology in the study of health disparities. Miller and his colleagues (2009) examined why children’s SES is associated with health (such as indicators of cardiovascular disease) decades later, in adulthood. Once again, and consistent with the two-stage historiography, this study began with epidemiological observations suggesting that low SES in childhood predicts coronary heart disease throughout adulthood even among people who attain high levels of SES later in life (Kittelson et al., 2006). They suggest that early adversity increases the likelihood of a “defensive phenotype,” the hallmark of which is exaggerated biological response (e.g., inflammation) to stressors. As stressors accumulate in the life course, people with this defensive phenotype will be increasingly prone to inflammatory diseases, including, for example, some types of cardiovascular and respiratory.
Consistent with this model, the authors observed that, controlling present SES, low SES during childhood (as indicated by parents’ occupations during the child’s first five years of life) was associated with several transcriptional patterns that imply a defensive phenotype at about age 34. These findings suggest that early socioeconomic experiences result in durable programming of the stress response system. As Shanahan and Hofer (2011, p. 143) observe, “although the use of occupational prestige to indicate ‘adversity’ deserves further consideration, the empirical findings raise the possibility that early experiences are capable of enduring biological programming (by way of transcription) in response to social circumstances.” The potential importance of early SES was likewise suggested by research showing that SES at age two was associated with the expression of GR mRNA during adolescence, a relationship that was not moderated by current SES (Miller and Chen, 2007). The authors conclude that “low SES operates most potently during this period [of early childhood] of immune-system priming… in a way that favors the emergence of a proinflammatory phenotype” (p. 408).
Two additional studies support the hypothesis that low SES early in life is associated with genetic transcription patterns consistent with G insensitivity and add nuance to this basic pattern. Chen et al. (2011) examined gene expression profiles among adults who grew up in low-SES households but experienced either high or low maternal warmth. The results suggested that low-SES children with warm mothers showed reduced bioinformatic indications of pro-inflammatory transcription factor activity and immune activating transcription factor activity compared to those who were low in SES early in life but experienced low maternal warmth. In other words, high maternal warmth served as a protective factor for children from low SES households with respect to transcription patterns related to immune response and inflammation. Consistent with Miller and Chen (2007), they also reported that current SES did not alter this pattern of findings.
Second, Chen et al. (2011) examined genome-wide transcription profiles for T lymphocytes of asthmatic children from low- or high-SES households. Low-SES children showed overexpression of genes regulating inflammatory and catecholamine processes, including some involved in chemokine activity, stress responses, and wound responses. Some of the observed differences, however, appeared to be mediated by perceived threat. That is, children from low-SES households were more likely to perceive threat in ambiguous situations, and this tendency appeared to activate neuroendocrine processes that eventuated in changes in inflammatory signaling pathways.
Social genomic studies of transcription are relatively new but have nevertheless led to a cohesive body of research that point to biological mechanisms by which social experiences can affect health. To date, evidence suggests that low SES in childhood and diverse contemporaneous social stressors alter genetic expression by adolescence and also in mid-adulthood. These alterations are consistent with a G resistance model such that prolonged exposure to stressors is thought to lead to insensitivity to Gs, which would otherwise up-regulate anti-inflammatory and down-regulate proinflammatory proteins. This resistance could, in turn, be associated with diverse inflammatory illnesses, including asthma, cardiovascular disease, and depression. Some evidence suggests that a sense of threat or heightened vigilance may serve as a critical social psychological mediator that links social stressors with transcription profiles.
At the same time, many of the discussed studies rely on small, possibly unrepresentative samples, and nonexperimental designs. Although efforts are made to statistically control alternative explanations (e.g., by accounting for correlations among social stressors), such studies likely can address these threats to causal inference in a limited fashion. However, experimental designs that examine these links constitute Stage Two research (e.g., Cole, Mendoza, and Capitanio, 2009) and, as these designs are increasingly used in conjunction with nonexperimental designs, causal inferences will be strengthened. Thus, Stage Three researchers should pay close attention to the extent to which nonexperimental results have replicated in experimental settings.
STAGE THREE: POPULATION-BASED MULTILEVEL STUDIES
This chapter has proposed a two-stage historiography according to which population-based research identified putative social risk factors (Stage One) and, drawing on this research, social genomic studies identified possible biological mechanisms by which these risk factors could eventuate in diminished health (Stage Two). Yet what are the implications of these studies for population research (Stage Three)? The Stage Two studies that have been reviewed raise a number of distinct opportunities and challenges, including (1) the collection of multilevel data, (2) the examination of diverse life course models, which require extensive longitudinal data, (3) the refinements of measures based on diverse measurement strategies (including observation and tasks, for example), (4) the specification of cause and effect, and (5) the use of comparative designs.
Collection of Multilevel Data
Mechanistic models of health and aging call for extensive information about the context, people’s psychological states and behaviors, and biological processes. In the context of the transcription studies discussed, for example, population-based studies should begin assessing gene expression profiles for cells associated with immune processes in peripheral blood. The initial body of research was based, quite understandably, on relatively small samples drawn nonprobabilistically from communities, which renders inference and the study of diversity difficult. However, gene expression patterns may provide critical evidence of linking mechanisms that connect social experiences with health. The collection of gene expression data in the context of demographic and epidemiological research may be possible in the near future, although there are presently practical barriers (e.g., peripheral blood draws require nontrivial processing in a timely manner) that make large-scale collection from a geographically dispersed population challenging and expensive. Until those barriers are addressed, however, several collection strategies may be strategic.
First, the collection and processing of peripheral blood from subjects participating in well-defined and characterized community samples is logistically plausible. Many community-based samples have been studied for decades, resulting in rich descriptions of the participants’ social experiences. Particularly when such studies begin before or shortly after the birth of the subjects and include valid, reliable measures of social, psychological, and medical assessments, the collection of expression data could be highly informative. Second, smaller purposive samples are also logistically feasible, which allow for the study of strategically defined groups. For example, as noted, Cole and his colleagues (2008a) studied expression profiles among adults who were pre-screened for social isolation, resulting in matched groups differing, apparently, only on this characteristic. These strategies could involve embedding, the selection of a subset of respondents from a larger, ongoing study, ideally with thorough matching on possible confounds. Researchers interested in health disparities might examine, for example, groups that are apparently resilient despite exposures to a risk factor (poverty, discrimination, etc.).
Whatever the research design and sampling, the resulting data will ideally include multiple assessments to address a series of life course problems. Such a proposal is not new in itself, with many studies attempting to collect multilevel data (e.g., famously, the National Health and Nutrition Examination Surveys, or NHANES).
Life Course Models
That health and well-being reflect life course processes has long been appreciated, although research is now accumulating that suggests the relevance of prenatal (and intergenerational) experiences to health throughout adulthood, realities that call for data covering at least the entirety of people’s lives, from conception to death.
At first glance, the results of gene transcription studies of SES appear largely consistent with a sensitive or perhaps critical period model. However, extant evidence is not decisive and indeed conceptual considerations suggest a hybrid model involving a sensitive or critical period followed by a “chain of insults,” with perhaps accumulating disadvantage. Extensively longitudinal data, ideally extending across generations, will be needed to resolve these issues.
A sensitive period model posits that a specific biological system is highly plastic (i.e., subject to change, also referred to as programming) at specific points in development; that the resulting biological change takes place in response to the environment; and that the biological change is durable, potentially creating stable biological and/or behavioral tendencies. The critical period model differs in that the period of plasticity is the only time during development in which the biological system is open to change. That is, in contrast, the sensitive period model suggests a time span during which the system has heightened sensitivity to programming, but it may change during other periods as well.
Adjudicating between these two models requires extensive longitudinal data that describe the biological system and the social factors that are thought capable of changing it. Such data would allow for the study of the purported sensitive or critical period but also “before” and “after” periods. Indeed, only data collection spanning “before-during-after” could inform whether a period is sensitive or critical; whether observed changes endure and, if so, for how long; and whether any enduring changes are in fact associated with biological and behavioral tendencies in later life. Presently, research suggests that socioeconomic experiences before age 5 are associated with gene transcription patterns perhaps as early as age 9 (Chen, Martin, and Mathews, 2006; Miller and Chen, 2007) and as late as age 40 (Miller et al., 2009). Indeed, one study raises the intriguing possibility that the one-to-two-year-old span is a sensitive period for GR, and the two-to-three-year-old span is a sensitive period for toll-like receptor 4, which are proteins of the innate immune system that recognize conserved features of potentially invasive microbes (Miller and Chen, 2007). That is, different age periods may be sensitive with respect to different aspects of the immune system.
At the same time, several opportunities are suggested by the complexities of the sensitive period model and the extant data. First, although socioeconomic status before the age of five is thought to be decisive, less is known about the “before” and “after” periods. The reviewed studies rely largely on retrospective measures of SES in childhood and control present SES. However, the role of SES patterns before birth and after age 5 have not been studied prospectively. With respect to the “before” period, a large body of evidence points to the possibly powerful roles of maternal experiences on fetal development (for an overview, see Godfrey, Gluckman, and Hanson, 2010) and of intergenerational transmission of gene expression patterns in response to social experiences of grandparents (for a review, see Morgan and Whitelaw, 2008). To the extent that these prenatal experiences are correlated with SES during infancy and toddlerhood, it is conceivable that they could play causal roles in shaping transcription patterns.
With respect to trajectories of SES after age 5, there are likely a limited number of life course trajectories of SES (Hallqvist et al., 2004; Rosvall et al., 2006), raising the possibility that, for example, few people with chronically low SES before age 5 experience high SES over the next five years. Thus, statistical control of present SES may not be entirely effective in simulating group comparisons between high and low SES toddlers, controlling subsequent SES trajectories. In any event, it may take very large samples to adequately study people with diverse longitudinal patterns of SES.
Second, although socioeconomic status in early childhood is hypothesized to be the causal contextual agent, it is unclear when altered transcription patterns emerge. The lag between the environmental exposure and these changes, or the induction period, is unknown. One possibility is that transcription patterns change very soon after, or during, the sensitive period, a possibility for which there is presently no evidence. An additional possibility is that the sensitive period model is characterized by a longer induction period, meaning that there is an appreciable interval of time between the environmental exposure and altered transcription. Extant data are presently consistent with an induction period extending somewhere between exposure at ages 1 to 3 and altered transcription perhaps as early as age 9. However, more data are needed.
Third, the “chains of risk model” posits that risks (such as low SES) increase the likelihood of subsequent disadvantages, creating a chain reaction of challenges, but very little extant data shed light on this possibility. Miller and his colleagues (2011) propose such a model, the “defensive phenotype model,” arguing that early chronic stressors such as low SES are associated with pro-inflammatory tendencies (as discussed above), but also vigilance and mistrust of others, diminished self-regulation, and a proclivity for risky behaviors. That is, GR insensitivity is integral to the defensive phenotype, but the latter is broader and includes psychosocial processes.
According to this perspective, early chronic stressors are also associated with heightened biological responses to other stressors, which accentuates the pro-inflammatory tendencies.
Thus, children growing up in low-SES households (i.e., subjected to chronic stressors) are characterized by a constellation of biological, psychological, and social challenges that, in turn, create yet more stressors, diminish their capacity to cope with stressors, and make them more responsive to the negative effects of stressors. The resulting chronic inflammation is then thought to lead, over many years, to inflammatory disease states, although pre-disease indications may be observable by late childhood (Koenig et al., 2011; Lupien et al., 2009). However, the types of social, psychological, and biological experiences that would connect early SES and later inflammatory gene expression patterns are not well studied. Indeed, very little is known about how children in low-SES settings may, through their behaviors, create stressors and impede effective coping and social supports. These considerations suggest a critical or sensitive period that, in turn, is accompanied by a chain of social, psychological, and biological risks with considerable positive feedback among the types of risk and over time.
Fourth, while life course epidemiology and demography recognize the chain of risk model (e.g., Hayward and Gorman, 2004; Kuh et al., 2003), life course sociology has proposed an additional, complementary form of risk accumulation. O’Rand (2006) proposes a cumulative disadvantage model, according to which early disadvantages (like those associated with low SES) initiate strongly path-dependent exposure to risks, a “chain of insults” that extends across the phases of life. In contrast, people with advantageous early circumstances encounter a strongly path-dependent sequence of enriched environments marked by high levels of social capital, interpersonal relationships that facilitate the attainment of goals and positive development. In keeping with a large empirical literature, O’Rand emphasizes the importance of the SES of the family-of-origin, which is highly influential with respect to lifelong patterns of social capital and social risks. The distinguishing feature of this model, however, is that differences attributable to initial disadvantage are magnified over time (analogous to compound interest) (DiPrete and Eirich, 2006). Some evidence suggests that this insight may be important in understanding health disparities in late adulthood (e.g., Dupre, 2007; Willson, Shuey, and Elder, 2007). That is, the “chains of risk” model refers to the accumulation of a risk factor or factors, and O’Rand’s model refers to how the effect of early risk is accentuated over time.
All of these considerations suggest a highly nuanced life course model: a sensitive or critical period, with a possibly short induction period followed by chains of social, psychological, and biological risks with extensive positive feedback among them; the child’s behaviors reflecting social
disadvantages but also creating stressors and undermining coping and social supports; predisease symptoms observable by late childhood; and disease states emerging in adulthood, perhaps according to a power function. Clearly, such possibilities call for multilevel data, extensively longitudinal data.
Refining Measures of Social Risk Factors
As noted, Stage Two can inform Stage Three by suggesting refinements in measures. There is impressive evidence linking SES with “flexible resources” (e.g., knowledge about health) by which people avoid risky behaviors and other threats to health, engage in health-promoting behaviors, and attempt to address symptoms and disease states (e.g., Phelan, Link, and Tehranifar, 2010). The multilevel research on gene expression, however, suggests two additional mechanisms by which SES could influence health, and these mechanisms suggest new avenues for the measurement of social context.
First, as noted, early pronounced, chronic stressors may lead people to view ambiguous situations as threatening, which in turn activates neuroendocrine processes that eventuate in changes in inflammatory signaling pathways (Irwin and Cole, 2011). The evidence for this link, between stressors and sense of threat and vigilance, is complex but hinges on changes in the corticolimbic circuitry, which is associated with memory and emotion (Miller, Chen, and Parker, 2011). In any event, a possible link between SES and the activation of the corticolimbic circuitry suggests the refinement of measures of the social environment to more directly assess contextual features that would foster a sense of threat, vigilance, and mistrust. Irwin and Cole’s (2011) review of connections between the SNS and threat suggests the importance of violence, hostility, aggression, interpersonal loss, trauma, and physical exhaustion. Thus, research could directly assess how specific aspects of SES and features that are strongly associated with SES induce threat, vigilance, and mistrust; SNS mechanisms; and gene expression.
Presently, some evidence supports this focus. Harsh, insensitive, and cold parenting likely fosters such reactions in children and indeed mediates links between SES and, for example, internalizing and externalizing symptoms (e.g., Conger and Donnellan, 2007). Many indicators of neighborhood disorganization and the built environment—crime, violence, safety, racism, sense of community, abandoned buildings, and dilapidated and disrepaired structures—likely breed vigilance and mistrust (Sampson, Morenoff, and Gannon-Rowley, 2002). However, extant studies apparently do not assess sense of threat, vigilance, and mistrust. Indeed, the assessment of threat in population-based studies may be difficult. One extant measure, CAUSE, developed and used by Chen and her colleagues, uses videos of
ambiguous situations (e.g., a clerk watching a customer in a store from a distance) to elicit interpretative remarks from the subject (e.g., “the clerk thinks the customer is going to steal something” or “the clerk wonders if the customer needs help”). Such a measure might be administered to large groups of people with the use of computers or personal digital assistants. In any event, it is unclear whether sense of threat or vigilance could be assessed by survey instruments, suggesting the need for behaviorally based assessments.
Less well-studied are aspects of daycares, preschools, schools, and racial discrimination that might aggravate these feelings in children. Also, social capital typically refers to ties between people that are characterized by trust and reciprocity. However, associations between networks (structural features and dynamics) and mistrust and vigilance have not been studied. Thus, a key unresolved issue is how SES is associated with specific features of families, neighborhoods, schools, and social networks that breed a sense of threat, mistrust, and vigilance. And, in turn, little is known about how sense of threat then creates more stressors for the person, contributing to the chain of risk model.
Second, early chronic stressors are thought to influence the cortiostriatal circuitry, which is central to the processing of reward-related information and self-regulation (including, for example, impulse control and goal-directed behaviors) (e.g., Miller et al., 2011). Very little is known about the specific features of social context that can account for this association, and whether such features reflect aspects of SES. Gianaros and his colleagues (2011) suggest that some of this relationship, once again, reflects turbulent family relationships that are traceable to socioeconomic need. However, they also propose and report evidence consistent with a cultural argument, according to which SES is associated with a “cultural-intellectual orientation” that stimulates social and intellectual skills. According to this perspective, such family-based activities as engaging in intellectual discussions and attending cultural events positively influences brain development, which in turn facilitates cognitive abilities and capacities for self-regulation. There are presently no standard measures of the features of communities, neighborhoods, families, schools, and social networks that would provide these experiences, however. And once again, it will be difficult to link such measures to behaviors associated with the corticostriatal activity (prime examples being impulsive behavior or discounting of future rewards), which typically are based on behavioral tasks (e.g., Eisenberg et al., 2007). Thus, a major pathway by which SES may affect inflammatory processes involves diminished reward-related information processing and self-regulation, but the specific features of social context that would link SES with these behaviors are not known.
Both of the discussed mechanisms suggest the development of new measures that focus on features of social context that are graded by SES and that heighten a sense of vigilance, mistrust, and threat and that diminish self-regulatory capacities. Further, it may be that behavioral measures will be needed to assess these reactions to settings. In addition to measures of context and these behaviors, research could also incorporate imaging technology to directly assess corticolimbic and striatal activity (e.g., Gianaros and Manucj, 2010). Given logistical considerations, such an effort would likely be embedded in a larger study, but would provide crucial evidence bearing on whether connections between social context and, for example, sense of threat did indeed reflect differences in the corticolimbic system.
Specificity of Causes and Effects
As more detailed data from different levels of analyses are collected, issues of causal specificity will be more readily addressable. The extant evidence presently suggests that chronic, pronounced stressors in early childhood are associated with GR resistance and other pro-inflammatory mechanisms. Indeed, the replication of the GR resistance model across multiple studies that examine different types of stressors is impressive and noteworthy in genetic research as an apparently robust pattern of replication. At the same time, these findings raise the issue of specificity in two respects. First, does a given inflammatory condition reflect one specific social risk factor, or even one specific “signature set” of risk factors (i.e., specificity of contextual cause; Shanahan et al., 2008)? It may be that a wide range of early, chronic stressors—stressors associated with SES—are functionally equivalent, meaning that they are substitutable contextual experiences that trigger the same biological pathways. Presently, the mediating role of the corticostriatal and limbic systems appear to be crucial, suggesting that any social experiences that could affect these systems could initiate causal chains that lead to pro-inflammatory patterns.
Viewed from the perspective of necessity and sufficiency, several possibilities cannot be ruled out. Because several different stressors appear to trigger GR resistance, it is unlikely that any one stressor is necessary and sufficient. Social isolation, parental stress due to a child’s severe illness, low SES, and child maltreatment have all been associated with the upregulation of pro-inflammatory and down-regulation of anti-inflammatory transcriptional pathways. Miller and his colleagues (2011) note that both low SES and child maltreatment share many common social, psychological, and biological consequences, perhaps because both are associated with the corticolimbic and corticostriatal processes discussed above.
At the same time, SES is not a stressor but rather many stressors are SES-graded. Indeed, childhood maltreatment is substantially correlated with
a wide range of SES-graded stressors, including poverty, family conflict, neighborhood disorganization, parental substance abuse, sibling hostility, geographic mobility, income instability, and parental psychopathology. This network of correlations among stressors is of particular concern, raising the issue of whether any one stressor is necessary and sufficient, unnecessary but sufficient, necessary but insufficient, or unnecessary and insufficient. The issue can only be resolved with large, diverse samples that assess a wide range of stressors. Moreover, the study of these possibilities may be facilitated by diverse statistical models, including methods well suited to the study of conjunctive and disjunctive patterns among possible predictors (e.g., Eliason and Stryker, 2009; Hastie, 2009).
Once environmental specificity with respect to transcriptional patterns associated with inflammatory pathways is addressed, a second type of specificity remains to be considered: whether a social risk factor or signature set of such factors predict only one or multiple outcomes (i.e., specificity of outcome). Research suggests that GR resistance and other pro-inflammatory pathways are associated with a range of diseases, possibly including cardiovascular disease, depression, asthma symptoms, and, in principle, other conditions (e.g., arthritis, allergies, and several cancers). Do risk factors that trigger GR resistance explain all of these disease states, or are there specific patterns of risk factors associated with specific inflammatory diseases? Answering this question will depend on data collection efforts that include a wide range of inflammatory symptoms and disease states. It may be that stressor-inflammatory symptom associations are characterized by multifinality (the same causal agents leading to different outcomes), equifinality (diverse causal agents leading to the same outcome), or both—types of complexity that are often not considered in empirical research.
Diverse, Mutually Informing Research Designs
The use of diverse, mutually informing research designs has been discussed as a way to study the full complexity of phenomena while allowing for strong causal inference whenever possible. From a demographic perspective, comparative designs also broaden the scope of inquiry by focusing attention to distinct social settings. In the case of social genomics, for example, how do distinct macro-social contexts trigger pro-inflammatory transcription patterns? A central issue is the extent to which low-SES children are exposed to settings that heighten a sense of threat and vigilance, and these comparative strategies may shed light on this problem. First, what policies protect low-SES children from these experiences and thus prevent or retard the emergence of vigilance, difference in gene expression, and inflammatory conditions? Comparisons of low-SES households in different regimes of transfer payments, political economies, and societies could
address this question. With respect to support for families and children, salient policy differences might include transfer payments to low-SES households, the provision of daycare and adequate health insurance, investments in schools and training opportunities, and the provision of safe, affordable housing. Indeed, gene expression profiles and related biological substrates (particularly biomarkers of the immune system) could be assessed in evaluation studies of specific policies, which could be especially informative when the policies have been applied to randomized groups.
With respect to political economies, for example, Esping-Andersen (1990) influentially distinguished among liberal (e.g., the United States), corporatist-statist (e.g., Germany), and social democratic (e.g., Sweden) regimes. Such distinctions produce variation in life course patterns of school, work, and family, and perhaps they also bear on the stress load created by low SES. The corporatist and social democratic regimes provide substantially more support to low-income households when compared to liberal regimes, but the social democratic society socializes costs associated with family life, and includes thorough welfare provisions for workers and the unemployed. With respect to other societal differences, not necessarily reflecting political economic policies, it may be that the distribution of SES conditions its effects on families. A very large body of research suggests that health is less favorable in societies where income differences are greater (Wilkinson and Pickett, 2006). To the extent that minority status is associated with discrimination (a potentially potent chronic stressor), societal distributions of race and ethnic groups may also bear on how much stress is suffered by low-SES children and their families.
Differences in policy settings, political economics, and demographic features of societies can also be studied by comparing migrants to a new setting and people who remained at the origin. Such a strategy has been used to study dietary changes and health, for example, and has the advantage of controlling, in the aggregate, for genetic factors that might otherwise distinguish, for example, low-SES children from two different countries. This strategy could be used to compare and contrast low-SES origin and destination groups and how their social location is associated with threat and self-regulation, and transcription patterns. Ideally, such a design would involve nonvoluntary migrants to control for selection to migration.
In any event, these comparative strategies could be used to study how the stress load created by low SES differs by social systems.
A large and complex body of research suggests that social experiences of early childhood may have lifelong implications for the immune system and the emergence of inflammatory diseases. This body of research began
with studies of social risk factors and health (Stage One) and then progressed to the study of how such risk factors could conceivably “get under the skin” (Stage Two). Broadly, early chronic stressors are associated with pro-inflammatory tendencies (e.g., due to GR resistance), but also vigilance and mistrust of others, diminished self-regulation, and a proclivity for risky behaviors. Children growing up in low-SES households (i.e., subjected to chronic stressors) are thus characterized by a constellation of biological, psychological, and social challenges that, in turn, create yet more stressors, diminish their capacity to cope with stressors, and make them more responsive to the negative effects of stressors. In this way, society may “get under the skin,” but behavior then creates a feedback to one’s social circumstances, creating yet more stressors (i.e., bidirectionality). The resulting chronic inflammation is then thought to lead, over many years, to inflammatory disease states, although pre-disease indications may be observable by late childhood.
This emerging model, in turn, suggests a symbiotic relationship between Stage Two and population-based studies of aging and health (Stage Three). On the one hand, Stage Two research provides evidence of linking mechanisms that may connect social experiences and health, mechanisms that are necessary for any convincing causal account of social risk factors and health. On the other hand, Stage Three studies can validate and extend Stage Two research. To date, logistic considerations have prohibited the application of genome-wide expression studies to large, representative samples. However, just such samples are needed to validate Stage Two studies and to study diverse patterns of social experiences and trajectories of symptoms and disease states.
Moreover, the results from Stage Two studies suggest unique challenges and opportunities for Stage Three research. Given that social experiences may be biologically embedded before age five, given the central role that chains of risk and processes of accumulation play in creating stress, and given the emergence of diseases over many decades, extensively longitudinal data are imperative. And, further, given that none of these relationships is determinative, extensively longitudinal research is also needed to study life course patterns that are associated with resilience and varying patterns of vulnerability. As noted, the basic model emphasizes the importance of very early experiences. At the same time, several studies reported GR resistance profiles among adults (e.g., Cole, 2008a; Miller et al., 2008), based on their contemporaneous experiences. The connections between these two sets of findings are unclear, but perhaps there are windows of vulnerability throughout life.
The Stage Two research also suggests new themes with respect to measurement and modeling. Given mechanisms suggested by Stage Two research, future studies could profitably focus on specific aspects of the
social context that heighten a sense of threat, vigilance, and mistrust, and that undermine self-regulatory capacities. Such refinements may be challenging because these behaviors are likely best assessed with behavioral measures and, ideally, would be accompanied by neuro-imaging studies. The reviewed studies also suggest very high levels of contingency among social experiences and the psychological and biological cascades that they initiate. That is, it may be that many different stressors are essentially substitutable, equally capable of instilling threat and increasing the likelihood of pro-inflammatory transcription patterns (i.e., equifinality). This challenge of contingency is compounded by the possibility that the same social experiences could produce different inflammatory symptoms and disease states (i.e., multifinality). Thus, methods that are sensitive to high levels of contingency—e.g., fuzzy set analysis, machine-learning techniques—will be appropriate, although their application may depend on further methodological developments that strengthen their inferential basis.
Most directly appealing to the traditional domain of demography, gene expression profiles can be studied in diverse comparative frameworks to understand how much stress a given social system generates. Such comparisons could involve different policies as they bear on the lives of low-SES households and different political economies, comparisons that may be especially informative when involving origin and destination groups of nonvoluntary migration. Finally, considerable attention has been devoted to risk behaviors (smoking, drinking, inactivity, poor diet, etc.) as crucial explanations for socioeconomic differences in health. Stage Two research appears to complement this focus, suggesting that stressors associated with low SES led to diminished self-regulation, which in turn may well be associated with a wide range of risk behaviors.
Population-based studies of health have traditionally had an admirably interdisciplinary quality. As the models that describe connections between social, psychological, and biological levels of analysis become increasingly complex, greater attention should be paid to how such teams are organized and encouraged. The payoff for such efforts will be increasingly thorough explanations of SES gradients in health and thus the scientific basis for effective prevention and intervention.
Cacioppo, J.T., Hawkley, L.C., Norman, G.J., and Bernston, G.G. (2011). Social isolation. Annals of the New York Academy of Sciences, 1,231, 17-22.
Chen, E., Martin, A.D., and Matthews, K.A. (2006). Socioeconomic status and health: Do gradients differ within childhood and adolescence? Social Science and Medicine, 62, 2,161-2,170.
Chen, E., Miller, G.E., Kobor, M.S., and Cole, S.W. (2011). Maternal warmth buffers the effects of low early-life socioeconomic status on pro-inflammatory signaling in adulthood. Molecular Psychiatry, 16(7), 729-737.
Cole, S.W. (2008a). Social regulation of leukocyte homeostasis: The role of glucocorticoid sensitivity. Brain, Behavior, and Immunology, 22(7), 1,049-1,055.
Cole, S.W. (2008b). Psychosocial influences on HIV-1 disease progression: Neural, endocrine, and virologic mechanisms. Psychosomatic Medicine, 70(5), 562-568.
Cole, S.W. (2009). Social regulation of human gene expression. Current Directions in Psychological Science, 18(3), 132-137.
Cole, S.W., Kemeny, M.E., Fahey, J.L., Zack, J.A., and Naliboff, B.D. (2003). Psychological risk factors for HIV pathogenesis: Mediation by the autonomic nervous system. Biological Psychiatry, 54(12), 1,444-1,456.
Cole, S.W., Hawkley, L.C., Arevalo, J.M., Sung, C.Y., Rose, R.M., and Cacioppo, J.T. (2007). Social regulation of gene expression in human leukocytes. Genome Biology, 8(9), R189. Available: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2375027 [August 5, 2013].
Cole, S.W., Mendoza, S.P., and Capitanio, J.P. (2009). Social stress desensitizes lymphocytes to regulation by endogenous glucocorticoids: Insights from in vivo cell trafficking dynamics in rhesus macaques. Psychosomatic Medicine, 71(6), 591-597.
Cole, S.W., Hawkley, L.C., Arevalo, J.M.G., and Cacioppo, J.T. (2011). Transcript origin analysis identifies antigen-presenting cells as primary targets of socially regulated gene expression in leukocytes. Proceedings of the National Academy of Sciences of the United States of America, 108(7), 3,080-3,085.
Conger, R.D., and Donnellan, M.B. (2007). An interactionist perspective on the socioeconomic context of human development. Annual Review of Psychology, 58, 175-199.
DiPrete, T.A., and Eirich, G.M. (2006). Cumulative advantage as a mechanism for inequality: A review of theoretical and empirical developments. Annual Review of Sociology, 32(1), 271-297.
Dupre, M.E. (2007). Educational differences in age-related patterns of disease: Reconsidering the cumulative disadvantage and age-as-leveler hypotheses. Journal of Health and Social Behavior, 48(1), 1-15.
Eisenberg, D.T.A., Mackillop, J., Modi, M., Beauchemin, J., Dang, D., Lisman, S.A., Lum, J.K., et al. (2007). Examining impulsivity as an endophenotype using a behavioral approach: A DRD2 TaqI A and DRD4 48-bp VNTR association study. Behavioral and Brain Functions, 3, 2.
Eliason, S.R., and Stryker, R. (2009). Goodness-of-fit tests and descriptive measures in fuzzyset analysis. Sociological Methods and Research, 38(1), 102-146.
ENCODE Project Consortium. (2011). A user’s guide to the encyclopedia of DNA elements (ENCODE). PLoS Biology, 9(4). Available: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001046 [September 13, 2012].
Esping-Andersen, G. (1990). The Three Worlds of Welfare Capitalism. Princeton, NJ: Princeton University Press.
Gianaros, P.J., and Manuck, S.B. (2010). Neurobiological pathways linking socioeconomic position and health. Psychosomatic Medicine, 72(5), 450-461.
Gianaros, P.J., Manuck, S.B., Sheu, L.K., Kuan, D.C.H., Votruba-Drzal, E., Craig, A.E., and Hariri, A.R. (2011). Parental education predicts corticostriatal functionality in adulthood. Cerebral Cortex, 21(4), 896-910.
Godfrey, K.M., Gluckman, P.D., and Hanson, M.A. (2010). Developmental origins of metabolic disease: Life course and intergenerational perspectives. Trends in Endocrinology and Metabolism, 21(4), 199-205.
Hallqvist, J., Lynch, J., Bartley, M., Lang, T., and Blane, D. (2004). Can we disentangle life course processes of accumulation, critical period, and social mobility? An analysis of disadvantaged socio-economic positions and myocardial infarction in the Stockholm Heart Epidemiology Program. Social Science and Medicine, 58(8), 1,555-1,562.
Hastie, T. (2009). The Elements of Statistical Learning: Data Mining, Inference, and Prediction (2nd ed.). New York: Springer.
Hayward, M.D., and Gorman, B.K. (2004). The long arm of childhood: The influence of early-life social conditions on men’s mortality. Demography, 41(1), 87-107.
Irwin, M.R., and Cole, S. (2011). Reciprocal regulation of the neural and innate immune systems. Nature Reviews Immunology, 11, 625-632.
Kittleson, M.M., Meoni, L.A., Wang, N.-Y., Chu, A.Y., Ford, D.E., and Klag, M.J. (2006). Association of childhood socioeconomic status with subsequent coronary heart disease in physicians. Archives of Internal Medicine, 166(21), 2,356-2,361.
Koenig, J.I., Walker, C.-D., Romeo, R.D., and Lupien, S.J. (2011). Effects of stress across the lifespan. Stress, 14(5), 475-480.
Kuh, D., Ben-Shlomo, Y., Lynch, J., Hallqvist, J., and Power, C. (2003). Life course epidemiology. Journal of Epidemiology and Community Health, 57(10), 778-783.
Lupien, S.J., McEwen, B.S., Gunnar, M.R., and Heim, C. (2009). Effects of stress throughout the lifespan on the brain, behaviour, and cognition. Nature Reviews Neuroscience, 10(6), 434-445.
Miller, G., and Chen, E. (2007). Unfavorable socioeconomic conditions in early life presage expression of proinflammatory phenotype in adolescence. Psychosomatic Medicine, 69(5), 402-409.
Miller, G.E., Chen, E., and Cole, S.W. (2009). Health psychology: Developing biologically plausible models linking the social world and physical health. Annual Review of Psychology, 60, 501-524.
Miller, G.E., Chen, E., Sze, J., Marin, T., Arevalo, J.M.G., Doll, R., Ma, R., et al. (2008). A functional genomic fingerprint of chronic stress in humans: Blunted glucocorticoid and increased NF-kappaB signaling. Biological Psychiatry, 64(4), 266-272.
Miller, G.E., Chen, E., Fok, A.K., Walker, H., Lim, A., Nicholls, E.F., Cole, S., et al. (2009). Low early-life social class leaves a biological residue manifested by decreased glucocorticoid and increased pro-inflammatory signaling. Proceedings of the National Academy of Sciences of the United States of America, 106(34), 14,716-14,721.
Miller, G.E., Chen, E., and Parker, K.J. (2011). Psychological stress in childhood and susceptibility to the chronic diseases of aging: Moving toward a model of behavioral and biological mechanisms. Psychological Bulletin, 137(6), 959-997.
Morgan, D.K., and Whitelaw, E. (2008). The case for transgenerational epigenetic inheritance in humans. Mammalian Genome, 19(6), 394-397.
O’Rand, A.M. (2006). Handbook of Aging and the Social Sciences (6th ed., pp. 145-162). San Diego, CA: Academic Press.
Phelan, J.C., Link, B.G., and Tehranifar, P. (2010). Social conditions as fundamental causes of health inequalities. Journal of Health and Social Behavior, 51(1 Supplement), S28-S40.
Rosvall, M., Chaix, B., Lynch, J., Lindstrom, M., and Merlo, J. (2006). Similar support for three different life course socioeconomic models on predicting premature cardiovascular mortality and all-cause mortality. BMC Public Health, 6, 203.
Sampson, R.J., Morenoff, J.D., and Gannon-Rowley, T. (2002). Assessing “neighborhood effects”: Social processes and new directions in research. Annual Review of Sociology, 28, 443-478.
Seeman, T.E. (1996). Social ties and health: The benefits of social integration. Annals of Epidemiology, 16, 95-106.
Shanahan, L., Copeland, W., Costello, E.J., and Angold, A. (2008). Specificity of putative psychosocial risk factors for psychiatric disorders in children and adolescents. Journal of Child Psychology and Psychiatry, 49(1), 34-42.
Shanahan, M.J., and S.M. Hofer (2011). Molecular genetics, aging, and the life course: sensitive periods, accumulation, and pathways models. In R.H. Binstock and L. George (Eds.), Handbook of Aging and the Social Sciences. New York: Elsevier.
Weinstock, G.M. (2007). ENCODE: More genomic empowerment. Genome Research, 17(6), 667-668.
Wilkinson, R.G., and Pickett, K.E. (2006). Income inequality and population health: A review and explanation of the evidence. Social Science and Medicine, 62(7), 1,768-1,784.
Willson, A.E., Shuey, K.M., and Elder, G.H. (2007). Cumulative advantage processes as mechanisms of inequality in life course health. American Journal of Sociology, 112(6), 1,886-1,924.