Brief Report: Subjective Social Status and Inflammatory Gene Expression

Abstract

Objective:

There exists a well-established link between low perceived social status and poorer health outcomes. However, the molecular mechanisms associated with this link remain unclear. This study begins to fill this gap by investigating the effects of low perceived subjective social status on health-related gene expression.

Methods:

Participants were 47 healthy heterosexual women (mean age 20.5 years) from a large American university. Participants gave 10 mL of peripheral blood and completed questionnaires assessing Subjective Social Status (SSS), perceived childhood socioeconomic status (SES), health, and relevant demographics. Putatively associated genes were subject to TELiS promoter-based bioinformatic analysis to assess activity of pro-inflammatory, anti-inflammatory, and antiviral transcription factors.

Results:

In analyses controlling for perceived childhood SES and other covariates, 84 transcripts showed > 1.5-fold difference in average expression across the range of SSS. TELiS bioinformatics analyses implicated the pro-inflammatory transcription factors, NF-κB and AP-1, in driving expression of genes that were up-regulated in low-SSS individuals. Results also indicated increased activity of CREB family transcription factors but no differential activity of the anti-inflammatory glucocorticoid receptor (GR) of interferon response factors (IRF). Transcript origin analysis implicated monocytes and dendritic cells as cellular mediators.

Conclusions:

In this first study examining the molecular correlates of SSS, experiences of low social status are associated with transcriptional effects similar to those previously observed for objective adversity conditions such as low SES, social isolation, and chronic stress.

Low subjective social status (SSS)—the subjective assessment that one is inferior to others in one’s social groups—is associated with increased risk for poor health outcomes, above and beyond the effects of objective markers such as socioeconomic status (SES; for review see Cundiff & Matthews, 2017). However, while there is growing evidence that SSS has an independent relationship with physical health outcomes, the biological mechanisms involved remain unclear.

Increased activation of pro-inflammatory genes has been theorized to contribute to the relationship between objective social status and the chronic diseases that dominate contemporary epidemiology (e.g., Miller, Chen, & Cole, 2009). Whereas acute inflammation is generally adaptive, prolonged inflammation contributes to many poor health outcomes, including arthritis, cardiovascular disease, cancer, and diabetes (Finch, 2010). Lower social status predicts higher levels of chronic inflammation in both human and non-human animals (e.g., Chen et al., 2009; Heidt et al., 2014; Powell et al., 2013; Snyder-Mackler et al., 2016). Ancestrally, higher inflammation may have been adaptive in low-status individuals by conferring anti-bacterial defenses against acute injury (e.g., Chen et al., 2004). However, the complexity of modern social systems may skew pro-inflammatory transcriptional responses and promote inflammation-related diseases in contemporary societies (Cole, 2013).

Chronic activation of pro-inflammatory genes is one component of a broader conserved transcriptional response to adversity (CTRA), which can be activated by threatening or stressful appraisals of the environment (Cole, 2013, 2014). CTRA gene expression profiles in immune cells can be evoked by activation of fight-or-flight pathways in the sympathetic nervous system (SNS; Heidt et al., 2014; Powell et al., 2013). SNS signaling directly stimulates the CREB family of transcription factors via beta-adrenergic receptors, and can also indirectly stimulate the activity pro-inflammatory transcription factors (e.g., NF-κB, AP-1) and indirectly inhibit the activity of interferon response factors (IRF) involved in antiviral responses (Cole, 2013, 2014). Under normal physiological conditions, inflammation triggers cortisol production by the hypothalamus-pituitary-adrenal (HPA) axis, which provides feedback inhibition of inflammation by stimulating the glucocorticoid receptor (GR) to suppress transcription of pro-inflammatory genes (Irwin & Cole, 2011). However, in people experiencing chronic stress, as in individuals of low SSS (Adler et al., 2000), the GR may become desensitized and thus fail to down-regulate inflammation (see Miller et al., 2009). This dynamic has been observed in studies examining objective indicators of low social status (e.g., income or years of education; Powell et al., 2013; Chen et al., 2009), but its relationship to SSS is not known.

In the present study, we investigated whether low SSS, net of perceived childhood SES and other potential confounders, is associated with increased expression of pro-inflammatory genes. We also tested whether the transcriptional effects of low SSS might be attributable to either increased activity of the SNS (as indexed by CREB transcription factor activity) or decreased feedback inhibition from the HPA axis (as indexed by reduced GR activity).

Method

Participants.

Participants were 47 heterosexual women (mean age = 20.5 years) recruited as part of a longitudinal study investigating the effects of new social bonds on transcriptional profiles (Murray, Haselton, Fales, & Cole, under review; see SOM). The study was approved by the UCLA IRB (#13–001122) and all participants provided written informed consent.

Online Questionnaire.

Participants completed the MacArthur Scale of SSS (Adler et al., 2000), which asked participants to indicate their perceived social status relative to individuals at UCLA on a ten-rung “ladder” (1 = least respected; 10 = most respected; M = 6.89, SD = 1.68, range = 1–8). SSS was treated as a continuous variable in all analyses. Women also reported their perceived childhood SES (Options: Lower, Lower-middle, Middle, Upper-middle, Upper; this measure was only modestly correlated with SSS, r = .32), ethnicity, height, weight, alcohol consumption, and other demographic parameters (see SOM).

Gene expression profiling and analyses.

Within two weeks of enrollment in the study, a 10 mL blood sample was collected for assessment of gene expression, from which peripheral blood mononuclear cells (PBMC) were isolated and stored at −70°C for subsequent transcriptome profiling. Total RNA was extracted from PBMC samples using an automated nucleic acid processing system (QIAcube; Qiagen), following the manufacturers’ standard protocol.

Gene transcription analyses.

Gene expression was assayed using Illumina HT-12 bead arrays following the manufacturer’s standard protocol. Quantile normalized data were log₂-transformed and differentially expressed genes were identified as those showing ≥1.5-fold difference in mean expression levels (as in previous research, e.g., Fredrickson et al., 2013) over the range from −2 SD to +2 SD relative to mean level of SSS in standard linear model analyses that additionally controlled for perceived childhood SES, age, race/ethnicity, BMI, and alcohol consumption. All inferential statistics reported below refer to set-based statistics testing hypotheses at the level of full genomic sets (not at the level of individual genes).

To identify the upstream signal transduction pathways that underlie differential gene expression, we conducted TELiS promoter-based bioinformatics analyses (Cole, Yan, Galic, Arevalo, & Zack, 2005). TELiS analyzes differential gene expression data in terms of the prevalence of transcription factor-binding motifs (TFBMs) within the promoters of all differentially expressed genes. Putatively associated genes were subject to TELiS promoter-based bioinformatic analysis to assess activity of pro-inflammatory transcription factors, NF- κB (using the TRANSFAC V$NFKAPPAB_01 transcription factor-binding motif weight matrix) and AP-1 (V$AP1FJ_Q2); the CREB family of transcription factors mediating β-adrenergic signaling responses to SNS catecholamine neurotransmitters (V$CREB_Q2); the anti-inflammatory glucocorticoid receptor mediating responses to cortisol from the HPA axis (V$GR_Q6); and IRF family factors involved in Type I interferon antiviral responses (V$ISRE_01). As in previous analyses (Cole et al., 2005; Powell et al., 2013), activity was quantified as the ratio of TFBM prevalence in up- versus down-regulated genes. Statistical inference was based on standard errors for the log2-transformed TFBM ratios that were derived from 200 cycles of bootstrap resampling of linear model residual vectors (i.e., controlling for any statistical dependence among genes). The study was approved by the UCLA Biomedical IRB.

Results

After statistically controlling for age, ethnicity, BMI, alcohol consumption, and childhood SES, a total of 84 transcripts showed a 1.5-fold or larger difference in gene expression (listed in Supporting Information Table 1). Results of these promoter-based bioinformatics analyses of differential gene expression (shown in Figure 1a) revealed that higher SSS was associated with a relative down-regulation of genes bearing response elements for the pro-inflammatory transcription factors, AP-1 (0.50-fold difference, SE = 0.10, p < .001), and NF-κB (0.45-fold difference, SE = 0.22, p = .05). Results also indicated down-regulation of genes bearing response elements for the SNS-related CREB transcription factor family (0.27-fold difference, SE = 0.14, p = .002). Lower SSS was not associated with any differential expression of genes bearing response elements for HPA axis-related GR (0.91-fold difference, SE = 0.25, p = .70) or the IRF family of transcription factors (1.48-fold difference, SE = 0.42, p = .56).

Figure 1a. Transcriptional activity in PBMC from people with high vs. low SSS.

Data represent (log2-transformed) ratio of transcription factor binding motifs (TFBM) for pro-inflammatory (NF-κB and AP-1), CREB, glucocorticoid (GR), and antiviral (IRF) transcription factors in the promoters of all genes showing ≥1.5-fold greater magnitude of activation over the continuous range of individual differences in SSS (−2SD vs +2SD in activation relative to the mean). (Note: conventional Bonferroni- corrected p threshold = .01). Error bars represent +/− 1 SE.

To determine whether the transcriptional correlates of SSS might occur within the same leukocyte subpopulations previously identified to mediate CTRA transcriptional effects of other chronic stressors (i.e., monocytes, dendritic cells, and B lymphocytes; Cole et al., 2012; Cole, Hawkley, Arevalo, & Cacioppo, 2011; Knight et al., 2016; Miller et al., 2008), we conducted transcript origin analysis (TOA; Cole et al., 2011). Results found downregulated genes to derive primarily from monocytes and dendritic cells (see Figure 1b). No specific leukocyte subpopulation was implicated in mediating expression of SSS-upregulated genes.

Figure 1b. PBMC cell type of origin as indicated by TOA cell-type diagnosticity.

Mean bootstrap values presented. (Note: conventional Bonferroni- corrected p threshold = .0083) Error bars indicate bootstrap SE.

Discussion

These data from genome-wide transcriptional profiling of circulating leukocyte samples in healthy young women suggest that low SSS is associated with increased expression of pro-inflammatory genes, and they implicate increased activity of the pro-inflammatory transcription factors NF-κB and AP-1 in driving these effects at the molecular level. These relationships emerged above and beyond the effects of perceived childhood SES and may contribute to the previously observed health risks associated with low SSS (see Cundiff & Matthews, 2017). Consistent with previous data demonstrating a causal effect of SNS neurotransmitter signaling in driving pro-inflammatory gene expression (Heidt et al., 2014; Powell et al., 2013), the present results also linked low SSS to increased activity of the CREB family transcription factors involved in mediating fight-or-flight β-adrenergic signaling from the SNS. However, these analyses found no evidence that desensitization of the anti-inflammatory GR contributed to the increased inflammatory signaling associated with low SSS. Although this study should be considered preliminary, these results do suggest that perceived low social status has transcriptional correlates similar to those previously observed for other forms of chronic adversity including low objective SES, social isolation, and chronic stress (Chen et al., 2009 Cole et al., 2007; Miller et al., 2008, 2014).

The finding of increased pro-inflammatory gene expression would be expected if low SSS individuals perceive the social world as threatening. Under such conditions, evolutionarily-conferred defensive programs are activated that serve to anticipate injury, and involve an increase in pro-inflammatory gene expression. The results add to a growing literature identifying differences in NF-κB by both SES and chronic stress (Chen et al., 2009; Miller et al., 2008, 2009; Powell et al., 2013). However, the present analyses did not find relationships between low SSS and other components of the CTRA dynamic, as no difference in IRF-related gene expression emerged. This discontinuity has been observed in some previous studies of SES (Chen et al., 2009; Miller et al., 2009) and may involve differences in the specific cell types that are most sensitive to the biological effects of social status. In the present analyses, low SSS is associated with increased expression of genes specifically expressed in monocytes, which play a predominant role in inflammation (e.g., Heidt et al., 2014; Swirski et al., 2009). By contrast, dendritic cells play a dominant role in producing the Type I interferon responses mediated by IRF family factors (e.g., Ferbas, Toso, Logar, Navratil, & Rinaldo, 1994) and the present data suggest only a modest contribution from this cell type (and one that acts in the opposite direction of the IRF down-regulation typically involved in the CTRA) (Cole et al., 2011). This study also found no relationship between the GR desensitization previously observed in some studies of chronic stress and low SES (Miller et al., 2009; Chen et al., 2009; Powell et al., 2013). It is possible that GR sensitivity is most directly affected by external socio-environmental conditions and is not substantially related to subjective experiences such as SSS. However, it is also possible that GR desensitization requires many years to develop and may thus be difficult to detect in this sample of healthy young women.

Although this study provides the first indication that pro-inflammatory molecular processes may link low SSS to increased health risk, these results should be interpreted with caution. Under the most strict family-wise error control methods, the transcriptional differences in NF-κB would only be marginally significant. However, the tests reported here were specified a priori and represent distinct substantive hypotheses; as such, they would not normally require correction for multiple testing (see Cao & Zhang, 2014). The present study is limited by its correlational design, a limited sample size, a restricted demographic variation, and absence of any health outcome data. Given this young female sample, the relevance of these results to older individuals and other populations remains to be determined in future research. However, the current results may offer a glimpse into the initial stages of a process whereby the health effects of chronic stressors accumulate over time. Despite these limitations, this study finds that SSS, net of the effects of perceived childhood SES, is associated with differences in inflammatory gene expression that may contribute to previously observed social gradients in health. These results provide an initial molecular framework for future research to more fully map the psychobiological pathways involved in the association between SSS and health.