|Home | About | Journals | Submit | Contact Us | Français|
Chronic interpersonal difficulties have a detrimental influence on mental and physical health, but little is known about the mechanisms underlying this phenomenon.
103 healthy young women (mean age = 17) were administered a structured interview to assess the degree of chronic interpersonal stress in their lives. At the same time blood was drawn to measure systemic inflammation, the expression of signaling molecules that regulate immune activation, and leukocyte production of the cytokine interleukin-6 following ex vivo stimulation with lipopolysaccharide. All of the immunologic assessments were repeated six months later.
To the extent subjects were high in chronic interpersonal stress at baseline, their leukocytes displayed greater increases in mRNA for the pro-inflammatory transcription factor nuclear factor-kappa B (NF-κB) over the next six months. They also showed larger increases in mRNA for inhibitor of kappaB, a molecule that sequesters NF-κB in the cytoplasm and minimizes its pro-inflammatory activities. Chronic interpersonal stress at baseline was unrelated to changes in biomarkers of systemic inflammation, but was associated with increasingly pronounced interleukin-6 responses to lipopolysaccharide. These associations were independent of demographics, lifestyle variables, and depressive symptoms.
These findings suggest that chronic interpersonal difficulties accentuate expression of pro- and anti-inflammatory signaling molecules. While this process does not result in systemic inflammation under quiescent conditions, it does accentuate leukocytes’ inflammatory response to microbial challenge. These dynamics may underlie the excess morbidity associated with social stress, particularly in inflammation-sensitive diseases like depression and atherosclerosis.
Considerable research indicates that socially integrated persons enjoy better mental and physical health than their more isolated peers 1,2. However, in a series of recent studies it has become evident that social ties can also have detrimental influences on health, especially when they are marked by conflict, mistrust, and instability 3–5. Among patients recovering from a mood disorder, for example, family tensions double the odds of a relapse occurring, and similar patterns are seen in patients with eating disorders and schizophrenia 6. Interpersonal difficulties are also associated with heightened susceptibility to respiratory infections, delayed healing of wounds, accelerated emergence of the metabolic syndrome, and increased morbidity and mortality from cardiovascular disease 7–11. These effects have been well-substantiated in a series of tightly controlled studies with rodents and non-human primates 12–14.
Despite this robust pattern of findings, little is known about the responsible underlying mechanism(s). One candidate hypothesis is that chronically abrasive relationships foster low-grade systemic inflammation, which then contributes to the evolution and/or expression of psychiatric, infectious, metabolic, and coronary diseases 15–17. This view received initial support in a study of married couples who had conflictual interactions in the laboratory; those who expressed greater hostility showed higher levels of the inflammatory cytokines interleukin-6 and tumor necrosis factor-α the next morning 8.
In this article we examine a broader spectrum of interpersonal relationships, including romantic relationships, friendships, and familial relationships, and consider whether their quality relates to two major biomarkers of systemic inflammation, C-reactive protein (CRP) and interleukin-6 (IL-6). To identify the molecular signaling pathways involved in these dynamics, we also assess the expression of mRNA for the chief pro-inflammatory transcription factor, nuclear factor-kappa B (NF-κB), and for the α and β isoforms of the glucorticoid receptor (GR). When it is ligated by cortisol, GR-α has potent anti-inflammatory properties, mediated by its inhibition of NF-κB signaling 18. This inhibition occurs through direct protein-protein interactions, as well as GR-mediated induction of inhibitor of kappa-b (IκB), a molecule that sequesters NF-κB in the cytosol and thereby prevents it from switching on pro-inflammatory genes. The functions of GR-β are not fully defined, but it is thought to inhibit the activity of GR-α and may, thus, facilitate pro-inflammatory signaling 19. Finally, to model the dynamics of these signaling pathways under conditions of immune challenge, we quantified expression of IL-6 by leukocytes that had been stimulated with bacterial product ex vivo.
These analyses were carried out within the context of a short-term prospective study, in which interpersonal difficulties were assessed at baseline, and inflammatory processes were measured at that time and again six months later. This design has a number of strengths compared to the cross-sectional analyses that are more common in psychoneuroimmunology. First, it allows changes over time in the outcomes of interest to be modeled, which is presumably more relevant to disease pathogenesis than one-time assessments. Second, it minimizes the likelihood that erroneous conclusions about directionality will be made, which is an important consideration here because inflammatory cytokines can have profound influences on social behavior 20. Finally, a design like this is able to capture stress-related changes that take time to evolve, such as those that result from gradual “wear and tear” on bodily systems. Based upon these considerations we hypothesized a prospective association between interpersonal difficulties and inflammatory processes. Specifically, to the extent that subjects were high in chronic interpersonal stress at study entry, we expected them to display greater activation of pro-inflammatory signaling pathways and higher levels of inflammatory biomarkers six months later.
These data were collected between October 2004 and December 2007 as part of a larger project on depression and atherosclerosis among young women at high-risk for mood disorders. Subjects were recruited from the Vancouver, British Columbia community through advertisements in local media. Eligibility criteria included being (1) female and 15–19 years old, (2) fluent in the English language, (3) free of acute and chronic medical conditions and standing medication regimens other than oral contraceptives, (4) without a lifetime history of psychiatric disorders, and (5) at high risk for developing an initial episode of depression. Subjects’ medical histories were ascertained through detailed interviews and laboratory testing. Psychiatric backgrounds were evaluated with the Structured Clinical Interview for DSM-IV Non-Patient Edition. High-risk was defined as having a first-degree relative with a history of affective disorder, and/or scoring in the top quartile of the population distribution on one of two indices of cognitive vulnerability to depression, the Dysfunctional Attitudes Scale or the Adolescent Cognitive Style Questionnaire 21.
This article focuses on 103 subjects who had been assessed at study entry and 6 months later. The sample had a mean age of 17.19 (SD=1.37) at study entry. Forty-five percent of subjects self-identified as East Asian, 45% as Caucasian, and the other 10% as East Indian, Aboriginal, or other. Participants came from homes in which parents averaged 14.98 years of education (SD = 3.63). The larger project was reviewed and approved by the University of British Columbia’s Research Ethics Board. Written consent was obtained from all subjects. For those who were younger than 18, a parent or guardian also provided consent.
To assess the extent of chronic interpersonal stress in subjects’ lives, we administered the UCLA Life Stress Interview - Adolescent Version 22 at study entry. This semi-structured instrument probes stressors in multiple domains of life. In each domain the interviewer asks a series of open-ended questions, and uses the data to rate the degree of chronic stress over the last six months. For this paper, we averaged ratings across the interview’s four social domains: romantic partner, closest friendship, other friendships, and family relationships. Ratings in these domains were modestly inter-related with correlations ranging from r = .10, p = .33 (romantic partner and closest friendship) to r = .56, p = .001 (closest friendship and other friendships). The average inter-domain correlation was r = .25. Though the inter-domain associations were somewhat modest, we elected to collapse across domains for two reasons. Conceptually, we did not have any a priori reason to believe that the domains would associate differentially with the project’s outcomes, so collapsing them into a broader index reflecting “abrasive social relations” seemed most appropriate. This strategy also made sense from a statistical perspective, because treating the domains separately would have quadrupled the number of analyses performed, and in doing so produced unacceptably high odds of Type 1 error. Scores on the final index could range from 1–5, with lower values reflecting warm, intimate, and supportive relationships, and higher values suggesting conflict, mistrust, and instability. The Life Stress Interview has been used widely in psychiatric research and there is robust evidence to support its reliability and validity in diverse populations like ours 22,23. In our project, interviewers showed excellent reliability on ratings, with agreement ranging from 89% (closest friendship) to 96% (romantic partner).
Expression of pro- and anti-inflammatory signaling molecules was quantified through real-time RT-PCR at study entry and six months later. Total RNA was extracted from leukocytes using PAXgene Blood RNA kits (Pre-Analytix, Hombrechtikon, Switzerland). RT-PCR reactions were carried out on a Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA), using one-step assays based on 5’ nuclease activity of FAM-labeled TaqMan probes (Applied Biosystems). For NF-κB and IκB, commercially available assays were used (Applied Biosystems #HS00765730_m1 and #HS00153283_m1) . For the GR isoforms, we developed a new TaqMan assay in collaboration with Applied Biosystems. The primer sequences were 5’-AGTGGTTGAAAATCTCCTTAACTATTGCT-3’ (forward) and 5’-GGTATCTGATTGGTGATGATTTCAGCTA-3’ (reverse) for GR-α and 5’-AGAAGATTATGTGCACTTCGTTGTCA-3’ (forward) and 5’-GGCACAGCTTCTTTTCCCATTTAAT-3’ (reverse) for GR-β. All assays used a standard thermal cycling protocol recommended by the manufacturer. As an internal control, 18S mRNA (for GR isoforms) or β-actin mRNA (for NF-κB and IκB) were quantified in parallel with target genes. The data were normalized using the ΔCT method (ΔCT = CT target – CT control). Results are expressed as relative quantities of each target, calculated by subtracting each patient’s ΔCT from the highest ΔCT in the distribution. Thus, higher relative quantities indicate greater expression of target genes.
Systemic inflammation was assessed using serum levels of CRP and IL-6 at study entry and six months later. CRP was measured using a high-sensitivity chemiluminescence technique on an Immulite 2000 (Diagnostic Products Corporation, Los Angeles, CA). This assay has an intra-assay variability of 2.2% and a minimum detection threshold of 0.20 mg/L. IL-6 was measured using commercially available high-sensitivity ELISAs (R&D Systems; Minneapolis, MN) with a minimum detection threshold of 0.039 pg/ml. Inter- and intra-assay variability were below 10%. To model the dynamics of inflammatory signaling pathways under immune challenge, we quantified leukocyte production of IL-6 following stimulation with lipopolysaccharide (LPS) at study entry and six months later. Whole blood was drawn into Lithium-Heparin Vacutainers and diluted 10:1 with saline, and then co-incubated with LPS at a concentration of 50 ng/ml (Sigma; Saint Louis, MO) for 6 hours at 37 C with 5% CO2. The supernatants were then harvested and frozen at −80 C until assayed for IL-6 by ELISA (DuoSet ELISA Development Systems; R&D Systems, Minneapolis, MN). These kits have a minimum detectable threshold of 0.7 pg/ml and inter- and intra-assay variability below 10%.
To determine whether behavioral and biomedical characteristics might be acting as confounders, we collected information regarding age, ethnicity, oral contraceptive use, socioeconomic status, smoking history, central adiposity, and strenuous physical activity. Each of these factors has been linked to interpersonal difficulties and/or immune functions in past work 1, 7, 8, 23. Socioeconomic status was assessed with the adolescent version of the MacArthur Scale of Subjective Social Status 24, and central adiposity was indexed as the ratio of waist to hip circumference. Strenuous physical activity was measured as minutes each week engaged in “regular activity akin to brisk walking, jogging, bicycling, etc, long enough to work up a sweat” 25. Because depressive symptoms can arise from chronic interpersonal difficulties 26, and themselves bring about systemic inflammation 27, we also administered the Beck Depression Inventory 28 to subjects at the 6-month visit.
To evaluate the study’s major hypotheses, we estimated a series of partial correlations between chronic interpersonal stress at baseline and inflammatory parameters 6 months later. Each analysis controlled for the potential confounding influences of age, ethnicity, oral contraceptives, socioeconomic status, central adiposity, and strenuous activity, as well as values of the outcome variable at baseline. (In other words, each of these covariates was held constant when the correlations were estimated.) Thus, significant associations indicate that chronic interpersonal difficulties at study entry presage changes in biological outcomes over the next six months, and do so in fashion that is independent of the demographic and biobehavioral covariates included in the model. Smoking was not included as a covariate because only 3 patients reported daily use of cigarettes. (The findings were similar regardless of whether these subjects were included or excluded from analyses, so we kept them in.) All results are based on two-tailed tests of significance.
Table 1 describes the sample’s characteristics and provides descriptive statistics for major predictors, covariates, and outcomes. As it shows, the study had a diverse sample of teenage females that mirrored the broader Vancouver population in terms of racial/ethnic background. Because of the project’s strict inclusion criteria, all were in good health, free of psychiatric disorders, and without standing medication regimens other than oral contraceptives. On the whole, the sample was rated as having modest levels of chronic interpersonal stress (2.36 on a 1–5 scale), but there was a good deal of variance around this average. Analyses of psychometric characteristics revealed that all of the study’s predictors and outcomes were distributed normally. The only exceptions to this rule were CRP and serum IL-6, both of which had positive kurtosis, with more observations at the left tail of the distribution than expected. To ensure that this did not affect the results of statistical analyses, we repeated all of them involving CRP and IL-6 with non-parametric techniques.
Analyses of the project’s various inflammatory parameters indicated that several of them were inter-correlated. Specifically, there were significant associations between GR-α and GR-β mRNA, NF-κB and IκB mRNA, CRP and IL-6, and GR-α mRNA and IL-6 production at both study entry and at follow-up (all r’s > .22, all p’s < .05; mean r = .60). None of the other pairs of outcomes displayed consistently significant associations. Analyses of stability over the six-month follow-up period revealed considerable variability across parameters. Whereas CRP, IL-6, and LPS-stimulated IL-6 were moderately stable over time (all r’s > .46, all p’s < .001; mean r = .51), none of the intra-cellular signaling molecules was (all r’s < .08, p’s > 43.) The one exception to this pattern was for NF-κB mRNA, which to our surprise was inversely correlated over time, r = -.26, p = .01.
Before conducting primary statistical analyses, we examined relations between predictors, outcomes, and covariates. Chronic interpersonal stress was higher among subjects with more central adiposity (r = .23, p = .02), and marginally higher among those low in socioeconomic status (r = −.18, p = .06), but did not vary by age, ethnicity, strenuous activity, or oral contractive use (all p’s > .10). None of the covariates was related to the project’s major outcome variables - changes over time in the various inflammatory parameters, all p’s > .09. We nonetheless elected to include all covariates in the models presented below. This approach safeguards against the possibility of spurious relations emerging between chronic stress and biological outcomes as a result of the variance they share with potential confounders.
At study entry there was a marginally significant inverse association between chronic interpersonal stress and NF-κB mRNA, r = −.19, p = .06. However, none of the other cross-sectional associations between chronic stress and inflammatory parameters reached statistical significance (all r’s < .17, all p’s > .09). (This was also true in non-parametric analyses of CRP and IL-6, p’s > .34).
By contrast, chronic interpersonal difficulties at study entry were related to changes over time in a number of outcomes. Table 2 describes the results of these analyses. As can be seen, to the extent that they had chronic interpersonal stress at the time of study entry, subjects displayed larger increases in LPS-stimulated production of IL-6 over the six-month follow-up (see Figure 1). Over the same time frame, subjects with interpersonal difficulties showed larger increases in mRNA for the pro-inflammatory molecules GR-β and NF-κB, and in mRNA for IκB. The latter molecule sequesters NF-κB in the cytoplasm and minimizes its pro-inflammatory activities. Chronic interpersonal stress at study entry was unrelated to changes in GR-α, serum CRP and serum IL-6 over the follow-up period, all p’s > .23. (Identical results emerged in non-parametric analyses of CRP and IL-6, p’s > .27).
All of the significant prospective associations were independent of age, ethnicity, socioeconomic status, oral contraceptives, central adiposity, and strenuous activity (middle row of Table 2). These associations also were independent of depressive symptoms at the 6-month assessment (bottom row of Table 2), suggesting that observed changes in inflammatory signaling were directly related to interpersonal difficulties, and not mediated by any consequent increases in dysphoria. (Incidentally, there was not a consistent pattern of cross-sectional or prospective associations between depressive symptoms and inflammatory parameters, all r’s < .14, all p’s >.17. The one exception to this was for IκB mRNA, which was positively associated with depression in both cross-sectional and prospective analyses, r’s > .20, p’s <.05.)
Mounting evidence suggests that chronic interpersonal stressors have a detrimental influence on mental and physical health, and one emerging hypothesis regarding the mechanism of these effects involves the ability of stress to alter physiologic inflammatory processes. The present results support that hypothesis in identifying significantly greater 6-month increases in expression of gene products involved in the transduction of inflammatory signals (NF-κB, GR-β, and IκB mRNA) in leukocytes from young people who experienced chronic interpersonal stress. These alterations in inflammatory gene expression appear to have significant consequences for leukocyte functional responses to stimulation. In response to a model bacterial stimulus (LPS), leukocytes from those showing higher levels of interpersonal difficulty showed greater increases over time in the production of the pro-inflammatory cytokine IL-6 than did cells from those experiencing low levels of interpersonal difficulty. These conflict-related changes in expression of inflammatory signaling pathway genes appear to predominately affect the leukocyte’s potential to respond to a pathogen challenge, rather than its basal production of pro-inflammatory cytokines. That is, subjects experiencing high levels of interpersonal difficulty at baseline did not show elevations in serum biomarkers of basal inflammatory activity (i.e., circulating IL-6 or CRP). Differences only emerged when inflammatory signal transduction pathways were actively engaged by a model stimulus (i.e., in LPS-stimulated production of IL-6 ex vivo).
These results imply that, at least in young healthy people with comparatively low levels of chronic inflammation, social conflict creates a potential for hyper-inflammatory responses that requires an exogenous immunological stimulus for realization. Thus, any health consequences of these dynamics, would likely involve a person x situation interaction in which interpersonal difficulties (person) acts to amplify the effects of a pathogenic insult (situation) to impact inflammation-related disease pathogenesis. Such dynamics could have relevance in a number of mental and physical illnesses that are known to be exacerbated by interpersonal difficulties. For example, depression is particularly sensitive to inflammation, and so are a number of infectious, metabolic, and coronary diseases 15–17,29.
The mechanisms responsible for stress-related changes in inflammatory signaling remain to be elucidated. Because of the project’s strict inclusion criteria and use of statistical controls, we can be reasonably confident that lifestyle variables, psychiatric conditions, medical illnesses, and demographic factors are not responsible for the observed relationship between interpersonal difficulties and altered inflammatory signaling. One plausible hypothesis is that sympathetic nervous system activation underlies this phenomenon. Acute bouts of social conflict provoke the release of norepinephrine 30, which can potentially increase NF-κB expression or enhance its DNA binding activity 31. However, the available evidence suggests that these dynamics unfold over fairly short periods of time and, as such, they would be expected to manifest in a “real-time” association between interpersonal difficulties and inflammatory parameters. By contrast, our data suggest that there is a time lag, of up to six months, between exposure to social stress and changes in leukocyte functions.
To some degree cortisol may help to explain these time-lagged patterns. Individuals without a regular schedule of warm social contacts tend to have unstable rhythms of HPA output 32,33. With repeated exposure to high doses of cortisol over time, such persons may eventually become resistant to the hormone’s influence. In the immune system the development of such resistance would enable inflammation to flourish without its usual hormonal constraints 23,34. In fact, chronic social stress has been shown to foster resistance to the anti-inflammatory properties of glucocorticoids in rodent models 13,35. And similar dynamics have been found in humans: recent genome-wide microarray experiments demonstrate that people who are chronically lonely or facing significant interpersonal stress., e.g. caring for a spouse with cancer, show heightened NF-κB activity and simultaneous impairment of cortisol-mediated signaling 34,36. Collectively, these results suggest that social difficulties may provoke HPA abnormalities, which over time foster resistance to glucocorticoids and expression of inflammatory mediators. The time required for this chain-of-events to unfold may help to explain why our data yielded evidence of prospective associations between interpersonal difficulties and inflammatory processes, but no cross-sectional relationships.
This study had several limitations worth noting. First, because the sample was chosen to be at high-risk for depression, they are not representative of the general population. Though this constrains the generalizability of the findings, it does not seriously complicate interpretation of them. Future research will need to be done, however, to substantiate the effects in the broader population. Second, the study quantified mRNA for inflammatory signaling molecules, but did not measure their associated proteins or directly assess the functional activity of each individual protein (although the integrated activity of the pathway as a whole was assayed using the LPS stimulation model). Assessment of individual protein alterations will need to be done in future research. Third, the study only had two points of inflammatory assessment. As a result, we are unable to specify how long the “incubation period” is between exposure to interpersonal difficulties and subsequent alterations in inflammatory processes. Without more points of assessment, we are also unable to specify if and when these alterations resolve, or what influence earlier social conflict had on biological processes captured at study entry. Multiwave studies will be needed to address these questions. Fourth, the sample was composed of young women who, on average, had fairly good social relations and limited amounts of systemic inflammation. The modest stress levels may have somewhat restricted the magnitude of associations we were able to observe, and the low CRP and IL-6 values may explain why these outcomes were unrelated to interpersonal difficulties in our sample, but have been linked to stress in other projects, e.g., 34,36. Finally, we did not measure any clinical outcomes in this project, so it remains unclear whether these dynamics have disease implications. It will be important for the next wave of studies in this area to do so, and determine whether inflammatory processes are the mechanism through which chronic interpersonal stress “gets under the skin” to undermine mental and physical health. In the meantime, these findings extend into humans a large corpus of animal research 37,38 suggesting that an organism’s physiology is intimately regulated by the social context in which s/he resides.
This work was supported by grants from the Canadian Institutes of Health Research, the Michael Smith Foundation for Health Research, the Heart and Stroke Foundation of Canada, the National Alliance for Research on Depression and Schizophrenia, and the German Research Foundation.
CONFLICTS OF INTEREST
None of the authors has a biomedical financial interest or a conflict of interest to declare related to this project.