Family-Based Analyses Reveal Novel Genetic Overlap Between Cytokine Interleukin-8 and Risk for Suicide Attempt

Abstract

Background

Suicide is major public health concern. It is imperative to find robust biomarkers so that at-risk individuals can be identified in a timely and reliable manner. Previous work suggests mechanistic links between increased cytokines and risk for suicide, but questions remain regarding the etiology of this association, as well as the roles of sex and BMI.

Methods

Analyses were conducted using a randomly-ascertained extended-pedigree sample of 1882 Mexican-American individuals (60% female, mean age=42.04, range=18-97). Genetic correlations were calculated using a variance components approach between the cytokines TNF-α, IL-6 and IL-8, and Lifetime Suicide Attempt and Current Suicidal Ideation. The potentially confounding effects of sex and BMI were considered.

Results

159 individuals endorse a Lifetime Suicide Attempt. IL-8 and IL-6 shared significant genetic overlap with risk for suicide attempt (ρ_g=0.49, p_FDR=7.67x10⁻³; ρ_g=0.53, p_FDR=0.01), but for IL-6 this was attenuated when BMI was included as a covariate (ρ_g=0.37, se=0.23, p_FDR=0.12). Suicide attempts were significantly more common in females (p_FDR=0.01) and the genetic overlap between IL-8 and risk for suicide attempt was significant in females (ρ_g=0.56, p_FDR=0.01), but not in males (ρ_g=0.44, p_FDR=0.30).

Discussion

These results demonstrate that: IL-8 shares genetic influences with risk for suicide attempt; females drove this effect; and BMI should be considered when assessing the association between IL-6 and suicide. This finding represents a significant advancement in knowledge by demonstrating that cytokine alterations are not simply a secondary manifestation of suicidal behavior, but rather, the pathophysiology of suicide attempts is, at least partly, underpinned by the same biological mechanisms responsible for regulating inflammatory response.

Introduction

A recent report by the Centers for Disease Control and Prevention indicated that in the United States alone suicide rates increased by more than 30% between 1999 and 2016¹ making suicide a leading cause of death. Moreover, this report does not account for the fact that suicide attempts are grossly underreported, with twenty or more attempts estimated for every completion ^2,3.

The need for biomarkers to identify suicide risk in an effective and timely manner has never been greater. Biomarkers are objective indicators of illness risk ⁴ and one potential biomarker of suicide is inflammation. There is strong evidence for altered levels of inflammation in individuals experiencing suicidality, particularly in pro-inflammatory cytokines TNF-α, IL-6 and IL-8 ^5,6. This association has been demonstrated in several psychiatric disorders, in both the periphery and central nervous system, in individuals that have either attempted or completed suicide, and in a plethora of centrally- and peripherally-measured cytokines ^7-23.

However, important questions remain regarding the relationship between TNF-α, IL-6 and IL-8 and risk for lifetime suicide attempt or current suicidal ideation. First, the etiologic origins of the relationship are unclear. Risk for suicide has been shown to be heritable ²⁴ as have levels of TNF-α, IL-6 ^25,26 and IL-8 ²⁷. However, no study has attempted to assess any potential overlap in the genetic factors that influence inflammation and suicide risk. Therefore, it is not known whether inflammation is directly involved in the pathophysiology of suicidal behavior, or arises as a secondary manifestation of these behaviors.

Second, despite consistent sex differences in both suicidal behavior and immune response, the effect of sex is rarely examined ²⁸. Males more often complete suicide than females (although recent trends indicate that that gap is narrowing ²⁹). Paradoxically, females make, on average, twice as many suicide attempts as males. This difference in mortality rates is thought to stem from males more often choosing violent and lethal suicide methods than females ³⁰. There are also sex differences in how the immune system responds to both foreign and self-antigens ³¹. In general, female immune responses tend to be stronger, resulting in greater immunity (e.g. enhanced vaccine response), but also increased susceptibility to inflammatory and autoimmune diseases ^28,32,34. A handful of studies on suicidal behavior have assessed the impact of sex on inflammation levels ^9,35-40, but no study, to date, has tested whether any genetic overlap between inflammation and risk for suicidal behavior varies as a function of sex. Evidence for genetic overlap in one sex but not the other might hint at important differences between males and females in the pathophysiology of suicidal behavior.

Third, increased adiposity is a well-established correlate of inflammation ⁴¹, but studies on the relationship between inflammation and suicide have generally not examined the effect of adiposity on this relationship, or treated it as a nuisance covariate ^9,11,39. In particular, IL-6 production is thought to increase with adiposity, which is typically measured using its proxy body mass index (BMI), with up to a third of circulating IL-6 originating from adipose tissue ^42-44. Evidence for a link between suicide and BMI is mixed ⁴⁵. Some studies report an inverse association between BMI and suicide completion ^46-48 and attempt ^49-51, while others report a positive association with completions ^52,53 and attempts ^54-56. Of the four studies on suicidal behavior that tested the influence of BMI on inflammation, three indicated that BMI had no significant effect ^13,37,38, and one showed that BMI was significantly associated with inflammation ⁵⁷. Thus, current evidence for an association between BMI, inflammation and suicide, is mixed.

Previous studies have been unable to provide answers to these questions. Family-based data (e.g. twins or extended pedigrees) are the gold standard for establishing the extent of genetic influence (i.e. heritability) on individual traits, as well as genetic overlaps (i.e. genetic correlation) between traits, but no study on inflammation and suicide has been conducted in families. Moreover, a sizable sample is required for such studies and while one of the more recent studies looking at the relationship between IL-6 and suicide had a relatively large sample (N = 784, ⁹), many studies prior to that had samples of less than or just over one-hundred individuals ⁶. These sample sizes and lack of statistical power may partly account for why the effects of sex and BMI have been difficult to pin down.

Using a large, multi-generational, multi-family sample of 1882 Mexican-American individuals combined with univariate and multivariate variance components methods we present the first study to disentangle shared genetic effects on suicidal behavior and cytokine levels. We chose to investigate the shared genetic influences of cytokines and suicide because of robust associations between risk for suicidal behavior and cytokine levels. In the blood-based assay available in the sample three cytokines were present at detectable levels for subsequent analysis to take place: IL-6, IL-8 and TNF-α. The aims of this study were threefold, to establish: (a) the degree of genetic overlap between both lifetime suicide attempt and current suicidal ideation and three different cytokines; (b) the extent to which the genetic overlap between IL-6, over and above IL-8, is confounded by BMI; and (c) whether the genetic overlap between risk for suicide and either interleukin was sex-specific.

Methods

Participants

The sample comprised 1882 individuals from extended pedigrees (83 families, mean size = 21.67 individuals, range 2-178 individuals). The sample was 60.43% female and had a mean age of 42.04 years (sd = 15.74 years, range = 18-97 years). Individuals in this Genetics of Brain Structure and Function Study (GOBS) have actively participated in research for over 18 years and were randomly selected from the community. Inclusion criteria were: 1) Mexican American ancestry, 2) part of a large family, and 3) live within the San Antonio region (see ⁵⁸ for recruitment details). All participants provided written informed consent on forms approved by the institutional review board at the University of Texas Health Science Center of San Antonio.

Suicide Assessment

Participants completed the semi-structured Mini-International Neuropsychiatric Interview (MINI;⁵⁹). Interviews were conducted by masters- and doctorate-level research staff who had established reliability for conducting suicide risk assessment (κ≥0.85). Suicidality was assessed in two ways in the present study, both assessments were taken from Section C of the MINI. Lifetime Suicide Attempt was assessed by item C6: ‘Did you ever make a suicide attempt?’. Our Lifetime Suicide Attempt index is therefore a dichotomous trait, with a score of 1 indicating a “yes” response and 0 indicating a “no” response. The MINI also includes a section on current suicidality (items C1-C5), asking whether in the past month the participant thought they would be better off dead (C1); wished that they were dead (C2); wanted to harm themselves (C3); thought about suicide (C4); or had a suicide plan (C5). These items were collapsed to generate a categorical index of Current Suicidal Ideation, with a score of 1 indicating that any item between C1-C5 was endorsed and a score of 0 indicating that none were endorsed. The presence of major depressive, anxiety, alcohol and substance use disorders was also evaluated as part of the MINI assessment. Subjects who reported possible pathology were discussed in case conference meetings with licensed psychologists and/or psychiatrists. Consensus diagnoses for major depressive disorder and a lifetime suicide attempt were determined using available medical records, the MINI, and the interviewer’s narrative.

Inflammation Measurement

Participants completed a venipuncture between the hours of 7am and 11am for the measurement of inflammation markers. Blood was drawn after a 12-hr fast. Blood samples were centrifuged for 20 min at 4°C, aliquoted in 50uL volumes (chiclets) of EDTA plasma and stored at −80°C within 4 h ofvenipuncture. Inflammation assays were run in plasma using a Milliplex MAP Human Adipokine Magnetic Bead Panel 2 (Millipore catalog #HADK2MAG-61K) on a Luminex 200 instrument. The cytokines run were TNF-α, IL-6 and IL8. Assays were performed in 96-well plates on plasma samples stored at −80°C. 10% of the samples were run in duplicate on separate plates. Intra-assay coefficients of variation were <10% for all markers. Minimum detectable concentration (pg/mL) was 0.2, 0.3, and 0.3 for IL-6, IL-8 and TNF-α respectively.

Quantitative Genetic Analysis

All genetic analyses were performed in SOLAR ⁶⁰. SOLAR implements a maximum-likelihood variance decomposition approach to determine the proportions of phenotypic variance that are due to genetic and environmental influences. The simplest such decomposition (univariate polygenic analysis) is one where the additive genetic contribution of a single trait is indexed by its narrow-sense heritability, denoted by h². We first estimated the heritability of all traits. All inflammation measures were normalized using an inverse Gaussian transformation. For all traits, age, age^2, sex and their interactions, as well as BMI were entered as covariates into univariate polygenic models.

Under a bivariate polygenic model, the phenotypic covariance between each inflammation phenotype and suicide attempt is decomposed into its genetic and environmental constituents to determine the extent to which they are influenced by shared genetic effects. Bivariate polygenic analysis was applied to pairs of inflammation and suicide traits, with relevant covariates (those deemed significant in univariate analysis, see above) for each trait. We also ran the bivariate polygenic models with potentially confounding diagnoses as covariates including: any anxiety disorder, current and lifetime substance and alcohol dependence disorders, bipolar disorder type I and type II, schizoaffective disorder and schizophrenia.

We also applied a coefficient-of-relatedness-analysis to lifetime suicide attempt and each cytokine. Briefly, this analysis leverages the many coefficients of relationship that exist in extended-pedigree data. The coefficient of relationship refers to the average number of alleles held in common between individuals. The coefficient of relatedness refers to the average number of alleles held in common between individuals; for example, first-degree relatives (e.g. full siblings or parents) share on average 50% of their alleles, second-degree relatives (e.g. grandparents or aunts/uncles) share 25%, third-degree relatives (e.g. great-grandparents or great-aunts/-uncles) share 12.5%, and so on. Thus, it is possible, given an individual with a disease, to index all other pedigree members by their degree of relatedness to that individual. This scalar can then be used to perform a fixed-effect single-degree-of-freedom test within the univariate polygenic analysis framework (outlined above), providing an estimate of the standardized genetic covariance between the potential endophenotype and illness risk ⁶¹. This analysis is useful because affected individuals are excluded and thus, confounding by factors such as medication status and alteration in inflammation levels as a consequence of suicide attempt is minimized.

To control for multiple testing of main effects, the false discovery rate (FDR) was set at 5% ⁶².

Results

Heritability of Risk for Suicide Attempt and Suicidal Ideation

Of the individuals assessed as part of GOBS, 159 endorsed having attempted suicide in their lifetime and 135 endorsed some level of current suicidal ideation. Of those individuals that had attempted suicide, 98 had a diagnosis of a past major depressive disorder. Table 1 provides descriptive statistics for these individuals. A greater proportion of females than males had attempted suicide and were experiencing current suicidal ideation, but age and BMI were fairly consistent across groups (Table 1).

Table 1.

Descriptive statistics (N, % female, mean age and BMI) of those individuals that have attempted suicide and those individuals that have current suicidal ideation.

Phenotype	Endorsement	N	% Female	Age (SD)	BMI (SD)
Suicide Attempt	Yes	159	70.44	41.09 (13.45)	32.65 (7.72)
	No	1725	58.72	42.23 (15.93)	31.57 (7.49)
Current Suicidal Ideation	Yes	135	74.07	42.73 (16.92)	31.42 (7.80)
	No	1749	58.60	42.09 (15.65)	31.67 (7.50)

Table 2 shows results of univariate polygenic analysis, including covariate effects, on risk for suicide attempt and current suicidal ideation. Both suicide attempt (h² = 0.46, p_FDR = 1.68x1O⁻⁴) and suicidal ideation (h² = 0.47, p_FDR = 1.15x1O⁻⁰³) were significantly heritable. Sex was a significant covariate of both suicide traits (p = 0.01 and p = 0.04, respectively) but neither age, nor age^2 (nor their interactions with sex), nor BMI were significant covariates of either suicide trait.

Table 2.

Univariate polygenic analysis results for risk for suicide attempt and current suicidal ideation and inflammatory markers, including h² (and standard error) estimates, p-values and X² (and p-values) for covariates. Significant covariates are highlighted in bold.

Trait	h2 (se)	p	age	ag^2	sex	age*sex	age^2*sex	BMI
Suicide Attempt	0.47 (0.13)	2.68x10−05	0.61 (0.43)	0.74 (0.39)	6.33 (0.01)	0.87 (0.35)	0.42 (0.51)	1.03 (0.31)
Suicidal Ideation	0.46 (0.16)	2.75x10−04	0.64 (0.42)	0.01 (0.91)	4.31 (0.04)	0.04 (0.85)	0.42 (0.52)	0.01 (0.93)
TNF-α	0.26 (0.05)	2.86x10−08	70.40 (4.84x10−17)	0.52 (0.47)	9.39 (2.18x10−03)	0.50 (0.48)	5.65 (0.02)	27.93 (1.26x10−07)
IL-6	0.17 (0.05)	5.46x10−05	30.24 (3.82x10−08)	2.32 (0.13)	2.00 (0.16)	1.59 (0.21)	0.19 (0.66)	123.55 (1.06x10−28)
IL-8	0.30 (0.05)	4.54x10−14	87.94 (6.73x10−21)	4.56 (0.03)	19.07 (1.26x10−05)	1.10 (0.29)	5.80 (0.02)	1.17 (0.28)

Heritability of Cytokines

The heritability estimates and covariate effects for the interleukin markers are shown in Table 2. TNF-α (h² = 0.26, p_FDR = 3.11x10⁻⁰⁸), IL-6 (h² = 0.18, p_FDR = 1.68x10⁻⁰⁴) and IL-8 (h² = 0.30, p_FDR = 7.15x10⁻¹³) were significantly heritable (Table 2) and were significantly associated with age. IL-8 and TNF-α were significantly associated with sex and Il-6 and TNF-α showed strong association with BMI.

Genetic Correlation between Cytokines with Risk for Lifetime Suicide Attempt and Current Suicidal Ideation

The results from bivariate polygenic analysis (including significant covariates from each trait according to univariate polygenic analysis, see above) applied to risk for lifetime suicidal attempt and current suicidal ideation paired with each cytokine marker are shown in Table 3. This analysis revealed that only IL-8 (ρ_g = 0.49, se = 0.17, p_FDR = 8.94x1O⁻⁰³) exhibited a significant genetic overlap with risk for lifetime suicide attempt. The genetic correlation between IL-8 and current suicidal ideation was not significant. Neither TNF-α nor IL-6 showed significant genetic correlation with risk for lifetime suicide attempt nor current suicidal ideation.

Table 3.

Bivariate polygenic analysis results each suicide trait (lifetime suicide attempt and current suicidal ideation) paired with an inflammatory marker (TNF-α, IL-6, IL-8). including ρ_g (and standard error) estimates and p-values.

Suicide Trait	Inflammation Marker	ρg(se)	pFDR
Lifetime Suicide Attempt	TNF-α	0.10 (0.10)	0.60
	IL-6	0.37 (0.23)	0.12
	IL-8	0.49 (0.17)	8.94x10−03
Current Suicidal Ideation	TNF-α	0.15 (0.20)	0.47
	IL-6	0.39 (0.23)	0.15
	IL-8	0.29 (0.16)	0.15

Figure 1 shows results of coefficient-of-relatedness analysis, which excluded affected individuals. While IL-8 (β = 0.54, se = 0.15, p_FDR = 1.25x1O⁻⁰³) levels varied as a function of genetic risk for suicide attempt, TNF-α (β = 0.20, se = 0.15, p_FDR = 0.25) and IL-6 (β = 0.12, se = 0.05, p_FDR = 0.47) levels did not (Figure 1). Altogether, these results suggest that IL-8, but not IL-6 or TNF-α, has genetic overlap with risk for suicide attempt, and also that IL-8, but not IL-6 or TNF-α, varies with genetic risk for suicide attempt in unaffected relatives. In sum, in this sample, IL-8 appears to be a better marker of suicide risk than IL-6.

Figure 1.

Average levels of IL-6 and IL-8 in suicide attempters, unaffected relatives, and unaffected unrelateds.

The genetic correlations between the inflammatory markers were all significantly different from zero but also significantly different from 1 indicating partial genetic overlap with specificity between markers (Table S1).

Comorbid psychiatric diagnoses derived from the MINI as well as self-reported somatic medical conditions and self-reported current prescriptions are listed in Table S2 and Table S3. The genetic correlation between lifetime suicide attempt and IL-8 remained significant when those comorbid conditions (any alcohol or substance use disorder, bipolar disorder, psychosis, diabetes) that were significant covariates of either trait were included as covariates in the bivariate polygenic model (ρ_g = 0.39, se = 0.19, p_FDR = 0.03).

The genetic correlation between lifetime suicide attempt and IL-8 remained significant when currently prescribed medications (antidepressants, antidiabetics, ACE inhibitor, antipsychotics, diuretics, opioids) that were significant covariates of either trait were included as covariates in the bivariate polygenic model (ρ_g = 0.48, se = 0.18, p_FDR = 7.13x10⁻⁰³).

The Effect of BMI on the Relationship Between Risk for Lifetime Suicide Attempt and IL-6

Given previous findings on IL-6 and suicide risk, it was surprising that IL-8, but not IL-6, demonstrated a significant genetic overlap with risk for lifetime suicide attempt. Since IL-6 was significantly associated with BMI, BMI (mean = 31.38, SD = 7.53, range = 15.53-78.37) was included as a covariate in the bivariate polygenic model between IL-6 and suicide attempt. Removing BMI as a covariate generated a significant genetic correlation (ρ_g = 0.53, se = 0.19, p_FDR = 0.02), suggesting that the effects of BMI should be carefully considered when investigating the relationship between inflammatory measures, particularly IL-6, and risk for suicide attempt. We also ran a trivariate model of IL-6, lifetime suicide attempt and BMI. Using this model we estimated the genetic correlations between pairs of traits in the same model rather than treating BMI as a confound or nuisance covariate. Inspection of Figure 2 shows that the genetic correlation between IL-6 and risk for lifetime suicide attempt was not significant when BMI was included as a third phenotype in the model.

Figure 2.

Trivariate polygenic model of IL-6, BMI and lifetime suicide attempt.

Removing BMI as a covariate in the bivariate polygenic model between TNF-α and suicide attempt did not alter the non-significant genetic correlation between the trait pair (ρ_g = 0.22, se = 0.20, p_FDR = 0.36).

The Effect of Sex on the Relationship Between Lifetime Suicide Attempt and IL-8

Since both suicide attempt and IL-8 exhibited effects of sex (Table 2), we also ran the bivariate polygenic model between suicide attempt and IL-8 separately in males and females. In females, the genetic correlation between IL-8 and risk for suicide attempt was significant (ρ_g = 0.56, se = 0.21, p_FDR = 0.02), but not in males (ρ_g = 0.27, se = 0.86, p_FDR = 0.36). However, the heritability of risk for suicide attempt in the bivariate model was much lower in males (h² = 0.11, se = 0.36) than in females (h² = 0.52, se = 0.17). This may be because there is an insufficient number of cases in males for h² to be reliably estimated. To circumvent this limitation, we constrained the threshold of risk for suicide attempt in males to be the same as that in females (= 1.27), resulting in a heritability of risk for suicide attempt similar to that in females (h² = 0.55, se = 0.37). However, the genetic correlation remained non-significant (ρ_g = 0.44, se = 0.41, p_FDR = 0.36). These analyses suggest that the relationship between suicide attempt and IL-8 in this sample is moderated by the sex, such that the relationship is driven largely by females and is not evident in males. This finding implies a strong sex-specific effect for IL-8 on lifetime risk for suicide attempt.

Does Major Depressive Disorder (MDD) Confound the Relationship Between Lifetime Suicide Attempt and IL-8?

Of the 159 individuals that had attempted suicide, 98 (62%) also had a lifetime diagnosis of MDD, making it possible that MDD confounded the relationship between IL-8 and lifetime suicide attempt. However, when the presence of any major depressive disorder was included as a covariate in the bivariate polygenic model the relationship between IL-8 and lifetime suicide attempt remained significant (ρ_g = 0.48, se = 0.20, p_FDR = 0.02). Moreover, in a trivariate model where the genetic correlations between trait pairs are estimated in the same model, the genetic correlation between risk for suicide attempt and IL-8 remained significant (ρ_g = 0.48, p_FDR = 0.01), the genetic correlation between risk for suicide attempt and MDD was also, unsurprisingly, significant (ρ_g = 0.92, p_FDR = 3.00x10⁻⁰⁴), and the genetic correlation between risk for MDD and IL-8 was significant (ρ_g = 0.40, p_FDR = 0.04). Of the 112 females that had attempted suicide, 72 (=64%) had been diagnosed with MDD. Running the trivariate model in females alone generated similar results to those from the entire sample. The genetic correlation between suicide attempt and IL-8 was significant (ρ_g = 0.52, p_FDR = 0.03), as was the genetic correlation between suicide attempt and MDD (ρ_g = 0.90, p_FDR = 1.92x10-03). However, the genetic correlation between IL-8 and MDD was not significant (ρ_g = 0.42, p_FDR = 0.07).

Discussion

There is clear evidence for inflammatory alterations in individuals who have engaged in suicidal behavior⁶. Using data from a large, multi-family, multi-generational sample we have advanced understanding of this relationship in several ways.

First, we tested the extent to which alterations in the inflammatory markers IL-6 and IL-8 shared a genetic overlap with risk for suicide attempt. Studies have already shown that suicide attempts are associated with increased inflammation ^5,6. However, no study has previously looked at whether this association is caused by the same genes affecting both traits, as opposed to environmental factors (e.g. inflammation levels increasing after a suicide attempt). We provide evidence that plasma-based IL-8 and risk for suicide attempt are driven, at least in part, by a common set of genes. This suggests that alterations in peripherally measured IL-8 do not arise merely as a consequence of suicidal behavior but that some of the same genes that affect inflammation levels also mediate risk for suicide attempt. That is, the biological mechanisms that regulate IL-8 are some of the same that impact risk for a suicide attempt. Future studies attempting to isolate genes that influence risk for suicide attempt might focus their attention on some of the well-characterized molecular pathways for IL-8 regulation.

Second, we tested whether the genetic overlap between IL-8 and risk for suicide attempt varied as a function of sex. In line with previous literature ³⁰, both suicidal ideation and attempts were significantly more common in women than in men in the present sample. By stratifying our analyses by sex, we showed that a genetic correlation between IL-8 and risk for suicide attempt was present in both males and females but that, overall, the effect appeared to be driven by females. This finding suggests important differences in the pathophysiology of suicide attempts in males and females. Specifically, in females, but not necessarily in males, the biological mechanisms associated with IL-8 are a key contributor to risk for suicide. This finding is in line with evidence that the immune response, including the production of cytokines and chemokines, is strongly affected by sex ³¹, and that women are more susceptible than men to inflammation-induced mood symptoms ⁶³. Recent findings showed that depressed males exhibited no differences in a range of serum-based pro-inflammatory markers compared to healthy controls, while depressed females showed higher levels on a number of markers, including IL-8 ⁶⁴. Our trivariate results in females showed that risk for suicide attempt, but not MDD, had shared genetic influences with IL-8, suggesting that suicidality, rather than other aspects of depressive symptomatology, drives the apparent relationship between MDD and inflammation. However, the non-significant genetic correlations between risk for suicide and MDD in females could also be due to a lack of power, particularly given that we observed separate and significant correlations in the larger sample. Future work might focus on what specifically drives these differences in males and females, for example the interaction between sex hormones, immune response, and suicidality ^65,66.

Third, we investigated the extent to which BMI, as a proxy of adiposity, confounded the relationship between IL-6 and risk for suicide attempt. IL-6 has been described previously as one of the most reliable peripheral biomarkers of MDD ^67-69. IL-6 is more often studied, than for example IL-8, in relation to suicidality ⁶. However, IL-6 is susceptible to a major confounder, adiposity, which is also often associated with disorders of mood ^70,71. Up to a third of circulating IL-6 originates from adipose tissue ⁴². Indeed, in the present study, the relationship between risk for suicide attempt and IL-6 was confounded by BMI since the genetic correlation between IL-6 and suicide attempt was only significant when BMI was not included as a covariate. Of course, BMI is an imperfect proxy for adiposity because it cannot differentiate between muscle and fat ⁷². Nevertheless, the present study suggests that BMI is an important source of bias in IL-6 levels and should be considered when assessing the association between IL-6 and suicide. This is not to say that BMI should be included as a covariate in all studies focusing on the relationship between mood and/or suicide and inflammation, however our results strongly suggest that the possibly influence of BMI on increased inflammation levels should be considered.

We showed that IL-8, but not IL-6 (when BMI was included as a covariate) or TNF-α, overlapped with risk for suicide attempt. This is intriguing since these are three belong to a diverse group of key communicators in the development of both acute and chronic inflammation ⁷³, called cytokines. However, each also has distinct structural morphologies and roles ⁷⁴. IL-6 and TNF-α are members of pleiotropic cytokines with roles related to the maturation of B cells into anti-body producing plasma cells, T cell activation, B-lymphocyte function and haemtopoiesis ^73,75,76. Chemokines (or, chemotactic cytokines) are a family of small cytokines that include IL-8, whose primary role is to induce chemotaxis in local cells, recruiting leukocytes to the site of infection or injury ⁷³. Thus, the two interleukins have quite different roles in the inflammatory response; while IL-6 is a cytokine that initiates an inflammatory response by triggering signaling cascades, IL-8 is a chemokine that initiates cell movement to recruit immune cells to the site of infection. To the extent that IL-6 and IL-8 have different roles it makes sense that one, but not the other, might be involved in the pathophysiology of risk for suicide. However, further work is required to delineate how chemotaxis, particularly in the central nervous system, might increase risk for suicidal behavior.

Despite a wealth of research dedicated to the topic, few, if any, clinically translatable biomarkers have been identified in psychiatry ⁷⁷. This study provides strong support for IL-8 as a biomarker of risk for suicide attempt. Recent national increases in suicidal behavior³, combined with the poor predictive power of traditional clinical risk factors ⁷⁸, mean that the need for robust markers of suicidal behavior has never been greater. This is particularly true when one considers that 54% of individuals who died by suicide in 2016 had no diagnosed mental illness ¹. IL-8 is already in routine use as a marker in various medical subspecialties for a number of clinical conditions, including prostatitis, non-Hodgkin’s lymphoma, bladder cancer, pulmonary infections and others ⁷⁹). Our findings suggest that IL-8 might also have utility in psychiatry, particularly in the detection of suicidal intent.

The heritability of TNF-α, IL-6 and IL-8 have been published previously. Heritability estimates of IL-6 from twin studies range between 0.17 and 0.61, though the majority of studies indicate estimates closer to 0.61^25,26,80-83. The classical twin design has been criticized for giving upwardly biased heritability estimates {^84-86. Indeed, a large extended pedigree study indicated that the heritability of IL-6 was 0.25, which is somewhat closer to the heritability found in the present study ⁸⁷. Moreover, one of the twin studies that accounted for BMI in their models, as we did in the present study, showed that the heritability of IL=6 was 0.26 ²⁶. TNF-α and IL-8 have featured less prominently in heritability studies of inflammatory traits. The heritability of TNF-α has been estimated between 0.26 and 0.53 ^25,26. Only one study has, to the best our knowledge, investigated the heritability of IL-8 in humans prior to the present one and that estimated heritability of IL-8 to be 1.00 ²⁷. The authors of that study acknowledge that the high heritability of IL-8 might be due to an unapparent infection running in families or possibly due to medication effects. Pedigree-based analyses in cows and baboons indicate that the heritability of IL-8 is 0.30 and 0.28 respectively ^88,89. Thus, in general the heritability of cytokines vary considerably in the literature which is likely due to the variety of research designs and inflammation assays employed. However, the estimates given in the present study fall within the limits of those published previously.

The present study has a number of limitations that might be addressed by future work. First, personality traits such as impulsivity or aggressiveness might confound or even mediate the relationship between inflammation and suicide attempt. Impulsivity and aggression have been cited as essential features of suicidal behavior ^90,91. Moreover, impulsivity and suicide attempt via a violent method are related to plasma-based IL-6 ³⁹. Second, the participants in the present study are of Mexican American ancestry and it is possible that the results we observe here are population specific. Subsequent work should focus on replicating our results in other ethnicities. Third, increased circulating cytokine levels have been observed in response to stress ⁹², infection ⁹³, trauma ⁹⁴, arthritis ^95,96, psoriasis ⁹⁷, atherosclerosis ⁹⁸ and obesity ⁹⁹. While we accounted for BMI other confounds were not accounted for, which might have affected our heritability (and by extension genetic correlation) estimates. Fourth, as authors of previous studies have noted common environmental influences, e.g. an unapparent infection running in a family, might increase inflammation and confound estimates of heritability (and by extension genetic correlations) ²⁷, however the families in question were nuclear families. The present study, on the other hand, used an extended pedigree design. It is a widely held view among genetics researchers that one of the benefits of extended pedigrees relative to smaller families is that there is reduced confounding of genetic and shared environmental effects because family members are usually distributed across multiple households ¹⁰⁰. Where in the present study the largest family comprised 178 individuals (mean size = 21.67) it is highly unlikely that entire pedigrees were subject to such common environmental effects. Fifth, inflammation levels vary diurnally ¹⁰¹. While in the present study all blood draws were completed within a narrow time window in the morning. Future work that might focus on tracking whether inflammation levels track with suicidal ideation, which are subject to change over a matter of hours ^102,103. This would be an advancement of the current study, which used a relatively crude index of suicide assessment.

It is not possible, in the present study, to draw causal inference from our results due to the cross-sectional nature of the study. Although, we found evidence for a strong and significant genetic correlation between risk for suicide attempt and IL-8 levels. A genetic correlation, in contrast to a phenotypic correlation, implies that both traits are influenced by partially overlapping genetic factors: here, directionality is of course guaranteed (genes caused the correlation between the traits). Causal evidence for the association between inflammation and depression has been provided by experiments using exogenously-induced inflammation ^104,105. After exposure to rhinovirus or influenza those individuals that became infected showed increase in IL-6 production that was associated with reduced positive (but not increased negative) affect ¹⁰⁶. The introduction of an endotoxin (Salmonella abortus equi) and concomitant increases in inflammatory markers was associated with increased sleepiness ¹⁰⁷ and also depressed mood ¹⁰⁸. Administration of the Salmonella typhi vaccine and associated increases in IL-6 resulted in increased negative mood ¹⁰⁹. Increases in IL-6 after receiving the influenza vaccine were associated with increases in depressed mood (though not with decreases in positive affect or increases in fatigue ¹¹⁰. However, few studies have focused on suicidality using these paradigms and subsequent work might focus on this.

In the present study, IL-8, but not IL-6, showed shared genetic influences with risk for suicide attempt. This genetic overlap appeared to be particularly pronounced in females, providing strong evidence for a sex-specific effect whereby the same biologic mechanisms that drive IL-8 production and metabolism also drive, at least to a certain extent, suicide attempts in women. Moreover, based on the results of this study, the effect of BMI on inflammation levels may have confounded results from previous studies that placed greater emphasis on IL-6 for suicide risk. These findings represents a significant advancement in knowledge by demonstrating that inflammation alterations are not simply a secondary manifestation of suicidal behavior, but rather, the pathophysiology of suicide attempts is, at least partly, underpinned by the same biological mechanisms responsible for regulating inflammatory response.

Highlights.

• IL-8 level and risk for suicide attempt are driven by a common set of genes

• The genetic correlation, derived via multivariate variance components methods. between IL-8 and suicide attempt was driven by females

• BMI is an important source of bias in IL-6 levels

• BMI should be considered when assessing the association between IL-6 and suicide

• Pathophysiology of suicide attempts is underpinned by the same biological mechanisms responsible for regulating inflammatory response