Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Nov 9.
Published in final edited form as: Brain Behav Immun. 2020 Apr 6;88:403–410. doi: 10.1016/j.bbi.2020.04.003

TNF-α and its soluble receptors mediate the relationship between prior severe mood episodes and cognitive dysfunction in euthymic bipolar disorder

CE Millett 1,2,*, J Harder 1,2,*, JJ Locascio 4, M Shanahan 1,4, G Santone 3, R Fichorova 3, A Corrigan 1, C Baecher-Allan 5,6, KE Burdick 1,2
PMCID: PMC8577222  NIHMSID: NIHMS1585636  PMID: 32272224

Abstract

Background:

Bipolar disorder (BD) is one of the most disabling mental health conditions in the world. Symptoms of cognitive impairment in BD contribute directly to occupational and social deficiencies and are very difficult to treat. Converging evidence suggests that BD patients have increased peripheral markers of inflammation. The hypothesis of neuroprogression in BD postulates that cognitive deficits develop over the course of the illness and are influenced by prior severe mood episodes, leading to wear-and-tear on the brain– however, there exists a paucity of data statistically testing a mediating role of immune molecules in cognitive dysfunction in BD.

Methods:

This is a cross-sectional study. We measured serum levels of tumor necrosis factor alpha (TNF-α), and soluble (s) TNF receptors one and two (sTNF-R1 and sTNF-R2) in 219 euthymic BD patients and 52 Healthy Controls (HCs). Structural equation modeling (SEM) was used for the primary purpose of assessing whether TNF markers (measured by the multiple indicators TNF-α, sTNF-R1 and sTNF-R2) mediate the effect or number of prior severe mood episodes (number of prior psychiatric hospitalizations) on cognitive performance.

Results:

BD and HC groups did not differ on circulating levels of TNF molecules in the present study. However, we found higher sTNF-R1 concentration in ‘late-stage’ BD illness (>1 prior psychiatric hospitalization) compared to those in early stage illness. In the subsequent SEM, we found that the model fits the data acceptably (Chi-square= 49.2, p=0.3), and had a ‘close fit’ (RMSEA= 0.02, PCLOSE= 0.9). Holding covariates constant (age, sex, premorbid IQ, education, and race), we found that the standardized indirect effect was significant, p=0.015, 90%CI [−0.07, −0.01], indicating that the estimated model was consistent with peripheral TNF markers partially mediating a causal effect of severe mood episodes on executive function.

Conclusions:

Our results indicate that circulating levels of TNF molecules partially mediate the relationship between prior severe mood episodes and executive function in BD. These results may implicate TNF variables in the neuroprogressive course of BD and could point to novel interventions for cognition.

Keywords: cognition, mood disorder, euthymia, inflammation

1. Introduction

Bipolar disorder (BD) is a chronic mental illness associated with functional decline, mortality, and significant health care costs (Ferrari et al., 2013). Significant cognitive impairment has been found in 40–60% of people with BD (Burdick et al., 2011), and such impairments are thought to underlie functional deficits seen in this disorder (Depp et al., 2016). Thus far cognitive impairment has not been successfully targeted by treatment. In recent decades, the immune system has emerged as an important candidate mechanism underlying the onset and progression of BD. BD has been categorized as a multisystem inflammatory disease due to the high rate of inflammatory illnesses associated with the disorder (Leboyer et al., 2012). Yet much is still unknown about the potential for inflammatory molecules as treatment targets in this debilitating mental illness. Elucidating the immunological markers underlying the disease that are associated with BD’s clinical features, may hold answers in the future for such understanding.

The term neuroprogression has been used to describe the pathological brain changes that occur over time in BD (F. Kapczinski et al., 2014a). Evidence suggests that BD is a neuroprogressive condition, with declines in baseline functioning observable in a large proportion of patients (López-Villarreal et al., 2019; Rosa et al., 2010). Cognitive and functional decline have been associated with course of illness (including number of mood episodes, duration of illness, and number of prior psychiatric hospitalizations) (Bourne et al., 2013; Martinez-Aran et al., 2007; Torrent et al., 2012) which is thought to reflect a stepwise cumulative effect, tied directly to the ‘wear and tear’ on the brain (F. Kapczinski et al., 2014b; Lewandowski et al., 2011). The biological mechanisms that drive this declining course are unknown, but recent evidence suggests a key role for inflammation.

Many peripheral inflammatory cytokines are upregulated during acute manic and depressive states in BD. These acute elevations typically return to “normal” levels (similar to those of HCs) during euthymic periods (Berk et al., 2011). Acutely elevated inflammatory markers rise to a level that is intermediate between the healthy state and the highly inflamed state of sepsis (Kauer-Sant’Anna et al., 2009; Panizzutti et al., 2015; Siwek et al., 2017a; Tatay-Manteiga et al., 2017), with attenuated recovery in later stages of the illness, resulting in low-grade chronic inflammation.

One marker that has been consistently implicated in late-stage BD is the pro-inflammatory cytokine tumor necrosis factor-α (TNF-α), which increases as the illness progresses (Kauer-Sant’Anna et al., 2009). TNF-α soluble receptors— sTNF-R1 and sTNF-R2— have also been linked to the pathophysiology of BD (Siwek et al., 2017b). A recent meta-analysis found that levels of TNF-α are significantly increased in mania and bipolar depression but not in euthymic periods (Rowland and Marwaha, 2018). Another study suggests that TNF-α levels may drop after an acute episode is treated, whereas sTNF-R1 and sTNF-R2 levels may be persistently elevated in the euthymic period (Doganavsargil-Baysal et al., 2013). Importantly, an inverse association between TNF family molecules and cognition in BD has been suggested (Hope et al., 2015a; Hoseth et al., 2016). TNF family molecules are found at higher concentrations in later stages of BD, which are marked by persistent cognitive impairment in euthymia (Kauer-Sant’Anna et al., 2009).

There remains a gap in our understanding as to whether and how TNF family molecules propagate cognitive deterioration in BD. Few studies have directly tested the mediating properties of the TNF family of molecules on the neuroprogressive pathway in BD. To directly test this question, we developed a structural equation model (SEM) to examine whether TNF family molecules exhibit a mediating effect on prior severe mood episodes and cognitive outcomes in BD.

2. Methods

2.1. Participants

Participants with BD-I and BD-II based on DSM-IV criteria (n=219, ages 18–65) were euthymic and recruited as outpatients from the Icahn School of Medicine at Mount Sinai Hospital (ISMMS). Criteria for exclusion for the BD participants included the inability to consent, current affective instability indicated by a score of >8 on the Hamilton Depression Rating Scale (HDRS)(Hamilton, 1960), and/or on the Young Mania Rating Scale (YMRS)(Young et al., 1978). Further exclusionary criteria included current diagnosis of mild cognitive impairment (MCI) or dementia, a history of CNS trauma, an active and unstable medical condition that interferes with cognition, neurological disorder, known learning disability or childhood onset ADHD, substance abuse or dependence within 3 months of study visit, medications that are known to reduce (i.e., topiramate, tricyclics, anticholinergics), or enhance cognition (e.g., dopamine agonists or amphetamine), and electroconvulsive therapy (ECT) in the last year.

The Healthy Control (HC) participants (n=52, ages 20–70) were recruited through advertisements. Exclusionary criteria were the same as those for the BD group, with the exception that the presence of an Axis-I disorder was exclusionary. The study was approved by the Institutional Review Board (IRB) at ISMMS, and all participants gave written informed consent prior to participating in the study.

Approximately 386 individuals were screened for the BD group. From this group, 149 were excluded for various reasons (consented but deemed ineligible (n=66), ‘no show’ for the study appointment (n=75), did not complete the study (n=6) or had compromised data (n=4)). Of the remaining participants, 224 had available cytokine data. 5 individuals were further excluded due to key missing clinical or neurocognitive data (final n=219).

2.2. Clinical Measures

The BD and HC participants were assessed for diagnostic eligibility using the Structured Clinical Interview for the DSM-IV (SCID-IV). Depressive and manic mood states were assessed with the HDRS and YMRS, respectively, and were administered within one week of neurocognitive testing. For those with a BD diagnosis, a trained staff member assessed and recorded illness features (derived from the SCID), including age at onset, number and type of prior episodes and prior psychiatric hospitalizations, BD and psychotic subtype, rapid cycling and comorbid axis I diagnoses (current and lifetime).

2.3. Neurocognitive Measures

We administered tests to assess executive function including the Controlled Oral Word Association Task (COWAT)(Benton et al., 1994), the Stroop color and word test (Golden, 1978) and the Wisconsin Card Sorting Task (WCST)(Grant and Berg, 1948). Z-scores were calculated for cognitive data in the BD group (normalized using the HC group), for subsequent analyses. We also assessed neurocognition for all participants using the MATRICS Consensus Cognitive Battery (MCCB) (Nuechterlein et al., 2008). All neurocognitive scores from the MCCB were standardized on a T scale score (mean=50, SD=10), and age and sex adjusted using published MCCB normative data.

2.4. Blood Biomarker Analysis

Samples were drawn via venous blood draw, in a serum separator tube by a trained staff member. Samples were stored at −80° C until used. Concentrations of TNFα, sTNF-R1 and sTNF-R2 were quantified at the Genital Tract Biology Laboratory at Brigham and Women’s Hospital (under accreditation by the College of American Pathologists) using an electrochemiluminescence (ECL) multiplex-based assay platform with a S600 Meso Scale Discovery (MSD) reader (MSD, Gaithersburg, MD). The ECL platform is highly sensitive with a 5-log scale of linearity, with lowest level of detection (LLD) for TNF 3.96 pg/ml, sTNF-R1 2.6 pg/ml and sTNF-R2 – 0.226 pg/ml. Samples were run undiluted for TNFα, and diluted to fit the linearity range of the soluble receptors – 4-fold for the sTNF-R1 and 10-fold and 50-fold for the sTNF-R2 assay. Two split quality control samples were run on each ECL assay plate to ascertain reproducibility of measurement across the linearity range of each analyte showing acceptable ≤10 % inter-assay coefficient of variation (4.59% and 8.69% for TNFα, 9.3 and 10% for sTNF-R1 and 9.16% and 9.99% for sTNF-R2).

2.5. Statistical Analysis

All analyses were performed using the statistical program SPSS (IBM SPSS Statistics version 24, 2016), SPSS Amos Graphics (version 26, 2019, IBM Corp), and R Studio (Version 1.1.442, 2008–2018). Normality of data was assessed using the Kolmogorov-Smirnov test. Demographic differences between groups were assessed using Student’s t tests and Chi-square tests for continuous and categorical variables, respectively. For non-normally distributed cytokine data, we used the non-parametric Mann-Whitney U test to compare groups. The Box-Cox power transformation approach was used to transform biomarker concentrations into normal distributions for structural equation modeling analyses (Box and Cox, 1964). The Box-Cox lambda for each TNF variable was determined in R software, and variable transformation was performed in SPSS software. For TNF-α, transformation did not result in normalization on the first try. Based on this, one extreme outlier was removed. Subsequent transformation successfully resulted in normality for each marker; Kolmogorov-Smirnov test: TNF-α, D(219)=0.06, p=0.06; sTNF-R1, D(219)=0.05, p=0.2; sTNF-R2, D(219)=0.04, p=0.2. We used the Box-Cox transformed concentrations in subsequent analyses.

2.5.1. Structural Equation Model (SEM)–

The SEM was performed using SPSS Amos Graphics (version 26, 2019, IBM Corp). The SEM was used for the primary purpose of assessing whether TNF markers mediate the effect of clinical variables related to neuroprogression on executive function in BD (Bollen, 1989). Our final model was informed by an initial non-significant model which tested for the mediation effect of TNF variables in a simplistic way (using single variables) between clinical variables of disease progression (total number of mood episodes and prior psychiatric hospitalizations) and executive function (initial evaluation of correlations among variables partly suggested the initial model). Based on the lack of fit of our initial model, changes were made based on statistical criteria (“modification indices”) and substantive reasonableness.

A subsequent SEM with both latent and manifest variables was then tested using maximum likelihood estimation. P- values were obtained through bootstrapping. The latent variables are shown in ovals and the observed variables are in rectangles in Figure 3. A latent variable, “Peripheral TNF” measured by the multiple indicators TNF-α, sTNF-R1 and sTNF-R2, was hypothesized to mediate the effect of number of prior psychiatric hospitalizations on the latent variable “Executive Function” assessed by a set of observed variables, namely, the neurocognitive tests: COWAT, WCST and Stroop. A separate direct effect of prior psychiatric hospitalization on executive function, apart from the mediation of TNF, was also allowed for in the model. We included five exogenous covariates in the model, namely, sex, age, education, race and premorbid IQ. Incorporation of correlations among exogenous covariates into the model was determined by their statistical significance. Non-significant path coefficients were removed from the model in an iterative process. One non-significant association was retained in the model, between race and executive function, because it reached trend level significance (p=0.06). Statistical identification of latent variables was achieved by setting the path coefficient for one of the multiple indicators of each of the latent variables to one (see Figure 3).

Figure 3.

Figure 3.

Open in a new tab

Structural Equation Model (SEM) showing standardized path coefficient estimates and correlations among exogenous covariates. Latent variables are shown as circles and manifest variables are shown as rectangles. Covariates (education, sex, age, premorbid IQ, race) are in grey. *=p<0.05, **=p<0.01, ***=p<0.001, ^p=.06.

2.5.2. Assessment of Model Fit –

Goodness of fit of the path model was assessed by examining the Chi-square test (which assesses the fit between the empirical and predicted correlation matrices), and the Root Mean Square of Error Estimation (RMSEA) and PCLOSE measure. A good model was defined by a non-significant p value in the Chi-square test (p>0.05), and a RMSEA value of <0.06.

3. Results

3.1. Clinical and demographic data –

Table 1 shows the clinical and demographic data for the participants with BD. The sample was almost half female (47%) and Caucasian (47%), and mostly non-Hispanic (~77%). On average BD participants had 14 years of education (equivalent to some college) and were overweight (BMI >25). Most participants were taking at least one psychotropic medication (~78%); the average number of meds being taken was 1.7 (±1.3). The average participant had been ill for about 22.0 (±11.5) years and had experienced about 24.0 (±21.2) mood episodes in their lifetime.

Table 1.

Clinical and demographic variables of BD and HC groups. Data represented as Mean (±SD) or N (%) unless otherwise noted.

VARIABLES BIPOLAR (N=219) HC (N=52) p- value
TNF-α (pg/mL)* 3.3 (IQR=1.4) 3.4 (IQR=1.5) 0.91
sTNF-R1 (pg/mL)* 3825.1 (IQR=1461.1)  3806.8 (IQR=1216.4) 0.88
sTNF-R2 (pg/mL)* 4224.9 (IQR=1739.5)  4185.3 (IQR=1842.0) 0.42
Age 43.6 (±11.9) 38.9 (13.4) 0.02
BMI N=37, 29.9 (±6.3) N=24, 26.1 (±5.9) 0.02
Education 14.2 (±2.5) 15.6 (±1.7) <0.001
Sex (female) 103 (47.0%) 30 (57.7%) 0.17
Ethnicity (Hispanic) 50 (22.8%) 10 (19.2%) 0.57
Race (Caucasian) 101 (46.1%) 15 (28.8%) <0.001
Premorbid IQ (WRAT-3) 100.7 (±15.3) 107.1 (±11.4) <0.01
BD type I 176 (80.4%) -- --
Not on medication 49 (22.4%) -- --
Duration of illness (years) 21.9 (±11.5) -- --
Manic episodes N=214, 7.7 (±9.6) -- --
Depressive episodes N=214, 13.4 (±14.2) -- --
# Mood episodes N=213, 24.1 (±21.2) -- --
Psychosis history 93 (42.5%) -- --
 Wks since mood episode N=217, 57.4 (±110) -- --
# Psych hosps 3.7 (±4.5) -- --
# Suicide attempts 0.9 (±2.2) -- --
# Psychotropic meds 1.7 (±1.3) -- --
Open in a new tab
*

data shown as median and interquartile range (IQR).

In a series of Mann Whitney U Tests, we found no significant main effect of diagnosis (BD and HC) in TNF family molecule concentrations (reported as mean and standard deviation): TNF-α (HC: 3.6 ±1.7 pg/mL; BD: 3.6 ±1.6 pg/mL; p=0.9), sTNF-R1 (HC: 3991.0 ±1193.9 pg/mL; BD: 4174.8 ±2406.0 pg/mL; p=0.9), and sTNF-R2 (HC: 4540.9 ± 2092.3 pg/mL; BD: 4796.2 ±2414.0 pg/mL; p=0.4).

3.2. Peripheral TNF in early and late stage BD–

Peripheral TNF concentrations (TNF-α, sTNF-R1, and sTNF-R2) were compared between early vs. late-stage BD (Figure 1). We used psychiatric hospitalizations as a proxy for “severe mood episodes” for three reasons: 1) inpatient hospitalizations for affective mood symptoms are markers for a more severe course of illness and have been shown to predict cognitive function in BD (Levy and Weiss, 2009); 2) Our SCID-based interview is inherently retrospective, the analysis of “total mood episodes” is limited based on patients’ memories. We propose that hospitalizations are a more reliable measure than total mood episodes alone; 3) An episode for which a patient is hospitalized is inherently severe in nature and may therefore contribute more directly to cognitive decline (and therefore be a more reliable estimate of prior course/stage).

Figure 1.

Figure 1.

Open in a new tab

Non-parametric Mann Whitney U tests on early versus late TNF family molecules. Early stage illness is defined as ≤1psychiatric hospitalizations and late stage illness is defined as >1 psychiatric hospitalization. *=p<0.05.

Early stage BD was defined as having one or fewer prior psychiatric hospitalizations for either mania or depression; late stage BD was defined as having greater than one prior severe mood episodes. It should be noted that validation of staging for people with BD is still a work in progress, and there is no single accepted clinical cutoff to define early or late stage of illness (F Kapczinski et al., 2014). However, we chose this cutoff based on prior literature which indicated that early or late stage of illness can be defined based on the presence of psychiatric hospitalizations (Cao et al., 2016; Lavagnino et al., 2015; Magalhães et al., 2012). Early and late stage groups did not differ by sex: Chi-square (1, 219)=2.6 p=0.1, race (dichotomized): Chi-square (1, 219) =0.35, p=0.6, education: t(217)=1.5 p=0.1, age: t(217)=−1.2, p=0.2, or BMI: t(35)=−1.7, p=0.09. The groups did not differ on duration of illness, t(217)=−1.5, p=0.1, nor number of psychotropic medications t(217)=−1.8, p=0.06. A series of non-parametric Mann Whitney U tests revealed that sTNF-R1 was significantly higher in late-stage BD (N= 130, 4440.0±2967.1 pg/mL) relative to early-stage BD (N=89, 3787.5±1086.1 pg/mL), p=0.047. We did not find a significant difference between early and late stage groups for TNF-α (p=0.5) or sTNF-R2 (p=0.09).

3.3a. Partial correlations between peripheral TNF variables and cognitive tests –

A series of partial correlational analyses (controlled for age and sex), performed prior to the design of the SEM, revealed significant negative associations between TNF-α and executive function measures, including Stroop (Pearson r=−0.2, p=0.005) and WCST (Pearson r=−0.2, p=0.03). sTNF-R2 was also significantly associated with Stroop (Pearson r=−0.2, p=0.005) (Figure 2). The relationship seemed to be specific to executive functioning, as we did not find a significant correlation between TNF variables and the MCCB composite: TNF-α (r=−0.006, p=0.9), sTNF-R1(r=0.1, p=0.2), sTNF-R2 (r=−0.03, p=0.6).

Figure 2.

Figure 2.

Open in a new tab

Partial Pearson correlations between TNF-family molecules and cognitive assessments of executive function (controlled for age and sex). COWAT (Controlled Oral Word Association Test), WCST (Wisconsin Card Sorting Task), Stroop color-word. *=p<0.05, **=p<0.01, #p=.13

3.3b. Relationship between TNF family molecule, psychiatric hospitalizations, and cognitive function –

The results of the SEM with estimated path coefficients (standardized partial regression coefficients) are shown in Figure 3. The Chi-Square value of 49.2 had a non-significant p-value of 0.3, indicating acceptable model fit. Further, the Root Mean Squared Error Approximation (RMSEA) of 0.02 and PCLOSE of .9 further supported that the model had a ‘close fit’ (Browne and Cudeck, 1989).

Holding covariates constant (age, sex, premorbid IQ, education, and race), the direct effect (c’ path) of prior psychiatric hospitalizations on executive function (EF) is −0.14 (standardized path coefficient), and the direct effect of soluble TNF receptor levels on EF in this model is −0.19 (b path). The indirect effect (a*b paths) of prior psychiatric hospitalizations on EF (using the tracing rule) mediated by TNF is (0.15 X −0.19) = −0.03; thus, the total effect (c path) of hospitalizations on EF is (−0.14 + (−0.03)) = −0.17. The path coefficients above are individually significant per the path diagram, and thus by inference the direct (c’) effect [std β =−0.138 (90%CI [−.26, −.03], p=0.025)], indirect (a*b) effect [std β =−0.028 (90%CI [−.07, −.01], p=0.015)] and total (c) effect [std β =−0.166 (90%CI [−.28, −.07], p=0.01) are each statistically significant. Thus, this estimated model is consistent with peripheral TNF variables partially (though not completely) mediating a causal effect of psychiatric hospitalizations on executive cognitive function.

4. Discussion

Here we have shown that although TNF family molecules (TNF-α, sTNF-R1, sTNF-R2) are not elevated in euthymic BD compared to HCs, sTNF-R1 is elevated in late-stage BD patients, and these molecules mediate the relationship between the number of severe mood episodes and cognition in BD, with executive function being particularly affected. These results suggest that TNF family molecules may play an important role in the pathophysiology of neuroprogression in BD.

We did not find group differences in soluble TNF family molecules between BDs and HCs. TNF family molecules have previously been reported to be altered in euthymic BD, yet the literature is mixed (Barbosa et al., 2014). Results from several studies suggests that sTNF-R1 is elevated in euthymic BD groups compared to HCs (Modabbernia et al., 2013; Munkholm et al., 2013) with one study also noting elevated sTNF-R2 (Teixeira et al., 2015). Notably, two recent meta-analyses reported either no significantly elevation of TNF-α in euthymic BD compared to HCs, or a “normalization” of TNF-α in the euthymic phase of illness. Although meta-analyses indicate an overall increase in sTNF-R1 in euthymic BD, a few individual findings are consistent with our results, reporting no significant difference between BD and HC groups in TNF family molecule concentrations.

Few studies have statistically tested for mediating properties of TNF in the neuroprogressive pathway using outcomes such as cognition as markers of progression; however, some have shown that TNF family molecules represent independent markers of illness stage in BD. A few studies have found sTNF-Rs to be positively associated with duration of illness (Barbosa et al., 2011) and number of psychiatric hospitalizations in BD groups (Hope et al., 2013), in line with our present findings. A study comparing levels of TNF-α in 30 early-stage BD (recent first manic episode) and 30 late-stage BD patients (minimum of 10 years after diagnosis), found significantly higher TNF-α in the late-stage group (Kauer-Sant’Anna et al., 2009). Another study of inflammatory biomarkers in 133 BD patients found increased sTNF-R (80kDa) in later-stage illness (Siwek et al., 2017b) suggesting it may represent a staging biomarker. Of note, these later stages of illness are characterized by persistent cognitive impairment in the inter-episode period (Kapczinski et al., 2009).

Our work fits into a burgeoning body of literature suggesting TNF-family molecules are associated with cognitive dysfunction in major psychiatric illnesses; however, results are mixed. In one study in non-euthymic BD (N=117) and patients with schizophrenia (SZ N=109), sTNF-R1 was negatively associated with performance on verbal memory and working memory learning and recall tests (Hoseth et al., 2016), a domain that has been shown to be affected by executive function (Chang et al., 2010). In another study, in which levels of sTNF-R1 were higher in non-euthymic patients (BD, N=111, SZ, N=121) than controls, sTNF-R1 was observed to be negatively associated with general cognitive functioning based on the WASI IQ test (Hope et al., 2015b). However, in subgroup analyses, sTNF-R1 was only associated with general cognitive ability in the SZ group, and the authors did not observe a significant association in the BD group, in contrast to our findings. Another group found that TNF-α was negatively correlated with delayed recall in 54 euthymic BD patients (Doganavsargil-Baysal et al., 2013). In contrast, one study of euthymic BD patients (N=25) found that TNF-α was positively correlated with inhibitory control, and sTNF-R2 levels were inversely correlated with motor programming (Guimarães Barbosa et al., 2012). It is worth noting that our sample of BD participants is twice as large as the largest study reporting these relationships to date, thereby allowing for enhanced power to detect these relationships.

Our results suggest that executive function is particularly influenced by TNF molecules, which is consistent with reports implicating TNF in BD-related changes in the lateral and medial prefrontal cortex (PFC). The relationship between TNF family molecules and cognitive performance observed in our study and others is in keeping with findings that there are regional variations in mood disorder patients in this family of molecules in the frontal cortex (Dean et al, 2013), the volume and thickness of which has been associated with executive functioning (Yuan & Raz, 2014). Further, in vivo studies in mood disorder patients have shown that increased peripheral TNF levels correlate with the presence of white matter hyperintensities, which are in turn associated with executive function impairment (Smagula et al., 2017). TNF levels also impact functional connectivity within the PFC in MDD (Liu et al., 2019).

The TNF family system, reflected in literature examining transcript levels (mRNA) and soluble and membrane-bound protein forms of TNF and its receptors, is a highly complex interplay of ligand and receptor molecules. Soluble TNF-Rs compete with the transmembrane receptors for TNF-α, and regulate TNF-α signaling (Aderka et al., 1992; Moelants et al., 2013). The manner in which sTNF receptors affect TNF activity depends on their concentration as well as on the concentration of TNF itself, such that these soluble receptors can either inhibit or augment the effects of TNF (Aderka et al, 1992). As such, elevation in sTNF-Rs may be insufficient to completely neutralize TNF-α activity in sites where transmembrane TNF-Rs are markedly increased (Cope et al., 1992). In diseases such as rheumatoid arthritis (RA) there is a correlation between elevated sTNF-R concentrations and clinical indicators of inflammatory disease activity (Cope et al., 1992). Further, levels of sTNF-Rs have been found to be indicative of disease activity in other inflammatory conditions, including Guillain-Barré syndrome (Créange et al., 1996) and juvenile chronic arthritis (Rooney et al., 2000). The story of this system is complicated further by tissue and cell type examined. Future work should aim to develop conceptual models which integrate this complex network of molecules.

There were a few limitations to the current study. One is the potential confounding factors that may impact TNF-family molecule levels which we did not measure in our participants, including level of physical activity, diet, and infections or low-grade inflammatory conditions that were not identified. TNF-family molecules are altered in various chronic health conditions such as cardiovascular disease (Ferrari, 1999) and autoimmune disorders like rheumatoid arthritis (Cope et al., 1992) and multiple sclerosis; however, beyond patient self-report of medical conditions we did not collect this information. Further, TNF-family molecules are known to fluctuate based on the time of day (Keller et al., 2009) and we did not standardize collection times nor systematically collect information related to the time of blood collection.

Based on current evidence, it stands to reason that chronic elevation in pro-inflammatory pathways may affect cognitive function in some people. TNF-α may directly affect the brain at the level of the blood-brain- barrier (BBB) where it binds to its receptors (TNF-R1 and TNF-R2) and is translocated to the basal side of the endothelium (Pan and Kastin, 2007). Subsequent pro-inflammatory cascades in the parenchyma may result in activation of resident immune cells in the brain, including microglia, which can also secrete TNF-α (Giridharan et al., 2019). It has been proposed that disruptions in the BBB in BD render the central nervous system more vulnerable to the heightened pro-inflammatory profile observed in the periphery (Patel and Frey, 2015). Further, TNF-α is known to activate endothelium, cause white blood cell recruitment and vascular leakage (Brietzke and Kapczinski, 2008); It can also activate apoptotic pathways, leading to cell death (Brietzke and Kapczinski, 2008). Further, TNF-α is expressed by macrophages, endothelial cells, neurons, astrocytes and microglia (Giridharan et al., 2019). In the brain, TNF-α can influence synaptic transport, synaptic pruning, neurogenesis and long-term potentiation (Giridharan et al., 2019).

TNF-α has already been identified as a target for clinical intervention in BD. Early trials have begun to use anti-inflammatory agents targeting TNF (infliximab) to treat aspects of BD (depression) (McIntyre et al., 2019; C. L. Raison et al., 2013). Infliximab reduces circulating TNF-α by mimicking sTNF-Rs, and is a medication typically used for treatment of RA (Moss et al., 2008). Early evidence suggests that the effectiveness of anti-inflammatory treatments is greatest within a subgroup of BD patients, i.e., those with an “inflammatory” illness type versus those without (Fillman et al., 2014). High baseline levels of CRP (>3–5mg/L) have been shown to predict antidepressant response to both lurasidone and infliximab in mood disorders (Raison et al., 2013; Raison et al., 2018). These results point toward inflammation as a primary treatment target, at least in a subgroup of BD patients. Our data suggests that if anti-TNFα treatment is to be tested in BD, measuring changes in cognition during the course of treatment may be useful to indicate responder/non-responder status.

Here we have shown evidence that 1) circulating TNF levels do not differ between BD and HC groups in our sample, 2) late-stage BD is marked by increased sTNF-R1 levels, 3) TNF-α and sTNF-R2 are negatively associated with executive function in BD, and 4) TNF family molecules partially mediate the relationship between number of severe mood episodes and executive function in euthymic BD. These results suggest a role for TNF family molecules in the pathophysiology of neuroprogression in affectively stable patients with BD. The partiality of mediation may point to a subgroup of patients (those with higher peripheral inflammation) who may benefit from an anti-inflammatory adjunctive agent during severe mood episodes in an effort to slow or abort cognitive decline. This idea is supported by a recent meta-analysis which indicated that only a subset (27%) of patients present with low-grade inflammation in depression (Osimo et al., 2019). Our prior work also supports the idea of “inflammatory heterogeneity” in BD (as described in our recent work Millett et al., 2019). Although SEM models can be useful to test for a (statistically) causal relationship among key measures; our data are cross-sectional. A true causal relationship cannot be established until longitudinal studies are done to address the exact nature and timing of these brain changes in BD.

  • We found no significant difference between BD and HC groups in circulating TNF family molecule concentrations.

  • Circulating soluble TNF receptor one (sTNF-R1) was significantly higher in late-stage BD compared to early-stage illness.

  • We found significant negative associations between TNF family molecules and executive function measures in BD.

  • TNF variables partially (though not completely) mediate a causal effect of psychiatric hospitalizations on executive cognitive function in BD.

Funding Sources:

The work is funded by R01MH100125 to KEB. CEM is partially supported through the Harvard University Sackler Scholars in Psychobiology Program.

Footnotes

Disclosures:

Dr. Burdick has served as an advisory board member to Dainippon Sumitomo Pharmaceutical Neuralstem, and Takeda-Lundbeck. All other authors have nothing to disclose.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

REFERENCES

  1. Aderka D, Engelmann H, Maor Y, Brakebusch C, Wallach D, 1992. Stabilization of the bioactivity of tumor necrosis factor by its soluble receptors. J. Exp. Med. 175, 323–329. 10.1084/jem.175.2.323 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Barbosa IG, Bauer ME, Machado-Vieira R, Teixeira AL, 2014. Cytokines in Bipolar Disorder: Paving the Way for Neuroprogression. Neural Plast. 2014. 10.1155/2014/360481 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Barbosa IG, Huguet RB, Mendonça VA, Sousa LP, Neves FS, Bauer ME, Teixeira AL, 2011. Increased plasma levels of soluble TNF receptor I in patients with bipolar disorder. Eur. Arch. Psychiatry Clin. Neurosci. 261, 139–143. 10.1007/s00406-010-0116-z [DOI] [PubMed] [Google Scholar]
  4. Benton AL, Hamsher K, Rey GL, Sivan AB, 1994. Multilingual Aphasia Examination. AJA Associates, Iowa City, IA. [Google Scholar]
  5. Berk M, Kapczinski F, Andreazza AC, Dean OM, Giorlando F, Maes M, Yücel M, Gama CS, Dodd S, Dean B, Magalhães PV, Amminger P, McGorry P, Malhi GS, 2011. Pathways underlying neuroprogression in bipolar disorder: focus on inflammation, oxidative stress and neurotrophic factors. Neurosci Biobehav Rev 35, 804–817. 10.1016/j.neubiorev.2010.10.001 [DOI] [PubMed] [Google Scholar]
  6. Bollen KA, 1989. Structural Equations with Latent Variables. John Wiley & Sons, NY, NY. [Google Scholar]
  7. Bourne C, Aydemir Ö, Balanzá-Martínez V, Bora E, Brissos S, Cavanagh JT, Clark L, Cubukcuoglu Z, Dias VV, Dittmann S, Ferrier IN, Fleck DE, Frangou S, Gallagher P, Jones L, Kieseppä T, Martínez-Aran A, Melle I, Moore PB, Mur M, Pfennig A, Raust A, Senturk V, Simonsen C, Smith DJ, Bio DS, Soeiro-de-Souza MG, Stoddart SD, Sundet K, Szöke A, Thompson JM, Torrent C, Zalla T, Craddock N, Andreassen OA, Leboyer M, Vieta E, Bauer M, Worhunsky PD, Tzagarakis C, Rogers RD, Geddes JR, Goodwin GM, 2013. Neuropsychological testing of cognitive impairment in euthymic bipolar disorder: an individual patient data meta-analysis. Acta Psychiatr Scand 128, 149–162. 10.1111/acps.12133 [DOI] [PubMed] [Google Scholar]
  8. Box GEP, Cox DR, 1964. An Analysis of Transformations, Journal of the Royal Statistical Society. Series B (Methodological). [Google Scholar]
  9. Brietzke E, Kapczinski F, 2008. TNF-α as a molecular target in bipolar disorder. 10.1016/j.pnpbp.2008.01.006 [DOI] [PubMed]
  10. Browne MW, Cudeck R, 1989. Single Sample Cross-Validation Indices for Covariance Structures. Multivariate Behav. Res. 24, 445–455. 10.1207/s15327906mbr2404_4 [DOI] [PubMed] [Google Scholar]
  11. Burdick KE, Goldberg TE, Cornblatt BA, Keefe RS, Gopin CB, Derosse P, Braga RJ, Malhotra AK, 2011. The MATRICS consensus cognitive battery in patients with bipolar I disorder. Neuropsychopharmacology 36, 1587–1592. 10.1038/npp.2011.36 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Cao B, Bauer IE, Sharma AN, Mwangi B, Frazier T, Lavagnino L, Zunta-Soares GB, Walss-Bass C, Glahn DC, Kapczinski F, Nielsen DA, Soares JC, 2016. Reduced hippocampus volume and memory performance in bipolar disorder patients carrying the BDNF val66met met allele. J. Affect. Disord. 198, 198–205. 10.1016/j.jad.2016.03.044 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Chang YL, Jacobson MW, Fennema-Notestine C, Hagler DJ, Jennings RG, Dale AM, McEvoy LK, 2010. Level of executive function influences verbal memory in amnestic mild cognitive impairment and predicts prefrontal and posterior cingulate thickness. Cereb. Cortex 20, 1305–1313. 10.1093/cercor/bhp192 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Cope AP, Gibbons D, Brennan FM, Feldmann M, Aderka D, Doherty M, Jones AC, Engelmann H, Wallach D, Maini RN, 1992. Increased levels of soluble tumor necrosis factor receptors in the sera and synovial fluid of patients with rheumatic diseases. Arthritis Rheum. 35, 1160–1169. 10.1002/art.1780351008 [DOI] [PubMed] [Google Scholar]
  15. Créange A, Bélec L, Clair B, Raphaël JC, Gherardi RK, 1996. Circulating tumor necrosis factor (TNF)-α and soluble TNF-α receptors in patients with Guillain-Barre syndrome. J. Neuroimmunol. 68, 95–99. 10.1016/0165-5728(96)00075-6 [DOI] [PubMed] [Google Scholar]
  16. Depp AC, Dev S, Eyler TL, 2016. Bipolar Depression and Cognitive Impairment: Shared Mechanisms and New Treatment Avenues. Psychiatr. Clin. North Am. 39, 95–109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Doganavsargil-Baysal O, Cinemre B, Aksoy UM, Akbas H, Metin O, Fettahoglu C, Gokmen Z, Davran F, 2013. Levels of TNF-α, soluble TNF receptors (sTNFR1, sTNFR2), and cognition in bipolar disorder. Hum. Psychopharmacol. Clin. Exp. 28, 160–167. 10.1002/hup.2301 [DOI] [PubMed] [Google Scholar]
  18. Ferrari AJ, Charlson FJ, Norman RE, Patten SB, Freedman G, Murray CJ, Vos T, Whiteford HA, 2013. Burden of depressive disorders by country, sex, age, and year: findings from the global burden of disease study 2010. PLoS Med 10, e1001547. 10.1371/journal.pmed.1001547 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Ferrari R, 1999. The role of TNF in cardiovascular disease. Pharmacol. Res. 40, 97–105. 10.1006/phrs.1998.0463 [DOI] [PubMed] [Google Scholar]
  20. Fillman SG, Sinclair D, Fung SJ, Webster MJ, Weickert S, 2014. Markers of inflammation and stress distinguish subsets of individuals with schizophrenia and bipolar disorder. Transl. Psychiatry 4, 365. 10.1038/tp.2014.8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Giridharan VV, Sayana P, Pinjari OF, Ahmad N, da Rosa MI, Quevedo J, Barichello T, 2019. Postmortem evidence of brain inflammatory markers in bipolar disorder: a systematic review. Mol. Psychiatry. 10.1038/s41380-019-0448-7 [DOI] [PubMed] [Google Scholar]
  22. Golden CJ, 1978. Stroop Color and Word Test: a manual for clinical and experimental uses. Stoelting, Co., Chicago, IL. [Google Scholar]
  23. Grant D, Berg E, 1948. A behavioral analysis of degree of reinforcement and ease of shifting to new responses in a Weigl-type card-sorting problem. J Exp Psychol 38, 404–411. [DOI] [PubMed] [Google Scholar]
  24. Guimarães Barbosa I, Rocha NP, Barreto Huguet R, Ferreira RA, Salgado V, Carvalho LA, Pariante CM, Lúcio Teixeira A, 2012. Executive dysfunction in euthymic bipolar disorder patients and its association with plasma biomarkers. 10.1016/j.jad.2011.12.034 [DOI] [PubMed]
  25. Hamilton M, 1960. A rating scale for depression. J Neurol Neurosurg Psychiatry 23, 56–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Hope S, Hoseth E, Dieset I, Mørch RH, Aas M, Aukrust P, Djurovic S, Melle I, Ueland Torill, Agartz I, Ueland Thor, Westlye LT, Andreassen OA, 2015a. Inflammatory markers are associated with general cognitive abilities in schizophrenia and bipolar disorder patients and healthy controls. Schizophr. Res. 165, 188–194. 10.1016/j.schres.2015.04.004 [DOI] [PubMed] [Google Scholar]
  27. Hope S, Hoseth E, Dieset I, Mrch HR, Aas M, Aukrust P, Djurovic S, Melle I, Ueland Torill, Agartz I, Ueland Thor, Westlye TL, Andreassen AO, 2015b. Inflammatory markers are associated with general cognitive abilities in schizophrenia and bipolar disorder patients and healthy controls. Schizophr. Res. 165, 188–194. [DOI] [PubMed] [Google Scholar]
  28. Hope S, Ueland T, Steen NE, Dieset I, Lorentzen S, Berg AO, Agartz I, Aukrust P, Andreassen OA, 2013. Interleukin 1 receptor antagonist and soluble tumor necrosis factor receptor 1 are associated with general severity and psychotic symptoms in schizophrenia and bipolar disorder. Schizophr. Res. 145, 36–42. 10.1016/j.schres.2012.12.023 [DOI] [PubMed] [Google Scholar]
  29. Hoseth EZ, Westlye LT, Hope S, Dieset I, Aukrust P, Melle I, Haukvik UK, Agartz I, Ueland T, Ueland T, Andreassen OA, 2016. Association between cytokine levels, verbal memory and hippocampus volume in psychotic disorders and healthy controls. Acta Psychiatr. Scand. 133, 53–62. 10.1111/acps.12467 [DOI] [PubMed] [Google Scholar]
  30. Kapczinski F, Dias VV, Kauer-Sant’Anna M, Brietzke E, Vázquez GH, Vieta E, Berk M, 2009. The potential use of biomarkers as an adjunctive tool for staging bipolar disorder. Prog. Neuro-Psychopharmacology Biol. Psychiatry 33, 1366–1371. 10.1016/J.PNPBP.2009.07.027 [DOI] [PubMed] [Google Scholar]
  31. Kapczinski F, Magalhães PVS, Balanzá-Martinez V, Dias VV, Frangou S, Gama CS, Gonzalez-Pinto A, Grande I, Ha K, Kauer-Sant’Anna M, Kunz M, Kupka R, Leboyer M, Lopez-Jaramillo C, Post RM, Rybakowski JK, Scott J, Strejilevitch S, Tohen M, Vazquez G, Yatham L, Vieta E, Berk M, 2014a. Staging systems in bipolar disorder: an International Society for Bipolar Disorders Task Force Report. Acta Psychiatr. Scand. 130, 354–363. 10.1111/acps.12305 [DOI] [PubMed] [Google Scholar]
  32. Kapczinski F, Magalhães PVS, Balanzá-Martinez V, Dias VV, Frangou S, Gama CS, Gonzalez-Pinto A, Grande I, Ha K, Kauer-Sant’Anna M, Kunz M, Kupka R, Leboyer M, Lopez-Jaramillo C, Post RM, Rybakowski JK, Scott J, Strejilevitch S, Tohen M, Vazquez G, Yatham L, Vieta E, Berk M, 2014b. Staging systems in bipolar disorder: an International Society for Bipolar Disorders Task Force Report. Acta Psychiatr. Scand. 130, 354–363. 10.1111/acps.12305 [DOI] [PubMed] [Google Scholar]
  33. Kapczinski F, Magalhães PVS, Balanzá-Martinez V, Dias VV, Frangou S, Gama CS, Gonzalez-Pinto A, Grande I, Ha K, Kauer-Sant’Anna M, Kunz M, Kupka R, Leboyer M, Lopez-Jaramillo C, Post RM, Rybakowski JK, Scott J, Strejilevitch S, Tohen M, Vazquez G, Yatham L, Vieta E, Berk M, 2014. Staging systems in bipolar disorder: an International Society for Bipolar Disorders Task Force Report. Acta Psychiatr. Scand. 130, 354–63. 10.1111/acps.12305 [DOI] [PubMed] [Google Scholar]
  34. Kauer-Sant’Anna M, Kapczinski F, Andreazza AC, Bond DJ, Lam RW, Young LT, Yatham LN, 2009. Brain-derived neurotrophic factor and inflammatory markers in patients with early- vs. late-stage bipolar disorder. Int. J. Neuropsychopharmacol. 12, 447. 10.1017/S1461145708009310 [DOI] [PubMed] [Google Scholar]
  35. Keller M, Mazuch J, Abraham U, Eom GD, Herzog ED, Volk HD, Kramer A, Maier B, 2009. A circadian clock in macrophages controls inflammatory immune responses. Proc. Natl. Acad. Sci. U. S. A. 106, 21407–21412. 10.1073/pnas.0906361106 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Lavagnino L, Cao B, Mwangi B, Wu M-J, Sanches M, Zunta-Soares GB, Kapczinski F, Soares J, 2015. Changes in the corpus callosum in women with late-stage bipolar disorder. Acta Psychiatr. Scand. 131, 458–464. 10.1111/acps.12397 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Leboyer M, Soreca I, Scott J, Frye M, Henry C, Tamouza R, Kupfer DJ, 2012. Can bipolar disorder be viewed as a multi-system inflammatory disease? J Affect Disord 141, 1–10. 10.1016/j.jad.2011.12.049 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Levy B, Weiss RD, 2009. Cognitive functioning in bipolar and co-occurring substance use disorders: A missing piece of the puzzle. Harv. Rev. Psychiatry. 10.1080/10673220902979870 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Lewandowski KE, Cohen BM, Ongur D, 2011. Evolution of neuropsychological dysfunction during the course of schizophrenia and bipolar disorder. Psychol. Med. 41, 225–41. 10.1017/S0033291710001042 [DOI] [PubMed] [Google Scholar]
  40. Liu K, Zhao X, Lu X, Zhu X, Chen H, Wang M, Yan W, Jing L, Deng Y, Yu L, Wu H, Wen G, Sun X, Lv Z, 2019. Effect of selective serotonin reuptake inhibitor on prefrontal-striatal connectivity is dependent on the level of TNF-α in patients with major depressive disorder. Psychol. Med. 49, 2608–2616. 10.1017/S0033291718003616 [DOI] [PubMed] [Google Scholar]
  41. López-Villarreal A, Sánchez-Morla EM, Jiménez-López E, Martínez-Vizcaíno V, Aparicio AI, Mateo-Sotos J, Rodriguez-Jimenez R, Vieta E, Santos JL, 2019. Progression of the functional deficit in a group of patients with bipolar disorder: a cluster analysis based on longitudinal data. Eur. Arch. Psychiatry Clin. Neurosci. 10.1007/s00406-019-01050-9 [DOI] [PubMed] [Google Scholar]
  42. Magalhães PV, Dodd S, Nierenberg AA, Berk M, 2012. Cumulative morbidity and prognostic staging of illness in the systematic treatment enhancement program for bipolar disorder (STEP-BD). Aust. N. Z. J. Psychiatry 46, 1058–1067. 10.1177/0004867412460593 [DOI] [PubMed] [Google Scholar]
  43. Martinez-Aran A, Vieta E, Torrent C, Sanchez-Moreno J, Goikolea JM, Salamero M, Malhi GS, Gonzalez-Pinto A, Daban C, Alvarez-Grandi S, Fountoulakis K, Kaprinis G, Tabares-Seisdedos R, Ayuso-Mateos JL, 2007. Functional outcome in bipolar disorder: the role of clinical and cognitive factors. Bipolar Disord 9, 103–113. 10.1111/j.1399-5618.2007.00327.x [DOI] [PubMed] [Google Scholar]
  44. McIntyre RS, Subramaniapillai M, Lee Y, Pan Z, Carmona NE, Shekotikhina M, Rosenblat JD, Brietzke E, Soczynska JK, Cosgrove VE, Miller S, Fischer EG, Kramer NE, Dunlap K, Suppes T, Mansur RB, 2019. Efficacy of Adjunctive Infliximab vs Placebo in the Treatment of Adults with Bipolar I/II Depression: A Randomized Clinical Trial. JAMA Psychiatry 76, 783–790. 10.1001/jamapsychiatry.2019.0779 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Millett CE, Perez-Rodriguez M, Shanahan M, Larsen E, Yamamoto HS, Bukowski C, Fichorova R, Burdick KE, 2019. C-reactive protein is associated with cognitive performance in a large cohort of euthymic patients with bipolar disorder. Mol. Psychiatry. 10.1038/s41380-019-0591-1 [DOI] [PubMed] [Google Scholar]
  46. Modabbernia A, Taslimi S, Brietzke E, Ashrafi M, 2013. Cytokine Alterations in Bipolar Disorder: A Meta-Analysis of 30 Studies. Biol. Psychiatry 74, 15–25. 10.1016/j.biopsych.2013.01.007 [DOI] [PubMed] [Google Scholar]
  47. Moelants EA, Mortier A, Van Damme J, Proost P, 2013. Regulation of TNF-α with a focus on rheumatoid arthritis. Immunol. Cell Biol. 91, 393–401. 10.1038/icb.2013.15 [DOI] [PubMed] [Google Scholar]
  48. Moss ML, Sklair-Tavron L, Nudelman R, 2008. Drug insight: tumor necrosis factor-converting enzyme as a pharmaceutical target for rheumatoid arthritis. Nat. Clin. Pract. Rheumatol. 4, 300–9. 10.1038/ncprheum0797 [DOI] [PubMed] [Google Scholar]
  49. Munkholm K, Vinberg M, Vedel Kessing L, 2013. Cytokines in bipolar disorder: A systematic review and meta-analysis. J. Affect. Disord. 144, 16–27. 10.1016/J.JAD.2012.06.010 [DOI] [PubMed] [Google Scholar]
  50. Nuechterlein KH, Green MF, Kern RS, Baade LE, Barch DM, Cohen JD, Essock S, Fenton WS, Frese FJ, Gold JM, Goldberg T, Heaton RK, Keefe RS, Kraemer H, Mesholam-Gately R, Seidman LJ, Stover E, Weinberger DR, Young AS, Zalcman S, Marder SR, 2008. The MATRICS Consensus Cognitive Battery, part 1: test selection, reliability, and validity. Am J Psychiatry 165, 203–213. 10.1176/appi.ajp.2007.07010042 [DOI] [PubMed] [Google Scholar]
  51. Osimo EF, Baxter LJ, Lewis G, Jones PB, Khandaker GM, 2019. Prevalence of low-grade inflammation in depression: A systematic review and meta-Analysis of CRP levels. Psychol. Med. 10.1017/S0033291719001454 [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Pan W, Kastin AJ, 2007. Tumor necrosis factor and stroke: Role of the blood-brain barrier. Prog. Neurobiol. 10.1016/j.pneurobio.2007.07.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Panizzutti B, Gubert C, Schuh AL, Ferrari P, Bristot G, Fries GR, Massuda R, Walz J, Rocha NP, Berk M, Teixeira AL, Gama CS, 2015. Increased serum levels of eotaxin/CCL11 in late-stage patients with bipolar disorder: An accelerated aging biomarker? J. Affect. Disord. 182, 64–69. 10.1016/j.jad.2014.12.010 [DOI] [PubMed] [Google Scholar]
  54. Patel JP, Frey BN, 2015. Disruption in the blood-brain barrier: The missing link between brain and body inflammation in bipolar disorder? Neural Plast. 10.1155/2015/708306 [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Raison C, Rutherford RE, Woolwine BJ, Shuo C, Schettler P, Drake DF, Haroon E, Miller AH, 2013. A randomized controlled trial of the tumor necrosis factor antagonist infliximab for treatment-resistant depression: the role of baseline inflammatory biomarkers. JAMA psychiatry 70, 31–41. 10.1001/2013.jamapsychiatry.4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Raison CL, Pikalov A, Siu C, Tsai J, Koblan K, Loebel A, 2018. C-reactive protein and response to lurasidone in patients with bipolar depression. Brain. Behav. Immun. 73, 717–724. 10.1016/j.bbi.2018.08.009 [DOI] [PubMed] [Google Scholar]
  57. Raison CL, Rutherford RE, Woolwine BJ, Shuo C, Schettler P, Drake DF, Haroon E, Miller AH, 2013. A randomized controlled trial of the tumor necrosis factor antagonist infliximab for treatment-resistant depression: The role of baseline inflammatory biomarkers. Arch. Gen. Psychiatry 70, 31–41. 10.1001/2013.jamapsychiatry.4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Rooney M, Varsani H, Martin K, Lombard1 PR, Dayer1 J-M, Woo P, 2000. Tumour necrosis factor alpha and its soluble receptors in juvenile chronic arthritis, Rheumatology. [DOI] [PubMed] [Google Scholar]
  59. Rosa AR, Reinares M, Michalak EE, Bonnin CM, Sole B, Franco C, Comes M, Torrent C, Kapczinski F, Vieta E, 2010. Functional impairment and disability across mood states in bipolar disorder. Value Heal. 13, 984–988. 10.1111/j.1524-4733.2010.00768.x [DOI] [PubMed] [Google Scholar]
  60. Rowland TA, Marwaha S, 2018. Epidemiology and risk factors for bipolar disorder. Ther. Adv. Psychopharmacol. 8, 251–269. 10.1177/2045125318769235 [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Siwek M, Sowa-Kućma M, Styczeń K, Misztak P, Nowak RJ, Szewczyk B, Dudek D, Rybakowski JK, Nowak G, Maes M, 2017a. Associations of Serum Cytokine Receptor Levels with Melancholia, Staging of Illness, Depressive and Manic Phases, and Severity of Depression in Bipolar Disorder. Mol. Neurobiol. 54, 5883–5893. 10.1007/s12035-016-0124-8 [DOI] [PubMed] [Google Scholar]
  62. Siwek M, Sowa-Kućma M, Styczeń K, Misztak P, Nowak RJ, Szewczyk B, Dudek D, Rybakowski JK, Nowak G, Maes M, 2017b. Associations of Serum Cytokine Receptor Levels with Melancholia, Staging of Illness, Depressive and Manic Phases, and Severity of Depression in Bipolar Disorder. Mol. Neurobiol. 54, 5883–5893. 10.1007/s12035-016-0124-8 [DOI] [PubMed] [Google Scholar]
  63. Smagula SF, Lotrich FE, Aizenstein HJ, Diniz BS, Krystek J, Wu GF, Mulsant BH, Butters MA, Reynolds CF, Lenze EJ, 2017. Immunological biomarkers associated with brain structure and executive function in late-life depression: exploratory pilot study. Int. J. Geriatr. Psychiatry 32, 692–699. 10.1002/gps.4512 [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Tatay-Manteiga A, Balanzá-Martínez V, Bristot G, Tabarés-Seisdedos R, Kapczinski F, Cauli O, 2017. Clinical staging and serum cytokines in bipolar patients during euthymia. Prog. Neuro-Psychopharmacology Biol. Psychiatry 77, 194–201. 10.1016/j.pnpbp.2017.04.028 [DOI] [PubMed] [Google Scholar]
  65. Teixeira AL, de Sousa RT, de Souza RT, Zanetti MV, Brunoni AR, Busatto GF, Zarate CA, Gattaz WF, Machado-Vieira R, 2015. Increased plasma levels of soluble TNF receptors 1 and 2 in bipolar depression and impact of lithium treatment. Hum. Psychopharmacol. 30, 52–6. 10.1002/hup.2450 [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Torrent C, Martinez-Arn A, del Mar Bonnin C, Reinares M, Daban C, Sol B, Rosa RA, Tabars-Seisdedos R, Popovic D, Salamero M, Vieta E, 2012. Long-term outcome of cognitive impairment in bipolar disorder. J. Clin. Psychiatry 73, e899–e905. [DOI] [PubMed] [Google Scholar]
  67. Young RC, Biggs JT, Ziegler VE, Meyer DA, 1978. A rating scale for mania: reliability, validity and sensitivity. Br J Psychiatry 133, 429–435. [DOI] [PubMed] [Google Scholar]

RESOURCES