Increased soluble tumor necrosis factor-α receptors in patients with major depressive disorder


Rodrigo Grassi-Oliveira, MS, MD, PhD, Av. Ipiranga, 6681, prédio 11, sala 933, Porto Alegre, RS 90619-900, Brazil. Email:


Aim:  Several lines of evidence suggest that major depressive disorder is associated with an inflammatory status. Tumor necrosis factor-α has been investigated as a potential molecular target in mood disorders. Tumor necrosis factor-α exerts its activity through binding to specific cell membrane receptors named as TNFR1 and TNFR2. The aim of the present study was to investigate soluble plasma TNFR1 (sTNFR1) and TNFR2 levels (sTNFR2) in major depressive disorder patients.

Methods:  Female outpatients with major depressive disorder (n = 30) were compared with a healthy control group (n = 19). Severity of depressive symptoms was evaluated on Beck Depression Inventory; post-traumatic stress disorder (PTSD) symptoms were evaluated on PTSD Checklist–Civilian Version; and childhood abuse and neglect on the Childhood Trauma Questionnaire. Plasma tumor necrosis factor-α and its soluble receptors were measured by ELISA.

Results:  Patients had no changes in tumor necrosis factor-α concentrations but did have increased sTNFR1 (P < 0.001) and sTNFR2 (P < 0.001) levels compared to controls. Plasma level of sTNFR1 was positively predicted by age (B = 0.25, P = 0.05) and PTSD-like symptoms (B = 0.41, P = 0.002) and plasma levels of sTNFR2 by depression severity (B = 0.67, P < 0.001).

Conclusions:  Soluble tumor necrosis factor-α receptors could be reliable markers of inflammatory activity in major depression.

RECENT STUDIES REPORT that patients with major depressive disorder (MDD) show activated inflammatory status, with increased pro-inflammatory cytokines, acute-phase proteins, and increased expression of chemokines and adhesion molecules.1–3 Tumor necrosis factor-α (TNF-α) has been investigated as a potential molecular target in mood disorders,4 and clinical improvement was associated with decreased serum levels.5,6 In addition, specific depressive symptoms (e.g. suicidal behavior) were associated with increased cytokine levels.7

Tumor necrosis factor-α exerts its main effects by binding two specific receptors, TNFR1 (55 kDa) and TNFR2 (75 kDa).8 The soluble forms of the TNF-α receptors (sTNFR1 and sTNFR2), which represent the extracellular portions of membrane-associated TNF-α receptors, play a role as modulators of the biological TNF-α activity.9 The binding of TNF-α to TNFR1 leads to recruitment of associated death domain protein-mediated apoptosis and nuclear factor-kappa B (NF-κB) activation.10 TNFR2 is associated only with NF-κB activation and seems to have a dominant role in suppressing TNF-mediated inflammatory responses.11 Specifically, soluble TNF-α receptors seem to prolong the half-life of TNF-α by protecting this cytokine against proteolytic degradation and ultimately influencing neuroprotective or neurotoxic effects.12,13

In humans, measurement of circulating levels of the two soluble receptors is useful to determine the overall production of TNF-α.14 Because TNF-α may be less stable than its soluble receptors, some authors suggest that sTNFR1 and sTNFR2 are more reliable markers of TNF-α activity and, as consequence, of inflammatory activity.15 There are only a few recent studies indirectly addressing an increase in sTNFR1 levels in depressive disorder16,17 and none addressing sTNFR2. In addition, post-traumatic stress disorder (PTSD) symptoms and early life stress are associated with a low-grade systemic pro-inflammatory state, enhancing TNF-α activity.18,19 Specifically childhood maltreatment has been associated with adult inflammatory status.20,21 The aim of the present study was to evaluate sTNFR1, sTNFR2 and TNF-α plasma levels in MDD female outpatients. A secondary goal of this study was to assess possible influences of PTSD symptoms and history of childhood maltreatment with sTNFR1 and sTNFR2 plasma levels.


Subjects and clinical assessment

Thirty MDD female outpatients in partial remission (20–55 years of age) were recruited from of an affective disorder program at the Hospital Presidente Vargas (Porto Alegre, Brazil). Diagnosis of MDD was confirmed using the Structural Clinical Interview for DSM-IV Axis I Disorders (SCID-I). Social status was evaluated on Hollingshead's Index of Social Status.22 Severity of depressive symptoms was evaluated on Beck Depression Inventory (BDI),23,24 post-traumatic stress disorder (PTSD)-like symptoms, on the PTSD Checklist–Civilian Version (PCL-C),25,26 and childhood abuse and neglect, on Childhood Trauma Questionnaire (CTQ).27,28 All patients had been on a stable dose of antidepressant monotherapy for at least 3 months and were free of other psychiatric and non-psychiatric drug use for 4 weeks before blood collection. Exclusion criteria were: Axis I comorbidities, severe or unstable clinical illness or illness associated with abnormal immunological parameters, neurological disorder, psychotic symptoms, or any psychoactive substance used in the last 30 days (excepting nicotine, caffeine and antidepressants).

Ninety-four healthy hospital employees (all female, 20–50 years of age) from low social status were evaluated on SCID-I, Hollingshead's Index, BDI, PCL-C and CTQ. All participants with past or current Axis I disorder, severe or unstable clinical illness or illness associated with abnormal immunological parameters, neurological disorder, moderate to severe childhood abuse and neglect or psychoactive substance use in the last 30 days (excepting nicotine and caffeine) were excluded. The remaining participants comprised the control group (n = 19). The present study was approved by Pontifical Catholic University of Rio Grande do Sul (PUCRS) Ethics Committee. Written informed consent was obtained from all participants.

Plasma sTNFR1 and sTNFR2

Whole blood was collected between 08.00 and 09.00 hours and participants had been instructed not to eat or take medication for at least 8 h before blood collection. Plasma was separated within 30 min and the supernatant was stored at −80°C for up to 6 months. Plasma TNF-α, sTNFR1 and sTNFR2 were measured according to the procedures supplied by the manufacturer and using ELISA kits for sTNFR1 and sTNFR2 (DuoSet, R&D Systems, Minneapolis, MN, USA) as routinely performed at our lab.29 All samples were assayed on duplicate. The detection limits were 25 pg/mL for TNF-α, and 12 pg/mL for both soluble receptors. Values below the detection limits were assumed to be zero. Concentration is expressed as pg/mL.

Statistical analysis

All variables were tested for normality of distribution by means of the Kolmogorov–Smirnov test. Student's t-test was performed to compare demographic and clinical characteristics of groups. Mann–Whitney U-test was used to assess group differences in variables that failed the normality tests. Exploratory correlation analyses between sTNFR1 and sTNFR2 levels and demographic and clinical parameters were performed using Spearman's correlation coefficient. Analysis of covariance (ancova) was used to compare group mean differences using factors that had significant association with each soluble TNF-α receptors. Multiple regression analysis was further performed to assess the independent weight of variables that could affect sTNFR1 and sTNFR2 concentration. The significance level was set at α = 0.05 (two-tailed). Statistical analyses were performed using SPSS 15.0 (SPSS, Chicago, IL, USA). All values are presented as mean ± SD.


The demographic and clinical characteristics of cohorts are shown in Table 1. Both groups were homogeneous regarding age, body mass index, education and social status. Patients with MDD reported 11.5 ± 5.58 years of illness duration and 7.56 ± 3.19 years of formal treatment 46.7% of patients were using tricyclic antidepressants (n = 14), 43.3% selective serotonin re-uptake inhibitors (n = 13), and 10% was currently using other antidepressants (n = 3). The type of antidepressant did not affect sTNFR1 (F = 1.74, P = n.s.) or sTNFR2 (F = 1.03, P = n.s.) plasma levels.

Table 1.  Subject characteristics
 Depression (n = 30)Control (n = 19)P
mean± SDmean± SD
  • Hollingshead's Index of Social Status.

  • Statistical analysis was conducted using Student's t-Test.

  • BDI, Beck Depression Inventory; BMI, body mass index; CTQ, Childhood Trauma Questionnaire; PCL-C, Post-Traumatic Stress Disorder Checklist–Civilian Version.

Age (years)39.218.5637.365.48n.s.
BMI (kg/m2)26.892.8625.542.50n.s.
Education (years)8.363.748.792.54n.s.
Social Status63.756.9562.574.79n.s.
PCL-C scores55.6614.2722.925.39<0.001
BDI scores28.5411.144.783.92<0.001
CTQ scores58.3618.3241.7110.62<0.001

Zero-order analyses indicated a significant correlation between sTNFR1 and sTNFR2 (r = 0.71, P < 0.001). Moreover, sTNFR1 was found to be associated with age (r = 0.36, P = 0.001), severity of depression (BDI; r = 0.47, P = 0.001) and PTSD-like symptoms (PCL-C; r = 0.48, P < 0.001). In addition, sTNFR2 was correlated with severity of depression (r = 0.67, P < 0.001), PTSD-like symptoms (r = 0.62, P < 0.001), and childhood trauma severity (CTQ; r = 0.37, P = 0.008). The TNF-α plasma levels were similarly low or under the detection limit in both the control and MDD group (U = 239.5, P = n.s.; Fig. 1).

Figure 1.

Plasma concentration of tumor necrosis factor-α (TNF-α) in outpatients with major depressive disorder (depression, n = 30) and healthy participants (control, n = 19).

In contrast, analyses of covariance were used to assess mean group differences controlling for all variables correlated with soluble TNF-α receptors. Plasma sTNFR1 and sTNFR2 were detectable in all control and MDD subjects. After adjusting for previously correlated variables, patients with recurrent MDD had significantly higher concentration of sTNFR1 (447.71 ± 118.96 pg/mL) in comparison to healthy controls (298.62 ± 95.99 pg/mL) (F = 6.13, P < 0.001; Fig. 2a). Similarly, concentration of sTNFR2 was greater in MDD patients (2223.98 ± 303.56 pg/mL) than in control subjects (1627.15 ±  272 pg/mL; F = 10.8, P < 0.001; Fig. 2b).

Figure 2.

Plasma concentration of (a) sTNFR1 and (b) sTNFR2 in outpatients with major depressive disorder (depression, n = 30) and healthy participants (control, n = 19). Horizontal lines indicate mean values. Statistical analysis (ancova) was conducted using as covariates: age, depression severity, childhood abuse and neglect and post-traumatic stress disorder-like symptoms.

Considering the results obtained on exploratory analyses, we built a theoretical model in which plasma sTNFR2 level was predicted by depression severity (BDI), PTSD-like symptoms (PCL-C) and by childhood trauma severity (CTQ). In contrast, the variability of sTNFR1 concentration was investigated with regard to age (years), depression severity and PTSD-like symptoms. Thus, multiple linear regression with the stepwise method was used to assess the independent role of each variable on sTNFR2 and sTNFR1 plasma levels. In regression models there was no evidence of high colinearity between selected variables, and all variance inflation factors (VIF) were <1. Using sTNFR2 as a dependent variable, the only selected variable (F ≤ 0.05) to be included in the equation was depression severity (B = 0.67, P < 0.001; R = 0.67, adjusted R2 = 0.44; F(1,47) = 39.26, P < 0.001). This variable was thus considered to be an important predictor of sTNFR2 concentration after adjusting the slope of the line for severity of PTSD-like symptoms and childhood maltreatment. In addition, level of sTNFR1 was independently predicted by PTSD-like symptoms (B = 0.41, P = 0.002) and age (B = 0.25, P = 0.05) after adjusting for depression severity (R = 0.54, adjusted R2 = 0.26; F(2,46) = 9.51, P < 0.001).


To our knowledge this is the first study that objectively assessed plasma soluble TNF-α receptors in recurrent MDD. The sTNFR1 and sTNFR2 plasma levels were significantly increased in MDD, despite unchanged TNF-α concentrations. Multivariate analysis indicated that sTNFR2 levels are influenced by depression severity. Moreover, levels of sTNFR1 were independently predicted by PTSD-like symptoms and age.

Tumor necrosis factor-α receptors seem to have a potential antagonistic function with respect to neuronal survival upon exogenous stress signals and/or tissue damage.30 There is some evidence suggesting that TNFR2 is involved with TNF-mediated anti-apoptotic pathways31 given the TNF-mediated apoptotic pathway via TNFR1.32 Therefore TNFR2 is related to neuroprotection and TNFR1 to neurotoxicity. Although the intracellular effects of TNFR2 are not completely known, TNFR1 depends largely on the nuclear translocation of the transcription NF-κB. The duration of NF-κB activation is critical to achieve significant tissue protection.33 Cellular toxicity was observed in experiments in which TNFR1 induced a rapid but transient NF-κB response and TNFR2 produced a more persistent response.31 In contrast, soluble forms of TNFα receptors are correlated with the modulation effect of TNF-α at the site of synthesis and the facilitation of its transport to distant organs, promoting its systemic effects.12 The hypothesis regarding the TNF-α neutralizing effect is not attractive because a 30–300-fold molar excess of TNF-α soluble receptors is required to inhibit the cytotoxic action of TNF-α.34 Thus it is suggested that the role of TNF-α soluble receptor is related to stabilizing the trimeric structure of TNF-α more than buffering the effects of this cytokine.12 In the present study the MDD patients had increased TNF-α soluble receptors despite similar TNF-α plasma levels compared with controls, indicating an increased TNF-α activity. Therefore it is possible that both pro-apoptotic and anti-apoptotic pathways could be enhanced. The activation of the same intracellular key molecule (NF-κB) could trigger a completely different signal output (neuroprotection vs neurotoxicity) due to a differential duration of the active state of this molecule.

Neurodegeneration, reduced neuroprotection and neuronal repair are common pathological features of major depression.35 The sTNFR1 and sTNFR2 could be considered as potential biological markers associated with such features, despite the fact that studies are needed to confirm such speculation. In contrast, severity of depression was positively related to sTNFR2 but not to sTNFR1. The underlying mechanisms involved with the triggering of TNF-α soluble receptors in depression are largely unclear. Previous studies, however, have reported that severity of depressive symptoms was positively associated with pro-inflammatory cytokines.36 Thus elevated sTNFR2 plasma levels could be related to this inflammatory status. Proteolytic cleavage (shedding) is a well-documented mechanism to downregulate the cell response to TNF-α.37 It has been reported that downregulation of its receptors by TNF-α is due to the shedding of sTNFR2 and the internalization and shedding of sTNFR1, leading some authors to suggest the involvement of a TNF-α-dependent mechanism for sTNFR1 and a different mechanism, non-directly related to TNF-α, for sTNFR2.12 Considering that its modulation has been related with other inflammatory pathway it was speculated that sTNFR2 could be more dependent on enhanced pro-inflammatory processes and subsequently more sensitive to depression severity.

Another interesting finding of the present study was that sTNFR1 levels were predicted by age and PTSD-like symptoms. The present data are in accordance with previous work indicating an age-related increase in sTNFR1.38 Moreover, the positive correlation observed between sTNFR1 and PTSD-like symptom severity is consistent with a low-grade systemic pro-inflammatory state directly related to PTSD symptom levels.19 We used PCL-C, however, to evaluate PTSD-like symptoms, and it is important to highlight that such a scale does not always specifically measure PTSD symptoms: instead depressive symptoms are also included. Therefore such findings could be the result of overlapping depressive symptoms that are part of the PCL-C scale. Future cross-sectional and longitudinal studies should take these variables into consideration to correctly assess soluble TNF-α receptors in MDD.

There are some limitations in the present study to be discussed. First, the stringent inclusion/exclusion criteria limited the sample size but yielded strictly healthy control subjects. It should be noted, however, that it was very difficult to recruit healthy controls from low socioeconomic status with no past or current psychopathology who did not report childhood maltreatment. The sample sizes, however, are in line with a previous study that measured TNF-α levels in MDD.1 Replication with large samples and longitudinal follow up is needed to overcome this limitation. Second, MDD participants were using antidepressants and it was not possible to discard the effects of medication on soluble TNF-α receptors or TNF-α levels. Indeed, the lack of detection of TNF-α could be related to high levels of soluble forms of its receptors present in plasma, which could bind circulating TNF-α and interfere with its detection.12 Taking into consideration that antidepressant treatment reduces TNF-α levels in MDD subjects,6 it is possible that this effect could be related to the increased level of TNF-α soluble receptors. Despite such considerations it should be noted that it is extremely difficult to investigate non-medicated patients who have recurrent MDD.


Taking into account the limitations described in the previous section, the present results suggest that sTNFR1 and sTNFR2 could be reliable markers of TNF-α activity in MDD. Moreover, particular roles of each soluble TNF-α receptors should be investigated. Specifically the complete understanding of mechanisms involved in the shedding of sTNFR2 and the internalization and shedding of sTNFR1, in addition to a better understanding of differential signals by TNFR1 and TNFR2 in neurons, is essential in order to develop new effective strategies. Although interesting and important studies have been performed on anti-inflammatory treatments for depressive symptoms, particularly using celecoxib,39,40 future studies targeting TNF-α activity through sTNFR1 and sTNFR2 may be promising.


We thank the patients and staff at the Hospital Presidente Vargas, and Professor Lilian Milnitsky Stein, PhD, for her support. This research was funded by independent CNPq awards to MEB (470747/2006-4 and 305155/2006-7) and by CAPES scholarships to RG-O and to RPL.