An inflamed subtype of difficult-to-treat depression

Background: Chronic low-grade inflammation may play a role in the pathophysiology of depression, at least in a subset of patients. High-sensitivity C-reactive protein (hs-CRP) has been used to define an inflamed subgroup of depression with specific clinical characteristics and symptoms. In this study we investigated biochemical and clinical characteristics in patients with difficult-to-treat depression with and without chronic low-grade inflammation. Method: We assayed plasma levels of interferon-gamma, tumor necrosis factor-alpha, Interleukin (IL)-10, IL-6, IL-8, and vitamin D in a clinically well-characterized sample of patients with difficult-to-treat depression ( n = 263) and healthy controls ( n = 46). Serum hs-CRP levels were available in the patient group and were used to define “ inflamed depression ” (hs-CRP > 3 mg/L). Based on previous studies correlating specific depressive symptoms to inflammatory markers, we calculated a composite score of inflammatory depressive symptoms (Infl-Dep score). A principal component analysis (PCA) was performed to identify patterns of variance in cytokines and vitamin D among patients. Results: Mean levels of IL-6 and IL-8 were significantly higher in depressed patients compared to controls, also after adjusting for sex, smoking, BMI, and age. None of the other inflammatory markers differed significantly between depressed patients and controls. Two components were extracted using PCA; one showed general cytokine elevations and one represented a pattern where IL-6 and IL-8 were inversely related to vitamin D (IL6-IL8-VitD component). The inflamed subgroup (hs-CRP > 3, n = 51) exhibited significantly higher BMI, higher Infl-Dep scores and higher IL6-IL8-VitD component scores than uninflamed patients (hs-CRP ≤ 3, n = 212). There were no significant differences in overall depression severity or suicidality between the inflamed and uninflamed groups. Conclusion: Our results support the hypothesis of an inflamed subgroup of depression as a meaningful construct. This subgroup may have certain biological and clinical characteristics and more studies are needed to determine potential clinical implications.

Some depressive symptoms, such as sleep or appetite alterations, fatigue, lethargy, and anhedonia may be more "inflammatory" than other symptoms (Dantzer et al., 2008;Frank et al., 2021;Fried et al., 2020). Interestingly, these inflammatory depressive symptoms overlap with so called sickness behavior, frequently observed in animals and humans in response to immune activation (Dantzer et al., 2008;Thylur and Goldsmith, 2022;Miller and Raison, 2016;Milaneschi et al., 2020). As recently reviewed elsewhere (Lucido et al., 2021), there is a vast body of evidence from preclinical and clinical studies linking anhedonia to inflammation. For instance, Haroon et al. (Haroon et al., 2020) identified a subgroup of depressed patients with elevated inflammatory markers, severe anhedonia, and a greater likelihood of poor treatment response. Despite solid evidence for a link between low-grade inflammation and depression, the results from several clinical trials testing the antidepressant effect of anti-inflammatories have been negative when evaluating the primary outcome, which has usually been general depressive symptoms (Lucido et al., 2021;McIntyre et al., 2019;Husain et al., 2020). However, there are some indications from secondary analyses that specific symptoms, such as anhedonia, are more responsive to anti-inflammatory drugs than placebo (Lee et al., 2020). Thus, sickness behavior-related symptoms, including anhedonia, appear to be a core feature of inflamed depression and these symptoms may be more responsive to certain anti-inflammatory interventions. Indeed, the association between inflammation and specific depressive symptoms has been confirmed in several population-based and one clinical cohort (summarized in Supplementary Table 1), where anhedonia, fatigue, sleep disturbance, and changes in appetite were associated with inflammation (Frank et al., 2021;Fried et al., 2020;White et al., 2017;Jokela et al., 2016).
There is currently no clinical consensus on how to define inflamed depression, but several investigators have used a cut-off level of 3 mg/L for high-sensitivity C-Reactive Protein (hs-CRP) (Osimo et al., 2019;Shin et al., 2016;Wium-Andersen et al., 2013;Felger et al., 2020;Foley et al., 2021). This is also an established cut-off for low-grade inflammation according to the Centers for Disease Control and Prevention and the American Heart Association (Ridker, 2003;Pearson et al., 2003), and approximately one third of all depressed patients have hs-CRP above this threshold (Osimo et al., 2019). Peripheral hs-CRP is an accessible and standardized biomarker that could be useful in identifying a subgroup of inflamed depression patients as it has been associated with several tentative brain biomarkers of depression including decreased corticostriatal connectivity (Felger et al., 2016), increased basal ganglia glutamate (Haroon et al., 2016), and cerebrospinal fluid (CSF) CRP (Lindqvist et al., 2013).
In the present study, we divided a clinically representative sample of patients with difficult-to-treat depression into two groups of inflamed and uninflamed depression, using previously established hs-CRP cutoffs. We defined difficult-to-treat depression as "depression that continues to cause significant burden despite usual treatment efforts" (McAllister-Williams et al., 2020). We then examined differences in clinical characteristics, symptom profile, other inflammatory markers, and vitamin D between depressed patients with and without hs-CRP elevations.

Study overview and ethics
The study sample was a part of the Genes Depression and Suicidality (GEN-DS) study for which the recruitment process has been described in more detail elsewhere (Asp et al., 2020;Asp et al., 2021). The study followed the ethical principles of the Declaration of Helsinki. Participants were given oral and written information about the study and provided written informed consent before inclusion. The GEN-DS study was approved by the Regional Ethical Review Board in Lund (reference # 2011/673).

Participants
Patients were referred to the GEN-DS from four secondary psychiatric clinics in southern Sweden between 2012 and 2021. Patients were referred to the study if they had ongoing depressive illness with insufficient treatment response. The exclusion criteria were BMI < 15, liver failure or pregnancy (Asp et al., 2020). The diagnostic protocol included semi-structured interviews as well as expert-and self-rating scales, as detailed below. Patients were diagnosed according to the DSM, 4th Edition (IV-TR).
Blood sampling was done on the day of the clinical assessment. Hs-CRP and cytokines were analysed in 266 patients. One patient was excluded due to prolonged time (1 month) between clinical assessment and blood draw. Two patients were excluded from the analyses due to extreme values (≥ + − 4 SD) of inflammatory markers, leaving 263 subjects for the final analyses. Vitamin D was analysed in 238 patients. None of the included patients reported any acute infectious or inflammatory illnesses.
Forty-six healthy controls were recruited through ads in social media and newspapers. Exclusion criteria for healthy controls were: any previous or present psychiatric illness (all were interviewed using the Mini International Neuropsychiatric Interview (MINI) 6.0), ongoing infection, psychotherapy or treatment with psychotropic drugs, severe or chronic somatic illness, immune-modulating drug therapy, pregnancy, or breast-feeding. Healthy controls received 500 SEK in compensation for participating in the study.

Diagnostic assessment and symptom severity ratings
All patients were assessed by a specialist in psychiatry or a specialistsupervised resident in psychiatry and underwent diagnostic interviews according to the MINI 6.0 and the Structured Clinical Interview for DSM-IV Personality Disorders (SCID-II). Severity of psychiatric symptoms was rated by the Comprehensive Psychopathological Rating Scale (CPRS) (Asberg et al., 1978). From the CPRS, we extracted the 9-item expert version of Montgomery-Åsberg Depression Rating Scale (MADRS) for evaluation of depressive symptoms. Several self-rating scales were filled out, including the Suicide Assessment Scale (SUAS).
In previous studies, several different rating scales have been used to identify depressive symptoms associated with low-grade inflammation. Based on these results (summarized in supplementary Table 1), we calculated a composite score of inflammatory depressive symptoms (Infl-Dep score). The symptoms most consistently associated with inflammation in previous studies (translated into corresponding CPRS items) were summarized into this composite score. These symptoms were: altered sleep patterns, appetite disturbances, tiredness, and anhedonia (corresponding to CPRS items, #5 Inability to feel, #14 Lassitude, #15 Fatiguability, #18 Changes in appetite, #19 Reduced sleep, #20 Increased sleep).

Blood sampling and analyses
At the day of the study visit, screening of routine blood tests, including hs-CRP, was conducted and additional plasma was sampled and stored in − 80 • C until future analyses. Patients were instructed to fast four hours before the blood-draw and avoid nicotine use and taking medications in the morning.
Serum hs-CRP levels were analysed in patients according to clinical standard at the Department of Clinical Immunology, Skåne University Hospital, Lund, Sweden. Limit of quantitation (LOQ) for hs-CRP was 0.60 mg/L. Samples below LOQ (37%) were imputed between 0.0 and 0.6 using same method as for the cytokines. Hs-CRP was not available from the healthy controls.

Statistics
All analyses were performed using the Statistical Package for the Social Sciences for Mac (SPSS version 27, IBM, Armonk, NY, USA). All variables were assessed for normal distribution and in case of nonnormality, according to histogram or Q-Q-plot, ln-transformation was performed for parametric tests or Mann-Whitney U was used for comparative analyses. Groupwise-comparisons of biomarkers were tested using t-tests with ln-transformed data, while group-wise differences in rating scale scores were tested using Mann-Whitney U tests. Since five inflammatory biomarkers were compared between-groups, p-values<0.01 (0.05/5) were considered statistically significant according to the Bonferroni method. All correlation analyses were conducted with Pearson's correlation.
We divided patients based on hs-CRP levels to define a subgroup of inflamed depression using a cut-off of 3 mg/L since it is a wellestablished indicator of chronic low-grade inflammation (Ridker, 2003) that has been used in previous studies (Osimo et al., 2019;Shin et al., 2016;Wium-Andersen et al., 2013;Felger et al., 2020;Foley et al., 2021). Hs-CRP levels ≤3 mg/L are referred to as uninflamed depression and hs-CRP levels >3 mg/L are referred to as inflamed depression.
To identify patterns of variations among available inflammatory biomarkers and vitamin D and compare such between patients with CRP ≤3 vs >3 mg/L, we performed a Principal Component Analysis (PCA) to avoid multiple comparisons. The PCA was conducted on ln-and ztransformed variables of IFN-γ, TNF-α, IL-10, IL-6, IL-8, and total vitamin D levels. We used Varimax rotation, however, to explore the effect of choosing an orthogonal rotation, analysis was repeated using direct-oblimin (oblique) rotation with similar outcome. Due to the sample size and average communalities in our data (<0.6), we did not use the pre-set extraction of factors with Eigenvalue>1 (Kaiser's criterion). Instead, we assessed the computed scree plot for inflection point and included all factors pre to that point (Field, 2013). Scores were saved for all components using Bartlett's method.

Demographic characteristics and inflammatory biomarkers in patients and controls
Demographic characteristics and biomarker levels in patients and controls are presented in Table 1. Patients had significantly higher mean BMI and were to a greater extent everyday smokers compared to controls. BMI was significantly correlated with TNF-a (r = 0.23, p < 0.001, n = 298), IL-6 (r = 0.43, p < 0.001, n = 298) and CRP (r = 0.60, p < 0.001, n = 252), but not with the other inflammatory biomarkers. Cytokines IL-6 and IL-8 were significantly elevated in patients compared to controls, also after adjustment for sex, smoking, BMI, and age (p < 0.01). Mean freezer storage time was longer for patient samples compared to control samples. We therefore separately adjusted for years in freezer and the between-group differences in IL-6 and IL-8 remained significant (p < 0.01).
As shown in Table 2, the most common affective disorders were recurrent unipolar depression (41.8%) and dysthymia/chronic depression (29.7%). After the in-depth diagnostic assessment 10.6% did not fulfill criteria for any current affective disorder as main diagnosis. Moreover, 93.9% of the patients were prescribed some type of psychopharmacological treatment. The most prescribed drug classes were antidepressants (81.4%), anxiolytics (53.2%), and mood stabilizers (25.9%). Psychiatric comorbidity was common (67.7% among all patients), with anxiety disorders (49.4%) and personality disorders (36.9%) being most frequent. Also, somatic comorbidity was commonly reported. Approximately 66% of the patients reported at least one somatic illness, where musculoskeletal, gastrointestinal, or endocrinological disorders were the most frequent.  Raw data for biomarkers. Independent samples t-tests were performed for comparison with controls, using ln-transformed values to ensure normal distribution. Data was missing for patients on BMI (n = 11), and smoking (n = 2). CRP was not available for healthy controls.

Demographics and symptom ratings in inflamed and uninflamed depression
The uninflamed depression group (hs-CRP ≤3 mg/L) comprised 212 patients, whereas 51 patients were included in the inflamed depression group (hs-CRP >3 mg/L). As presented in Table 3 and Fig. 3, inflamed depression cases had significantly higher mean BMI (p < 0.001) and higher scores on the composite Infl-Dep score (p = 0.022) compared to the uninflamed patient group. There were no significant differences in age, smoking, gender, ethnicity, or sampling season between uninflamed and inflamed cases. There were no significant between-group differences in mean scores of MADRS or SUAS.

Correlations between peripheral inflammatory markers and vitamin D levels
Ln-transformed cytokine and vitamin D-levels to be included in PCA were assessed for correlations among patients included in the analysis, as presented in Fig. 1.

PCA
A PCA was performed on all subjects with full data on vitamin D and inflammatory biomarkers included in the analysis (n = 238). The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was 0.7 and Bartlett's test of sphericity showed significance level < 0.001, indicating an acceptable sample size and correlations for the analysis. A solution of two components was chosen, both with Eigenvalues>1. The individual loadings of biomarkers on extracted components are presented in Fig. 2. The extracted components 1 (C1) and 2 (C2) explained 34.27% and 22.36% of variance after rotation, respectively. All cytokines, but not vitamin D, loaded positively >0.3 on C1. Vitamin D loaded negatively on C2. IL-6 and IL-8 loaded positively on both C1 and C2, although their loadings were higher on the latter component. We interpret C1 as a general cytokine elevation component and C2 as a component characterized by low vitamin D and higher IL-6 and IL-8, hereafter referred to as the general cytokine elevation-component and the IL6-IL8-VitD component, respectively.

Component scores in inflamed and uninflamed depression
Mean levels of the general cytokine elevation-component (C1) did not significantly differ between uninflamed or inflamed subjects (t (60.8) = − 1.1, p = 0.27). The inflamed depression group, however, Diagnoses, treatment categories and comorbidities. Data were missing for affective disorders (n = 1), ongoing treatment (n = 1), comorbidities (anxiety disorder, n = 1, personality disorder, n = 4, substance use disorder, n = 1, neuropsychiatric disorder, n = 2, and somatic conditions, n = 1). a Main diagnosis, if not affective disorder, were anxiety disorders (n = 15) and neuropsychiatric disorder (n = 2). 11 patients were assessed as in remission from an affective disorder. b Several patients had more than one comorbidity and more than one class of medication.   showed significant higher levels of the IL6-IL8-VitD component compared to the low inflammation group (t(62.0) = − 3.6, p < 0.001), as illustrated in Fig. 3.

Sensitivity analyses
Sensitivity analyses excluding patients with no current affective disorder (n = 28) and missing data for the main diagnosis (n = 1) were performed and did not significantly change the main findings. Nor did exclusion of patients with chronic inflammatory illnesses (e.g. rheumatic disorders or inflammatory bowel disease, n = 18) and patients treated with antibiotics (n = 5) change the main results. We compared levels of all inflammatory markers between patients taking antidepressants (n = 168), antipsychotics (n = 4) or a combination (n = 46). There were no significant differences between these subgroups in inflammatory marker levels (all p > 0.05, Bonferroni adjusted).

Discussion
In this large-scale cohort of clinically well-characterized patients with difficult-to-treat depression, we found that patients with inflamed depression (defined according to hs-CRP >3 mg/L) had higher BMI and greater severity of some, but not all, depressive symptoms. Moreover, we found evidence for a biochemical distinction between these two groups with a stronger link between the inflamed group and the factor associated with IL-6, IL-8 and vitamin D.
In all depressed patients, regardless of hs-CRP levels, IL-6 and IL-8 were significantly higher compared to controls. The elevated IL-6 levels are consistent with most previous studies on depression, summarized in meta-analyses (Köhler et al., 2017;Haapakoski et al., 2015;Howren et al., 2009;Lynall et al., 2020;Goldsmith et al., 2016). Previous studies have, however, been more inconsistent with regards to IL-8, reporting either unchanged (Dowlati et al., 2010;Köhler et al., 2017;Goldsmith et al., 2016) or elevated (Osimo et al., 2020) levels in depressed patients compared to controls. Important confounders in the relationship between depression and IL-8 may be BMI (Osimo et al., 2020) and medication status (Çakici et al., 2020). Though, in the present study, IL-6 and IL-8 levels remained significantly higher in depressed cases also after adjustment for BMI, age, gender, and smoking.
Depressive disorders are heterogenous conditions, and most previous studies have mainly focused on DSM-based categorical definitions of depression. It is possible that this definition has hampered our understanding of the pathophysiological processes underlying depression, including inflammation (Thylur and Goldsmith, 2022). In the present study, we defined a subtype of inflamed depression, according to hs-CRP levels, in order to study clinical and biochemical characteristics of this subgroup. Although there is no clear consensus how to best categorize  inflamed depression, hs-CRP has frequently been used to define this subgroup . This biomarker holds several important advantages including clinical accessibility, standardized assays, and low diurnal variation (Meier-Ewert et al., 2001). Furthermore, peripheral hs-CRP has been shown to correlate with several other important inflammatory markers, both in plasma and CSF . In our sample, 19% of the patients had inflamed depression according to a hs-CRP cut-off of 3 mg/L which is largely in accordance with previous reports (Osimo et al., 2019).
While there were no between-group differences in overall severity of depressive symptoms, we found that patients in the inflamed depression group had more severe symptoms previously related to inflamed depression (Frank et al., 2021;Fried et al., 2020). To test this hypothesis, we calculated a composite score summarizing symptoms related to sleep and appetite disturbance, fatigue, lethargy and anhedonia. This is consistent with several previous reports linking peripheral hs-CRP levels to specific depressive symptoms such as tiredness or lack of energy (Frank et al., 2021;Fried et al., 2020;White et al., 2017;Jokela et al., 2016), sleep disturbance (Frank et al., 2021;Fried et al., 2020;White et al., 2017;Jokela et al., 2016), appetite changes (Frank et al., 2021;Fried et al., 2020;Jokela et al., 2016), and anhedonic symptoms such as inability to feel and lassitude (Frank et al., 2021;White et al., 2017). Most of these studies analysed population-based cohorts (Frank et al., 2021;White et al., 2017;Jokela et al., 2016), while we included patients from specialized psychiatric care with long-lasting, clinically relevant and more severe psychiatric symptoms. These findings of specific inflammatory depressive symptoms seem to be consistent across studies including patients with various treatments and degrees of illness severity. In the future, these findings may have important clinical implications, as some depression trials using anti-inflammatory medications have reported a specific effect on some of these symptoms such as anhedonia (Lee et al., 2020). In the present study, the inflamed group had significantly higher BMI compared to the uninflamed group which is in line with a bidirectional link between obesity and depression as reviewed elsewhere (Shelton and Miller, 2011). Obesity is a driver of systemic low-grade inflammation via secretion of cytokines and chemokines from adipose tissue. This may be an important pathophysiological mechanism not only behind inflamed depression but also behind several somatic conditions, often co-morbid with depression, such as diabetes and cardiovascular disease (Tilg and Moschen, 2006). In the current study, we did not factor out BMI when comparing symptoms and biomarkers between the inflamed and the uninflamed depression groups. The reason for this is that some have argued that BMI should be viewed as a mediator rather than a confounder in the relationship between inflammation and depression, since obesity may be an integral part of inflamed depression pathophysiology (Lamers et al., 2020). Future studies, specifically powered for this hypothesis, should investigate the mediating effects of obesity on inflamed depression.
Some studies have suggested that suicidal depression may be more tightly linked to inflammation (Brundin et al., 2015), but we did not find any differences in degree of suicidality between the uninflamed and inflamed depression groups. An important difference is that some previous reports linking suicidality to inflammation included recent suicide attempters (Lindqvist et al., 2009;Janelidze et al., 2011), while we defined suicidality according to SUAS total score which includes items dealing with several different aspects of suicidality. While we divided patients into two groups based on inflammation, some investigators have suggested that a dichotomous division of depressed patients into an inflamed and uninflamed group may be an over-simplistic approach (Felger and Miller, 2020). Lynall et al. (Lynall et al., 2020) identified four subgroups among 206 depressed with unique biological characteristics. Two of these groups were interpreted as inflamed and displayed different immune cell profiles; one lymphoid-and the other myeloid-biased. The latter inflamed subgroup consisted of 18% of the patients and had higher CRP levels, BMI, and anhedonic symptom scoring compared to uninflamed cases, which is partly in line with our results. Future studies should take the complexity of inflammation into account when designing studies investigating the link between inflammation and depression.
We found that the inflamed subgroup was associated with a specific biological factor of elevated IL-6, IL-8 and decreased vitamin D. Based on previous findings from our group of an inverse relationship between vitamin D and inflammation in depressed individuals, vitamin D might be of relevance for inflamed depression (Grudet et al., 2014;Grudet et al., 2020). Lower absolute levels of vitamin D have been associated with depression in some (Ronaldson et al., 2020;Milaneschi et al., 2014;Parker et al., 2017) but not all (Dana-Alamdari et al., 2015;Pan et al., 2009) studies. Vitamin D receptors are present in many cell types (Sassi et al., 2018;Geng et al., 2019) and vitamin D has modulating effects on several biological systems (Wacker and Holick, 2013). This includes immunomodulatory effects that can shortly be described as 'anti-inflammatory', where vitamin D, directly and indirectly, promotes a shift in CD4+ T helper cell differentiation from the pro-inflammatory Thelper 1 (Th-1)/Th-17 immune cell response to the anti-inflammatory Th-2/regulatory T cell (Tregs) immune cell response (Sassi et al., 2018;Aranow, 2011;Calton et al., 2015). Our findings of a link between low vitamin D and inflamed depression further strengthens the hypothesis that interactions between vitamin D and inflammation may be important in the pathophysiology of certain cases of depression.
Our study has several strengths. As mentioned above, the detailed diagnostic procedure is an important strength and a unique feature compared to many other studies within the field. Moreover, our patients, many of whom had comorbidities and were taken concurrent medications, most likely reflect patients seen in psychiatric clinical practice. Although the inclusion of "real-life patients" may be considered a strength, it may also lead to bias since some of these variables mentioned above may affect biomarker levels. For instance, almost all the patients were treated with some type of psychotropic medication, hence we were unable to adjust for this factor in the groupwise comparisons between depressed patients and healthy controls. Previous studies suggest that antidepressants, antipsychotics as well as mood stabilizers may influence cytokine levels (Baumeister et al., 2016;Rapaport and Manji, 2001). Therefore, we cannot rule out that concomitant medications may have had an impact on some of our findings, but on the other hand there are reports of cytokine elevations also in unmedicated depressed patients compared to controls (Lindqvist et al., 2017). There are also other limitations of our study. Firstly, the current patient sample is relatively young which may limit the generalizability. Also, we were not able to determine temporal changes in inflammation status and symptom profile due to the cross-sectional study design. Finally, we used the more broad and less well-defined definition of difficult-to-treat depression to recruit our sample, as opposed to other operationalized methods, and this may also be seen as a limitation of the study.
In conclusion, our findings add support to the hypothesis of a clinically meaningful subtype of inflamed depression with a specific symptom profile and a biochemical signature. If replicated in other studies, these findings could inform the design of future treatment studies testing the antidepressant effects of anti-inflammatory compounds (Lucido et al., 2021;Osimo et al., 2019).

Ethics approval statement
The study followed the ethical principles of the Declaration of Helsinki. The GEN-DS study was approved by the Regional Ethical Review Board in Lund (reference # 2011/673).

Data availability
In accordance with the Swedish Public Access to Information and Secrecy Act, data cannot be made freely available. Requests of data can be sent to registrator@lu.se.