Advancing an Inflammatory Subtype of Major Depression

Abstract

Chronic inflammation plays a prominent role in multiple medical disorders including psychiatric diseases such as major depression. Exposure to inflammatory stimuli leads to changes in neurotransmitter systems and neurocircuits in the brain that are associated with depressive symptoms. Blockade of inflammatory cytokines can reduce depressive symptoms in medically ill and medically healthy depressed individuals. Increased biomarkers of inflammation are associated with an overrepresentation of neurovegetative symptoms including anhedonia, fatigue, and psychomotor slowing and can predict response to antidepressant treatments. Importantly, however, increased inflammatory biomarkers only occur in a subgroup of depressed individuals. Thus, there appears to be a subset of depressed patients with a unique symptom presentation and treatment response whose disease is primarily driven by inflammation. Further identifying and characterizing this inflammatory subtype of depression can foster the development of treatments targeting the immune system and its effects on the brain. Moreover, by using this mechanism-based approach to parsing the heterogeneity of depression, we can refine our diagnostic nosology and model a strategy for precision medicine and targeted therapeutics in psychiatry.

Depression is a common and devastating disorder, being a leading cause of disability as well as death by suicide (1). A hallmark of depression is its heterogeneity, making a one-size-fits-all approach to its treatment often ineffective (2). A better understanding of the mechanisms of depression can lead to personalization of its treatment and ultimately improved outcomes. One mechanism that contributes to depression is inflammation (3). Inflammation has effects on neurotransmitter systems and neurocircuits in the brain leading to alterations in behaviors related to neurovegetative function and threat sensitivity. Moreover, a significant percentage of depressed patients exhibit increased inflammatory biomarkers, which in turn can predict response to antidepressant treatments including conventional antidepressant therapies such as serotonin reuptake inhibitors as well as ketamine and electroconvulsive therapy and treatments targeting the immune system. These data suggest that there is an inflammatory subtype of depression with a unique symptom presentation and response to treatment. Further instantiating this depressive subtype can improve our diagnostic nosology while unraveling pathophysiologic mechanisms and leading to targeted treatments, thereby supporting precision psychiatry (Figure 1).

Figure 1. The Inflammatory Subtype of Major Depression.

Understanding the mechanisms of psychiatric disorders can allow us to revise our nosology and reduce heterogeneity. Identifying an inflammatory subtype of major depression based on biomarkers and symptoms will further refine the pathology and elucidate treatment targets. Tailoring treatment to pathology and patient preference will personalize care and promote precision medicine. dACC: dorsal anterior cingulate cortex; KYN: kynurenine; QUIN: quinolinic acid; SMA: supplementary motor area; sx: symptoms; vmPFC: ventromedial prefrontal cortex

Making the Case

Inflammation occurs in a subgroup of depressed patients.

Multiple meta-analyses have substantiated reliable increases in mean peripheral blood concentrations of inflammatory biomarkers including tumor necrosis factor (TNF), interleukin (IL)-1beta, IL-6 and the acute phase reactant C-reactive protein (CRP) in depressed patients versus control subjects (3–6). These findings are robust with increases in peripheral blood CRP for example being apparent in large population samples, even after adjusting for relevant covariates such as age, sex, race, body mass index, and smoking status (7). Evidence of increased inflammation is also found in postmortem brain samples of depressed patients including increased inflammatory signaling molecules in brain parenchyma (8) and the presence of macrophages and activated microglia (9), the immune cells of the brain as well as increased inflammatory biomarkers in cerebrospinal fluid (10). In addition, increased inflammatory mediators such as TNF, IL-1beta, and Signal Transducer and Activator of Transcription (STAT) 3 (a pivotal IL-6 signaling molecule) are primary upstream regulators of molecular signatures of depression in multiomics analyses of tissues and cells from the brain and periphery of depressed subjects (11). Nevertheless, not all depressed patients exhibit increased biomarkers of inflammation.

Based on a meta-analysis of 30 studies, only approximately 25% of depressed patients exhibit evidence of chronic low grade inflammation as indexed by a peripheral blood CRP >3mg/L (12). CRP >3mg/L is one of several cut-points used to identify elevated inflammatory risk for inflammation-related disorders such as cardiovascular disease with CRP<1mg/L representing low risk; CRP 1–3mg/L moderate risk; and CRP >3mg/L high risk (13). Thus, while mean differences in inflammatory biomarkers exist between depressed and control samples, the effect is accounted for by a subgroup of depressed individuals. Of note, chronic inflammation has been characterized in multiple other psychiatric disorders including anxiety disorders, post-traumatic stress disorder, and schizophrenia, thereby potentially representing a transdiagnostic pathophysiology. However, depression has been the most studied to date and therefore has received the most attention.

Although early theories regarding the immunology of inflammation in depression emphasized the primacy of myeloid cell populations as outlined in the “macrophage theory of depression” (14), more recent work has indicated that inflammation in depression may be more complex, potentially involving multiple immune cell subsets. For example, in one large study using an unsupervised analysis that obviated binarizing patients into low or high inflammation groups, depressed patients with increased peripheral blood biomarkers of inflammation including CRP and IL-6 segregated into two populations, one characterized by a predominance of cells of myeloid lineage (neutrophils and monocytes/macrophages); the other being dominated by lymphoid cells (T and B lymphocytes) (15,16). These findings are consistent with data suggesting that highly inflammatory T helper 17 cells as well B cell production of brain reactive antibodies may contribute to depressive-like behavior in laboratory animals and humans (17,18). Moreover, these findings suggest that there are either distinct immune pathways to inflammation in depression or that there are stages of immunopathology that evolve over time, progressing from early innate immune responses of a myeloid nature to subsequent engagement of more adaptive immune responses characterized by T and B lymphocytes (16). Of note, the latter has been observed in other inflammation-related disorders such as diabetes (16).

Inflammation is associated with specific symptoms.

Increased inflammatory biomarkers including CRP and IL-6 in depressed patients have been associated with a cluster of neurovegetative symptoms including anhedonia, lack of motivation, fatigue, and psychomotor slowing as well as disruptions in sleep and appetite (19–21). Neurovegetative symptoms in the context of inflammatory exposures such as infectious diseases were initially described as “sickness behavior” (22,23). Sickness behavior refers to a coordinated set of behavioral responses believed to subserve evolutionary imperatives of conserving resources for the substantial energy demands of an activated immune response and its containment of pathogens as well as wound healing (24,25). Inflammation is also associated with symptoms related to threat sensitivity including anxiety, irritability, and symptoms of arousal such as in PTSD (21,26,27). These symptoms align with the adaptive nature of hypervigilance against subsequent attacks in a sick or wounded animal and are considered a requisite component of a survival response (24,25). Suicidal ideation has also been linked to increased inflammation as evidenced by peripheral and central inflammatory biomarkers including measures of the translocator protein (TSPO) using positron emission tomography (PET) (28,29). TSPO binding in the brain is a putative marker of activated microglia, although its expression on multiple other cell types in the brain has limited interpretation of imaging results (30).

The relationship between inflammatory markers and specific symptoms also appears to include systemic and cellular markers of metabolic function. Indeed, depressed patients with increased inflammation plus anhedonia (versus those with increased inflammation without anhedonia) exhibit activation of gene expression pathways in peripheral blood immune cells related to a shift in cellular metabolism from the energy efficient oxidative phosphorylation to the more energy expedient (and costly) glycolysis (31). This shift in cellular metabolism, which involves signaling pathways including mammalian target of rapamycin (mTOR), 5’ AMP-activated protein kinase (AMPK), phosphoinositide 3 kinase (PI3K) and protein kinase B (akt), is a critical component of immune cell activation, allowing for rapid proliferation and release of inflammatory mediators (32). Transcriptomic signatures of psychomotor slowing in depressed patients also reveal an overrepresentation of glycolysis-related genes (33). Of note, these metabolic pathways have become novel targets for inhibiting inflammatory responses in a variety of autoimmune and inflammatory disorders (34). Interestingly, systemic measures of immunometabolism including lipids and measures of glucose metabolism have also been associated with neurovegetative symptoms including anhedonia, fatigue, hypersomnia, and hyperphagia (19,20,35,36). Indeed, the concept of immunometabolic depression which integrates dysregulation of both immune and metabolic systems is gaining increasing traction and may reflect the full complement of pathology represented in the inflammatory subtype of depression (19,36).

Inflammation is associated with treatment response.

Increased baseline peripheral blood inflammatory and metabolic biomarkers predict a poor response to conventional antidepressant medications especially selective serotonin reuptake inhibitors (SSRIs) in ambulatory depressed patients and females (37–39). Of the largest studies to date, two of these included over 100 patients and each showed a preferential response to a catecholaminergic agent (nortriptyline or bupropion) compared to the SSRI escitalopram in depressed patients with increased inflammation as indexed by a CRP >1mg/L (an indicator of at least moderate inflammatory risk)(40,41). In addition, peripheral blood expression of relevant inflammatory genes including mRNA levels of TNF, IL-1beta, and macrophage inhibitory factor predict a poor response to conventional antidepressant treatment along with polymorphisms in several immune genes including IL-1beta and STAT3 as well as genes that support T cells (42–44). Depressed patients with a history of antidepressant treatment non-response also exhibit increased peripheral blood inflammatory biomarkers including TNF, IL-6, and CRP (45,46). Moreover, increased inflammation is associated with a poor response to psychotherapy (47).

Interestingly, increased inflammation including peripheral blood IL-6 has been shown to predict a positive response to both ketamine and electroconvulsive therapy (48–50). Several studies also indicate that increased concentrations of inflammatory markers including CRP and the expression of inflammatory genes as well as markers of immunometabolic function (e.g. lipid and glucose metabolism) predict a positive response to treatments that target inflammation and its effects on the brain including anti-inflammatory drugs such as the TNF antagonist infliximab and minocycline as well as the dopamine precursor levodopa (see below)(51–58). Taken together, these data suggest that inflammatory biomarkers and the inflammatory subtype of depression may have clinical utility in predicting response to available antidepressant treatments and may ultimately help guide therapy with novel treatment strategies targeting immune and metabolic pathways in depression.

Identifying Treatment Targets

Mechanisms by which Inflammation Affects the Brain and Behavior

Cause and Effect

It has been suggested that the presence of increased immune biomarkers in certain depressed patients may reflect an epiphenomenon of behavioral and lifestyle factors associated with depression that are also known to increase inflammation including psychological stress, sedentary behavior, poor diet, obesity, sleep disruption, and the use of alcohol and tobacco. Relevant in this regard, multiple longitudinal studies have demonstrated that elevated CRP predicts future depressive symptoms, whereas prospective associations of depression with CRP are less robust (59). Although these data support the notion that inflammation can cause subsequent depressive symptoms, more definitive data comes from the administration of inflammatory stimuli to otherwise non-depressed individuals. For example, treatment with the inflammatory Type I interferon (IFN), IFN-alpha, for infectious diseases or cancer is associated with the development of significant depressive symptoms in 30–50% of patients depending on the dose, with 45% of patients meeting symptom criteria for major depression following high dose IFN-alpha treatment for malignant melanoma (60,61). Neurovegetative symptoms including anhedonia, fatigue, and psychomotor retardation tend to predominate early during IFN-alpha therapy and predict later cognitive symptoms (60,62). Similar depressive symptoms including anhedonia and motivational deficits as well as psychomotor slowing have been observed following laboratory exposures to inflammatory stimuli including endotoxin and vaccination with typhoid or influenza (63–66). Reversal of depressive symptoms has also been demonstrated following administration of drugs that block inflammatory cytokines (i.e., anti-cytokine antibodies) in patients with autoimmune and inflammatory disorders as well as in otherwise healthy depressed patients with increased inflammatory biomarkers (e.g., CRP) (57,67–69). Taken together, these data provide robust evidence of a cause and effect relationship between inflammation and depression.

Neurotransmitters and Neurocircuits

Much attention has been paid to the neurotransmitter systems and neurocircuits that are affected by inflammation and mediate its effects on the brain and behavior (Figure 1). Early studies focused on patients administered IFN-alpha as well as endotoxin and typhoid vaccination. Using functional magnetic resonance imaging (fMRI) and PET, these investigations revealed that subcortical brain regions including the basal ganglia were primary targets of inflammatory stimuli (36,70). A consistent finding across stimuli and across laboratories is that inflammation has reproducible effects on reward processing including reduced activation of the ventral striatum during reward anticipation, which in turn is associated with reduced motivation (66,71,72). Of note, data suggests that these effects may be more prominent in females (73). Alteration in neural activity of the substantia nigra in association with psychomotor slowing and increased IL-6 has also been described as has decreased connectivity between subgenual anterior cingulate cortex (ACC) and nucleus accumbens in association with depressed mood following typhoid vaccination (63,64). Studies in depressed patients have confirmed these effects of inflammation on corticostriatal networks. Indeed, increased peripheral blood CRP was found to correlate with decreased resting state functional connectivity between ventral striatum and ventromedial prefrontal cortex, which in turn correlated with increased anhedonia (74). Increased CRP as well as increased peripheral blood immune cell reactivity in depressed patients have also been associated with decreased ventral striatal responses to reward anticipation (75,76). Of note, inflammation-related disruption of corticostriatal pathways in depressed patients correlates with psychomotor slowing as well as with gene and protein markers of immunometabolic dysfunction (74,77). Consistent with the effects of inflammation on corticostriatal pathways and symptoms of anhedonia, treatment trials in depressed patients with anti-cytokine therapies have shown preferential improvement in symptoms of anhedonia as compared to overall depressive symptom severity (57,68,69). Moreover, a recent study that was focused on inflammation’s effects on motivational anhedonia showed that compared to placebo, a single dose of the TNF antagonist infliximab significantly increased objective measures of effort-based motivation in depressed patients with a CRP >3mg/L in association with changes in a network of brain regions including the ventral striatum (78). Interestingly, data from laboratory animals indicate that inflammatory cytokines including IL-6 may have direct access to ventral striatal brain regions including the nucleus accumbens during stress (79). These effects, which are associated with stress-induced depressive-like behavior in these animals, are believed to be related to breaches in the blood-brain-barrier (BBB) due to decreases in Claudin-6, which plays an important role in BBB integrity (79). Decreases in Claudin-6 expression have also been found in the nucleus accumbens in post-mortem brain samples of depressed patients (79).

In terms of neurotransmitter systems, both dopamine and glutamate have been implicated in the effects of inflammation on subcortical brain regions (3,36). Early studies in both laboratory animals and humans revealed decreased dopamine availability and release in the striatum following exposure to inflammatory stimuli including IFN-alpha and IL-6. For example, using in vivo microdialysis, non-human primates administered IFN-alpha exhibited decreased dopamine both at rest and after stimulation with potassium or amphetamine, an effect that could be reversed by administration of the dopamine precursor levodopa via reverse microdialysis (80,81). Decreased dopamine release was in turn associated with decreased effort-based motivation in these animals (80). Similar results were found in rodents administered IL-6 intraperitoneally (82). Further confirming the central role played by dopamine in inflammation’s effects of the brain and behavior is a study demonstrating that administration of levodopa to depressed patients reverses decreased functional connectivity in corticostriatal reward circuitry in individuals with high but not low inflammation as indexed by CRP (54). Moreover, repeated administration of levodopa across a range of doses to depressed patients with increased inflammation (CRP >2mg/L) was found to not only increase functional connectivity in reward circuitry but also increase effort-based motivation and reduce symptoms of anhedonia and overall depression severity (53). These data provide compelling support for a unique treatment response profile to a dopaminergic agent in the inflammatory subtype of depressed patients.

Alterations in glutamate metabolism in the basal ganglia also occur in the context of inflammatory exposure (83). Increased basal ganglia glutamate measured by magnetic resonance spectroscopy is seen following 4 weeks of IFN-alpha administration as well as in depressed patients with increased CRP (84,85). Increased basal ganglia glutamate in both IFN-alpha-treated and depressed patients is in turn associated with symptoms of anhedonia and decreased psychomotor speed (84,85). Depressed patients with both increased CRP and basal ganglia glutamate also exhibit multiple brain regions of decreased regional homogeneity (ReHo), a measure of local coherence of neuronal activity (86). Decreased ReHo is associated with decreased functional connectivity in corticostriatal reward circuitry and anhedonia. Glutamate alterations in the context of inflammation are believed to be related to the direct effects of inflammatory cytokines including TNF and IL-1beta on the function of astrocytes that play a pivotal role in controlling extracellular glutamate concentrations, which can be excitotoxic (83). Inflammatory cytokines are known to decrease the expression of excitatory amino acid transporter 2, limiting the reuptake of glutamate from the synapse (87,88). Excess glutamate can spill out of the synapse and bind to extrasynaptic N-methyl-d-aspartate (NMDA) receptors, which can decrease neuronal growth factors such as brain derived neurotrophic factor, while activating adjacent neurons, leading to chaotic, incoherent, and imprecise local neuronal signaling (87,88).

Increased inflammatory cytokines can also activate indoleamine 2,3 dioxygenase, leading to production of kynurenine (KYN) and its metabolites (89). KYN is taken up into the brain by the large neutral amino acid transporter and is converted into quinolinic acid (QUIN) by activated microglia and macrophages. Indeed, in depressed patients, plasma inflammatory mediators including TNF are strongly associated with KYN and its metabolites in the periphery which are in turn strongly associated with KYN metabolites including QUIN in the brain (90). QUIN can directly activate extrasynaptic NMDA receptors while also stimulating glutamate release and blocking glutamate reuptake by astrocytes (89). In patients with major depression, increased plasma QUIN has been associated with increased basal ganglia glutamate and together with cerebrospinal QUIN is associated with decreased functional connectivity in corticostriatal networks and anhedonia (91). Of note, another metabolite of KYN, kynurenic acid (KYNA), is an NMDA antagonist that is believed to have neuroprotective effects in the brain (89). Decreased plasma KYNA is associated with increased basal ganglia glutamate (91), and a decreased blood KYNA/QUIN ratio is correlated with increased anhedonia (92). Taken together, these data suggest a convergence of inflammatory pathways to pathology in depression involving KYN and its metabolites and glutamate, which characterize the inflammatory subtype of depression.

In addition to effects on subcortical neural networks involved in mood regulation, inflammation also engages cortical circuits involved in anxiety, arousal, and alarm as measured using fMRI (24). For example, administration of IFN-alpha was found to increase activation of the dorsal ACC in association with increased sensitivity to task-related errors (93). Moreover, increased peripheral blood concentrations of the soluble receptor for TNF (sTNFR2) in response to a psychosocial laboratory stressor were associated with increased activity of the dACC and insula during a social rejection task (94). Administration of IFN-alpha as well as endotoxin has been associated with increased activation of the amygdala in response to sad faces (95,96), whereas administration of the TNF antagonist infliximab reduced amygdala reactivity during sad face exposure (96). Finally, in a sample of highly stressed inner city subjects, plasma CRP was associated with decreased functional connectivity between the amygdala and the ventromedial prefrontal cortex which in turn was associated with anxiety symptoms (97). These data suggest that inflammation impacts multiple cortical and subcortical brain regions and circuits leading to a host of symptoms that may align more closely with an inflammatory subtype of depression as opposed to the currently conceived and overinclusive nosology reflected in what is now diagnosed as major depression.

Clinical Trials

Lessons Learned

Given the detailed understanding of the pathways from the immune system to the brain and ultimately behavior, there is an unprecedented opportunity for the development of treatments for the inflammatory subtype of depression (Figure 1). Unfortunately, results from clinical trials using anti-inflammatory drugs to treat otherwise healthy depressed individuals have been underwhelming and inconsistent (98), with two of the largest clinical trials to date showing no effect of either celecoxib or minocycline, two of the most commonly used anti-inflammatory agents to treat depression (99,100). Not surprisingly, most clinical trials have used outdated clinical trial designs that have taken a “one size fits all” approach with heterogenous populations of depressed patients and non-specific outcome variables such depressive symptom severity as represented for example by the Hamilton Depression Rating Scale (HAM-D)(101,102). Moreover, whether the treatment has hit the target (e.g., reduced inflammatory biomarkers in the brain or periphery) has rarely been assessed. Of note, when inflammatory biomarkers such as CRP or IL-6 have been used to stratify depressed samples, as noted above, treatment effects have been observed for infliximab, minocycline, and levodopa and especially for symptoms related to anhedonia (52,54,57,68). Albeit in studies with relatively small sample sizes, these data reinforce the notion that mechanism-based clinical trials will require more tailored designs that are aligned with the populations and outcomes of interest as well as the target of treatment.

Designs for the Future

Several components are inherent to mechanism-based clinical trials for the inflammatory subtype of depression: enrichment, outcome alignment, and target engagement.

Enrichment:

Strategies for enrichment may include the use of biomarkers for immune and/or metabolic pathways that characterize the inflammatory subtype of depression. Based on guidelines provided by the US Food and Drug Administration (FDA)(103), these biomarkers should be developed within a context of use (COU) where a “predictive” biomarker can be used to enrich for a subgroup of depressed patients (i.e., inflammatory subtype) who are more likely to respond to a particular therapeutic agent (e.g., anti-inflammatory drug). Biomarkers must go through a rigorous process of validation including the establishment of reliable measurement parameters and a qualification process that links the biomarker with biological mechanisms and clinical endpoints. It should be noted that for FDA approval, a drug tied to a biomarker should be effective in patients with the biomarker and not in those without. Establishing such efficacy in a match/mismatch design can add considerable complexity and cost to the drug development process. There are several candidate biomarkers that have received attention for the inflammatory subtype of depression including notably CRP. CRP is measured under standardized conditions throughout the US and has been linked to multiple aspects of the impact of inflammation on the brain and behavior in depression as noted above. Moreover, peripheral blood CRP concentrations are associated with other inflammatory biomarkers both in the peripheral blood and cerebrospinal fluid (104). Thus, CRP may be a good starting point to enrich for depressed patients with the inflammatory subtype. Of note, establishing a diagnostic subgroup within major depression (e.g., by using an inflammatory specifier) could potentially obviate the need for enrichment and allow drug development efforts to specially focus on an inflammatory subtype (105).

Outcome alignment:

Like with biomarkers, the FDA has a formal regulatory process for establishing clinical outcome assessments (COAs) that are well-defined, validated, and reliable assessments of patients’ symptoms, overall mental state, or function (106). It should be noted that according to the FDA, the qualification pathway for a novel COA “is a multi-year process that requires a high degree of commitment” (106). Nevertheless, given that inflammation has rather specific effects on neurotransmitters and neurocircuits in the brain, it is unlikely that the currently FDA accepted, non-specific measures of depressive symptom severity as reflected by scales such as the HAM-D will be sufficient for novel drug development for the inflammatory subtype (101). Indeed, there are no currently FDA-approved measures for neurovegetative symptoms or anhedonia. Thus, the development of novel COAs that align with the effects of inflammation on the brain is imperative (101).

Target engagement:

Although not part of the regulatory process, a major advantage of developing therapeutics for a mechanism-based nosology is that it can be readily determined early in the drug development process whether a treatment is “hitting the target.” In the case of the inflammatory subtype, virtually all of the treatment targets can be measured as proximal endpoints, thus obviating the reliance on behavioral outcomes in the early phases of drug development. Such proximal endpoints can be highly specific with limited variance, therefore requiring a much smaller sample size for proof-of-concept. A recent example was the use of functional connectivity within reward circuitry as a proximal endpoint to assess the efficacy of levodopa in patients with high versus low inflammation (53). Demonstration of improved connectivity following a single dose of levodopa in depressed patients with high inflammation led to a follow-up clinical trial indicating sustained improvement in effort-based motivation, anhedonia, and HAM-D scores following repeated administration of levodopa to patients with high inflammation (53).

Conclusion

A tremendous amount of progress has been made in detailing the pathways leading from an activated immune response to changes in behavior that can contribute to diseases such as major depression. Recognition that this is but one mechanism leading to pathology allows a modernization of our current nosology of depression and ushers in the exciting new era of precision medicine. Nevertheless, in order for the field to embrace these opportunities, all stakeholders will need to fully commit to accelerating the incorporation of biomarkers and novel clinical outcome assessments into our clinical trial designs. Such a commitment will lead us into the next generation of targeted therapeutics.