Is there Progress? An Overview of Selecting Biomarker Candidates for Major Depressive Disorder

Major depressive disorder (MDD) contributes to a significant worldwide disease burden, expected to be second only to heart disease by 2050. However, accurate diagnosis has been a historical weakness in clinical psychiatry. As a result, there is a demand for diagnostic modalities with greater objectivity that could improve on current psychiatric practice that relies mainly on self-reporting of symptoms and clinical interviews. Over the past two decades, literature on a growing number of putative biomarkers for MDD increasingly suggests that MDD patients have significantly different biological profiles compared to healthy controls. However, difficulty in elucidating their exact relationships within depression pathology renders individual markers inconsistent diagnostic tools. Consequently, further biomarker research could potentially improve our understanding of MDD pathophysiology as well as aid in interpreting response to treatment, narrow differential diagnoses, and help refine current MDD criteria. Representative of this, multiplex assays using multiple sources of biomarkers are reported to be more accurate options in comparison to individual markers that exhibit lower specificity and sensitivity, and are more prone to confounding factors. In the future, more sophisticated multiplex assays may hold promise for use in screening and diagnosing depression and determining clinical severity as an advance over relying solely on current subjective diagnostic criteria. A pervasive limitation in existing research is heterogeneity inherent in MDD studies, which impacts the validity of biomarker data. Additionally, small sample sizes of most studies limit statistical power. Yet, as the RDoC project evolves to decrease these limitations, and stronger studies with more generalizable data are developed, significant advances in the next decade are expected to yield important information in the development of MDD biomarkers for use in clinical settings.

Hypothalamic–Pituitary–Adrenal Axis (DST, DEX/CRH, Cortisol Response, Hypocretin) HPA-axis hyperactivity has been associated with a spectrum of neuropsychiatric disorders due to its deleterious effects on the nervous system including dendritic process atrophy, decreased neurogenesis and neuroplasticity, and neuronal losses ( 18 , 19 ); consequently, a wide range of biomarkers may be disrupted by HPA-axis dysfunction, such as disturbed adrenocorticotropic hormone (ACTH) regulation, dysfunctional corticosteroid receptor signaling, and glucocorticoid (GR) excess ( 18 ). Furthermore, mutations in genetic regions involved in abnormal HPA-axis function (such as the FKBP5 allele) have also been associated with an increased risk for depression, and are similarly associated with abnormal plasma cortisol and ACTH concentrations ( 20 – 23 ). However, beyond genetic factors, epidemiologic and clinical studies have determined that disturbances in HPA axis function have been consistently associated with biological changes in depression ( 24 , 25 ). For example, one facet of depression history that is associated with HPA axis changes is early life stress. Early life stress (e.g., maltreatment or abuse) was found to result in HPA axis dysfunction during childhood and adolescence and contributed to an increased risk of developing MDD later in life ( 26 ). Moreover, diminished cortisol suppression following dexamethasone (DEX) administration was observed in MDD patients with metabolic abnormalities of prefrontal and hippocampal regions, areas often related to MDD pathology ( 27 ). Other studies found that antidepressant treatment often resulted in decreased cortisol levels and a return to normal HPA axis function ( 28 , 29 ). Originally, as corticotrophin-releasing hormone (CRH) has been reported to be associated with increased depressive symptoms such as anhedonia and reduced appetite ( 30 ), a combined DEX/CRH test was thought to be capable of increasing diagnostic power over the Dexamethasone Suppression Test (DST) ( 12 , 31 ). However, abnormal DEX/CRH results also occur in other psychiatric disorders resulting in lack of specificity as a diagnostic biomarker for major depression ( 12 ). Measuring cortisol levels is a more direct and accurate method of assessing HPA axis activity in depressed patients ( 32 ). Additionally, more recent studies focusing on cortisol measurements have demonstrated a link between cortisol levels and depression severity or depressive subtypes. A recent meta-analysis ( 33 ) reports a significant association between HPA-axis hyperactivity as measured by elevated cortisol levels and the presence of melancholic or psychotic depression while lower cortisol levels were characteristic of depression with atypical features. For example, a longitudinal study of adolescents with depressive symptoms found that male adolescents with high morning salivary cortisol levels and increased depressive symptoms were more susceptible to the development of MDD demonstrating a sex-linked differentiation ( 34 ). Another study also reported that persistent increases in cortisol awakening response (CAR) in adolescents more strongly correlated with higher levels of depressive symptoms than with anxiety symptoms ( 35 ). Lastly, a large cohort study confirmed increased CAR and dynamic cortisol secretion in depressed patients compared to controls in both current MDD and remitted MDD subjects, indicating that both measurements reflect an inherent risk in the development of depression ( 29 ). These studies suggest the use of morning salivary cortisol as a trait-like biomarker for developing preventative measures for high-risk populations, especially in asymptomatic individuals with possible genetic risks ( 36 ). However, a recent study revealed that increased CAR in healthy female adolescents significantly correlated with higher magnitudes of Profile of Mood States (POMS) subscale scores for “Tension-Anxiety,” “Depression-Dejection,” “Fatigue,” and “Confusion” ( 37 ) suggesting that morning salivary cortisol levels may also be descriptive of mood states and episodic depressive symptoms rather than characteristic of a purely trait marker for MDD. Such findings suggest variability in the use of salivary cortisol as a depression biomarker. However, it is important to consider how these contrasting conclusions may be affected by methodological heterogeneity and differences in subject populations among these studies. Another possible biomarker source includes hypocretin, a neuropeptide that plays a role in sleep and arousal. Recently, it has been suggested that decreased numbers and size of hypocretin-containing neurons may be associated with the development of depressive symptoms including eating/drinking behaviors and disrupted sleep ( 38 , 39 ). One study found that hypocretin levels in the CSF of MDD-diagnosed patients with high suicidal ideation were significantly lower than those of patients with dysthymia and adjustment dis

There also exists an abundance of evidence that elevated proinflammatory cytokine concentrations and an increased immune response are associated with depression diagnosis, symptomatology, and severity ( 65 – 69 ). Reflective of this significant proinflammatory response in depressive disorders, a recent proteomic study found elevated levels of acute phase reactants (e.g., ferritin, serotransferrin, Haptoglobin-related protein, ceruloplasmin) and proinflammatory markers (e.g., IL-16, MIF, Tenascin-C, and EN-RAGE) in drug-naive MDD patients, indicating a disorder-related increase in immune processes ( 70 ). These findings are supported by neuroimaging and animal studies, which have demonstrated that alterations in neuroplasticity promote manifestations of depressive phenotypes as a result of cytokine-induced neural apoptosis and metabolic dysregulation ( 71 ). Further, inflammatory cytokines have been described to alter basal ganglia processes leading to common depressive phenotypical characteristics including anhedonia, fatigue, and psychomotor retardation ( 72 – 75 ). A recent review also reported increased neopterin levels in a number of studies of depressed patients, specifically in melancholic subtypes of depression ( 76 ). Additionally, inflammatory markers IL-6 and soluble intercellular adhesion molecule (sICAM) have been associated with sleep disturbances in depressed patients ( 77 ). IL-8 and TNF-α have also been reported to remain elevated in certain subsets of depressed patients after antidepressant therapy, indicating possible trait characteristics ( 78 ). C-reactive protein (CRP) and interleukin (IL)-6, specifically, have been found to exhibit trait characteristics (i.e., gender effects, impact of early life adverse events) as an inflammatory biomarker for depressive pathology ( 79 – 82 ). In their meta-analysis, Valkanova and colleagues ( 83 ) found that these two putative analytes had a small but significant association with the development of depressive symptoms, indicating the presence of raised inflammatory markers preceding the development of MDD. However, the authors cautioned that their results might have limited significance due to heterogeneity (i.e., of depression, methodologies, populations, etc.) across studies. For instance, one study ( 84 ) found both higher and lower levels of different inflammatory markers in major depressed patients depending on the presence or absence of melancholic features, indicating that that the overall characteristics of depressive symptoms were more associated with the composition of inflammatory profiles and less so on concentrations of individual markers. Moreover, a study of an elderly population found that when controlling for age-related chronic diseases, CRP was not a statistically significant marker associated with the presence of MDD or sub-threshold depression ( 85 ). Lastly, one group reported significantly lower levels of IL-6 in subjects with high self-reported depressive symptoms while showing no significant differences of IL-8, IL-10, and TNF-α levels when compared to controls ( 86 ). As a whole, these studies demonstrate the complexities of relying on individual inflammatory marker concentrations to characterize generalized depression. Yet, there is reasonable evidence that suggests inflammatory responses are more prominent in certain subsets of MDD than others. A study evaluating biomarker associations with depressive subtypes found that increased inflammatory markers (i.e., CRP, IL-6, TNF-α) were significantly associated with atypical depression as compared to typical or melancholic depression ( 87 ). Consistent with these results, a more recent study ( 88 ) reported that elevated IL-6 levels were consistently higher in patients with atypical depression. Similarly, a recent study has detected consistently increased CRP levels in depressed patients with comorbid diabetes mellitus ( 89 ). Moreover, these studies reinforce clinical evidence that both inflammatory diseases and depression are often associated with comorbid illnesses like metabolic disorders ( 87 , 90 ), especially in more elderly subjects ( 91 – 93 ). It is possible that in many cases, the predisposition to depression in patients with elevated inflammatory biomarker concentrations is affected by a number of outlying factors often present before the emergence of the first depressive symptoms. Therefore, in addition to their lack of specificity to MDD, the diagnostic value of individual inflammatory biomarkers could be hindered by some inherent heterogeneity of depression. Although they may be useful as MDD biomarkers in a research environment, their low sensitivity and specificity ( 6 ) prevent them from being utilized in the majority of clinical settings. In contrast to the relatively limiting findings of individual inflammatory analyte concentrations as biomarkers for depression, inflammatory markers have potential use as state markers by characterizing treatment response to

Neurotrophins [i.e., nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), neurotrophin (NT)-3, -4, and -5] are homodimeric proteins that interact with the tropomyosin receptor kinase (Trk) family of receptors through which they mediate the processes of neurogenesis and neural plasticity in both the peripheral and central nervous systems ( 127 ). Several connectomic studies have increasingly indicated the disruption of integral whole-brain structural networks in MDD, suggesting the presence of abnormal neuronal synapse formation within certain populations of depressed patients ( 128 , 129 ). In fact, there is evidence that disrupted neurogenesis may be a characteristic of MDD pathophysiology, especially of the hippocampus ( 130 , 131 ). Due to the role of NTs in neuroplastiticty, their use as potential biomarkers has often been reviewed. Of these, the most researched is BDNF, with studies finding its downregulation in the limbic structures of chronic-stress exposed rats and reports of decreased peripheral levels in MDD patients ( 132 – 136 ). Significantly, a recent study ( 137 ) has shown serum BDNF may also have significant potential as a discriminatory diagnostic tool for first major depressive episode (MDE) patients, prompting the need for more expansive studies concerning its use in clinical settings. While there have been conflicting results concerning correlations of depression severity with BDNF levels ( 138 ), BDNF concentrations have been reported to increase after antidepressant therapy with more prominent elevations in patients with higher baseline depression severity ( 139 – 145 ). Several studies have also reported elevated BDNF levels in responders to antidepressant treatment compared to non-responders that continued to demonstrate lower BDNF concentrations after pharmacologic management ( 146 , 147 ). These findings suggest that BDNF may be utilized as a state marker to assess psychopharmacological therapy and prognosis of individual MDD patients ( 148 ), although the effect on BDNF levels may vary between different classes of antidepressants ( 149 ). Ultimately, due to its intrinsic function in influencing the development and maintenance of a patient’s cognitive abilities, BDNF could have potential for evaluating other therapy effects involving learning, memory, and executive functions ( 134 , 150 ). Alternatively, de Azevedo Cardoso and colleagues ( 151 ) have suggested that BDNF may also have trait-like properties. For example, differences between male and female BDNF levels have been associated with contrasting antidepressant effects between the two genders ( 145 ). Additionally, there is evidence that BDNF levels are more negatively affected in patients with chronic depression who have experienced more adverse life events ( 152 ). Supporting this theory, several groups have illustrated how BDNF genotypic variations were associated with risk for depression ( 151 , 153 – 157 ). BDNF DNA methylation patterns have also been associated with depression severity, and the presence of suicidal ideation in MDD subjects ( 158 – 160 ). Consequently, the potential for BDNF to be a trait and state-like marker makes it one of the more versatile biomarker candidates being researched today. In contrast, fewer studies have focused on other NTs, with the notable exception of NGF, which has been found to be increased during circumstances that cause anxiety or anticipation of anxiety ( 161 ). Regarding NGF’s relationship to depressive symptoms, reports of NGF concentrations have yielded conflicting results ( 151 , 162 , 163 ). Additionally, due to its association with other affective disorders such as bipolar disorder (BPD) ( 164 ); it is unlikely to be specific to MDD. These limitations, however, should not preclude it from further research.

Epigenetic mechanisms have been used to explain how early life exposures to toxic or stressful stimuli may contribute to the predisposition or development of mental illness ( 195 ). For depressive disorders, histone modification at the amino (N)-terminal tails and DNA methylation has been the most studied in determining how epigenetic factors affect the progression, severity, symptomatology, and treatment response of depression ( 196 , 197 ). Significantly, these epigenetic modifications may affect expression of certain receptors (e.g., GR receptors in the hippocampus), which leads to either an increased or decreased risk for depression in the future ( 196 ). This is supported by animal studies that show antidepressant-like effects of histone deacetylase inhibitors ( 195 , 196 , 198 – 200 ), which are thought to induce histone acetylation in certain regions of the brain. Overexpression of DNA methyltransferases also leads to an increase in DNA methylation and has been associated with abnormal dendritic spine plasticity and alterations in behavioral responses. Supporting this, one group found site-specific hypermethylation of TrkB-T1 to be increased in suicide completers ( 201 ), suggesting a pattern of methylation abnormalities in subjects with depressive phenotypes. Recently, epigenome-wide association studies have demonstrated several genes with methylation associations in depressed subjects compared to controls. As these studies are mostly array-based, they have had the advantage of investigating the entire genome, but replication studies are currently lacking ( 202 ). One recent genome-wide study ( 203 ) was able to separate medication naive MDD subjects from controls by observing differences at 363 CpG sites that differed from the pattern they observed in their schizophrenia patients ( 204 ) indicating disease-specific patterns. Furthermore, several candidate gene studies involving DNA methylation have been investigated and include genes that have been previously implicated in depression. Among these genes, SLC6A4, BDNF, and NR3C1 have been the most studied, with BDNF methylation having the most consistent data concerning associations between DNA methylation and depressive symptoms/antidepressant response ( 202 ). This study demonstrated a significant association between depression and methylation levels of BDNF at specific CpG sites. Notably, the authors have shown that such robust biomarkers may come from easily obtainable specimens such as buccal samples. Although epigenetic research is still in its infancy, these epigenetic mechanisms and resulting patterns in chromatin remodeling are becoming established as a basis by which chronic social defeat, early life stress, variability of maternal care, and antidepressant therapy may influence the progression or resolution of depressive symptoms ( 197 , 202 , 205 , 206 ). Further studies and elaboration on these mechanisms will likely lead to significant advances in the development of an epigenetic model from which MDD biomarkers may be retrieved. Please see Nestler et al. ( 195 ), Tsankova et al. ( 197 ), and Januar et al. ( 202 ) for further review.

Though there have been a number of studies analyzing the various neurobiological features persistently found in depressed patients, no specific marker from a single biological system has been capable of significantly improving upon the current diagnostic criteria set for MDD patients. As several of the aforementioned biomarkers seem necessary but not individually sufficient, multiplex assays are currently the most promising to contribute consistent results to aid in further standardizing MDD diagnosis and research. As past studies have demonstrated, depression pathology is influenced by disruption from multiple systems including the HPA axis, oxidative pathways, inflammatory processes, and neurotrophic homeostasis. Collectively measuring the putative analytes of each system will likely increase the power of any diagnostic panel developed for MDD. This concept is supported by studies that used multiple analytes of different origins and considered to be potential biological markers of depressive disorders to increase specificity and sensitivity in diagnosing MDD. One study worth noting achieved high sensitivity (above 90%) and specificity (above 80%) in distinguishing MDD patients from healthy controls ( 218 ). The authors used nine biomarkers from different biological sources such as inflammatory and oxidative indices (α1 antitrypsin, apoplipoprotein CIII, myeloperoxidase, soluble TNFα receptor type II); the HPA axis (epidermal GF, cortisol); neurogenesis (BDNF); and metabolism (prolactin, resistin) to develop an algorithm that produces a score that could potentially be used for an objective diagnosis of MDD. In addition to achieving high sensitivity and specificity in their pilot study, Papakostas et al. also produced a similar performance in their replication study. This group further refined their model algorithm by factoring in gender, BMI, and normalized cortisol levels ( 219 ). Another group used CSF concentrations of multiple analytes including inflammatory biomarkers (IL-6), serotonin metabolites (5-hydroxyindoleacetic acid), dopamine metabolites (homovanillic acid), and HPA axis biomarkers (hypocretin) to detect severe suicidal behavior and increased risk of completing suicidal attempts in MDD patients ( 220 ). Likewise, CSF protein biosignatures were found to be capable of discriminating depressed, bipolar, and schizophrenic patients from healthy controls ( 221 ). These markers included proteins involved in neurogenesis (e.g., neuronal growth regulator 1, neural proliferation differentiation and control protein); neurotransmission (seizure-related 6 homolog protein); and oxidative damage (GPX 3). However, Maccarrone and colleagues have indicated difficulties differentiating between individual psychiatric disorders and controls, as only a few proteins of their CSF biosignatures were found competent enough to distinguish between disease groups. They have reported high accuracy rates of distinguishing bipolar, depressed, and schizophrenic patients (i.e., 83.3% for MDD). Other multiplex studies that have been discussed in previous sections have also shown significant inflammatory/oxidative features ( 70 ) and epigenetic variations ( 203 ) in MDD subjects. Due to their inherent sophistication and more comprehensive analysis relative to individual markers, these multiplex assays have the potential to reduce inconsistent data that develop due to differences in study populations and methods seen in past, single biomarker studies ( 219 ). However, it is currently imperative to conduct future studies that focus on replicating and confirming such findings that yield increased MDD diagnostic accuracy using these methods.

Multiple biological pathways are robust sources of tissue-based MDD biomarkers with trait and state characteristics. However, individual biomarkers currently impart limited clinical utility. In the future, multiplex assays comprised of putative depression biomarkers may improve upon the clinical evaluation of MDD, assess treatment efficacy, and serve to standardize discharge criteria. However, independent replication studies with large sample sizes are needed to fully substantiate the validity of such panels. Furthermore, the use of these markers are limited by high costs and confounding factors associated with each component of prospective diagnostic constituents ( 225 ). If these markers become reproducible and translate into readily available diagnostic tools with ease of access, low cost, rapid formulation, and high sensitivity/specificity, the implications for clinical use would be tremendous. After decades of investigations and several promising markers falling into obscurity, it is difficult to say whether we are getting closer or farther away from one of the holy grails of diagnostic biomarkers for depression. Suffice it to say, every study that contributes to the development of such biomarkers will assuredly be needed if such a goal is to be achieved. As the RDoC project and current technology evolve to lessen the limitations of past studies, future large-scale MDD biomarker studies will be necessary to yield advances that will hopefully have utility in the clinical setting.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.