Introduction

Major depression disorder (MDD) is a leading cause of disability worldwide and ranks as the fourth leading cause of morbidity, according to the World Health Organization [1]. Estimates of the lifetime prevalence of depression range from 7 to 20% [2] and the 12-month prevalence can be as high as 5% [1].

Major guidelines recommend pharmacotherapy as first-line treatment for moderate to severe MDD [3, 4]. Antidepressants are among the most commonly prescribed medications worldwide [5] and have proven efficacy in reducing depressive symptoms. Currently, clinicians have access to several treatment options belonging to different classes of antidepressants, but none of them has clear evidence of superiority over the others and the remission rates with antidepressant therapy are still concerningly low [6]. Indeed, only  ~ 30% of patients achieves remission after the first treatment, and even after four trials of treatment the percentage of remission does not reach 70% [7, 8]. After each treatment failure, the chance of response to the next antidepressant decreases and the risk of treatment-resistant depression (TRD), which is commonly defined as the lack of response to at least two treatments, increases [9, 10]. The trial-and-error process of finding an effective antidepressant can be prolonged and demoralizing, leading to delayed recovery and potentially contributing to the chronicity of the disease. Additionally, lack of treatment response exposes patients to a range of distressing and debilitating side effects [11, 12], which can undermine the therapeutic alliance.

Therefore, understanding the individual factors that influence treatment response in MDD is of primary importance to improve outcomes related not only to depressive symptoms, but also quality of life and overall functioning. Treatment response has a hereditary basis—for example, a concordance of antidepressant response- has been demonstrated in affected members of the same family [13] and a significant single nucleotide polymorphism (SNP)-based heritability (h2SNP) was found by genome-wide association studies (GWASs) [14]. Clinical and socio-demographic variables also contribute to treatment outcomes [15, 16]. Certain risk factors, such as suicidality and comorbid anxiety, may have a genetic basis overlapping with that of treatment response; conversely, socio-demographic variables and clinical factors such as duration and severity of the depressive episode, may exert effects independent from the genetics involved in treatment outcomes [17].

Although MDD is conceptualized as a single disorder, its diagnosis is formulated from a combination of symptoms that present with considerable variability, as over 1000 unique symptom combinations can be observed [18]. This phenotypic variability likely reflects the biological and environmental heterogeneity among patients, and may be partly linked to the heterogeneity observed in treatment response. Strong evidence shows that certain symptoms tend to co-occur, identifying subtypes of depression [19]. This is the case, for example, of atypical depression, which is characterized by the presence of mood reactivity and reversed neurovegetative symptoms (e.g. increased appetite/weight and hypersomnia). These symptoms likely have similar pathophysiological correlates and share common polygenic liabilities [20, 21]. On the other hand, distinct symptom domains are expressions of the heterogeneous genetic architecture of depression [22]. Consistently, specific symptom profiles tend to exhibit greater responsiveness to particular medications because of their pharmacodynamic profiles, in line with recommendations by the Canadian Network for Mood and Anxiety Treatments (CANMAT) guidelines [4]. Furthermore, certain depressive symptom dimensions such as anxiety, hypersomnia, anhedonia, suicidality, are associated with poorer treatment response, suggesting shared genetics between specific depressive symptom profiles and treatment outcomes [18].

Therefore, the aims of this narrative review are to (1) summarize the genetic factors associated with treatment efficacy outcomes in MDD; and (2) to explore the overlap between these genetic factors and those involved in specific symptom dimensions previously associated with treatment outcomes in MDD. This approach can provide information useful to identify specific biomarkers for treatment personalization, with implications for future research and clinical practice. To reach the second aim, we first reviewed the literature to identify the genetic signals associated with MDD treatment outcomes, focusing on GWASs and whole exome sequencing studies (WESs), given the instability of results from candidate gene studies [23, 24]. Then, we investigated whether these genetic associations overlap with those of MDD treatment outcomes. The rationale of this approach is to contribute to interpreting the existing literature in terms of genetic factors associated with treatment outcomes that may also be linked to specific symptoms, thereby helping to dissect the biological and clinical heterogeneity of poor treatment response and suggesting new treatment targets specific to certain symptom dimensions rather than MDD overall.

Methods

We searched PubMed, GWAS catalog (https://www.ebi.ac.uk/gwas/), GWAS atlas (https://atlas.ctglab.nl/), and medRxiv (https://www.medrxiv.org/) for original research articles on the following topics: (1) GWASs or WESs that investigated efficacy outcomes in MDD (response, remission, symptom improvement, or TRD, as defined in each original study); (2) the association between MDD treatment outcomes and specific depressive symptoms/dimensions; (3) GWASs that investigated the genetics of the symptom dimensions identified in (2).

When more GWASs were available in (3) for the same trait, we focused on the largest study demonstrating a significant h2SNP. We considered association signals at locus, gene, and gene set level. As the main topic of this work was to review the genetics of treatment outcomes in MDD, for studies in (1) we decided to present both genetic associations surviving multiple testing correction and suggestive association signals (p-value  <  5 × 10–8 and 5 × 10–8  ≤  p-value < 5 × 10–6 at locus level, respectively). Although this was not a systematic review, we aimed to provide a comprehensive consideration of GWASs and WESs relevant to MDD treatment outcomes.

Finally, based on the overlap between genes identified (either significant or suggestive associations) from the SNP- and gene-level analyses of treatment outcomes and those significatively associated with the clinical dimensions in (3), we conducted a gene-drug interaction analysis via www.DGIdb.org [25].

Results

Genetic associations with treatment outcomes

A total of 13 GWASs and one WESs were included in this review, and their main characteristics are summarized in Table 1. Most studies had relatively small sample size (i.e., < 10 K), with only three exceptions, including one large study with 154,433 participants [26]; only 6 studies demonstrated a significant h2SNP of the outcome(s) of interest, with values generally around 0.8–0.10.

Table 1 Characteristics of the genetic studies included for treatment outcomes (A) and associated clinical dimensions (B)
Full size table

At locus level, 11 genome-wide significant associations were identified, spanning across multiple genes, including ITGA9, NRXN3, UST, MECOM, FTO, and MCHR1 (Table 2A). Additional suggestive signals were identified, including for example LINGO2, CACNA1C, PRG3, ITGA1, EPHB1 and SLC27A1 genes (Supplementary Table 1). However, we underline that these non-significant results have unclear relevance and were not replicated across studies.

Table 2 Genome-wide significant variants (A), genes (B), and gene sets (C) associated with AD treatment outcome
Full size table

A total of 25 genes were associated with the outcomes in the gene-level analyses, including LZTS3, PRNP, OR4K2, PPFIBP1, and GPHA2 (see Table 2B for all results). Suggestive associations were identified, including ADGRG5, MAP3K2, the solute carrier genes SLC17A4 and SLCO3A1, the glutamate receptor gene GRM3,, and genes encoding for many zinc finger proteins (Supplementary Table 2). However, none of these were replicated across studies.

Noteworthy, both suggestive and significant association findings had some consistency in terms of biological processes involved at the pathway level, showing an involvement of the immune system (e.g.: NCR3, LST1, and LCN2), synaptic transmission and dendritic spine formation (e.g.: PPFIBP1, PRNP, LZTS3, and NRXN3), neurogenesis and differentiation (e.g.: NRXN3, MECOM, CGREF1 and MAP1A), and regulation of transcription and gene expression (e.g.: DHX8, MECOM, ETV4, MEPCE, and PFAS). These observations are supported by the findings from gene set enrichment analyses (GSEAs). In particular, while no enrichment for specific gene sets was replicated across studies, GSEAs showed that many gene-sets are involved in or regulate similar biological processes. For example, signal transduction (Rhodopsine-like receptors A/1 R-HSA-373076 and Calcium-activated potassium channel activity GO:0015269); gene expression and nuclear functions (Chromosomal part GO:0044427 and Chromosome pathway GO:0005694), neurotransmission and synapse activity (Neuronal action potential GO:0019228, Transmission of nerve impulse GO:0019226, Long term potentiation hsa04720), immune function (Lymphocyte mediated immunity GO:0002449) ( Table 2C and Supplementary Table 3).

Genetic associations with symptom dimensions

Nine symptom dimensions were identified as associated with poor treatment outcomes in MDD and studied by previous GWASs showing significant h2SNP, namely: anxiety symptoms, neuroticism (including symptoms of apathy, worthlessness, guilt, loneliness, and excessive worry), anhedonia, cognitive functioning, melancholia, suicide attempt, psychotic symptoms, sleep symptoms, and sociability. No GWASs showed significant h2SNP for other symptoms also associated with poor therapeutic outcomes, such as irritable mood, inner tension, dissociative symptoms, and reverse or typical neurovegetative symptoms (with the exception of sleep-related symptoms) [15, 16, 27, 28].

The main characteristics of the included studies are summarised in Table 1, whilst full details and results are reported in the Supplementary Tables 1, 2, and 3. The selected GWASs were performed on large samples, in the range of 500 K participants, and all reported multiple significant genetic associations with the traits of interest (Table 1 and Supplementary Tables 1 and 2).

Given the focus of this review, a comprehensive description of these results is beyond our aims, while we were interested in describing the potential overlap with the genetic associations found for MDD treatment outcomes. Among the genes significatively associated with treatment outcomes, four overlapped with those associated with symptom dimensions, more precisely with anxiety (CGREF1), neuroticism (MCHR1) and sleep (FTO, NRXN3) (Table 3, in bold).

Table 3 Genes associated with AD treatment outcomes overlapping with genetic signals associated with symptom dimensions
Full size table

When considering also genes suggestively associated with treatment outcomes, additional signals were found in common with anxiety, anhedonia, executive functions, and sociability, as well as more overlapping genes for neuroticism and sleep (Table 3).

At gene set level, exact matches were found only with sleep, involving gene sets related to synaptic activity (hsa04730, hsa04730), G alpha signalling (R-HSA-418594), taurine/hypotaurine metabolism (hsa00430), immune response (BIOCARTA_VEGF_PATHWAY, hsa04612), and Alzheimer’s disease (hsa05010). However, even if no exact gene set overlaps were found between MDD treatment outcomes and other symptom dimensions, we identified patterns of possible overlap in some biological mechanisms involved. For example, gene sets associated with neuroticism, executive functions, and suicide attempts are predominantly related to neurogenesis and neurotransmission. Similarly, also other biological processes are involved both in MDD treatment outcome and some of the clinical dimensions, like synaptic plasticity (executive functions), immune response (suicide attempt, sleep), and nucleic acid and gene expression (suicide attempt) (Supplementary Table 3).

New therapeutic targets for treating symptom dimensions associated with poor response?

Based on the overlap discussed in the previous paragraph, we conducted a gene-drug interaction analysis. Of the four significant overlapping genes, no compounds were found to interact with either NRXN3 or CGREF1, while interactions were found only for the melanin concentrating hormone receptor 1 (MCHR1) and the fat mass and obesity associated gene (FTO, also known as ALKBH9). MCHR1, also known as SLC1, is a G-protein coupled receptor (GPCR) which binds the melanin-concentrating hormone (MCH, or PMCH) and inhibits cAMP accumulation while stimulating intracellular calcium influx. MCH is likely involved in the regulation of feeding behaviour, mood, sleep–wake cycle and energy balance [29]. Four still non-approved compounds resulted to interact with MCHR1, of which SNAP-7941 is the lead compound of MCHR1-inhibitors and displayed promising anxiolytic, antidepressant, and anorectic effects, even though not replicated in clinical trials [30]. Another MCHR1 antagonist, BMS-830216, is currently in phase 2 for the treatment of obesity [31].

Concerning FTO, eight compounds were found, all already approved as antineoplastics (INFɑ-2A, INFɑ-2B, mercaptopurine, and bisantrene), antiviral (ribavirin), antiarrhythmic (atenolol), antihypertensive (atenolol, hydrochlorothiazide), and disease-modifying antirheumatic drugs (azathioprine). FTO’s exact physiological function is yet to be uncovered; however, it is a non-heme iron enzyme located in the nucleus and likely related to growth, development, BMI, obesity, and type 2 diabetes mellitus [32].

When considering the overlap with genes associated with poor treatment outcome at a suggestive level, interesting gene-drug interactions were found for GRM3 with risperidone and two selective mGluR2/3 agonists (LY2969822 and LY404039, and the corresponding prodrug of the latter LY2140023); and CACNA1C with haloperidol, citalopram, valproate, and gabapentinoids. For all gene-drug interactions, see Table 3.

Discussion

The identification of genetic factors modulating MDD treatment outcomes has been challenging and led to few clinical applications, limited to genes involved in drug metabolism [33]. In the 50 years after the first evidence of a substantial heritability coming from family genetic studies [33], many studies focused on candidate genes, with minor and mainly unreplicated findings. In the last 10 years, GWASs produced more interesting findings, due to a more extensive coverage of the genome and larger samples, as discussed in this review. To date, it has been estimated that genetics may account for up to 60% of the variance in treatment resistance according to pedigree-based heritability [34].

However, the polygenic nature of MDD treatment outcomes and the relatively limited size of most samples resulted in scattered results which do not generally overlap, at least at SNP or gene level. The redundancies among the pathways modulating antidepressant outcomes and the heterogeneity of depressive symptoms are likely involved in the discrepancy of results at SNP and gene level [35,36,37,38]. The approach used in the present review aimed to partially overcome these issues, by integrating the genetic signals associated with antidepressant outcomes and specific symptom dimensions of clinical relevance, and by extending the analysis to pathways, pointing out potential mechanisms involved in treatment resistance and possible treatment targets.

Both at variant/gene level and gene set level, treatment outcomes were linked to gene expression regulation, central nervous system (CNS) development, synaptic plasticity, and immune system activity. One speculative interpretation of these results is that anomalies in the regulation of gene expression concur with abnormal CNS development (from tissue differentiation to synapse formation) and with an aberrant brain-immunity interplay, resulting in increased risk of developing more severe and less treatment-responsive MDD.

We identified four genes that were significantly associated with both treatment outcomes and the clinical dimensions of interest, namely CGREF1 (anxiety), MCHR1 (neuroticism), FTO and NRXN3 (sleep). NRXN3 (neurexin 3) encodes for a surface protein acting as cell adhesion molecule-receptor and it is likely involved in synaptic plasticity [39]. Other than with treatment response, it was also associated with sleep health (Supplementary Tables 1 and 2), suggesting a protective function on brain physiological activity. Polymorphisms in this gene have been linked to obesity and substance use disorders [40, 41], suggesting an influence on impulse control. This increases the interest towards this gene and the possibility to target/modulate it by future therapies, despite currently there are no known compounds/molecules targeting NRXN3.

FTO (fat mass and obesity-associated protein) and MCHR1 (melanin concentrating hormone receptor 1) were both significantly associated with TRD and with sleep and neuroticism, respectively. Both are involved in energy homeostasis, but FTO acts more on a cellular level regulating adipogenesis and fat mass [32, 42,43,44], while MCHR1 is involved in the regulation of feeding behaviour and energy balance and has known effects on mood and sleep–wake cycle [45]. Variants in MCHR1 have been linked to a reduced expression in the dorsolateral prefrontal cortex [46] and are in linkage disequilibrium with another variant recently linked to an increased risk of bipolar disorder [47]. Several drugs interact with FTO but, as far as we know, no specific compound has been developed or studied specifically. Interestingly, FTO also maps to a genomic region showing significant local genetic correlation between MDD and type 2 diabetes mellitus and obesity, suggesting that FTO could be implicated in the shared etiopathogenesis between MDD and insulin resistance-related conditions [48]. We found FTO to interact with antihypertensive drugs, i.e. hydrochlorothiazide (a diuretic) and atenolol (a β-blocker). Angiotensin agents, calcium-channel blockers, and β-blockers (but not diuretics) have been recently associated with decreased rates of depression [49]. On the other hand, MCHR1 has been investigated and promising results were found for some compounds [50, 51], but results were not replicated [30].

CGREF1 (cell growth regulator with EF-hand domain 1) inhibits cell proliferation, despite its function has not been fully elucidated [52]. We found that this gene was associated with remission to antidepressants and anxiety (Supplementary Table 2), but it was implicated in other traits as well by previous GWASs, such as brain measures, risk-taking behaviours, wellbeing, and educational attainment [53,54,55,56,57]. Therefore, CGREF1 may be involved in the modulation of multiple but likely connected phenotypes, which are relevant for antidepressant effects. Currently, there are no known compounds that target CGREF1.

Other genes were not significantly associated with antidepressant outcomes, however suggestive findings were reported (Supplementary Tables 1–3). Among these genes, we outline GRM3, [58], SLCO3A1, LINGO1 and LINGO2. EPHB1 (Supplementary Table 1) regulates chemotaxis and proliferation of neural progenitors in the hippocampus [58], a well known region for mood disorders physiopathology and AD efficacy [59, 60]. It was first identified as potentially involved in antidepressant response by one of the first GWASs in this field, which reported a suggestive association with a SNP located downstream of this gene [61]; however, this GWAS was not formally included in the present review, as it included both unipolar (MDD) and bipolar depression, while this work was focused on MDD. This gene was also associated with several symptom dimensions of interest, including neuroticism, anhedonia, and executive functions. EPHB1 has a key role in axon guidance and, with other ephrin-B receptors, is involved in the development and maturation of dendritic spine and synapse formation [58]. To date, no specific drug exists to selectively target EPHB1, however it interacts with progesterone [25], which has been proved to modulate the expression of γ-aminobutyric acid type-A receptors (GABAAR) via its metabolite allopregnanolone [62]. Noteworthy, brexanolone, a synthetic allopregnanolone analogous, is approved for the treatment of postpartum depression [63].

GRM3 represents an interesting finding. Unlike NMDA (N-Methyl-D-aspartic acid) and AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptors, glutamate metabotropic receptors are G-protein coupled receptors (GPCR) with more complex effects, involved in synapse plasticity, for example long-term depression (LTD) at excitatory synapses [64]. GRM3 was associated with neuroticism and executive functions (Supplementary Table 2). Thus, compounds acting at this site may hypothetically be beneficial for patients experiencing a depressive episode with a sense of guilt and worthlessness, interpersonal sensibility, and/or cognitive symptoms. To date, the only drug that seems to interact with GRM3 is risperidone, a second-generation antipsychotic (SGA) used as adjunctive therapy in psychotic depression or as augmentation to antidepressants in TRD [65]. Other compounds targeting glutamate receptors are under study for the treatment of schizophrenia [66] and depressive disorders, such as MGS-0210 [67], which is a selective metabotropic glutamate receptor antagonist with antidepressant-like activity [68].

Even if only suggestive, the association of the organic anion transporter SLCO3A1 with remission after treatment may suggest a role of endogenous organic anions, vasopressin and prostaglandins in MDD outcome [69, 70]. Noteworthy, SLCO3A1 is significatively expressed in the CNS, especially in oligodendrocytes [70, 71] but also in neurons and grey matter glial cells [69, 72]. However, SLCO3A1 activity is not yet fully understood and new interactions have been found with other exogenous compounds [73], such as modulation of its expression levels by valproic acid [74].

The role of energy metabolism in mood disorders is well known, as discussed above for FTO and MCHR1, and there is an increasing consensus on the involvement, for example, of glucidic metabolism in brain disorders [75, 76]. We observed significant association with many genes involved in feeding behaviour and sleep–wake cycle. LINGO1 and LINGO2 were associated with symptom remission and neuroticism, and are likely involved in synapse assembly [77], other than being associated with BMI [78]. No drug exists to date targeting these genes, but they are modulated by resveratrol and vitamin D [79] and they may represent putative targets for future complementary treatments, for example in patients with higher levels of neuroticism and BMI [80, 81].

This review provides a comprehensive overview of GWASs and WESs on treatment outcomes in MDD, also leveraging an innovative, clinically-oriented approach to explore the complex genetics of MDD treatment outcomes. By integrating genetic signals associated with MDD treatment outcomes and specific depressive symptom dimensions, our approach may pave the way for developing targeted treatments for non-responsive patients exhibiting specific symptom profiles. However, several limitations should be acknowledged. Firstly, we did not perform statistical analyses to test the genetic overlap between treatment outcomes and the symptom dimensions of interest; however, this was beyond the aims of this paper, being this work intended as a review of the literature. A common limitation of the included GWASs on MDD treatment outcomes was the relatively small sample size, and the resulting limited power to detect genome-wide significant associations and to replicate findings across studies. Last but not least, pharmacogenetic findings and targets identified by gene-drug interactions need functional validation to assess their potential clinical relevance and applicability. This consideration outlines the importance of using complementary and integrated research approaches, for example in vitro/in vivo models evaluating compound properties and activity.

In conclusion, this review presents significant insights into the genomics of treatment outcomes in MDD, highlighting the existence of genetic factors overlapping with specific clinical dimensions that are in turn associated with poor treatment outcomes. We prioritised four genes, including CGREF1, MCHR1, FTO, and NRXN3, which are linked to both MDD treatment outcomes and relevant clinical dimensions. These findings highlight the potential for developing new treatments that target specific depressive symptom dimensions, contributing to the advancement of precision psychiatry.