We searched PubMed using search terms ‘antidepressant treatment response’ or ‘SSRI treatment response’ AND ‘gene x environment’ or ‘genetic’ or ‘GWAS’. We included papers on human subjects published in English language before 2014 February. Next, the retrieved abstracts were investigated for quality of evidence. We also included papers from the reference lists of included studies. The main focus of the present review was large-scale GWA-studies including STAR*D, GENDEP, and MARS. Regarding candidate gene studies, detailed reviews are available [9, 14]; therefore, we only discussed genes that are likely to show gene-by-environment interactions (G × E). Namely, the serotonin transporter gene (SLC6A4), as the gene encoding the main target of SSRIs, and the most investigated genes involved in neuroplasticity/neurodevelopment and inflammation that also have major roles in antidepressant response based on animal studies . Thus, the present review was not intended to exhaustively summarize the literature.
Candidate gene studies
Serotonergic system: 5-HTTLPR
Lesch et al. in 1996described a SLC6A4 gene-linked polymorphic region (5-HTTLPR) upstream of the 5-HTT coding sequence (SLC6A4), having a short and long variant which robustly influences transcriptional activity. In this study, carriers of two long alleles synthesized 1.4 to 1.7 times more mRNA compared to those who carried the short variant at least in one copy . Lesch et al. have found an association between 5-HTTLPR and the neuroticism personality trait as measured by NEO-PI-R, which has been identified as a risk factor in the development of MDD. Subsequently, several studies attempted to clarify the link between MDD and 5-HTTLPR, with contradictory results. One of the most recent meta-analyses concluded that it has a small but significant effect; however, the authors warn about publication bias that might interfere with these findings .
The main site of action of SSRIs is the 5-HTT protein, so extensive research started in order to find the link between 5-HTTLPR and SSRI response. Studies in mixed-race samples yielded contradictory findings, however, in Caucasian patient samples there is suggestive evidence pointing to better SSRI response associated with presence of the long allele .
Besides the independent effects of 5-HTTLPR, its interactions with environmental factors (G × E interactions) were also reported. Short allele carriers who also experienced childhood maltreatment and recent life events showed increased vulnerability to stress, increased severity of depression, and reported persistent MDD rather than single, time-limited episodes [7, 8, 19]. Depression severity is a well-known risk factor of unfavorable treatment response , and childhood maltreatment is one of the main environmental predictors of severe depression . Thus, it can be assumed that childhood maltreatment in interaction with the 5-HTTLPR can influence SSRI efficacy, although it has not been tested as yet. However, 5-HTTLPR significantly interacted with recent stressful life events on escitalopram efficacy . This interaction suggests that genetic polymorphisms alone have rather poor predicting capabilities but with their specific environmental interactions, a more reliable predictor complex can be formed.
Besides the 5-HTTLPR, further polymorphisms within the SLC6A4 gene were identified and in some cases, found to be associated with SSRI response. For example, in a large-scale study, Stin4 showed similar interaction with stressful life events and modified SSRI response independent of 5-HTTLPR. Furthermore, an SNP has been identified within the long allele of 5-HTTLPR which could abolish the greater expression rate of the long variant, making it functionally equal to the short variant . Therefore, we can conclude that 5-HTTLPR has important interactions with other genetic variants within the serotonin transporter gene (SLC6A4) and possibly with other genes and environmental stressors, so examining its effect independently from all other factors seems to be an ineffective approach. Rather, focusing on the interactions of 5-HTTLPR already produced valuable information concerning SSRI response, and could possibly aid biomarker research for personalized treatment in the future.
Brain-derived neurotrophic factor (BDNF) is the most common neurotrophin in the human brain. It has a confirmed role in brain development and synaptic plasticity, and via these biological processes, it modulates personality development and cognitive functions . According to early animal studies, BDNF showed promising features to be a probable biomarker for depression. BDNF infusion to the midbrain reduced pain sensitivity and learned helplessness in rat models of depression , while other reports found elevated BDNF levels in SSRI-treated animals .
Indeed, physiological relationships have been identified between 5-HT and BDNF in neurogenesis and neuroplasticity. They activate intracellular cascades leading to the activation of the same set of transcriptional factors, (e.g. cAMP response element- binding protein(CREB),CREB-binding protein (CBP), forkhead box protein O (FOXO), nerve growth factor-inducible factor A (NGFI-A)) which can prevent cell death or induce neuronal growth . Experimental evidence is available about the facilitating effect of SSRIs on cognitive functions, spatial memory, hippocampal neurogenesis, and brain recovery after injury possibly through increased BDNF production .
A common SNP (rs6265) in the coding exon of the BDNF gene leading to a valin (Val) to methionin (Met) change has been identified, and examined comprehensively with the minor, Met coding variant showing reduced activity and contributing to the development of MDD, especially in elderly patients; while the higher activity Val allele was associated with bipolar depression, and substance abuse according to a recent meta-analysis . The Val66Met polymorphism alters the regulated release of BDNF, while the constitutive release remains intact. In addition, Met66 carriers often exhibit poorer episodic memory, and abnormal hippocampal activity .
Many studies investigating the efficacy of SSRIs yielded inconsistent results about the Val66Met polymorphism, and the hypothesis of Val66 variant facilitating better response is not confirmed [14, 27]. Some reports even found the Met66 allele associated with better response to SSRIs . However, it is important to note that in animal models, overexpression of BDNF in the hippocampus causes antidepressant-like effect while in the ventral tegmental area or nucleus accumbens, it is associated with depression-like behavior [29, 30]. These results suggest that proper BDNF function which is a crucial mediator of activity-dependent neuronal plasticity is important for a favorable SSRI response. In addition, other factors should be taken into consideration in the case of BDNF’s effect on SSRI efficacy.
A recent meta-analysis found a significant interaction between stressful life events and BDNF Val66Met polymorphism. Met66 carriers showed a higher probability to develop depression if stressful life events were present . Another study reported a more complex interaction, where the Met66 variant with other SNPs in the BDNF-TRkB-CREB1 pathway showed interaction with childhood adversity and increased the risk for depression, while the Val66 variant among other SNPs was associated with a maladaptive cognitive response style for life stress, called rumination, which is also a risk factor for depression . It seems likely that both alleles of the Val66Met polymorphism of BDNF have their own spectrum of vulnerability by modulating consequences of specific environmental factors, and therefore, they cannot be generalized as either favorable or unfavorable alleles. Moreover, it is important to note that introducing intermediate phenotypes, like rumination, hippocampal volume or stress hormone response, can uncover novel mechanisms that are responsible for contradictory findings in gene-main effect studies. Intermediate phenotypes are genetically determined traits that in neuropsychiatry represent neurobiological processes with a causal role in the disease pathway [33, 34]. Considering these, a comprehensive analysis of BDNF G × E interaction effects on SSRI response and using intermediate phenotypes of antidepressant response could explain more variability than previous studies focusing on the independent effects of the SNPs and only one outcome measure. Unfortunately, there is no such study available yet, partially due to the difficulty of reliably measuring complex environmental factors.
Another important confounding factor is the dependence of BDNF gene expression on an enormous number of other genes, and their polymorphisms in the neuroplasticity pathway. For example, in rat models, SSRI treatment could not increase BDNF levels and exert its behavioral effect in the absence of the CREB gene . Nonetheless, a large-scale mRNA expression study which measured changes in SSRI-treated mice found that the signaling pathway including BDNF-TrkB-CREB1 showed the best association with SSRI treatment . Thus, searching for SNPs associated with the BDNF pathway in interaction with environmental factors seems now a fruitful approach for candidate gene studies to produce useable results.
Despite the fact that the monoaminergic theory is the currently accepted hypothesis of MDD, there is undeniable evidence about the existence of several other factors playing the role in the background of depression including inflammation-associated processes. In the wide comorbidity spectrum of MDD, there are many disorders associated with strong underlying inflammatory mechanisms . Moreover, in MDD patients compared to controls a significantly higher concentration of proinflammatory cytokines, such as TNF-α, IL-1β, and IL-6 can be measured . In addition, administration of TNF-α antagonist etanercept to patients with psoriasis was reported to reduce MDD symptoms. In contrast, IFN-α and IL-2 treatment induces depressive symptoms as a common side effect . Based on these observations, we can conclude that the inflammation pathway has a significant effect in depression. However, antidepressant treatment also has an effect on inflammatory processes, namely they lowered the levels of IL-1β, and specifically SSRI treatment reduced the IL-6 levels in MDD patients in a recent meta-analysis .
Another recently proposed possible biochemical link between 5-HT and inflammation may also explain the relationship between depression and inflammation. 5-HT synthesis from tryptophan can be altered by proinflammatory cytokines, through the increased activation of indolamine-2,3-dioxygenase (IDO). IDO activation promotes the production of kynurenine from tryptophan instead of 5-HT, and kynurenine has two NMDA agonist metabolites with neurotoxic properties . There is experimental evidence that polymorphisms in the IDO1 and IDO2 genes significantly modulated citalopram response possibly by decreasing the availability of neurotoxic kynurenine pathway products [41, 42].
There are several possible mechanisms proposed besides the effect on tryptophan bioavailability and neurotoxic kynurenine metabolites which can cause inflammation-associated MDD. Adaptive immune functions are impaired in MDD according to several reports, so the balance between proinflammatory and anti-inflammatory cytokines is disturbed. The prolonged pro-inflammatory effects cause increased glucocorticoid production and increased hypothalamic-pituitary-adrenal (HPA) axis activity. These consequently damage hippocampal neurons, and contribute to the development of MDD. Pro-inflammatory cytokines also disturb neurogenesis and synaptic plasticity through microglial activation, which negatively influence the survival rate of new hippocampal neurons .
Considering G × E interactions, a polymorphism (rs1360780) in the FKBP5 gene which codes a protein with a major role in HPA axis regulation interacts with childhood maltreatment on depression severity and development . Although this gene influenced antidepressant response in the STAR*D study, there is no data published whether it is modulated by life stresses . Glucocorticoid receptor gene (NR3C1) polymorphisms were also found to be predictors of response to both tricyclic and SSRI antidepressants on a large sample of MDD patients . These findings suggest that preventing production of proinflammatory cytokines can possibly increase efficacy of antidepressants. For example, celecoxib combined with SSRIs showed significantly better efficacy than placebo plus SSRI , although some contradictory findings were also published [46, 47].
In summary, SSRIs down-regulate proinflammatory processes while anti-inflammatory treatment may enhance SSRI efficacy. Nevertheless, the role of inflammation on the outcome of antidepressant treatment is usually neglected, despite the fact that several well-established lines of evidence suggest its importance. Taking biomarkers of inflammation into account in genetic association studies on antidepressant treatment, similarly to other environmental effects such as childhood maltreatment or stressful life events, may reveal several new aspects of SSRI efficacy. However, there is no such study published as yet to the best of our knowledge. A possible difficulty of these studies is that ‘classic’ environmental factors such as stressful life events already include conditions which have a unique and well-described relationship with inflammation, like different medical illnesses. Therefore, a more precise measurement of environmental effects is required to improve the future G × E interaction studies considering inflammation-associated gene regions.
Other potential mechanisms
There are many candidate gene studies which provided promising results about genes and polymorphisms related to antidepressant response, for example in cannabinoid, glutamate, dopamine, or norepinephrine pathways [48, 49]. Comprehensive discussion of these pathways is beyond the scope of the present paper which focuses on G × E interactions. Furthermore, we only aimed to represent findings of antidepressant treatment response, not depression as a disorder, on which a thorough discussion can be found in other recent papers, e.g., by Mandelli and Serretti .
Genome-wide association studies (GWAS) use entirely different methods than previously presented candidate gene studies. GWAS do not rely on a set of genes expected to predict the outcome. The inclusion criterion is merely the minor allele frequency (MAF), which describes the occurrence rate of the polymorphism. Therefore, GWAS scan hundred thousands of SNPs in an appropriately large sample of patients. The high volume of genotyped samples give immense statistical power to GWAS studies, however, way more rigorous significance levels (p ≤ 5 × 10-8) have to be employed in order to avoid false-positive results. In the following three sections, we would like to review the three biggest GWAS studies on antidepressant efficacy.
The Genome-based Therapeutic Drugs for Depression (GENDEP) study examined 706 MDD patients treated with nortriptyline or escitalopram for 12 weeks. On a genome-wide significance level, the gene coding uronyl-2-sulphotransferase (UST) was associated with nortriptyline response. On a suggestive level, IL-11 expression was associated with escitalopram response, and two intergenic regions on chromosome 1 and 10 were found to be linked with response to both medications. On a lower level of significance, IL-6 appeared as a noticeable modulator, which further supports the significance of the role of the proinflammatory pathway in SSRI efficacy and the inflammation theory of depression. The other identified predictor regions were pretty unexpected. UST is an enzyme necessary to produce oversulfated proteoglycans, influencing neurogenesis, and neuronal migration. The intergenic polymorphism found on chromosome 1 is close to the coding sequence expressing zinc finger protein 326 (ZNF326). On chromosome 10, the closest gene is the plexin domain containing 2 (PLXDC2). They are both involved in the structural changes of chromatin, but the exact mechanism on SSRI effect is unknown. However, these findings suggest that polymorphisms in non-coding areas which possibly influence the transcription rate of other gene-coding regions have the same or even more power to modulate SSRI efficacy .
Besides genetic factors, age, depression severity, and stressful life events showed significant association with SSRI response. Younger patients with the lowest baseline depression severity showed better response to escitalopram. Interestingly, subjects with one or more stressful life events also responded better to escitalopram, suggesting the role of the serotonergic system in the development of life stress-induced depression .
In the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study, all 1,493 patients were treated initially with citalopram in a naturalistic design, with no placebo control or serum level assessments to filter out no real medication effect and non-compliance . There was no genome-wide significant association found among the 430,198 SNPs tested. The top two associated polymorphisms were in non-coding areas, one is between the ubiquitin protein ligase E3C gene (UBE3C) and motor neuron and pancreas homebox 1 (MNX1 or HLXB9), and the other is near the bone morphogenic protein 7 (BMP7) gene. UBE3C signals proteins for degradation, and MNX1 has a role in embryonic brain development. Just like in GENDEP, the top findings are in non-coding regions, close to non-expected genes . Comparing the genetic properties of sustained responder groups to those of non-sustained responders, the top 25 SNPs were again intronic or intergenic. Although there is a lower chance of placebo effect or non-compliance in sustained responder than in non-sustained responders, the STAR*D top findings were non-replicable in the GENDEP sample .
Apart from genetic features, minority status, socioeconomic disadvantage, comorbid conditions, lower functioning and quality of life, and anxious or melancholic features were found to be unfavorable on the outcome . Among the sociodemographic predictors, simultaneous presence of higher education, higher income, not living alone, and good employment status were highly correlated with remission .
The Munich Antidepressant Response Signature (MARS) included 339 patients, 88.8% with MDD and 11.2% with BPD, all of whom were Caucasian, and 85.1% from German descent .The CDH17 gene polymorphism rs6989467 showed the most significant association with treatment outcome. The genome-wide testing was also accomplished on an independent German sample (n = 361), and an intergenic SNP between AK090788 and PDE10A genes showed the strongest association. None of these findings could achieve genome-wide significance, and the authors concluded that these genes have a rather modest effect on SSRI response. Suspecting multiple genes, pathway analysis was performed, which identified three gene clusters. The first cluster is centered on fibronectin 1 (FN1), including transcriptional factors and the substrate, and receptor genes of ephrin-A5, which has a role in late-stage nervous system development, and is associated with hippocampal neural plasticity. The second cluster has risk genes for cardiovascular and metabolic disorders, which show a high comorbidity with depression, and the third is related to GABA and glutamatergic neurotransmission . Signaling pathway analysis is a progressive approach to identify genes with cumulating effects on complex phenotypes like depression, and could be a very fruitful way to search for potential target molecules in the future.
In the same sample, HPA axis hyperactivity was tested by the combined dexamethasone suppression/CRH stimulation (DEX-CRH) test. As proposed earlier, HPA hyperactivity most likely caused by impaired negative feedback control in MDD patients was associated with depression severity. After 5 weeks of antidepressant treatment, normalization in HPA hyperactivity was observed in remitters, but not in non-remitters. A single DEX-CRH in males and repeated tests in females were able to predict antidepressant treatment outcome . Therefore, testing HPA axis function, as a biomarker of the body response to environmental stressors, seems to be an interesting approach to predict SSRI efficacy prior to treatment.
The three studies above were summarized by a meta-analysis, which could not find a genome-wide significant association in case of any of the 1.2 million analyzed polymorphisms . Although the authors concluded that there are no reliable predictor gene regions or a set of common SNPs, valuable conclusions could be made about the relevance and capabilities of GWAS studies. Methodological problems such as definition of responders/non-responders, or choosing the inclusion criteria, and determining the study design can cause more interference with the results if they are expected to meet the GWAS level of significance . Individual differences in symptom trajectories were found to be important predictors of treatment outcome itself further supporting that MDD is a heterogeneous disorder. For example, patients with less activity and lack of interest showed worse response to antidepressant treatment . In contrast, cognitive symptoms were found to be more responsive to SSRI treatment . Based on this experience, the more accurate declaration of our expectations on SSRI response should help find more significant associations. Moreover, integrating measures of non-subjective intermediate phenotypes in the symptom trajectory seems to be a promising approach. For example, neuroimaging-based intermediate phenotypes can help us better understand the brain processes behind symptom trajectories . Although it is not possible to scan enough patients with neuroimaging techniques for a GWA study, these results can help us refine our phenotype-measuring system and develop more biology-related categories to work with.
Another crippling deficiency of the GWAS approach is ignoring environmental factors, while in many cases, they show better association with the response than genetic polymorphisms . In the future, new more powerful statistical methods, like the Bayesian network analysis, must be developed in order to take environmental factors into consideration [61, 62].