The endogenous and reactive depression subtypes revisited: integrative animal and human studies implicate multiple distinct molecular mechanisms underlying major depressive disorder

Genome-wide expression profiling of the hippocampus (HIP) from two studies from the rodent arm of the Genome-based Therapeutic Drugs for Depression (GENDEP) study [15] was used to inform candidate gene selection in a comparable human post-mortem, case-control study on MDD from the Stanley Brain Consortium. The GENDEP project is a large-scale, multi-center human pharmacogenomics study that also includes a series of large-scale studies using animal models and in vitro experiments. The GENDEP project was designed to allow for integrative analysis of the results of the transcriptomics and proteomics on the samples from the human, the rodent and the in vitro studies, in order to gain further insight into the molecular mechanisms of MDD and identify biomarkers of antidepressant drugs (AD) treatment response. The mouse study used 144 animals from four strains of well-characterized inbred mice to model individual variation in humans. The mice were subjected to one of two stress protocols and a control condition (maternal separation (MS) - ‘early stress’, unpredictable chronic mild stress (UCMS) - ‘late stress’ - or the control condition (ENV)) to model ‘reactive’ depression. Litters of each strain were randomly allocated to the MS, UCMS or control group. Findings from the mouse study were cross validated in a parallel rat study that compared HIP mRNA differences between Flinders Sensitive and Flinders Resistant rat lines as models of ‘endogenous’ depression. Finally, genes differentially expressed in response to both stress protocols in the mouse study and in the rat study were used to inform probe set selection in comparable mRNA expression study in humans.

A total of 144 male and female mice (72 of each sex) from four different strains ((129S1/SvImJ, C57LB/6 J, DBA/2 J and FVB/NJ) were bred in the barrier unit at the Institute of Psychiatry, London, UK. Weaning took place when the animals were 21 to 28 days old. Animals were group-housed under standard conditions with a 12:12 h light:dark cycle, 22°C ± 11°C, food and water ad libitum. A total of 144 animals were sacrificed by cervical dislocation. Animals used for the transcriptomic study were not behaviorally tested. The hippocampus, liver and spleen were dissected following previously published protocols [16, 17]. All housing and experimental procedures were carried out in accordance with the UK Home Office Animals (Scientific Procedures) Act, 1986.

A total of 39 animals from two cohorts of Flinders Sensitive Lines and Flinders Resistant Lines (22 FRL and 17 FSL) were bred and maintained at Karolinska Institutet (Stockholm) and housed under standard room temperature (22 ± 1°C), relative humidity (45 to 55%) and a 12 h light:dark schedule (light on at 07:00 a.m.). Food and water were available ad libitum. The study was conducted as part of a parallel GENDEP investigation. The Stockholm's Ethical Committee for Protection of Animals approved the study and all procedures were conducted in conformity with the Karolinska Institutet's guidelines for the care and use of laboratory animals, which follows the European Communities Council Directive of 24 November 1986. Additional information on the rat study is available elsewhere [18].

The Rankprod method was used to identify the most robustly differentially expressed genes between ‘late’ (UCMS) stressed animals and control and between ‘early’ (MS) stressed animals and control. We considered only those genes that show consistency in the direction of change across all four strains. Inconsistency in the direction of change indicates Stress x Strain interaction effects, which are not specific to our research question. The results of this analysis uncovered 406 probe sets altered in response to UCMS across all four strains. These probes tag the expression of 370 known genes in mice. A summary of genes uncovered from this analysis with a previous association with stress response or MDD is presented in Table 1. The results reveal a number of genes previously associated with UCMS protocols and believed to play a role in the pathogenesis of MDD. The same analysis was repeated to compare the maternally separated animal (MS) and control. The results from this analysis revealed 396 probe sets differentially regulated in response to the maternal separation protocol. These probe sets could be mapped to 350 known genes in mice. A summary of the top genes differentially expressed in response to maternal separation protocols is presented in Table 2. We then explored the number of altered genes in response to either ‘early’ or ‘late’ stressors as well as the genetic overlap between the two conditions (Figure 2). There were remarkably few. Only 67 genes, less than 8.8% of significantly altered genes were in common between mice exposed to early and late stress paradigms.

The minimal gene expression overlap suggests that the biological mechanisms underpinning ‘reactive’ depression caused by early and late stressors differs considerably. In order to gain further understanding into these differences, genes differentially expressed for each of the two models were analyzed using IPA [35]. This allowed us to uncover gene networks showing the molecular relationship between the genes and evaluate networks according to the fit of significant genes in each dataset [12]. First, we explored gene networks associated with ‘late’ UCMS protocols. A total of 350 genes from our reference list were found on the IPA database. The top two functional networks identified by IPA have a score >42, with 29 reference molecules included in the first network and 23 in the second. Both networks were significantly associated with stress signaling response. The most significant transcriptional regulators included ELK1/2/4 TFIIA, SMARCB, CREB1 and THRB (see Additional file 1: Figure S1 and Additional file 2: Figure S2). We repeated the pathway analysis with genes differentially expressed in response to ‘early’ (MS) stressors. A total of 347 genes from our reference list were found on the IPA knowledge database. IPA returned three networks with a score >40. The associated functions of the top networks include mRNA post-transcriptional modification, protein synthesis and cellular development. The networks are associated with developmental and neurological disorders, which is a good match to the “early” stress protocol used. The top-ranking network (see Additional file 3: Figure S3) includes 29 focus molecules from our reference gene set. The most prominent interacting genes within this network are with the Yhwaz and Yhwag. These genes are of particular interest as they have been systematically uncovered across several proteomic and transcriptomic studies from the GENDEP project in both mice and rats [12, 13, 17, 22]. These genes show a direct interaction with the STK25 kinase, which plays a role in stress response. The second network (Additional file 4: Figure S4) is composed of 27 molecules from our reference dataset. This network is centered on the NF-κB complex. The nuclear factor-κB (NF-κB) is a ubiquitous transcription factor involved in the regulation of gene expression and cell stress response and cell proliferation. Interestingly, NF-κB can be activated by different stimuli, including cytokines (such as TNF-α and IL-1): this finding is congruent with the inflammation hypothesis for MDD [41‐45].

To gain an understanding of the similarities between stress-induced ‘reactive’ depression and a congenital ‘endogenous’ model of depression, we compared genes differentially regulated in response to early (MS) and late stress (UCMS) with mRNA differences between Flinders sensitive and resistant lines. Flinders lines are a genetic animal model of depression that allows us to cross-validate stress-altered genes within a parallel, independent mRNA study where depressive-behaviors occur in the absence of environmental stressors. The RankProd algorithm and conservative cut-offs described previously was used to evaluated mRNA differences between Flinders Sensitive and Flinders Resistant lines. The results revealed 715 down-regulated and 1,145 up-regulated probe sets. To obtain a list of gene names, probe sets obtained were subsequently annotated using PANTHER [34]. The probe sets tagged the expression of 501 down-regulated and 727 up-regulated genes. A total of 1,228 genes were used for cross-validation with the mouse study. First, we explored the genomic overlap between maternally deprived animals and Flinders line rats. From a total of 350 genes differentially regulated in maternally deprived mice, a total of 65 genes (19%) were also differentially regulated in rats. The same comparison was performed with genes differentially expressed in mice exposed to UCMS. A total of 52 genes (11%) were differentially expressed between ‘late’ stress animals and Flinders rats. A compelling finding is that 21 genes are differentially expressed in rats and in both early and late stressed mice (Figure 3). This is an important genetic overlap given that only 67 genes were commonly expressed between early and late stressed animals in the first place. Validations in an independent, methodologically different study using a genetic model of depression point to an important genetic overlap between stress-related and syndrome-related mechanisms. In order to gain further biological insight, genes significantly altered in response to both stresses in mice and between Flinders Sensitive and Resistant lines were carried forward for analysis using Ingenuity’s IPA system. All 21 genes were found in the Ingenuity reference database. A significant network with a score >40 consisting of over 55% of the reference molecules (12/21) was revealed (Figure 4). Among the genes in the pathway, four genes (Ywhaz, Ppm1a, Nkfb and Mapk1) are of particular interest as they have been previously reported across different “omics” GENDEP investigations [12, 17, 46‐48].

From a total of 21 genes differentially expressed across all three rodent studies, 15 human orthologs were found. The expression of these 15 genes is tagged by 21 probe sets on the Affymetrix HU95A oligonucleotide array. The RankProd algorithm was used to evaluate expression between post-mortem cases and controls. Out of a total of 15 genes, the VAMP-2 is significantly down-regulated after correcting for the number of multiple non-independent tests using a prediction of false discovery rate of PFP <0.05 (Table 3). In our study, the Vesicle-Associated Membrane Protein 2 (VAMP-2: Synaptobrevin2) gene is significantly altered across all rodent studies and in the human study.

Consistent with our hypothesis and with previous findings, the gene-expression profiles of both of our ‘reactive’ models of depression were largely distinct from our ‘endogenous’ model. Interestingly, this differed according to which ‘reactive’ depression paradigm was compared. For the early or stress ‘reactive’ model 19% of genes overlapped with the endogenous model, while for the late stress ‘reactive’ model the overlap was considerably lower at just 11%.

Surprisingly, the genomic signatures of our two ‘reactive’ models were more distinct from one another than the ‘endogenous’ model with fewer than 9% of genes shared between the two paradigms. This suggests that the two different models result in depressive-like behavior in mice through distinct biological mechanisms. Interestingly, gene pathway analysis returned plausible functional networks, with the more significant network for ‘early’ stressed animals associated with neurodevelopmental disorders and those of ‘late’ stressed animals associated with cell stress response and cell-signaling. Taken together, our results suggest that early exposure to stress modulates the expression of genes belonging to pathways associated with neurodevelopmental mechanisms. These changes may condition an individual’s exposure to later stresses and response to pharmacological and behavioral interventions later in life in yet unclear ways. Conversely, late onset stresses may act primarily on brain neurochemistry with neurostructural changes occurring via the cascading effects of neurochemically-related mechanisms, including neurogenesis and apoptosis [49].

While each of our three animal models of depression showed largely distinct gene-expression profiles, a set of genes were differentially regulated across all three paradigms. Pathway analysis of these genes revealed a gene network which included Ppm1a, Ywhaz, Nkfb and Mapk.

All four genes have each been implicated in both the etiology of MDD and the response to treatment and may, therefore, represent a final common pathway to the disorder. We previously reported that the expression ppm1a was significantly modulated by the antidepressant nortriptyline. We have also shown that several single nucleotide polymorphisms in the human ortholog of this gene (PPM1A) predict a response to the same drug in a parallel human pharmacogenetic study [46]. Tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein (Ywhaz) has systematically been uncovered across several GENDEP studies and plays a role in cell proliferation and neurogenesis, which is a current explanatory model of MDD [50‐54]. Moreover, it interacts with IRS1 protein and the MAPK pathway by modulating the activation of JNK1 and p38 MAPK both of which have been systematically associated with depressive mechanisms [52‐56]. Lastly, the Nfkb gene has extensively been associated with peripheral inflammation and is consistent with the inflammation hypothesis for MDD [57].