Altered miRNA landscape of the anterior cingulate cortex is associated with potential loss of key neuronal functions in depressed brain

Abstract

MicroRNAs (miRNAs), a family of non-coding RNAs, have recently gained a considerable attention in neuropsychiatric disorders. Being a pleiotropic modulator of target gene(s), miRNA has been recognized as central to downstream gene regulatory networks. In the recent past, reports have suggested their role in changing the epigenetic landscape in brain of subjects with major depressive disorder (MDD). Anterior cingulate cortex (ACC) is a brain area implicated in several complex cognitive functions, such as impulse control, emotion, and decision-making and is associated with psychopathology associated with mood regulation. In this study, we examined whether MDD is associated with altered miRNA transcriptome in ACC and whether altered miRNA landscape is associated with modifications in specific gene network(s) at the functional level. Using next generation sequencing, it was observed that that 117 miRNAs (4.61%) were significantly upregulated and 54 (2.13%) were downregulated in MDD subjects (n=22) compared with non-psychiatric controls (n=25). Using 24 most significantly upregulated miRNAs in the MDD group, we determined functional enrichment of target genes and found them to be associated with long-term potentiation, neurotrophin signaling, and axon guidance. Intra- and inter-cluster similarities of enriched terms based on overrepresented gene list showed neurobiological functions associated with neuronal growth and survival. Web centric parameters and ontology enrichment functions identified two major domains related to phosphatidyl signaling, GTPase signaling, neuronal migration, and neurotrophin signaling. Our findings of altered miRNA landscape along with a shift in targetome relate to previously reported morphometric changes and neuronal atrophy in ACC of MDD subjects.

1. Introduction

Major Depressive Disorder (MDD) is one of the debilitating mental disorders worldwide and the lifetime prevalence is 10.8% (Lim et al., 2018). Despite concerted efforts, the neurobiology of depression is not well understood, which is also reflected in the poor outcome of the current medication for MDD (Blackburn, 2019). Anterior cingulate cortex (ACC) is a brain area responsible for psychopathology associated with emotion regulation (Stevens et al., 2011). Neuroimaging and electrophysiological studies have strongly suggested the role of ACC in processing of emotional and sensory cues from various brain regions implicated in mood regulation (Pizzagalli et al., 2001; Wu et al., 2016). The critical role of ACC in mood regulation is further substantiated by the findings of significant changes in its morphology in MDD subjects (Drevets et al., 2008). Morphometric changes along with decreased gray matter volume in the subgenual region of ACC have been associated with both unipolar and bipolar depression (Wise et al., 2017). The largest worldwide study conducted by ENIGMA MDD Working Group detected gray matter loss in the orbitofrontal cortex, anterior and posterior cingulate, insula and temporal lobes (Schmaal et al., 2017). Additionally, smaller gray matter volume of ACC and bilateral insula have been reported in MDD subjects compared to bipolar subjects as well as healthy controls (Wise et al., 2017). Several rodent studies have also shown that gray matter reduction is strongly associated with dendritic atrophy and reductions of glia cells under the stressful condition (Radley et al., 2008; Wellman, 2001). The ACC has several functions such as attentional orienting, affective processing, and behavioral selection (Bush et al., 2000). As an integration center, ACC includes several important processing modules which help in controlling the cognitive and emotional network with afferent and efferent projections. Past studies have also suggested ACC related GABAergic deficiency in both adult and adolescent MDD brains (Gabbay et al., 2012; Price et al., 2009). Glutamatergic dysfunctionality has also been shown in ACC of adult MDD subjects (Cotter et al., 2002). These changes could be associated with altered gene functionality, as has previously been shown in other brain regions of MDD subjects (Malki et al., 2015; Roy et al., 2017b). A strong possibility of epigenetic regulation of aberrant gene expression cannot be ruled out given that environmental risk factors are critical in the complex behavioral outcomes associated with mood disorders (Wang et al., 2018). Moreover, several studies have examined MDD in the context of suicidal behavior and found to be driven by adverse environmental influence. It is important to mention that emotional turmoil in survivors of patients who died by suicide may last for a long time, and in some cases, may end with their own suicide. Thus, it is fundamental to understand the bereavement process after the suicide of a significant other in order to provide proper care, reduce stigma, and improve the outcomes. Although the role of miRNAs is not well understood in the bereavement process or in their link to MDD leading to suicidal behavior (Pompili et al., 2013), postmortem brain studies provide critical information regarding completed suicide, which may help explain the molecular changes happening in the brain under such conditions. Recently, numerous studies have shown that non-coding RNAs, including microRNAs (miRNAs), can have significant epigenetic influence in the development of psychiatric disorders such as MDD (Briggs et al., 2015; Kocerha et al., 2015; Qureshi and Mehler, 2013; Roy et al., 2017a; Salta and De Strooper, 2012; Sartor et al., 2012). A recent qPCR-based study using a set of 29 miRNAs identified downregulation of two miRNAs (miR-184 and miR-34a) in the ACC of MDD subjects (Azevedo et al., 2016). However, the involvement of miRNA-mediated epigenetic regulation of altered ACC functions in MDD has not been evaluated in greater detail at transcriptome-wide level. In the present study, we examined transcriptome-wide differential regulation of miRNAs in a relatively large number of ACC samples from non-psychiatric controls (n=25) and MDD subjects (n=22) using miRNA-sequencing. Subsequently, the KEGG based pathway and gene ontology (GO) analysis were done based on predicted targets of top significantly altered miRNAs in the MDD group. The results helped us to comprehend the functional role of ACC-specific miRNA changes in MDD pathogenesis.

2. Experimental procedures

2.1. Subjects

The study was approved by the Institutional Review Board of the University of Alabama at Birmingham (UAB) under exemption 4. The study was performed in ACC obtained from the New South Wales Brain Tissue Resource Centre at the University of Sydney. ACC from a total of 25 non-psychiatric controls (hereafter referred as controls) and 22 MDD subjects were used. Detailed demographic and clinical characteristics of subjects are shown in Table 1. Demographic data included age, gender, race, postmortem interval (PMI), brain pH, cause of death, history of drug/alcohol abuse, and antidepressant toxicology at the time of death. There were no significant differences in age, PMI, brain pH between MDD and control subjects (Table 1). MDD group had 12 males and 10 females whereas control group had 15 males and 10 females. Within the MDD group, 10 subjects died by suicide and 12 died by causes other than suicide. Also, out of 22 MDD subjects, 4 had a history of alcohol abuse and 18 showed positive antidepressant toxicology at the time of death.

Table 1.

Demographic and clinical characteristics of control and MDD subjects

		Control	MDD	p-value
Number of subjects		25	22	N/A
Age (Year)		50.8 ± 10.6	51.6 ± 12.9	0.809
Gender	Male	15	12	N/A
	Female	10	10
Postmortem interval (Hour)		28.5 ± 15.3	29.5 ± 9.7	0.781
Brain pH		6.5 ± 0.4	6.6 ± 0.3	0.328
Alcohol abuse		0	4	N/A
Antidepressant drugs		0	18	N/A
Suicide		0	10	N/A
Cause of death		Cardiac arrest, respiratory failure	Cardiac arrest, cancer, trauma, hepatic abnormalities, Ischemic bowel	N/A
Method of suicide		N/A	Drug toxicity, hanging

Values denote mean ± standard deviation. MDD: Major Depressive Disorder, N/A: not applicable.

Detailed tissue dissection, preservation, quality control and detailed in earlier publications (Dedova et al., 2009; Sheedy et al., 2008). Briefly, the cerebrum, cerebellum and brainstem are divided in the sagittal plane. The cerebellar hemisphere and brainstem were removed from each cerebral hemisphere by transversely sectioning the brainstem through the rostral midbrain at the level of the superior colliculi. Each cerebral hemisphere and its contralateral cerebellar hemisphere and brainstem were assigned a “fresh” or “fixed” status on a random basis. The fresh hemisphere was cut into 1 cm coronal slices and several areas were dissected including ACC and digitally photographed. The tissue blocks and were frozen on precooled trays at a temperature of −80°C and were bagged into air-tight plastic bags. Detailed neuropathological examination was done for several brain regions. Blood samples taken at the time of autopsy were used for toxicology screens for therapeutic drugs. toxicology including neuroleptics, paracetamol, and anti-depressants. Several assessments including, diagnostic characterization, review of agonal status, measurement of brain tissue pH, histological assessment of post-mortem autolysis and a full neuropathological to ensure the tissue integrity. Brain tissue pH was measured in cerebellar tissue (Harrison et al., 1995).

The post-mortem clinical diagnosis of each case was determined through extensive review of medical records by one of two independent clinicians. All psychiatric cases were confirmed by a psychiatrist retrospectively using the Diagnostic Instrument for Brain Studies – Revised (DIBS-R). The DIBS is semi-structured instrument designed specifically for postmortem psychiatric assessment using medical records and informants where available. This instrument is compliant with the Diagnostic and Statistical Manual of Mental Disorders. The DSMIV and DSMV Diagnostic Criteria were used in all psychiatric disorder cases.

2.2. Total RNA isolation from ACC

TRIzol ® (Invitrogen, Grand Island, NY, USA) was used to isolate RNA. The detail method has been described earlier (Roy et al., 2017a). The RNA purity was checked by Nanodrop (260/280 nm; cutoff ⩾ 1.8) and integrity by agarose gel electrophoresis. It was made sure that RNA Integrity Number (RIN) for all samples were within acceptable range of 7–8.

2.3. Library preparation and sequencing of miRNAs using total RNA from ACC

The isolated total RNA from ACC of all 47 subjects were sequenced using Genomic Core facility at University of Alabama at Birmingham following small RNA library preparation method. In brief, the RNA quality was checked on the Agilent BioAnalyzer for the presence of a strong 5S peak to assure the miRNA fraction. Total RNA was used in an initial ligation reaction to attach 3’ oligo with RNA ligase. A subsequent ligation was performed to add 5’ oligo, also with RNA Ligase. A primer specific for the 3’ oligo was used to generate 1st strand cDNA in a standard reaction. Double stranded cDNA was generated via PCR. The resulting library was checked for size and quality using the DNA High Sensitivity Chip on the Agilent BioAnalzyer. The libraries were quantitated via qPCR (KapaBiosystems), normalized to 3 nM, and loaded onto Illumina NextSeq 500 and sequenced with 75bp single end reads as per the manufacturer’s protocol. The miRNA profiling was conducted using Illumina NGS platform, libraries were created using Qiagen’s miRNA kit. The resulting libraries were standardized for concentration and sequenced on the NextSeq500 using 75 bp single end reads.

2.4. Analysis of miRNA sequencing data

Reads were trimmed, unique molecular identifier (UMI) sequences were identified and were aligned to miRbase and the human hg38 genome. Next, the count reads and UMIs were assigned to each miRNA. The UMIs were then normalized in Qiagen’s GeneGlobe Data Analysis Center (https://geneglobe.qiagen.com/us/analyze/). Based on normalized miRNA expression values across 25 control and 22 MDD subjects, the expression heat map was generated following a hierarchical clustering method. In the expression heat map, the miRNAs with high expression values are shown in green color and miRNAs with low expression values are shown in red (Fig. 1A). For the clustering purpose, the Euclidean method was used to measure the distance, and the average linkage algorithm was applied to calculate the average pairwise distance between all pairs of points. To determine the miRNA expression changes in a biologically meaningful way, fold-change (FC) was calculated using the normalized miRNA expression in each test sample divided by the normalized miRNA expression in the control sample. For the geNorm and Total Molecular Tag Count normalization methods, the p values were calculated based on Student’s t-test of the replicated normalized miRNA expression values for each miRNA in the control group and the MDD group. For the DESeq2 and Trimmed Mean of M (edgeR) normalization methods, the p values were determined by the respective Bioconductor software packages.

Fig. 1: miRNA expression heatmap and differentially expressed miRNAs.

A) All 2540 annotated miRNAs and their normalized values between MDD and control (n = 22 vs 25) subjects were plotted in an expression heatmap with hierarchical clustering. Green and red color in the map show higher and lower expression values respectively. For the clustering purposes, the Euclidean method was used to measure the distance, and the average linkage algorithm was applied to calculate the average pairwise distance between all pairs of points. Following the average linkage clustering algorithm, the dendrogram was constructed to demonstrate the expression similarities. B) MDD specific numerical distribution of significantly altered miRNAs was presented with a bar of pie diagram. The pie plot shows the distribution of differentially expressed 171 annotated miRNAs and 59 predicted miRNAs based on their significance level (p<0.05). The 117 up and 54 downregulated miRNAs from the list of 171 annotated miRNAs were plotted with a bar diagram.

2.5. miRNA target gene prediction

GO analysis was done independently based on the predicted targets of miRNAs which were found to be significantly upregulated (FC≥ 2.0) in ACC of MDD subjects. To obtain a consensus list of predicted targets, in-silico prediction algorithm from eight different prediction programs (i.e., miRWalk, Microt4, miRanda, miRDB, Pictar2, PITA, RNA22, and Targetscan under miRWalk version 2.0 software package) was used. Next, a common list of targets that were shared by any five of them were used for further functional prediction analysis.

2.6. Prediction based pathway and GO analysis

The consensus list of predicted targets prepared from miRWalk (version 2.0) software was used to predict gene ontology (GO) terms using the ShinyGO program and KEGG (Kyoto Encyclopedia of Genes and Genomes) based pathway analysis using WEB-based GEne SeT AnaLysis Toolkit (WebGestalt, http://www.webgestalt.org/). The GO prediction analysis was done following an FDR corrected p-value cutoff 0.05 to determine the gene set enrichment in biological process (BP) and cellular component (CC) category separately. In BP, 30 most significant terms were used to plot the network with edge cutoff 0.05, whereas, in the CC, the connected nodes are presented with 40 most significant terms with edge cutoff 0.03. In each category, the enriched terms were used to create networks where nodes are presented with terms and connected with edges. If two nodes are connected, then they share 20% (default) or more genes. Bigger nodes represent larger gene sets. Thicker edges represent more overlapped genes. In WebGestalt, Over-Representation Analysis (ORA) with KEGG pathway was used to find the relevant pathways. Further, the gene list obtained from WebGestalt was filtered based on the parameters defining neurobiological functions associated with neuronal growth and survival. A short list of genes was further fed into Metascape program (http://metascape.org/gp/index.html#/main/step1) which uses 40 independent knowledge bases to integrate system level information and creates comprehensive gene list with designing features including gene annotation, pathway and functional enrichment. For creating the color-coded bar plot, the algorithm first identified all statistically enriched terms (Zhou et al., 2019). Next, accumulative hypergeometric p-values and enrichment factors were calculated and used for filtering them. Remaining significant terms were then hierarchically clustered into a tree based on Kappa-statistical similarities among their gene memberships. The tree was casted into terms using an optimized kappa threshold (0.3).

2.7. Statistical analysis

Statistical analyses were conducted using SPSS 25.0 software (IBM, Chicago, IL, USA). The Shapiro-Wilk test was used to assess the normality of the data. The average difference of age, PMI, and brain PH, was assessed by Student’s t test. Differences in miRNA expression based on gender, alcohol abuse, antidepressants, and suicide were analyzed by Student’s t test. Correlation between miRNA expression and age, brain pH, and PMI were evaluated by Pearson Product Moment analysis. Statistical significance was set at the 95% level (p = 0.05).

3. Results

3.1. MDD specific differential miRNA expression profile in ACC

From the differential expression analysis between control and MDD groups, a total of 2540 fully annotated miRNAs regardless of their statistical significance were detected. Using the normalized values of all miRNAs expressed across all 47 subjects, a hierarchically clustered expression heat map was generated (Fig. 1A) to show the differential expression profile between MDD and control subjects. Green and red color in the map show higher and lower expression values respectively. Following the average linkage clustering algorithm, the dendrogram was constructed to demonstrate the expression similarities. These miRNAs included both up and downregulated groups. Besides that, a set of 423 predicted (novel) miRNAs were also found to be altered in the MDD group. However, filtering the list of differentially regulated miRNAs following statistical significance (p≤0.05) resulted in the identification of a total 230 miRNAs including both annotated (171) and predicted (59) ones (Fig. 1B). For annotated miRNAs, 117 (4.61%) were upregulated and 54 (2.13%) were downregulated. From the list of 59 novel miRNAs, 50 miRNAs (11.8%) were found to be upregulated and 9 miRNAs (2.13%) were downregulated. In the significantly altered annotated miRNA group, sorting the list with ≥2 fold change (FC) value identified 24 highly upregulated miRNAs in the MDD group. Two separate tables have been presented to show all significantly (p ≤ 0.05) altered miRNAs (Table 2). Topmost significantly expressed (extremely upregulated, FC cut off >2.0; and downregulated, FC cut off <0.5) miRNAs in the MDD group are highlighted with bold fonts in Table 2.

Table 2.

Significantly altered miRNAs in ACC of MDD group (p≥ 0.05)

miRNA	miRBase ACC No.	Fold change	p-value	Chromosomal location
Up-regulated
hsa-miR-6077	MIMAT0023702	2.59	1.9E-05	chr1: 148388282–148388363 [+]
hsa-miR-4632-3p	MIMAT0019688	2.69	2.2E-05	chr1: 12191713–12191773 [+]
hsa-miR-6789-3p	MIMAT0027479	2.49	0.00017	chr19: 2235829–2235926 [−]
hsa-miR-648	MIMAT0003318	2.41	0.00025	chr22: 17980868–17980961 [−]
hsa-miR-4498	MIMAT0019033	2.83	0.00049	chr12: 120155434–120155499 [−]
hsa-miR-6084	MIMAT0023709	2.22	0.0007	chr1: 20633679–20633788 [+]
hsa-miR-4433a-3p	MIMAT0018949	2.15	0.00095	chr2: 64340759–64340839 [+]
hsa-miR-4638-5p	MIMAT0019695	2.04	0.00128	chr5: 181222566–181222633 [−]
hsa-miR-4258	MIMAT0016879	2.39	0.00142	chr1: 154975693–154975783 [+]
hsa-miR-6850-5p	MIMAT0027600	2.01	0.00168	chr8: 144791931–144791991 [−]
hsa-miR-4510	MIMAT0019047	1.62	0.00171	chr15: 35926856–35926923 [+]
hsa-miR-3651	MIMAT0018071	2.47	0.00173	chr9: 92292458–92292547 [−]
hsa-miR-4497	MIMAT0019032	2.73	0.00243	chr12: 109833348–109833436 [+]
hsa-miR-668-5p	MIMAT0026636	2.62	0.00244	chr14: 101055258–101055323 [+]
hsa-miR-6800-3p	MIMAT0027501	1.72	0.00291	chr19: 49832018–49832099 [+]
hsa-miR-7108-3p	MIMAT0028114	2.81	0.00474	chr19: 2434914–2435000 [−]
hsa-miR-4761-3p	MIMAT0019909	2.96	0.00478	chr22: 19963753–19963834 [+]
hsa-miR-4426	MIMAT0018941	2.56	0.00509	chr1: 192716328–192716390 [+]
hsa-miR-4720-3p	MIMAT0019834	1.93	0.00573	chr16: 81385018–81385093 [+]
hsa-miR-4281	MIMAT0016907	1.78	0.00656	chr5: 176629439–176629500 [−]
hsa-miR-1273h-3p	MIMAT0030416	1.36	0.0066	chr16: 24203116–24203231 [+]
hsa-miR-4736	MIMAT0019862	1.84	0.00694	chr17: 58335976–58336022 [−]
hsa-miR-3155a	MIMAT0015029	1.49	0.00736	chr10: 6152196–6152277 [+]
hsa-miR-1227-5p	MIMAT0022941	1.81	0.00751	chr19: 2234062–2234149 [−]
hsa-miR-6802-5p	MIMAT0027504	1.5	0.00762	chr19: 55239912–55239976 [−]
hsa-miR-572	MIMAT0003237	2.21	0.00776	chr4: 11368827–11368921 [+]
hsa-miR-1470	MIMAT0007348	2.5	0.00876	chr19: 15449548–15449608 [+]
hsa-miR-520f-3p	MIMAT0002830	2.11	0.01011	chr19: 53682159–53682245 [+]
hsa-miR-92b-3p	MIMAT0003218	1.36	0.01019	chr1: 155195177–155195272 [+]
hsa-miR-5572	MIMAT0022260	1.68	0.01052	chr15: 80581103–80581239 [+]
hsa-miR-4746-3p	MIMAT0019881	2.18	0.01059	chr19: 4445978–4446048 [+]
hsa-miR-550b-2-5p	MIMAT0022737	1.99	0.01082	chr7: 30289794–30289890 [−]
hsa-miR-518e-5p	MIMAT0005450	1.84	0.01137	chr19: 53729838–53729925 [+]
hsa-miR-6781-5p	MIMAT0027462	1.66	0.01184	chr17: 42823880–42823943 [−]
hsa-miR-6890-3p	MIMAT0027681	1.91	0.01185	chr3: 49099854–49099914 [−]
hsa-miR-760	MIMAT0004957	1.34	0.01196	chr1: 93846832–93846911 [+]
hsa-miR-516a-3p	MIMAT0006778	1.75	0.01338	chr19: 53756741–53756830 [+]
hsa-miR-4695-5p	MIMAT0019788	1.99	0.01347	chr1: 18883202–18883275 [−]
hsa-miR-501-3p	MIMAT0004774	1.34	0.014	chrX: 50009722–50009805 [+]
hsa-miR-1180-5p	MIMAT0026735	1.6	0.01427	chr17: 19344506–19344574 [−]
hsa-miR-4728-3p	MIMAT0019850	1.86	0.01456	chr17: 39726495–39726561 [+]
hsa-miR-523-5p	MIMAT0005449	1.79	0.01576	chr19: 53698385–53698471 [+]
hsa-miR-1193	MIMAT0015049	2.45	0.01676	chr14: 101030052–101030129 [+]
hsa-miR-4318	MIMAT0016869	1.84	0.01778	chr18: 37657135–37657215 [+]
hsa-miR-5580-5p	MIMAT0022273	1.82	0.01786	chr14: 53948427–53948484 [−]
hsa-miR-7112-5p	MIMAT0028121	1.75	0.01981	chr8: 144262673–144262737 [−]
hsa-miR-6805-3p	MIMAT0027511	1.52	0.02005	chr19: 55388181–55388242 [+]
hsa-miR-5587-5p	MIMAT0022289	1.46	0.02055	chr16: 535316–535368 [+]
hsa-miR-6851-5p	MIMAT0027602	1.61	0.02114	chr9: 33467869–33467935 [−]
hsa-miR-6814-3p	MIMAT0027529	1.53	0.02122	chr21: 41746772–41746841 [−]
hsa-miR-4663	MIMAT0019735	1.56	0.02128	chr8: 123215788–123215863 [−]
hsa-miR-6075	MIMAT0023700	2.54	0.02134	chr5: 1510762–1510856 [−]
hsa-miR-1343-3p	MIMAT0019776	1.37	0.02178	chr11: 34941837–34941920 [+]
hsa-miR-6737-5p	MIMAT0027375	2.61	0.2188	chr1: 153962351–153962420 [−]
hsa-miR-6879-3p	MIMAT0027659	1.52	0.0229	chr11: 65018505–65018570 [+]
hsa-miR-4701-3p	MIMAT0019799	235	0.02293	chr12: 48771975–48772037 [−]
hsa-miR-92b-5p	MIMAT0004792	1.6	0.02334	chr1: 155195177–155195272 [+]
hsa-miR-6753-5p	MIMAT0027406	1.7	0.02367	chr11: 68044794–68044957 [+]
hsa-miR-4485-5p	MIMAT0032116	1.55	0.02503	chr11: 10508270–10508326 [−]
hsa-miR-6769b-5p	MIMAT0027620	1.65	0.02522	chr1: 206474803–206474864 [+]
hsa-miR-519b-3p	MIMAT0002837	1.84	0.02561	chr19: 53695213–53695293 [+]
hsa-miR-4430	MIMAT0018945	1.66	0.02604	chr2: 33418516–33418564 [+]
hsa-miR-6511b-5p	MIMAT0025847	1.38	0.0264	chr16: 2106669–2106753 [−]
hsa-miR-595	MIMAT0003263	1.72	0.02696	chr7: 158532718–158532813 [−]
hsa-miR-3591-5p		1.82	0.02726
hsa-miR-6892-5p	MIMAT0027684	1.48	0.0278	chr7: 143382686–143382800 [+]
hsa-miR-6769a-5p	MIMAT0027438	1.66	0.02784	chr16: 4671318–4671390 [+]
hsa-miR-6815-3p	MIMAT0027531	1.51	0.02888	chr21: 45478266–45478326 [+]
hsa-miR-6082	MIMAT0023707	1.6	0.02979	chr4: 171186184–171186292 [+]
hsa-miR-4466	MIMAT0018993	1.8	0.03022	chr6: 156779678–156779731 [−]
hsa-miR-4758-5p	MIMAT0019903	1.56	0.0304	chr20: 62332487–62332557 [−]
hsa-miR-622	MIMAT0003291	1.43	0.03112	chr13: 90231182–90231277 [+]
hsa-miR-6851-3p	MIMAT0027603	1.6	0.03117	chr9: 33467869–33467935 [−]
hsa-miR-521	MIMAT0002854	1.6	0.03178
hsa-miR-521-1				chr19: 53748636–53748722 [+]
hsa-miR-521-2				chr19: 53716594–53716680 [+]
hsa-miR-636	MIMAT0003306	1.73	0.03187	chr17: 76736450–76736548 [−]
hsa-let-7c-3p	MIMAT0026472	1.12	0.03194	chr21: 16539828–16539911 [+]
hsa-miR-6810-3p	MIMAT0027521	1.46	0.03235	chr2: 218341911–218341980 [+]
hsa-miR-1247-5p	MIMAT0005899	1.58	0.03266	chr14: 101560287–101560422 [−]
hsa-miR-3648	MIMAT0018068	1.77	0.0332
hsa-mir-3648-1				chr21: 8208473–8208652 [+]
hsa-mir-3648-2				chr21: 8986999–8987178 [+]
hsa-miR-484	MIMAT0002174	1.31	0.03341	chr16: 15643294–15643372 [+]
hsa-miR-3621	MIMAT0018002	1.8	0.03348	chr9: 137169186–137169270 [−]
hsa-let-7e-5p	MIMAT0000066	1.18	0.03408	chr19: 51692786–51692864 [+]
hsa-miR-519d-3p	MIMAT0002853	1.43	0.03453	chr19: 53713347–53713434 [+]
hsa-miR-486-3p	MIMAT0004762	1.43	0.03481	chr8: 41660441–41660508 [−]
hsa-miR-3189-3p	MIMAT0015071	1.48	0.03544	chr19: 18386562–18386634 [+]
hsa-miR-5189-5p	MIMAT0021120	1.69	0.03563	chr16: 88468918–88469031 [+]
hsa-miR-4522	MIMAT0019060	1.44	0.03575	chr17: 27293910–27293996 [−]
hsa-miR-302c-5p	MIMAT0000716	1.48	0.03583	chr4: 112648363–112648430 [−]
hsa-miR-328-3p	MIMAT0000752	1.44	0.03667	chr16: 67202321–67202395 [−]
hsa-miR-1-5p	MIMAT0031892	1.22	0.03697	chr20: 62554306–62554376 [+]
hsa-miR-1234-3p	MIMAT0005589	1.39	0.03734	chr8: 144400086–144400165 [−]
hsa-miR-6826-3p	MIMAT0027553	1.4	0.03748	chr3: 129272146–129272243 [+]
hsa-miR-378c	MIMAT0016847	1.15	0.03837	chr10: 130962588–130962668 [−]
hsa-miR-30d-5p	MIMAT0000245	1.11	0.03859	chr8: 134804876–134804945 [−]
hsa-miR-7156-3p	MIMAT0028223	1.32	0.03873	chr1: 77060143–77060202 [+]
hsa-miR-519e-5p	MIMAT0002828	1.42	0.03896	chr19: 53679940–53680023 [+]
hsa-miR-6869-5p	MIMAT0027638	1.57	0.03987	chr20: 1392900–1392961 [−]
hsa-miR-3689f	MIMAT0019010	1.74	0.04079	chr9: 134850742–134850807 [−]
hsa-miR-4512	MIMAT0019049	1.55	0.04117	chr15: 66496958–66497034 [−]
hsa-miR-6857-5p	MIMAT0027614	1.4	0.04123	chrX: 53405673–53405765 [−]
hsa-miR-6889-3p	MIMAT0027679	1.47	0.04153	chr22: 41252992–41253050 [−]
hsa-miR-6131	MIMAT0024615	1.37	0.04303	chr5: 10478037–10478145 [+]
hsa-miR-6510-5p	MIMAT0025476	1.53	0.04315	chr17: 41517164–41517217 [−]
hsa-miR-4505	MIMAT0019041	1.66	0.04334	chr14: 73758747–73758819 [+]
hsa-miR-4685-3p	MIMAT0019772	1.4	0.04387	chr10: 98431292–98431360 [−]
hsa-miR-4530	MIMAT0019069	1.82	0.04423	chr19: 39409623–39409678 [−]
hsa-miR-6794-5p	MIMAT0027488	1.56	0.04427	chr19: 12852260–12852327 [+]
hsa-miR-125b-2-3p	MIMAT0004603	1.17	0.04465	chr21: 16590237–16590325 [+]
hsa-miR-4538	MIMAT0019081	1.54	0.04479	chr14: 105858165–105858242 [−]
hsa-miR-663b	MIMAT0005867	1.8	0.0453	chr2: 132256966–132257080 [−]
hsa-miR-3609	MIMAT0017986	1.42	0.04669	chr7: 98881650–98881729 [+]
hsa-miR-2861	MIMAT0013802	1.76	0.04701	chr9: 127785918–127786007 [+]
hsa-miR-4783-5p	MIMAT0019946	1.4	0.04721	chr2: 127423537–127423618 [−]
hsa-miR-671-3p	MIMAT0004819	1.26	0.04791	chr7: 151238421–151238538 [+]
hsa-miR-3928-5p	MIMAT0027037	1.19	0.04812	chr22: 31160062–31160119 [−]
hsa-miR-6821-5p	MIMAT0027542	1.61	0.04932	chr22: 49962866–49962939 [+]
hsa-miR-6853-3p	MIMAT0027607	1.24	0.0499	chr9: 35732922–35732995 [+]
Downregulated
hsa-miR-136-5p	MIMAT0000448	0.73	0.00047	chr14: 100884702–100884783 [+]
hsa-miR-551b-3p	MIMAT0003233	0.73	0.00117	chr3: 168551854–168551949 [+]
hsa-miR-367-3p	MIMAT0000719	0.45	0.00137	chr4: 112647874–112647941 [−]
hsa-miR-5590-5p	MIMAT0022299	0.46	0.00221	chr2: 134857820–134857873 [+]
hsa-miR-337-5p	MIMAT0004695	0.84	0.00244	chr14: 100874493–100874585 [+]
hsa-miR-1251-5p	MIMAT0005903	0.81	0.00248	chr12: 97491909–97491978 [+]
hsa-miR-543	MIMAT0004954	0.83	0.00361	chr14: 101031987–101032064 [+]
hsa-miR-138-5p	MIMAT0000430	0.8	0.00399	chr16: 56858518–56858601 [+]
hsa-miR-20b-3p	MIMAT0004752	0.47	0.00427	chrX: 134169809–134169877 [−]
hsa-miR-495-5p	MIMAT0022924	0.85	0.00475	chr14: 101033755–101033836 [+]
hsa-miR-6808-3p	MIMAT0027517	0.62	0.00586	chr1: 1339650–1339708 [−]
hsa-miR-4705	MIMAT0019805	0.88	0.00765	chr13: 102045934–102046004 [−]
hsa-miR-376a-3p	MIMAT0000729	0.81	0.00766	chr14: 101040782–101040849 [+]
hsa-miR-4796-3p	MIMAT0019971	0.73	0.00775	chr3: 114743445–114743525 [−]
hsa-miR-495-3p	MIMAT0002817	0.75	0.01027	chr14: 101033755–101033836 [+]
hsa-miR-154-5p	MIMAT0000452	0.84	0.01082	chr14: 101059755–101059838 [+]
hsa-miR-496	MIMAT0002818	0.84	0.01097	chr14: 101060573–101060674 [+]
hsa-miR-548ar-3p	MIMAT0022266	0.46	0.01173	chr13: 114244505–114244561 [+]
hsa-miR-670-3p	MIMAT0026640	0.75	0.01195	chr11: 43559656–43559753 [+]
hsa-miR-488-3p	MIMAT0004763	0.85	0.01201	chr1: 177029363–177029445 [−]
hsa-miR-29b-2-5p	MIMAT0004515	0.9	0.01285	chr1: 207802443–207802523 [−]
hsa-miR-497-5p	MIMAT0002820	0.86	0.01297	chr17: 7017911–7018022 [−]
hsa-miR-335-5p	MIMAT0000765	0.87	0.01304	chr7: 130496111–130496204 [+]
hsa-miR-487a-3p	MIMAT0002178	0.82	0.01611	chr14: 101052446–101052525 [+]
hsa-miR-544a	MIMAT0003164	0.8	0.01635	chr14: 101048658–101048748 [+]
hsa-miR-342-5p	MIMAT0004694	0.89	0.01789	chr14: 100109655–100109753 [+]
hsa-miR-1197	MIMAT0005955	0.82	0.01812	chr14: 101025564–101025651 [+]
hsa-miR-153-5p	MIMAT0026480	0.81	0.01966	chr7: 157574336–157574422 [−]
hsa-miR-6844	MIMAT0027589	0.78	0.0209	chr8: 124508515–124508576 [−]
hsa-miR-758-3p	MIMAT0003879	0.85	0.02115	chr14: 101026020–101026107 [+]
hsa-miR-22-5p	MIMAT0004495	0.87	0.02205	chr17: 1713903–1713987 [−]
hsa-miR-29b-3p	MIMAT0000100	0.85	0.02327	chr7: 130877459–130877539 [−]
hsa-miR-377-3p	MIMAT0000730	0.83	0.02501	chr14: 101062050–101062118 [+]
hsa-miR-153-3p	MIMAT0000439	0.8	0.02614	chr2: 219294111–219294200 [−]
hsa-miR-2277-5p	MIMAT0017352	0.9	0.02674	chr5: 93620696–93620788 [−]
hsa-miR-9-3p	MIMAT0000442	0.9	0.02713	chr1: 156420341–156420429 [−]
hsa-miR-1307-5p	MIMAT0022727	0.83	0.03086	chr10: 103394253–103394401 [−]
hsa-miR-548a-3p	MIMAT0003251	0.72	0.03186	chr6: 18571784–18571880 [+]
hsa-miR-494-3p	MIMAT0002816	0.9	0.03362	chr14: 101029634–101029714 [+]
hsa-miR-22-3p	MIMAT0000077	0.89	0.03656	chr17: 1713903–1713987 [−]
hsa-miR-3689a-5p	MIMAT0018117	0.43	0.03897	chr9: 134849487–134849564 [−]
hsa-miR-146a-5p	MIMAT0000449	0.9	0.04076	chr5: 160485352–160485450 [+]
hsa-miR-4477b	MIMAT0019005	0.81	0.04126	chr9: 41233755–41233835 [−], chr9: 63819574–63819654 [+]
hsa-miR-582-3p	MIMAT0004797	0.88	0.04397	chr5: 59703606–59703703 [−]
hsa-miR-7-1-3p	MIMAT0004553	0.9	0.04513	chr9: 83969748–83969857 [−]
hsa-miR-376c-3p	MIMAT0000720	0.83	0.04613	chr14: 101039690–101039755 [+]
hsa-miR-135b-5p	MIMAT0000758	0.85	0.04615	chr1: 205448302–205448398 [−]
hsa-miR-376a-5p	MIMAT0003386	0.86	0.04712	chr14: 101040782–101040849 [+]
hsa-miR-128-2-5p	MIMAT0031095	0.62	0.04776	chr3: 35744476–35744559 [+]
hsa-miR-376c-5p	MIMAT0022861	0.85	0.04836	chr14: 101039690–101039755 [+]
hsa-miR-5582-3p	MIMAT0022280	0.63	0.04923	chr11: 46753125–46753192 [−]
hsa-miR-548ac	MIMAT0018938	0.53	0.04978	chr1: 116560024–116560111 [−]
hsa-miR-3182	MIMAT0015062	0.7	0.04988	chr16: 83508346–83508408 [+]
hsa-miR-548a-5p	MIMAT0004803	0.72	0.04997	chr8: 104484369–104484465 [−]

Bold face letter miRNAs show fold change >2.0 (upregulated) or <0.5 (downregulated)

3.2. Effects of covariates toward extremely upregulated 24 miRNAs (fold change cut off >2.0)

The extremely upregulated 24 miRNAs were used for target prediction and gene ontology. These miRNAs were further evaluated for their association with covariates such as age, gender, brain pH, PMI, alcohol abuse, antidepressant toxicology, and suicide as cause of death. Detailed results are shown in the supplementary section (Table S1). None of the changes in miRNAs were associated with any of the covariates except age was negatively correlated with miR-4701–3p (p=0.043, r=−0.407) and brain pH was positively correlated with miR-6737–5p (p=0.031, r=0.432), miR-6075 (p=0.040, r=0.423), miR-4701–3p (p=0.031, r=0.433) in the control group. Within the MDD group, alcohol abuse had significant impact on the expression of miR-6077 (p=0.012), miR-6789–3p (p=0.034), miR-648 (p=0.027), and mir-4746–3p (p=0.009). There were no significant differences in miRNA expression when MDD subjects who died by suicide were compared with those who died by causes other than suicide (Table S1).

3.3. Target gene prediction

Owing to the repressive nature of miRNAs in regulating target gene expression, top 24 significantly upregulated miRNAs (FC >2.0) were used for target gene prediction analysis following conservative prediction algorithm. As mentioned in methods section, a common list of targets that were shared by any five of the prediction algorithms was finalized. A total of 1092 predicted target genes were finally used for subsequent GO and KEGG based pathway analyses.

3.4. miRNA target gene-based gene ontology and pathway prediction

Following GO enrichment analysis with the predicted targets of significantly upregulated 24 miRNAs in ACC of MDD subjects, we found three separate clustering of GO terms under the BP category (p value cutoff <0.05 [FDR], Edge Cutoff: 0.3) (Fig. 2A). The clustered biological functions were plotted into networks to represent their relatedness. The nodes representing the GO terms are only connected through edges when they share 20% or more genes. The bigger nodes represent larger gene sets. Thicker edges represent more overlapped genes. Gross functional enrichments of genes for nucleic acid and protein metabolisms were identified as shown in the two left most clusters. The right most cluster in the network map demonstrates the presence of four enriched terms; all of which were central to neuronal growth and development (e.g., neurogenesis, nervous system development, neuron differentiation and generation of neurons). Next, another network map was created following the GO term enrichment under CC category (p value cutoff <0.05 [FDR], Edge Cutoff: 0.3) (Fig. 2B). The map showed predominant enrichment of terms correlated with somatic (neuronal cell body), somatodendritic and synaptic (pre, post synaptic, and synaptic vesicle) functions in central nervous system. Additionally, the targeted enrichment of genes was mapped translating to glutamatergic synapses which have earlier been suggested to be important in MDD associated ACC functions.

Fig. 2: Gene ontology and pathway prediction.

GO terms for A) biological process and B) cellular component based on predicted 1,092 target genes. The GO prediction analysis was done following an FDR corrected p-value cutoff 0.05 to determine the gene set enrichment in biological process and cellular component category separately by ShinyGO program. Each category the enriched terms were used to create networks where nodes are presented with terms and connected with edges. As shown in the graphs, if two nodes are connected, then they share 20% (default) or more genes. Bigger nodes represent larger gene sets. Thicker edges represent more overlapped genes. C) A pathway-based network map (p value cutoff <0.05 [FDR], Edge Cutoff: 0.2; most significant display terms: 50) was prepared showing the predicted involvement of various neuro-molecular pathways, those are earlier reported to be important in MDD pathogenesis.

A pathway-based network map (p value cutoff <0.05 [FDR], Edge Cutoff: 0.2; most significant display terms: 50) was separately prepared showing the predicted involvement of various neuro-molecular pathways (Fig. 2C) earlier reported to be quite important in MDD pathogenesis. These include ErbB, Wnt, MAPK, mTOR, and phosphatidylinositol signaling. A significant enrichment of pathways related to immuno-inflammatory functions, mostly pro-inflammatory cytokines (IL-17 and TNF) and their regulating factors (NF-kappa B signaling) were also noted.

WebGestalt program was additionally used to further validate ShinyGO based findings based on putative targets of 24 significantly upregulated miRNAs. In both the KEGG pathway and GO analysis, top 10 pathways are listed according to significance level as a large number of pathways were hit (KEGG pathway, 271; BP, more than 2,000; CC, 779; molecular functions [MF], 1158) (Table 3 and 4). From the KEGG pathway, long-term potentiation (LTP, hsa04720) had the highest enrichment ratio (3.4). From the BP results, several GO terms were found that were primarily related to the central nervous systems (CNS) such as neuron projection development (GO:0031175), neuron development (GO:0048666), generation of neurons (GO:0048699), neuron differentiation (GO:0030182), and neurogenesis (GO:0022008). From the CC results, CNS related GO terms such as synapse (GO:0045202), synapse part (GO:0044456), neuron part (GO:0097458), and neuron projection (GO:0043005) were found. In terms of MF, RNA polymerase II regulatory region DNA binding (GO:0001012) had the highest enrichment ratio (1.95). A summary of the significantly annotated (FDR corrected) pathways with a focus on neuronal morphogenesis and regulation are presented as a bar diagram showing the enrichment of genes under each term (Fig. 3).

Table 3.

KEGG pathway results from extremely upregulated 24 miRNAs (fold change cut off >2.0)

Gene Set	Description	Enrichment ratio	p-value	FDR	Candidate genes
hsa04720	Long-term potentiation	3.4039	0.00016382	0.0055154	ADCY1, CALM1, GRIA2, GRIN1, MAPK1, PPP1CB, PRKCA, PRKCB, PRKCG, RAP1A, RPS6KA1, RPS6KA3
hsa05220	Chronic myeloid leukemia	3.2509	0.00014340	0.0055154	CBL, CDKN1B, GAB2, GADD45A, HDAC2, MAPK1, MYC, NFKB1, NFKBIA, PIK3CD, POLK, RELA, SHC4
hsa04550	Signaling pathways regulating pluripotency of stem cells	2.7345	0.000035245	0.0041494	ACVR1B, ACVR2B, BMPR1A, FGFR2, FZD10, FZD3, FZD5, IGF1, IL6ST, KAT6A, MAPK1, MEIS1, MYC, ONECUT1, OTX1, PCGF5, PIK3CD, SMAD9, WNT3A, ZFHX3
hsa04722	Neurotrophin signaling pathway	2.7150	0.00014818	0.0055154	ARHGDIA, CALM1, IRAK2, IRAK4, MAPK1, NFKB1, NFKBIA, NTRK2, NTRK3, PIK3CD, RAN1A, RELA, RPS6KA1, RPS6KA3, RPS6KA5, SHC4, TP73
hsa04360	Axon guidance	2.4978	0.000040804	0.0041494	ARHGEF12, CFL2, CXCL12, DPYSL5, FZD3, GNAI2, MAPK1, NRP1, PAK2, PAK3, PARD6G, PIK3CD, PLXNA1, PLXNC1, PRKCA, RASA1, ROBO2, ROCK1, SEMA3A, SEMA3C, SEMA5A, SRGAP2, SRGAP3
hsa04014	Ras signaling pathway	2.3756	0.000010929	0.0035628	ARF6, CALM1, CSF1, FGF1, FGFR2, GAB2, GNGT2, GRIN1, IGF1, KIT, KSR2, MAPK1, NFKB1, NTRK2, PAK2, PAK3, PIK3CD, PRKCA, PRKCB, PRKCG, RAB5B, RAP1A, RASAL2, RASGRF2, RELA, SHC4, STK4, TEK
hsa05202	Transcriptional misregulation in cancer	2.3501	0.00010641	0.0055154	AFF1, CDKN1B, FOXO1, GADD45A, H3F3B, HDAC2, HMGA2, IGF1, IL2RB, KLF3, MEF2C, MEIS1, MMP3, MPO, MYC
hsa04144	Endocytosis	2.1030	0.00018487	0.0055154	AGAP1, ARF6, ARFGEF1, CAPZA1, CBL, CHMP4C, CHMP7, CYTH4, FGFR2, IL2RB, IQSEC2, KIF5C, NEDD4, PARD6G, PIP5K1C, PML, PRKCI, RAB10, RAB11FIP4, RAB31, RAB35, RAB5B, RABEP1, SH3GLB1, SMURF1, SNX12, STAM2
hsa04010	MAPK signaling pathway	1.9971	0.00016207	0.0055154	CACNA1E, CACNB4, CSF1, DUSP16, FGF1, FGFR2, CADD45A, IGF1, IRAK4, KIT, MAPK1, MEF2C, MKNK2, MYC, NFKB1, NLK, NTRK2, PAK2, PRKCA, PRKCB, PRKCG, RAP1A, RASA1, RASGRF2, RELA, RPS6KA1, RPS6KA3, RPS6KA5, STK4, TAOK1, TEK
hsa05200	Pathways in cancer	1.7704	0.000050912	0.0041494	ADCY1, ADCY2, ARHGEF12, ARNT, CALM1, CBL, CCDC6, CDKN1B, COL4A2, CXCL12, DLL4, EDNRB, ESR1, FGF1, FGFR2, FOXO1, FZD10, FZD3, FZD5, GADD45Am GNAI2, GNGT2, HDAC2, HEY2, IGF1, IL2RB, IL6ST, ITGA2, KIT, LAMC3, MAPK1, MYC, NFKB1, NFKBIA, NOTCH2, PIK3CD, PML, POLK, PRKCA, PRLCB, PRKCG, RELA, ROCK1, RPS6KA5, SP1, STK4, TPM3, TRAF4, WNT3A

Table 4.

Gene ontology results from extremely upregulated 24 miRNAs (fold change cut off >2.0)

Gene Set	Description	Enrichment ratio	p-value	FDR
Biological Processes (BP)
GO:0007156	Homophilic cell adhesion via plasma	3.2174	1.1770e-8	0.000035666
GO:0098742	Cell-cell adhesion via plasma-membrane	2.5358	1.1691e-7	0.00017713
GO:0030099	Myeloid cell differentiation	2.1868	4.9157e-7	0.00044689
GO:0031175	Neuron projection development	1.7059	4.3918e-7	0.00044689
GO:0048666	Neuron development	1.6533	4.7477e-7	0.00044689
GO:0048699	Generation of neurons	1.6481	3.8315e-9	0.000034832
GO:0030182	Neuron differentiation	1.6467	2.6161e-8	0.000047566
GO:0045944	Positive regulation of transcription by RNA	1.6351	2.6580e-7	0.00034519
GO:0022008	Neurogenesis	1.6031	1.0945e-8	0.000035666
GO:0006468	Protein phosphorylation	1.5305	1.9251e-8	0.000043753
Cellular Components (CC)
GO:0045202	Synapse	2.2821	0	0
GO:0044456	Synapse part	2.1256	3.8531e-	8.6364e-9
GO:0030054	Cell junction	1.8931	1.8015e-	3.0239e-8
GO:0097458	Neuron part	1.8893	1.0347e-	6.0790e-11
GO:0043005	Neuron projection	1.8638	4.3436e-	6.3796e-8
GO:0098805	Whole membrane	1.8445	3.5398e-	1.3864e-9
GO:0005887	Integral component of plasma membrane	1.8253	1.6646e-	4.8899e-9
GO:0031226	Intrinsic component of plasma membrane	1.7831	4.4101e-	8.6364e-9
GO:0044463	Cell projection part	1.7663	3.2440e-9	3.8118e-7
GO:0120038	Plasma membrane bounded cell projection	1.7663	3.2440e-9	3.8118e-7
Molecular Functions (MF)
GO:0001012	RNA polymerase II regulatory region	1.9494	2.4542e-	0.0000094567
GO:0000977	RNA polymerase II regulatory region	1.9123	8.2815e-	0.000022206
GO:0016773	Phosphotransferase activity, alcohol	1.9041	3.4398e-	0.000010761
GO:0044212	Transcription regulatory region DNA	1.8801	7.5765e-	0.0000068572
GO:0001067	Regulatory region nucleic acid binding	1.8759	8.4200e-	0.0000068572
GO:0000976	Transcription regulatory region sequence-	1.8594	1.3786e-	0.000028751
GO:1990837	Sequence-specific double-stranded DNA	1.8351	1.3181e-	0.000028751
GO:0003690	Double-stranded DNA binding	1.7775	1.6863e-	0.000031653
GO:0000981	DNA-binding transcription factor activity,	1.6004	2.5191e-	0.0000094567
GO:0003700	RNA -binding transcription factor activity	1.5945	1.0960e-	0.0000068572

Fig. 3: Significantly annotated top 10 biological pathway.

GO terms for biological process was predicted by WEB-based GEne SeT AnaLysis Toolkit (WebGestalt, http://www.webgestalt.org/). The significantly annotated (FDR corrected) pathways with a focus on neuronal morphogenesis and regulation are presented as a bar diagram showing the enrichment of genes under each term.

Next, in order to create an integrated visualization network to determine involvement of neuronal morphogenesis, growth and survival factors, a select list of 55 genes from WebGestalt analysis were used in the Metascape program. The enrichment visualization network resulted from the analysis was quite interesting and for the first time show intra- and inter-cluster similarities of enriched terms based on overrepresented gene list. As presented in Fig. 4A, the network was mapped with 20 enriched terms connected by two major domains and three smaller subdoamins. In particular, the terms associated with two major domains in the network were critically enriched (based on calculation using accumulative hypergeometric p-values and enrichment factors) for their roles in regulating phosphatidyl signaling, GTPase signaling, neuron migration and NTRK based neurotrophin signaling. The network is colored by cluster ID, where nodes that share the same cluster IDs are typically close to each other. In Fig. 4B, the same enrichment network has its nodes colored by p-value. The darker the color, the more statistically significant the node is (see legend for p-value ranges). The network result is also presented in a color-coded bar plot in Fig. 4C. Each bar in the plot shows enriched terms across input gene lists, and color coded with respective p-values.

Fig. 4: The enrichment visualization network based on overrepresented gene list.

A) The network was mapped with 20 enriched terms connected by two major domains and three smaller sub-domains. B) Representation of same enrichment network with colored coded (nodes) p-value. The darker the color, the more statistically significant the node is (see legend for p-value ranges) and C) the same enrichment terms presented listed up as a color-coded bar plot.

4. Discussion

This is the first study to examine transcriptome-wide miRNA expression changes in ACC of MDD subjects using next-generation miRNA sequencing. Based on sequencing results, we detected 2540 annotated miRNAs and 423 novel miRNAs. Of those, 230 miRNAs were significantly altered in MDD subjects compared to control subjects. When directionality of expression was considered, we found that 117 miRNAs (4.61%) were significantly upregulated and 54 (2.13%) were downregulated in MDD subjects (n=22) compared with non-psychiatric controls (n=25). Our web centric parameters and ontology enrichment functions (based on predicted target genes) identified two major domains related to phosphatidyl signaling, GTPase signaling, neuronal migration, and neurotrophin signaling.

In a previous qPCR-based study of 29 miRNAs, Azevedo et al (2016) found decreased expression of miR-34a and miR-184e in ACC of MDD subjects, however, these changes could not survived multiple correction testing (Azevedo et al., 2016). Owing to the repressive nature of miRNAs in regulating target gene expression, we selected top 24 significantly upregulated miRNAs (>2 fold) in ACC of MDD subjects for their possible role in MDD pathogenesis. Our prediction based results helped in finding many target genes that are important in glutamatergic system (GRIN1 and GRIA2) and neurotrophin signaling (NTRK2 and NTRK3) as well neuronal morphogenesis. The upregulated 24 miRNAs that were used in the prediction analysis were relatively new and their functions have not been extensively characterized. Nevertheless, we conducted gene set enrichment analysis to determine relevant neurobiological functions on the basis of stringent statistical parameters. Intra- and inter-cluster similarities of enriched terms based on overrepresented gene list showed neurobiological functions associated with neuronal growth and survival.

The network enrichment map based on ShinyGO results showed predominance of ontological terms that were correlated with somatic (neuronal cell body), somatodendritic, and presynaptic, postsynaptic, and synaptic vesicle functions. Additionally, we mapped the targeted enrichment of genes translating to glutamatergic synapse which has earlier been shown to play important role in ACC functions associated with MDD pathogenesis. Other reports have previously suggested the role of specific biological factors and their dysregulation in complex behavioral manifestations including MDD and MDD associated suicidal behavior (Pandey, 2013). Several of them have been tested for their prognostic values in predicting complex suicidal behavior (e.g., suicide attempts); however, only a handful of them have proved to be reliable in clinical practices. For instance, the levels of prolactin and thyroid hormone have been found to be associated with suicide attempts in psychiatric patients (Pompili et al., 2012). Moreover, showing the predicted involvement of various other neuro-molecular pathways such as ErbB, Wnt, MAPK, mTOR, and phosphatidylinositol signaling made these findings much more relevant. These pathways have been suggested to play critical roles in maintaining neuronal health and activity under normal conditions but are significantly modified in MDD. We also identified significant enrichment of pathways related to immunomodulatory functions, primarily underlying pro-inflammatory functions mediated through cytokines such as interleukin-17 and tumor necrosis factor (TNF) and their regulatory factors NF-kappa B (NF-kB). This is the first report showing miRNA-mediated regulation of pro-inflammatory genes/pathways in ACC of MDD subjects. This is in line with a previous report showing similar inflammatory activation in ACC by peripheral cytokines where inflammation-associated mood deterioration was significantly correlated with enhanced activity of subgenual anterior cingulate cortex (sACC) (Harrison et al., 2009). NF-kB signaling pathway (NFKB1) is important in MDD pathogenesis from neurotrophin point of view. It has been suggested that the p75NTR, a proBDNF receptor, activates the NF-kB signaling pathway, which results in the alterations of transcription of multiple genes involved to promote neuronal death such as toll-like receptors (TLRs) and tumor necrosis factor-alpha (TNF-α) (Napetschnig and Wu, 2013; Sugiyama et al., 2016). It appears that there is an integrated tripartite relationship between miRNAs, inflammation and neurotrophic activity, which might hold the possibility for deprivation of neuronal growth and functioning in ACC under MDD pathogenesis.

Our results from WebGestalt analysis suggested functional deficits associated with LTP, neurotrophin signaling, and axon guidance. It has earlier been suggested that the postsynaptic form of LTP in the ACC synapse depends on the activation of N-methyl-D-aspartate receptors (NMDARs), a type of glutamate receptor (Li et al., 2017; Zhao et al., 2005). In addition, decreased levels of gamma-aminobutyric acid (GABA) made by glutamate with glutamate decarboxylase (GAD), has also been reported in the ACC of MDD subjects (Price et al., 2009; Schur et al., 2016). Altogether, this shows that in ACC, the LTP modulation in MDD subjects could be associated with upregulation in certain miRNAs via NMDA receptor dysregulation. On the other, axon guidance is known as a key determinant in the formation of neuronal networks. As mentioned above, miRNA-based prediction results found several target genes that play key roles in axon and dendrite formation (Cariboni et al., 2011; Cobos et al., 2007; Dickson and Gilestro, 2006; Erskine et al., 2011; Mastick et al., 2010). Therefore, our miRNA findings may highlight that MDD associated morphometric deficits in ACC could be due to altered dendritic formation and their connecting networks.

As mentioned in results section, using Metascape program, we prepared enrichment visualization networks. The outcomes were very interesting and showed for the first time intra- and inter-cluster similarities of enriched terms based on overrepresented gene list defining neurobiological functions that either were associated with neuronal growth or survival. With the help of web centric parameters and ontology enrichment functions (based on calculation using accumulative hypergeometric p-values and enrichment factors), we were able to identify two major domains in the network. Interestingly, they built the most cohesive part of the network enrichment with terms associated with phosphatidylinositol signaling, GTPase signaling, neuron migration and most importantly NTRK (receptor for BDNF) based neurotrophin signaling. These results reconfirmed our findings from other two web centric prediction analysis programs (ShinyGo and WebGestalt). This is highly encouraging given that the Metascape analysis is based on 40 independent knowledgebase to integrate system level information and creating a comprehensive gene list. Moreover, the enrichment scores for creating the networks were infringed with rigorous statistical clustering.

Altogether our results identified significant changes in miRNA expression in ACC of MDD subjects. Functional annotation from all three prediction programs consistently highlighted the role miRNAs central to neuronal growth, morphogenesis and differentiation. Precisely, the neuronal atrophy and reduction in volume of ACC, which is hallmark of MDD pathology (Koolschijn et al., 2009), could be considered as potential loss of gene function due to the changes in miRNA landscape. Our comparison with MDD suicide and non-suicide groups suggest that these miRNAs were not specifically associated with suicide as has been reported earlier in dlPFC, where some miRNAs were explicitly related to suicidal behavior (Wang et al., 2018). Interestingly, miRNAs that were altered in ACC of MDD subjects, did not overlap with miRNAs that were dysregulated in other brain areas such as dlPFC (Smalheiser et al., 2012) and locus coeruleus (Roy et al., 2017b). This suggests that specific miRNAs may be regulating gene network(s) or functions that could be critical in ACC functioning and relevant to MDD pathogenesis. In the future, gene expression studies need to be performed to examine precise miRNA functioning in ACC of MDD subjects.

From gene regulation perspective, miRNA acts as repressor of target gene expression, thus large scale miRNA expression profiling can inform systems-level changes in gene regulation. Measurement of the relative abundance of a cohort of miRNAs, ranging from a few significantly altered miRNAs to a group of several miRNAs closely adhered to a specific neurobiological pathway (e.g., phosphatidyl signaling, GTPase signaling and neurotrophin signaling) could help underpinning the causal association. This may help strengthen strategize precision treatment towards a set of targets.

Highlights.

• Expression levels of miRNAs were determined in anterior cingulate cortex of MDD subjects by miRNA-seq.

• A significant number of miRNAs were differentially regulated in MDD subjects.

• Predicted target-based gene ontology suggested that altered miRNAs may be related to abnormalities in neuronal functions, atrophy and volume loss in ACC of MDD subjects.