Blood miR-144-3p: A Novel Diagnostic and Therapeutic Tool for Depression

Abstract

Major depressive disorder (MDD) is the leading cause of disability worldwide. There is an urgent need for objective biomarkers to diagnose this highly heterogeneous syndrome, assign treatment, and evaluate treatment response and prognosis. MicroRNAs (miRNAs) are short non-coding RNAs, which are detected in body fluids that have emerged as potential biomarkers of many disease conditions. The present study explored the potential use of miRNAs as biomarkers for MDD and its treatment. We profiled the expression levels of circulating blood miRNAs from mice that were collected before and after exposure to chronic social defeat stress (CSDS), an extensively validated mouse model used to study depression, as well as after either repeated imipramine or single-dose ketamine treatment. We observed robust differences in blood miRNA signatures between stress-resilient and stress-susceptible mice after an incubation period, but not immediately after exposure to the stress. Furthermore, ketamine treatment was more effective than imipramine at re-establishing baseline miRNA expression levels, but only in mice that responded behaviorally to the drug. We identified the red blood cell-specific miR-144-3p as a candidate biomarker to aid depression diagnosis and predict ketamine treatment response in stress-susceptible mice and MDD patients. Lastly, we demonstrate that systemic knockdown of miR-144-3p, via subcutaneous administration of a specific antagomir, is sufficient to reduce the depression-related phenotype in stress-susceptible mice. RNA-sequencing analysis of blood after such miR-144-3p knockdown revealed a blunted transcriptional stress signature as well. These findings identify miR-144-3p as a novel target for diagnosis of MDD as well as for antidepressant treatment, and enhance our understanding of epigenetic processes associated with depression.

Introduction

Major depressive disorder (MDD) is a highly heterogeneous syndrome characterized by mood disturbances, anhedonia, and alterations in physiological functions, cognition, and psychomotor activity (1). Current diagnosis for MDD is based solely on clinical observation, relying on the presence of specific signs and symptoms without any objective biological readouts (2). Antidepressant (AD) drugs, of which the vast majority act on the brain’s monoamine systems, are the front-line treatment for depression (3). However, commonly used ADs have a slow onset of action, with <50% of individuals showing complete remission (4). As a result, a great number of individuals remain treatment-resistant (5). Recently, ketamine, a non-competitive NMDA receptor antagonist among other actions, was approved by the U.S. FDA as a rapid-onset treatment for individuals with treatment-resistant depression (6, 7). To improve diagnostic precision and refine both treatment assignment and assessment of treatment response, there is a great need for objective biological biomarkers for depression.

MicroRNAs (miRNAs) have emerged as important players in depression pathogenesis (8). These small non-coding RNA molecules (~22 nucleotides) act as potent epigenetic post-transcriptional regulators of gene expression. MiRNAs bind to the 3’ untranslated region (UTR) of particular target mRNAs, which results in mRNA destabilization or translational repression, ultimately reducing the overall level of the encoded proteins. Notably, a single miRNA can target hundreds of transcripts, and individual mRNAs can be the target of numerous miRNAs, thereby forming complex transcriptional regulatory networks (9). MiRNAs are detected in the circulation, and evidence suggests a correlation between specific circulating miRNA levels and several disease states, including depression (10-13). Furthermore, several studies have found alterations in circulating miRNA levels following AD treatment (14, 15), suggesting that such treatments may exert their effect, at least partially, by influencing miRNAs. However, to date, no studies have comprehensively probed circulating miRNA expression in a pre-clinical investigation or compared how miRNA levels are altered by mechanistically divergent AD treatments.

In animal models, manipulating the expression levels of specific miRNAs in the brain has been shown to causally regulate stress responses and depression-related measures. For instance, knockdown (KD) of miR-16 using an antagomir in the raphe nucleus and locus coeruleus of mice blocked depressive-like behavioral abnormalities induced by chronic stress (16). Similarly, KD of miR-218 using an antagomir in the medial prefrontal cortex (mPFC) increased susceptibility to chronic social defeat stress (CSDS), whereas its overexpression using an adeno-associated virus promoted stress resilience (17). Similar results have been observed using genetic approaches, in that mice that express high or low levels of miR-135 in raphe nuclei displayed alterations in anxiety- and depression-like behaviors (14). While these studies show considerable promise, they involved invasive intracranial surgery which limits their translational potential. Therefore, identification of a circulating miRNA, the manipulation of which alters stress-related behavior, would provide a critical link for translational miRNA therapeutics for MDD.

Here, we comprehensively explored the potential use of circulating miRNAs as biomarkers for MDD and the prediction and assessment of treatment response. We first profiled the expression levels of mouse blood miRNAs collected at multiple time points before and after CSDS as well as after subsequent treatment with either of two ADs: repeated treatment with imipramine (a standard monoamine-based AD) or a single dose of ketamine. This rich dataset delineated the blood miRNA fingerprint of susceptibility versus resilience to CSDS, as well as the hallmarks of treatment response versus non-response to the two ADs. Our bioinformatic analysis identified miR-144-3p as a candidate biomarker for stress susceptibility and predicting ketamine treatment response in mice. We then demonstrated that this finding has transitional relevance in a human cohort as a predictor of MDD severity and ketamine treatment response. Finally, we showed that systemic KD of miR-144-3p using an antagomir is sufficient to attenuate depression-related phenotypes in mice following stress. RNA-sequencing (RNA-seq) analysis of blood samples after CSDS and miR-144-3p KD revealed that the KD also attenuated the molecular signature induced by CSDS on the blood transcriptome. Taken together, this study refines our understanding of epigenetic changes in response to stress and identifies a candidate miRNA that can be utilized both as a circulating biomarker for depression and its treatment as well as a potential treatment target itself.

Methods

Animals

C57BL/6J male mice (8-weeks old) (Jackson Laboratory, Bar Harbor) were used for all experiments and habituated to the animal facility for 1 week before experimental manipulations. Mice were housed in groups of 5 at room temperature (24 ± 1 °C) under a 12-hr light/dark cycle (lights on from 7:00 A.M.) with ad libitum access to water and food. Experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) at Mount Sinai.

Human subjects

Male and female participants between the ages of 18–65 were recruited through the Depression and Anxiety Center for Discovery and Treatment at the Icahn School of Medicine at Mount Sinai for a ketamine study (18). The psychiatric diagnosis was assessed using the Structural Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (SCID-IV) and fifth edition (SCID-5) (19, 20). Participants met diagnostic criteria for a current major depressive episode of at least moderate severity as measured by the Clinical Global Impression-Severity scale (CGI-S) (21), as well as a lifetime history of non-response to at least two trials of an AD (not ketamine) according to the Antidepressant Treatment History Form (22). Demographic information (Suppl. Table 1), as well as additional psychiatric evaluation, took place within 28 days before the initial blood draw. Exclusion criteria included a lifetime history of a psychotic disorder, bipolar disorder, alcohol or other substance abuse in the previous six months (confirmed by urine toxicology), any unstable medical condition, any systemic inflammatory or autoimmune disease, current AD use, pregnancy or nursing. Depression severity at baseline was measured using the Montgomery-Åsberg Depression Rating Scale (MADRS) (23) after which blood samples were collected. Thereafter, participants received a single ketamine I.V. infusion of 0.5 mg/kg over 40 min. One day after the infusion, both depression severity scores and blood samples were collected again. All participants provided informed consent to have blood samples and psychosocial data collected. The Program for the Protection of Human subjects at Mount Sinai approved the protocol, consent forms, and study procedures. Participants were enrolled in one of the following ketamine studies registered on clinicaltrials.gov: NCT01507181, NCT00548964, NCT01880593, NCT03102736.

Chronic social defeat stress (CSDS) procedure

In this procedure, C57BL/6J mice were introduced into the cage of resident CD1 retired breeder mice, which were pre-screened for aggression (24). The procedure was repeated daily over 10 days including 5-min of physical interaction with a novel CD1 mouse, followed by housing across a perforated Plexiglas divider for the remaining 24 hr to allow sensory contact. Control mice were housed with another control cage mate with a perforated divider between them and rotated to a different cage daily. Both control and defeated mice were single-housed during and after CSDS for behavioral testing and AD treatments.

Behavioral testing

Social interaction (SI)

The SI test for social avoidance behavior (24) was repeated at different time points. During this 5-min test, the animal’s exploratory behavior was assessed in an open-field arena containing a wire enclosure that was empty for the first 2.5-min of the test (no target present) and subsequently held a novel CD1 aggressor in the second 2.5-min (target present). Between the two sessions, the experimental mouse was removed from the arena and placed back into its home cage. The experiment was conducted under red light conditions. The immediate area around the wire enclosure is labeled as the interaction zone. Social avoidance was measured using the social interaction ratio calculated by dividing the time spent in the interaction zone when the target mouse is present over the time spent in the zone when the target mouse is absent. Using the SI ratio, mice were subdivided into either susceptible (SI ratio < 0.8) or resilient (SI ratio > 1.2) subgroups, a measure that has been shown to be highly predictive of numerous other behavioral sequelae after CSDS (25). SI data and group classifications for the miRNA discovery part of the study were published previously (26).

Sucrose preference

This test probes reduced preference for sweetened drinks as a proxy for anhedonia. In the sucrose preference test, mice were given the choice between two bottles with either water or a 2% sucrose solution, which were measured by daily weighing of the bottles over three days (25, 27). Sucrose preference was calculated by determining the percentage of total sucrose solution consumption divided by total liquid consumption.

Marble burying

In this test, twenty glass marbles (14 mm diameter) were evenly placed on top of corncob bedding in a standard plexiglass mouse cage. Thereafter, mice were placed individually in the cage for a 15-min period. The experiment was conducted under red light conditions. The number of marbles partly or fully buried was manually scored by a researcher blind to the experimental groups. Mice that have higher levels of anxiety-like behavior tend to bury larger numbers of marbles (28).

Elevated plus maze

The elevated plus maze assesses anxiety-like behavior in rodents (29). Under red light conditions, mice were placed in an elevated plus-shaped maze (1 m above floor level), which consisted of two open and two enclosed arms. All arms measured 12 x 50 cm, and the walls of the closed arms were 40 cm high. During the 6-min period, the number of entries in the closed arms, the distance traveled, and the time spent in the open arms were automatically scored using a video tracking system (Noldus, Wageningen, The Netherlands). A greater avoidance of the open arms is interpreted as elevated levels of anxiety-like behavior.

Antidepressant treatment

Control and resilient mice were treated with daily I.P. injections of 300 μl saline for 14 days. Susceptible mice were injected daily with either saline (300 μl), 20 mg/kg imipramine, or 13 days of 300 μl saline followed by a single dose of 10 mg/kg ketamine 24 hr before the second SI test.

Blood collection and RNA purification

For mice, total submandibular blood was collected using 20-G hypodermic needles into EDTA coated tubes (Microvette CB 300) (Sarstedt, Nümbrecht, Germany), mixed and stored on ice. The tubes were centrifuged at 1000xg for 20 min at 4°C to separate the blood cellular fraction from the plasma. A total of 100 μl of the blood cell fractions were put in RNAprotect animal blood tubes (Qiagen, Germantown, MD) and stored at −80°C. Total RNA, including miRNAs, was extracted using RNeasy Protect animal blood kit (Qiagen, Germantown, MD) according to the manufacturer’s instructions. Concentration and quality were assessed using NanoDrop (Thermo Fisher Scientific, Waltham, MA).

For human subjects, intravenous blood was collected prior to ketamine infusion and 24 hr after treatment. All participants fasted for 8 hr prior to blood draws. Blood was collected using Vacutainer K₂ EDTA tubes (Becton Dickinson, Franklin Lakes, NJ). Whole blood was removed and aliquoted without any additional additives or anticoagulants. All tubes were stored in a −80°C freezer. Total RNA, including miRNAs, was extracted using the total RNA Purification Kit (Norgen Biotek, Thorold, ON, Canada), according to the manufacturer’s instructions.

NanoString

A total of 100 ng of RNA was used to quantify expression levels of 599 circulating miRNAs using a probe-based miRNA expression assay (NanoString Technologies, Seattle, WA), in addition to 4 housekeeping genes and 12 assay controls (6 positives and 6 negatives) (Suppl. Table 2). Briefly, specific tags to target miRNAs were annealed to create multiplex probe libraries from all samples. Each miRNA was identified by a ‘color code’ that was generated by six ordered fluorescent spots present on a reporter probe and read by the nCounter Prep Station Digital Analyzer. Control RNA was included to monitor ligation efficiency and specificity through each step of the reactions. For NanoString analysis, raw data were imported into NSolver Analysis Software v2.0 and normalized to the average counts of all endogenous genes. Data were further normalized using internal positive spike-in controls to account for variability in the hybridization process. Normalized data were imported into MultiExperiment Viewer (MeV) software v.4.9 for further analysis.

Quantitative real-time polymerase chain reaction (RT-qPCR)

A total of 300 ng of RNA was converted to cDNA using Taqman^™ MicroRNA reverse transcription kit (Thermo Fisher Scientific, Waltham, MA). Several miRNAs were pulled in each RT reaction (Suppl. Table 3). qPCR reactions were run in triplicate using QuantStudio 7 qPCR machine (Thermo Fisher Scientific, Waltham, MA) using TaqMan^™ Universal PCR Master Mix without AmpErase (Thermo Fisher Scientific, Waltham, MA) and TaqMan probes (Suppl. Table 3). The expression levels of selected miRNAs were analyzed in mice and humans (Suppl. Table 3) in comparative ΔΔCt method and normalized to the small nucleolar RNA RNU6B.

Antagomirs

A custom Locked Nucleic Acid (LNA) oligonucleotide with a sequence designed to target miR-144-3p (anti-miR-144-3p) was used to downregulate the expression levels of miR-144-3p in the circulation of mice (Qiagen, Germantown, MD). As a control, a scrambled LNA oligonucleotide sequence was used (scrambled control). Anti-miR-144-3p and the scrambled control were separately dissolved in sterile PBS and delivered via a daily subcutaneous injection at a final concentration of 30 mg/kg for 3 consecutive days.

RNA-sequencing

Blood RNA from 6–7 individual mice per group was used to prepare sequencing libraries by Genewiz (South Plainfield, NJ). Total RNA samples were quantified using Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA) and RNA integrity was checked using 4200 TapeStation (Agilent Technologies, Palo Alto, CA). rRNA depletion along with globin depletion was performed using QIAGEN FastSelect rRNA HMR/Globin Kit (Qiagen, Hilden, Germany). RNA-seq library preparation used NEBNext Ultra II RNA Library Prep Kit for Illumina (NEB, Ipswich, MA) by following the manufacturer’s recommendations. Briefly, enriched RNAs were fragmented for 15 min at 94°C. The first and second cDNA strands were subsequently synthesized. cDNA fragments were end-repaired and adenylated at 3’-ends, and a universal adapter was ligated to cDNA fragments, followed by index addition and library enrichment with limited cycle PCR. Sequencing libraries were validated using Agilent Tapestation 4200 (Agilent Technologies, Palo Alto, CA) and quantified using Qubit 2.0 Fluorometer (Invitrogen, Carlsbad, CA), as well as by q-PCR (Applied Biosystems, Carlsbad, CA).

Sequencing libraries were multiplexed and clustered on the Illumina HiSeq instrument according to the manufacturer’s instructions. The samples were sequenced using a 150 paired end configuration. Image analysis and base calling were conducted by the HiSeq Control Software. Raw sequence data (.bcl files) generated from Illumina HiSeq were converted into fastq files and de-multiplexed using Illumina's bcl2fastq 2.17 software. One mismatch was allowed for index sequence identification.

Bioinformatics

For RNA-seq differential expression analysis, raw reads were aligned to the mouse genome (HISAT2 mm10) (30). Counts of reads mapping to genes were obtained using feature Counts software of Subread package against Ensembl v90 annotation (31). Quality control was performed using FastQC (32) and principal component analysis was used to identify outliers. Normalization and differential expression analysis were performed using DESeq2 (33).

Predicted targets of differentially expressed miRNAs in mice were aggregated using the miRNAtap package in R (34). The package utilizes 5 databases: DIANA (35), Miranda (36), PicTar (37), TargetScan (38) and miRDB (39). Targets had to be present in a minimum of 3 sources, and the union of all targets found for a list of differentially expressed miRNAs was combined in a table. The code was modified to provide additional information such as the number of times the gene was targeted within each comparison, an average rank score and a self-calculated rank score. Predicted targets of human miR-144-3p were identified using three miRNA target prediction databases: TargetScan (38), miRDB (39) and miRWalk (40). Heatmaps were generated using Morpheus (https://software.broadinstitute.org/morpheus), and gene ontology (GO) and KEGG pathway analysis were performed using Enrichr (41). Ingenuity Pathway Analysis (IPA) (Qiagen, Germantown, MD) was performed to identify predicted upstream regulators and canonical pathways.

Statistical analysis

GraphPad Prism 8.0 software package was used for analyzing the behavioral and qPCR data. For all experiments comparing two groups, between-subjects t-tests or repeated-measures ANOVA with subsequent Holm-Šidák test were used. Two-way ANOVAs were used for experiments with a 2x2 design, with subsequent post-hoc Tukey tests. Pearson correlation coefficients were calculated using GraphPad Prism or Excel. Outliers were detected and removed according to the Grubbs outlier test. p < 0.05 was used as a cutoff for significance.

Results

The present study first aimed to identify potential miRNAs in a preclinical mouse model of chronic stress that might serve as biomarkers for MDD and the prediction and assessment of treatment response. To that end, we profiled the expression levels of circulating miRNAs from the blood of male mice that was collected at three time points: before and after exposure to CSDS, as well as after AD treatment (Fig. 1A-C). One day after the last defeat (post-stress), mice were analyzed in the social interaction (SI) test and classified into stress susceptible or stress resilient groups. Control and resilient mice were treated with saline daily for 14 days (Fig. 1C) to control for the effects of chronic injections. Susceptible mice were treated daily with saline for 14 days as a control, with 20 mg/kg imipramine for 14 days, or with saline for 13 days followed by 10 mg/kg ketamine on day 14. All animals were analyzed 24 hr later in a second SI test. Based on the results in the second SI (post-treatment), susceptible mice were further subdivided into treatment responders and non-responders. We reported the behavioral data and transcriptional brain signatures of this experiment in a former publication (26), and now turned to analyze the blood miRNA profile (Fig. 1D).

Figure 1. Schematic outline of experimental design.

(A) We collected blood before and after chronic social defeat stress (CSDS), as well as after treatment with either saline, ketamine, or imipramine. Mice were analyzed in the social interaction (SI) test 24 hr after 10 days of CSDS and re-tested after treatment. (B) Schematic representation of blood collection processing and miRNA profiling. (1) Submandibular blood collection was followed by (2) separation of blood cells by centrifugation and RNA purification from blood cellular fraction. (3) Probe-based miRNA profiling was carried out using NanoString technology. (C) Schematic representation of experimental groups. Male mice were assigned to either non-stressed control (n = 10), or CSDS, which were subdivided into resilient (n = 8) or susceptible (n = 32) based on their behavior in the first SI test. Control and resilient mice received chronic saline injections. Susceptible mice were treated with either saline (n = 6), imipramine (n = 12), or ketamine (n = 14), and further subdivided into treatment responders or non-responders, based on their behavior in the post-treatment SI test. (D) Schematic overview of the global experimental approach. The schematic described above was used to identify circulating miRNAs regulated by CSDS, ketamine, or imipramine. Putative miRNA biomarkers were then validated in human depressed subjects. Finally, we demonstrate that altering the levels of a circulating target miRNA in mice can block CSDS susceptibility.

Resilient and susceptible mice show robust differences in circulating miRNA expression profiles

First, we analyzed the effects of CSDS on circulating miRNA expression levels by comparing the profile of susceptible (SUS), resilient (RES), and control (CON) mice at three time points: pre-stress, post-stress, and 20 days after stress (Suppl. Fig. 1A). These mice did not receive AD treatment but were injected daily with saline (post-saline). The largest number of differentially expressed miRNAs (cutoff p < 0.05 and fold-change > 30%) were observed directly after stress, rather than before stress (when all mice are essentially the same) or after chronic saline treatment (Suppl. Fig. 1B). Notably, as we are interested in describing broad patterns of regulation, we did not correct for multiple comparisons for these analyses. Union heatmaps (Suppl. Fig. 1C-E) revealed similar miRNA regulation in the RES versus CON and SUS versus CON comparison before stress (Suppl. Fig. 1C), suggesting only minor pre-existing differences in circulating miRNAs between mice that will later show behaviorally opposing responses to stress. Interestingly, shortly after stress, there were similarities in the miRNA regulation between SUS and RES animals, despite the divergent behavioral phenotype (Suppl. Fig. 1D). However, at the longer-term time point, the overlap between the SUS and RES signatures was lost, demonstrating a long-lasting effect of individual differences in stress responses on the blood miRNA signature that builds (incubates) over time (Suppl. Fig. 1E). Notably, this lasting signature diverges from the pre-stress miRNA profile (Suppl. Fig. 1F-I).

Next, we bioinformatically identified predicted target genes (PTGs) of differentially expressed miRNAs (Suppl. Table 4) as a way to provide insight into the molecular pathways associated with the altered blood miRNAs levels linked to CSDS exposure. We compared the PTGs of RES to that of SUS mice before and after stress, and at the longer-term time point. Whereas we found only one overlapping miRNA across conditions at the three different time points (Suppl. Fig. 1J), we observed many overlapping PTGs (Suppl. Fig. 1K). We performed Ingenuity Pathway Analysis (IPA) to identify predicted transcriptional upstream regulators of the PTGs of the differentially expressed miRNAs at each timepoint; this analysis recognized several miRNAs as predicted upstream regulators, including miR-30, miR-30c-5p, miR-17 and miR-8 (Suppl. Fig. 1L). This is reassuring, since indeed the miR-30 family was differentially expressed between RES and SUS mice within our dataset. Furthermore, we identified ESR1 (estrogen receptor α), as the top upstream regulator of the miRNAs’ PTGs associated with stress resilience. KEGG pathway analysis identified axon guidance as the top pathway both before stress and at the longer-term time point, as well as the dopaminergic synapse pathway directly after CSDS (Suppl. Fig. 1M), thus identifying brain-associated functions despite the fact that these analyses were from blood. Together, the PTGs of differentially expressed miRNAs following CSDS are associated with known stress-related molecular pathways.

Mice that respond to ketamine treatment display normalized miRNA levels

Next, we sought to identify miRNAs that are associated with AD treatment response. We compared blood miRNA signatures from susceptible mice which were treated with chronic imipramine (IMI) or acute ketamine (KET) to control mice treated with saline (CON). Mice receiving ADs were subdivided into responders (RESP) or non-responders (NON) based on their behavior in a second SI test performed one day after the last treatment. By comparing SUS animals treated with ADs to stress-naïve control animals, our goal was to identify miRNAs that show a reversion to a normal, unstressed state.

We first compared SUS animals treated with chronic imipramine to CON animals treated with chronic saline (Suppl. Fig. 2A). Our findings reveal little effect on the blood miRNA signature after imipramine treatment, as reflected in a small number of miRNAs changed, with the largest amount of differentially expressed miRNAs observed directly after stress (Suppl. Fig. 2B). Additionally, we found little preexisting or long-lasting changes in miRNA signature both in treatment responders and non-responders (Suppl. Fig. 2C, D). We compared IMP-RESP to IMP-NON to discover miRNAs that could predict treatment response and noted a pattern of predominantly upregulated miRNAs in RESP versus NON (Suppl. Fig. 2E), mostly in the post-stress time point. The only miRNA that was differentially expressed at more than one timepoint was miR-1896, which was downregulated in RESP versus NON both before and after stress, with no difference seen following imipramine treatment. We found little overlap among the differentially expressed miRNAs between the different time points for IMI-RESP versus IMI-NON (Suppl. Fig. 2F), and after performing analysis of PTGs we noted some overlap between the time points (Suppl. Fig. 2G). Similar to our previous comparison, we identified several miRNAs as upstream regulators of the PTGs of differentially expressed miRNAs between IMI RESP versus NON (Suppl. Fig. 2H). Notably, the upstream regulator miR-125b-5p was significantly upregulated during the pre-stress condition. Following imipramine treatment, we identified STAT4 as the top upstream regulator of the PTGs (Suppl. Fig. 2H) and KEGG pathways such as autophagy and T-cell receptor signaling (Suppl. Fig. 2I). These findings suggest that repeated imipramine treatment is associated with inflammatory-related responses in the periphery.

Next, we compared SUS animals treated with a single dose of ketamine to CON animals (Fig. 2A). In contrast to imipramine-treated animals, we found that the largest number of regulated miRNAs in the KET-RESP group were already detected before stress (Fig. 2B), while in the KET-NON mice it was after stress. This suggests pre-existing miRNA expression differences both prior to stress and treatment that relate to subsequent ketamine response. Blood miRNA expression patterns were disrupted by stress exposure for both KET-NON and KET-RESP, but restored to baseline levels to a larger extent in KET-RESP (Fig. 2C, D). When directly comparing KET-RESP to KET-NON, we found several miRNAs that were differentially expressed during multiple timepoints (Fig. 2E). Mir-144-3p and miR-142-3p were both downregulated in RESP versus NON before stress and after ketamine treatment, while miR-381 was upregulated in the same comparisons. Additionally, miR-108-2 and miR-1274a were both upregulated in RESP versus NON after CSDS but downregulated after ketamine treatment. We found some overlap in differentially expressed miRNAs linked to ketamine treatment between the different time points (Fig. 2F), but little overlap in their PTGs (Fig. 2G). Interestingly, we identified several genes previously linked to the pathophysiology of depression, such as Bdnf and Creb1 (42, 43), as upstream regulators of the PTGs following ketamine treatment (Fig. 2H). Lastly, PTGs of differentially expressed miRNAs in RESP versus NON were associated with KEGG pathways linked to axon guidance and cocaine addiction (Fig. 2I). This suggests again that the PTGs of altered miRNAs in blood following CSDS and ketamine treatment are associated with known depression-related molecular pathways.

Figure 2. Characterization of circulating miRNAs in ketamine-treated susceptible mice.

(A) Control mice (CON), ketamine responders (KET-RESP) and ketamine non-responders (KET-NON) were included in this analysis. (B) Count table displaying the number of differentially expressed miRNAs (cutoff: fold change > 30% and p < 0.05) for each pairwise comparison. Darker green colors indicate increasing numbers of differentially expressed miRNAs. The largest number of miRNAs were regulated in the KET-NON vs. CON comparison in the post-stress time point. (C-E) Union heatmaps comparing differential expression patterns of KET-RESP vs. CON (C), KET-NON vs. CON (D) and KET-RESP vs. KET-NON (E) between pre-stress (top lane), post-stress (middle lane) and post-ketamine treatment (bottom lane). *p < 0.05. (F) Venn diagram showing low overlap in the lists of differentially expressed miRNAs between KET-RESP vs. KET-NON between the before and after stress, as well as after ketamine treatment. (F) Venn diagram indicating some overlap in the lists predicted target genes (PTGs) between the KET-RESP vs. KET-NON comparison at the three different time points tested. (G, H) Top predicted upstream regulators (G) and KEGG pathways (H) of the PTGs regulated by the differentially expressed miRNAs at different time points for the KET-RESP vs. KET-NON comparison.

MiR-144-3p is a predictor of MDD severity and of ketamine response in humans

Next, we validated whether blood miRNAs identified in our mouse stress study have clinical predictive value for depressed human subjects. To that end, we focused on miR-144-3p which is upregulated after CSDS in SUS mice compared to controls, with its levels decreased after ketamine in responders only (Fig. 3A). Interestingly, while we observed a difference between KET-RESP and KET-NON both before stress and after treatment, we only observed a significant increase in expression levels of miR-144-3p in IMI-RESP versus IMI-NON after stress with no change at the other time points. We bioinformatically analyzed human PTGs of miR-144-3p using three miRNA target prediction databases: Targetscan (38), miRDB (39) and miRWalk (40) (Fig. 3B). Gene ontology (GO) analysis of the common 80 PTGs of miR-144-3p revealed biological processes related to transcription and the cellular response to stress, supporting a potential role in stress and depression (Fig. 3C). In line with these findings, KEGG pathway analysis of the PTGs identified PI3K-AKT signaling, a pathway that has been implicated in the pathophysiology of several psychiatric illnesses, including depression (44, 45) (Fig. 3D).

Figure 3. Circulating miR-144-3p levels predict ketamine response in mice.

(A) Heatmap representing differential expression levels of blood miR-144-3p in different comparisons. MiR-144-3p is downregulated prior to stress and following ketamine treatment in KET-RESP only. *p < 0.05, **p < 0.01, ***p < 0.001. (B) Venn diagrams showing the overlap in the lists of predicted target genes (PTGs) of miR-144-3p based on three online databases (TargetScan(38), MiRDB (39) and MiRTarBase (40)). (C-D) Top enriched GO ontology terms (C) and KEGG pathways (D) of the PTGs of miR-144-3p.

We investigated the translational relevance of our findings on miR-144-3p as a predictor of depression severity and treatment response in a human cohort of depressed subjects before and after acute ketamine treatment (Fig. 4A). We report patient demographics in Suppl. Table 1. Using qPCR, we found that higher baseline levels of miR-144-3p were correlated with more severe depression symptoms as measured by the MADRS (r = 0.34, p = 0.0393) (Fig. 4B). Males and females are combined for this analysis, as we found no differences between the sexes (t₍₃₇₎ = 0.1409, p = 0.8887). Analysis of each sex separately showed a trend for males (r = 0.1655, p = 0.0839) but not for females (r = 0.0164, p = 0.5909) (Suppl. Fig. 3A, B). Furthermore, similar to our findings in male mice, we found that a decrease in expression levels of miR-144-3p correlated with ketamine response as measured by a reduction in MADRS score after treatment in males only (r = −0.54, p = 0.0160) (Fig. 4C), with no change seen in females (r = −0.06, p = 0.8172) (Fig. 4D).

Figure 4. Circulating miR-144-3p is associated with MDD severity and ketamine treatment response in MDD patients.

(A) Schematic representation of experimental design including blood collection, Montgomery-Åsberg Depression Rating Scale (MADRS) score measurement and ketamine delivery in MDD patients. (B) Higher baseline levels of miR-144-3p correlate with depression severity as measured by the MADRS in male and female MDD patients (n = 38). (C, D) Decrease in miR-144-3p expression levels correlated with decreased in MADRS score after ketamine treatment in male (n=19) (C) but not female (n = 19) (D) MDD patients. (E) Heatmap indicating mostly negative correlations between the change in blood expression levels of miR-144-3p PTGs and the change in MADRS scores following ketamine treatment (n = 26). Similarly, there are mainly negative correlations between the expression levels of blood miR-144-3p and the expression levels of its PTGs (n = 21). FC, fold change. *p < 0.05.

To probe the regulation of miR-144-3p PTGs in the blood of depressed subjects, we leveraged an RNA-seq dataset of blood from the same subjects (18). We found a significant overlap between the human PTGs of miR-144-3p that are expressed in blood (n = 72) with genes regulated in MDD subjects by ketamine (6 overlapping genes, Fisher exact test: 2.6x, p < 0.026). These six genes included: MAP2K8, PANK1, ELL2, MORC3, ACSL4 and UBN2. Additionally, we found a pattern of negative correlation between the regulation of PTGs of miR-144-3p in the blood of MDD subjects by ketamine and change in MADRS scores following ketamine treatment (Fig. 4E and Suppl. Table 5). This finding suggests that clinical improvement due to ketamine is associated with the upregulation of PTGs of miR-144-3p, consistent with normalizing levels of this miRNA. Next, we correlated the change in miR-144-3p expression levels with the change in expression of its PTGs following ketamine treatment. Again, we found a negative correlation between expression levels of blood miR-144-3p and the expression levels of its PTGs (Fig. 4E, Suppl. Table 5). Such negative correlation between miR-144-3p and its putative targets matches the prediction of a repressive role for miRNAs.

Systemic inhibition of miR-144-3p rescues the depression-related phenotype in susceptible mice

We probed whether systemic manipulation of miR-144-3p has any have therapeutic potential. Since lower levels of miR-144-3p are associated with ketamine response, we hypothesized that KD of miR-144-3p in the circulation might be sufficient to reduce the depression-related phenotype in chronically stressed mice. To that end, we used a LNA-modified antisense oligonucleotide (ASO) that was designed to target miR-144-3p with high specificity (anti-miR-144-3p), or LNA-ASO containing a scrambled sequence as a control (scrambled control). To confirm efficacy, stress-naïve male mice received daily subcutaneous (SC) injections for 3 days of either the anti-miR-144-3p (30 mg/kg) or the scrambled control (30 mg/kg), and blood was collected at various times after the last injection for qPCR analysis of miR-144-3p levels (Fig. 5A). A subtle knockdown of miR-144-3p was noted two days after the last injection (t₍₁₉₎ = 1.418, p = 0.0862) (Suppl. Fig. 4A), however, after an incubation of an additional 9 days, anti-miR-144-3p led to ~60% downregulation compared to the scrambled control (t₍₂₀₎ = 1.913, p = 0.0351) (Fig. 5B). Notably, the expression levels of miR-144-5p, the inactive passenger strand of miR-144-3p, were not affected, confirming the specificity of our antagomir.

Figure 5. Systemic knockdown of miR-144-3p reverses behavioral abnormalities exhibited by susceptible mice.

(A) Schematic design of the validation experiment. Locked nucleic acid (LNA)-modified antisense oligonucleotide (ASO) for miR-144-3p (anti-miR-144-3p) or a scrambled control was subcutaneously administered to male mice for three subsequent days at a dose of 30 mg/kg. Submandibular blood was collected at two time points and analyzed using RT-qPCR. (B) Bar graph showing that anti-miR-144-3p led to a significant downregulation of miR-144-3p in blood compared to scrambled control. No change was seen in the passenger strand miR-144-5p expression levels (n = 11–12). (C) Schematic representation of the experimental design of the rescue experiment. Male mice were exposed to CSDS, followed by a social interaction test (SI1), injections of either the scrambled control or anti-miR-144-3p, behavioral testing including repeated SI test (SI2) and sucrose preference, and terminal blood collection. (D) Anti-miR-144-3p treatment increased social interaction in the SI2 test compared to scrambled control. (E) Anti-miR-144-3p treatment increased sucrose preference (SP) compared to the scrambled control group. (F, G) Significant correlations between miR-144-3p levels in the circulation following its inhibition with SI2 (F) and SP (G) scores. FC, fold change. *p < 0.05.

We then investigated whether systemic miR-144-3p KD is sufficient to rescue the deleterious effects of CSDS. Male mice were subjected to CSDS followed by an SI test to classify mice into SUS and RES groups in addition to the CON non-stressed group. Half of the mice received three daily SC injections of anti-miR-144-3p, while the other half received the scrambled control, after which all mice were behaviorally re-tested nine days following the last dose (Fig. 5C). As predicted, SUS mice receiving the anti-miR-144-3p displayed significantly less avoidance during the second SI test (Fig. 5D) (F_(1,26) = 4.704, p = 0.0394; post hoc Holm-Šidák test: Scrambled control_SI2 versus anti-miR-144-3 p_SI2, p = 0.0119) as well as exhibited higher sucrose preference (t₍₂₆₎ = 1.873, p = 0.0362) (Fig. 5E) compared to mice that received the scrambled control. We also found a significant correlation between the second SI scores and sucrose preference values (r = 0.56, p = 0.0026) (Suppl. Fig. 4B), which together indicates that downregulation of miR-144-3p in the circulation rescues the depression-related phenotype in SUS mice. We observed no difference between the groups in the marble burying or elevated plus maze tests (Suppl. Fig. 4C, D). We next measured the levels of miR-144-3p in the circulation of susceptible mice and found a negative correlation with the second SI scores (r = −0.40, p = 0.0457) (Fig. 5F), and a trending negative correlation with sucrose preference scores (r = −0.35, p = 0.0767) (Fig. 5G), suggesting that the degree of improvement in these tests is linked to the extent of miR-144-3p inhibition. By contrast, the same manipulation induced no appreciable effect in CON or RES mice in either of the tests (Suppl. Fig. 4E-P). Taken together, these results indicate a potential AD role for anti-miR-144-3p, specifically on depression-like behaviors in SUS mice only.

Anti-miR-144-3p dampens stress-induced transcriptional signatures in blood

Finally, we used RNA-seq to explore the molecular mechanisms by which anti-miR-144-3p alters depression susceptibility. We surveyed genome-wide transcriptional changes in the blood of SUS male mice after CSDS and injections of either anti-miR-144-3p (SUS-anti-miR-144-3p) or scrambled control (SUS-scrambled) (Fig. 6A). Analysis of differentially expressed genes compared to CON (unstressed) animals which received the scrambled control (baseline) (DEGs; cutoff: p < 0.05 and fold-change > 30%) showed that miR-144-3p KD in the circulation of SUS mice reduced the number of DEGs by half (Fig. 6B) compared to SUS-scrambled group. The union heatmap presented in Fig. 6C confirms that the transcriptional effects of CSDS in the circulation are dramatically blunted by miR-144-3p KD. Gene ontology (GO) analysis revealed that the unique DEGs in the scrambled control comparison were enriched for biological processes such as mitochondrion disassembly and autophagy of mitochondrion. The terms for the unique DEGs in the anti-miR-144-3p comparison were related to known functions of miR-144-3p, including hematopoietic stem cell differentiation and hematopoietic progenitor cell differentiation (Fig. 6D) (46). Upstream regulator analysis identified NFE2L2, a transcription factor and regulator of detoxification and antioxidant responses (47), as the top upstream regulator for the scrambled control versus baseline DEG list, and the transcription factor TCF7, which is implicated in hematopoietic development (48), as the top predicted upstream regulator for the anti-miR-144-3p versus baseline DEG list (Fig. 6E). In line with these findings, the canonical pathways associated with the scrambled control comparison are related to oxidative stress response, whereas the canonical pathways for the anti-miR-144-3p comparison are associated with immune response (Fig. 6F). Finally, we compared this blood RNA-seq dataset to our previously published brain data of the same ketamine-treated mice that we used for blood miRNA analysis reported above (26). The blood signature of anti-miR-144-3p mice showed the strongest overlap with the transcriptional changes in the amygdala of both KET-NON and KET-RESP animals (Fig. 6G). In contrast, DEGs from the SUS-scrambled mice showed the most overlap with the transcriptional signature of the hippocampus from KET-NON mice. Together, these results suggest that miR-144-3p KD in peripheral blood alters behavioral responses to stress, possibly through modulation of immune responses.

Figure 6. Transcriptional alterations in blood induced by CSDS are dramatically blunted after miR-144-3p KD.

(A) RNA-seq of blood cells was performed on the following experimental groups: control mice which were treated with the scrambled control (baseline), susceptible mice which were treated with the scrambled control (SUS-scrambled) and susceptible mice which were treated with anti-miR-144-3p (SUS-anti-miR-144-3p) (n = 6–7/group). (B) Venn diagrams showing the overlap in the lists of differentially expressed genes (DEGs) between susceptible animals who received the scrambled control (SUS-scrambled) or the anti-miR-1443p (SUS-anti-miR-144-3p) (cutoff: fold change > 30% and p < 0.05). (C) Union heatmap comparing DEG patterns between the SUS-scrambled vs. baseline (top lane) or SUS-anti-miR-144-3p vs. baseline (bottom lane). The transcriptional effects of CSDS in the circulation are dramatically blunted by miR-144-3p KD. (D-F) Top enriched GO ontology terms (D), upstream regulators (E) and canonical pathways (F) of the genes uniquely regulated in either the scrambled control or anti-miR-144-3p injections compared to baseline. (G) Overlap between the lists of genes regulated in the blood by anti-miR-144-3p to those regulated in the brain by ketamine. The table indicates p-value (text) and Fisher’s exact value (warmer colors indicating increasing Fisher’s exact value) representing the degree of the overlap between the blood and brain lists (26). PFC, prefrontal cortex; NAc, nucleus accumbens; AMY, amygdala; HIP, hippocampus. *p < 0.05.

Discussion

The present study explored the potential use of miRNAs as biomarkers for MDD diagnosis and for prediction of treatment response as well as the ability to manipulate a circulating miRNA to rescue depression-like phenotypes. We show that CSDS induces robust differences in blood miRNA signatures between susceptible and resilient mice in the long term, but not directly after stress. Furthermore, treatment with ketamine, to a greater degree than imipramine, re-established baseline miRNA expression levels in mice that showed AD-like responses to the drug, but not in non-responders. We bioinformatically identified miR-144-3p in susceptible mice as a candidate blood biomarker for stress susceptibility and predictor of treatment response. Consistent with these findings in our preclinical mouse model, we show that blood-based miR-144-3p was also shown to predict depression severity and treatment response in a human cohort of depressed subjects before and after acute ketamine treatment. Lastly, we demonstrate that systemic KD of miR-144-3p using an antagomir approach is sufficient to reduce depression-like behaviors in susceptible mice following CSDS. RNA-seq analysis of blood samples after CSDS and miR-144-3p manipulation revealed that the KD attenuated the molecular signature of CSDS on the blood transcriptome.

Several studies point to circulating miRNAs as potential indicators of stress exposure and depression (10, 14, 17). Here, we report the most comprehensive analysis to date by wide-scope profiling of miRNAs at different time points within the same stressed animals using NanoString analysis, which has been shown to be a suitable method for analyzing circulating miRNAs (49). This probe-based method requires less starting material compared to small RNA-seq and has high precision (50). We collected blood from mice both before and after exposure to CSDS, as well as after treatment with imipramine or ketamine – two pharmacologically distinct AD drugs. Imipramine, a tricyclic AD, inhibits the reuptake of serotonin and norepinephrine, and influences several monoaminergic and cholinergic receptors (51). Ketamine is a non-competitive NMDA glutamate receptor antagonist among other actions, which was recently FDA-approved as a rapid-onset AD for treatment-resistant patients (6, 7). Interestingly, we found that the largest effect on circulating miRNAs levels occurs following ketamine treatment, while our former report on the brain transcriptomics of these same mice noted a stronger effect on protein-coding gene expression following imipramine treatment (26). Together, these findings suggest divergent molecular processes both in the brain and in the circulation associated with effective AD treatment in a mouse model.

Our analysis of upstream regulators and pathways associated with miRNAs regulated in blood supports the hypothesis of an interplay between the circulation and the brain in depression. We identified ESR1 as the top upstream regulator of the PTGs of miRNAs identified in the RES versus SUS comparison after CSDS. This is in line with previous findings from our group, which demonstrated ESR1 as a key regulator of transcriptional changes in the nucleus accumbens that drive behavioral resilience to CSDS in male mice (52). Furthermore, we identified several well-known depression and AD response-related molecules such as BDNF (42, 53) and CREB1 (43, 54) as upstream regulators following ketamine treatment in susceptible mice. The blood miRNA and brain gene expression interactions are in line with former studies reporting, for example, a correlation between levels of circulating miRNAs, brain miRNAs and their target genes (14, 17). Interestingly, the most re-occurring KEGG pathway within our different comparisons was axon guidance – despite the analysis of blood. Axon guidance cues are being increasingly recognized as potential targets for neurological and psychiatric disorders, including depression (55-57). Notably, axon guidance molecules play important regulatory roles in inflammation processes (58), which we find to be altered upon our miR-144-3p systemic manipulation. Collectively, our analysis of putative upstream regulators controlling PTGs of differentially expressed blood miRNAs indicate intricate crosstalk between the circulatory system and brain function following stress.

We highlighted miR-144-3p, which is highly enriched in red blood cells (59, 60), as a potential biomarker for stress susceptibility and as a predictor of ketamine treatment response in mice. Specifically, increased blood levels of miR-144-3p are associated with susceptibility to CSDS in male mice. Moreover, these levels are reduced in mice that responded behaviorally to ketamine, but not in non-responders. In line with the possibility that mir-144-3p downregulation is important for ketamine responsiveness, blood expression levels of miR-144-3p were positively correlated with depression symptom severity in MDD patients at baseline, and decreased in association with lower MADRS scores after ketamine treatment. This indicates that increased levels of circulating miR-144-3p are associated with higher depression symptom severity, while downregulation following ketamine treatment is associated with clinical improvement. Consistent with our results, college students subjected to chronic academic stress showed elevated blood levels of both miR-144-3p and miR-144-5p (61, 62). Furthermore, the sperm of corticosterone-treated mice revealed elevated levels of miR-144, and their subsequent offspring displayed anxiety-like behavior (63). However, one study in rats reported significantly lower levels of miR-144-3p in hippocampus following chronic unpredictable mild stress (64), and another study reported decreased plasma levels of miR-144-5p, the inactive passenger strand of miR-144-3p, in human depressed patients compared to healthy controls (65). In mice, miR-144-3p overexpression in the basolateral amygdala rescued impaired fear extinction and protected against fear renewal (66). However, studies of miR-144-3p in the brain should be viewed with caution since the miRNA is not appreciably expressed in any brain cell type beyond erythrocytes present in the brain’s vasculature (67). In any event, these reports provide additional support for the role of circulating miR-144 in stress and depression.

We hypothesized that manipulation of miR-144-3p in the circulation might alter behavioral responses after CSDS. Indeed, we found that LNA-ASO-mediated miR-144-3p KD in the circulation of susceptible male mice blocked behavioral susceptibility to stress. While manipulation of specific miRNAs within discrete brain regions has been shown to alter stress responses (14, 17), no study to date has reported altered stress responses upon manipulation of a circulating miRNA. Moreover, as miR-144-3p KD did not alter the behavior of either resilient or control animals, our data indicate that this effect is specific to mice with a depressive-like phenotype. A distinct molecular signature within susceptible mice, that is not observed in resilient or control animals, could account for this divergent behavioral response. For example, susceptible, but not resilient, mice showed increased levels of circulating IL-6 acutely after stress exposure (68). Similarly, a subset of MDD patients demonstrated higher peripheral expression levels of inflammatory markers (69, 70). A previous study showed that chronic peripheral administration of an miR-144-3p mimic in mice was associated with promoting pro-inflammatory cytokine production (IL-1ß, IL-6, TNF-α) (71). Thus, if miR-144-3p is indeed associated with inflammatory responses, this could help explain why our findings were limited to susceptible mice only.

Our RNA-seq experiment on blood samples of susceptible mice treated with anti-miR-144-3p or scrambled control examined putative molecular changes underlying its AD effects. This analysis revealed that miR-144-3p KD suppresses CSDS-induced gene expression changes in susceptible mice, potentially indicative of a reversal of a peripheral molecular signature of depression. We found that KD of miR-144-3p, which is highly expressed in erythrocytes, leads to DEGs associated with hemopoietic differentiation and negative regulation of T-cell cytokine production. Cytokines are important mediators of innate and adaptive immunity, and are able to activate the HPA axis (72). The release of specifically pro-inflammatory cytokines (e.g., TNF-α, IL-6, IL-1β and interferon-α) has been associated with social withdrawal and decreased exploratory behavior in rodents (72). Similarly, clinical studies report higher levels of circulating pro-inflammatory cytokines in depressed patients (73, 74). Moreover, TCF7, which we identified as an upstream regulator of the unique transcriptional changes following miR-144-3p KD, is predominantly known for its regulatory role in T cells (75). We found downregulation of Tcf7 following anti-miR-144-3p administration in susceptible animals. Notably, other members of the TCF family, such as TCF4, were genetically linked to neuropsychiatric disorders including depression (76). We speculate that Tcf7 downregulation might suppress the peripheral inflammatory response induced by chronic stress and thereby contribute to resolution of depression-like symptoms. Additionally, TCF7 plays an important role in hematopoietic development (48). Studies report that higher depression scores are associated with a lower red blood cell count (77), and increased red cell distribution width is associated with negative clinical outcomes in patients with depressive disorders (77, 78). Subjects with anemia are also at increased risk of developing psychiatric disorders such as depression and anxiety disorders (79, 80). Further studies are needed to evaluate the contribution of TCF7 downregulation, suppressed inflammatory responses, altered hematopoietic differentiation and many other potential mechanisms by which miR-144-3p knockdown in the circulation of susceptible mice reverses chronic stress-induced behavioral abnormalities.

While we show that miR-144-3p might serve as a potential diagnostic and therapeutic tool for MDD, it is unknown how this circulating miRNA affects the brain. LNA-ASOs do not cross the blood-brain barrier; therefore, the observed effects of miR-144-3p and of its antagomir on behavior are presumably mediated by changes in erythrocyte function. This could occur via changes in oxygenation or miR-144-3p-dependent changes in immune responses, as noted above. Another possibility is that levels of miR-144-3p in erythrocytes control the release of other substances into the blood which then penetrate the brain. Supporting this explanation are our data showing an overlap in the transcriptional signature of systemic anti-miR-144-3p KD to that of ketamine treatment. Alternatively, there are reports of high expression of miR-144-3p in circulating extracellular vesicles (81), and we can detect appreciable levels of this miRNA in plasma (data not shown). Others have demonstrated that CSDS alters the blood-brain barrier integrity and promotes cytokine passage into the brains of susceptible, but not resilient, mice (82). These findings suggest that circulating miR-144-3p might conceivably enter the brain—via such vesicles or another mechanism—and then alter brain gene expression and function, however, such an action remains entirely speculative. Further experiments are needed to study these and additional potential mechanisms by which levels of miR-144-3p in erythrocytes control the deleterious effects of stress and their reversal with AD treatment.

To conclude, we identified a blood miRNA the levels of which are associated with depression-related measures and their treatment by ketamine in both mouse models and human MDD patients. The identification of miR-144-3p as a biomarker may ultimately allow for syndrome or treatment stratification of MDD patients in clinical settings, which could in turn lead to improved patient care. Future studies are needed to validate our findings in larger human cohorts, as well as to assess the long-term effects of ketamine on miR-144-3p expression levels. We also report that direct manipulation of this blood miRNA alters depression-like behavior in mice, suggesting a tangible path for future therapeutics. More generally, this work expands our knowledge of the epigenetic processes accruing in the circulation in response to stress, depression, and AD treatment.