Exploring the role of neuronal-enriched extracellular vesicle miR-93 and interoception in major depressive disorder

Major depressive disorder (MDD) is associated with interoceptive processing dysfunctions, but the molecular mechanisms underlying this dysfunction are poorly understood. This study combined brain Neuronal-Enriched Extracellular Vesicle (NEEV) technology and serum markers of inflammation and metabolism with Functional Magnetic Resonance Imaging (fMRI) to identify the contribution of gene regulatory pathways, in particular micro-RNA (miR) 93, to interoceptive dysfunction in MDD. Individuals with MDD (n = 44) and healthy comparisons (HC; n = 35) provided blood samples and completed an interoceptive attention task during fMRI. EVs were separated from plasma using a precipitation method. NEEVs were enriched by magnetic streptavidin bead immunocapture utilizing a neural adhesion marker (CD171) biotinylated antibody. NEEV specificities were confirmed by ow cytometry, western blot, particle size analyzer, and transmission electron microscopy. NEEV small RNAs were purified and sequenced. Results showed that: (1) MDD exhibited lower NEEV miR-93 expression than HC; (2) within MDD but not HC, those individuals with the lowest NEEV miR-93 expression had the highest serum concentrations of interleukin (IL)-1 receptor antagonist, IL-6, tumor necrosis factor, and leptin; and (3) within HC but not MDD, those participants with the highest miR-93 expression showed the strongest bilateral dorsal mid-insula activation. Since miR-93 is regulated by stress and affects epigenetic modulation by chromatin reorganization, these results suggest that healthy individuals but not MDD participants show an adaptive epigenetic regulation of insular function during interoceptive processing. Future investigations will need to delineate how specific internal and external environmental conditions contribute to miR-93 expression in MDD and what molecular mechanisms alter brain responsivity to body-relevant signals.


Introduction
Interoception refers to the nervous system's ability to sense, interpret, and integrate internal bodily signals across conscious and unconscious levels 1 . Individuals with major depressive disorder (MDD) often report dysregulated interoceptive awareness, or abnormal experiences of their internal body states [2][3][4][5] . It has been argued that, within the context of depression, the brain is rigidly insensitive to interoceptive signals and as a result fails to e ciently predict and regulate the body's metabolic needs 6 . The insular cortex likely integrates interoceptive signals with emotionally salient information during conscious as well as unconscious comparisons of anticipated versus experienced bodily states, a process which relies on the management of interoceptive prediction and prediction-error signals, embodied emotional signals, and self-related signals [7][8][9] . Moreover, a recent functional magnetic resonance imaging (fMRI) meta-analysis identi ed blunted activity of the left dorsal mid-insula in patients with MDD relative to healthy individuals across interoceptive awareness paradigms, suggesting that disrupted mid-insular activation may represent a neural marker and a putative target for novel interventions in depression 10 .
Although the cellular and molecular processes resulting in altered interoception in MDD are only beginning to be understood 11 , several hypotheses have been advanced. Among them is the notion that elevated levels of pro-in ammatory cytokines in some depressed individuals result in a decoupling of afferent interoceptive input from interoceptive predictions, leading to increased prediction-error signals [12][13][14] . Dysregulation of both the innate and adaptive immune systems is consistently described in depressed patients 15 and increased peripheral in ammatory markers have been linked to disturbed brain function in regions including the insula 16 . In addition to immuno-in ammatory activation, neuroendocrine regulators of energy metabolism such as leptin and insulin are involved in homeostatic adjustments and brain circuits integrating homeostatic and mood regulatory responses 17 . Leptin, a peptide hormone with proin ammatory properties that functions to maintain energy homeostasis 18 , can interact with neural circuitry to increase the likelihood of developing MDD 19 . These ndings suggest that mechanistic alterations of interoception in MDD may be related to elevated pro-in ammatory cytokine levels, immune system dysregulation, and disturbances in insula activity, as well as involvement of neuroendocrine regulators of energy metabolism and mood regulation.
Measuring neuronal-enriched extracellular vesicles (NEEVs) enables one to examine non-invasively speci c molecular processes in the brain. These molecular ndings can be integrated with systems-level signals such as fMRI to gain a better understanding of the biological dysfunction related to speci c cognitive or affective processes. Extracellular vesicles (EVs), including exosomes (biogenesis occurring through multivesicular bodies inside the cell; size range ~ 40nm to 160nm) and ectosomes (vesicles that pinch off the surface of the plasma membrane by outward budding; size range ~ 50nm to 1µm in diameter) are released by many cell types and can be associated with immune responses, among others 20 . EVs contain proteins, metabolites, and nucleic acids that can be delivered into recipient cells for intercellular communication, thereby effectively altering their biological responses related to regulation of central and peripheral immunity, and/or metabolic reprogramming 20 . Moreover, NEEVs cross the bloodbrain barrier from both directions and are involved in central nervous system (CNS) regulation by micro ribonucleic acid (RNA) (MiR) transmission. MiRs are a class of small non-coding RNAs functioning as key post-transcriptional regulators of gene expression through the destabilization of messenger RNA (mRNA); MiRs are enriched in the CNS and more distinct MiRs are expressed in the brain than in any other tissue 21 .
Neuronal MiRs account for 70% of all MiRs in our body that are involved in regulating neurogenesis and neuroplasticity 22 . It is important to note that MiRs carried from the brain may enter into blood circulation during major depressive episodes 23 . Identifying MiRs related to neuronal development, in ammation, and metabolic pathways can elucidate body-brain connections and how they differ as a function of MDD presence versus absence.
One such MiR, microRNA-93-5p (miR-93), is a member of the miR-106b-25 cluster, located on chromosome 5 24 , and has been implicated in the regulation of neural progenitor cell proliferation and neuronal differentiation 25 . miR-93 can regulate neurogenesis and neuronal growth 26 and target mRNAs involved in metabolic signaling 27 . miR-93 reduces in ammatory cytokine expression, including interleukin (IL)-1 beta (IL-1β), tumor necrosis factor (TNF), and IL-6 though signal transduction/activation of the transcription 3 (STAT3) signaling pathway 28,29 . Another pathway regulated by miR-93, the toll-like receptor 4 (TLR4) in ammatory pathway, has been postulated as one of the key players implicated in the increased in ammatory response in depressed individuals 30 . For instance, research indicates that miR-93 regulates the TLR4/NF-kB pathway 31,32 and EV-derived miR-93 protects lipopolysaccharide (LPS)induced cell injury by inhibiting TNF activation 33 . There is also some evidence showing the link between   dysregulated miR-93 and other processes perturbed in depression such as insulin resistance 34, 35 ,   adipogenesis 36 , which could make it a therapeutic target for obesity and the metabolic syndrome 37, 38 , and chromatin remodeling 39,40 .
Given the roles of miR-93 in regulating neuronal axogenesis, in ammation, and metabolism, the current study aimed to link these molecular processes to previously described mid-insula dysfunction during interoceptive processing in depression, by integrating brain fMRI and NEEV technology. A sub-sample of MDD and healthy comparison (HC) participants from the Tulsa 1000 (T1000) project 41 completed an interoceptive awareness task during fMRI and provided blood for isolation and enrichment of NEEVs and immunoassays. We evaluated the role of miR-93 to investigate whether MDD patients differed from HC in: (1) NEEV miR-93 expression; (2) the relationship between NEEV miR-93 expression and in ammatory and metabolic markers; and (3) the relationship between NEEV miR-93 expression and brain activity during interoceptive versus exteroceptive attention.

Participants
Participants in this study were drawn from the rst 500 individuals who completed baseline assessments as part of the T1000 project, a naturalistic longitudinal study of 1000 individuals including healthy and treatment-seeking individuals with mood, anxiety, substance use and eating disorders 41 . The T1000 study was conducted at the Laureate Institute for Brain Research (LIBR) in Tulsa, Oklahoma, United States. Baseline assessments occurred between 1/01/2015 and 12/21/2017. The T1000 project was approved by the Western Institutional Review Board and performed in accordance with the Declaration of Helsinki To determine whether individuals with depression show evidence for altered cellular processing related to interoception, only MDD and HC subjects were included in the present analysis. Our previous study using participants from this sample suggested that there was no statistically signi cant evidence for blood oxygen level-dependent (BOLD) signal differences between un-medicated MDD and MDD with use of selective serotonin reuptake inhibitors (SSRI) on the VIA task 44 ; therefore, both unmedicated and SSRImedicated subjects were included in this analysis. MDD subjects taking selective norepinephrine reuptake inhibitors (SNRI), or taking various other antidepressants were excluded from the current analysis. Participants were also excluded if they had in ammation or metabolic related disease (e.g., autoimmune disease, in ammatory bowel disease, or diabetes), or they were taking anti-in ammatory or anti-diabetic drugs. In addition, subjects were excluded if they had poor quality or missing VIA fMRI data. Finally, 41 MDD and 35 HC participants remained for data analysis (see Table 1). Adapters from 3' end were trimmed from the raw reads after a quality check, bases trimmed from both ends, and then aligned to the human genome hg38 using Bowtie alignment. Next, the aligned reads were quanti ed against the human miRbase mature microRNAs version 22 and reads from miR genes were normalized and scaled to reads per million for statistical data analysis.

Neuroimaging
Each participant completed a structural MRI scan followed by fMRI scanning while performing an interoceptive awareness task.

Interoceptive awareness task
The Visceral Interoceptive Attention (VIA) task was comprised of two eight-minute runs, each containing interoceptive and exteroceptive conditions. During the interoceptive conditions, the words "heart" and "stomach" cued participants to attend to sensations from that part of the body. During the exteroception (i.e., control) condition, participants attended to the word "target" as it alternated between black and varying shades of grey. Trials lasted 10 seconds, and half of trials were followed by a 5-second period for participants to rate stimulus intensity (0 = 'no sensation' to 6 = 'extreme sensation'). Each run included 6 trials per condition (intertrial interval range 2.5-12.5 s). The VIA task has been previously shown to be effective at mapping the neural signal associated with interoceptive attention, including in depressed individuals 5, 50-54 .

fMRI data preprocessing
Single-subject image pre-processing was performed using Analysis of Functional NeuroImages (AFNI) software (http://afni.nimh.nih.gov/afni) 55 . The anatomical scan was registered to the rst volume of the EPI time-course and then aligned to Montreal Neurological Institute (MNI) space via a ne transformation, saving the transformation parameters for application to the EPI data. The rst three TRs were discarded from each EPI time-course to allow the fMRI signal to reach steady state, followed by despiking; slicetiming correction and co-registration to anatomical volumes. Motion correction and spatial transformation to MNI space of the EPI data were implemented in a single transformation. The EPI data were then smoothed with a 4mm Gaussian full-width at half-max smoothing kernel, and signal intensity normalized to re ect percent signal change from each voxel's mean intensity across the time-course. All images were resampled to 2 x 2 x 2 mm 3 isometric voxels.

Subject-level fMRI imaging analysis
Each subject's functional imaging data were analyzed using a voxelwise general linear model analysis. Block regressors were convolved with a canonical hemodynamic response function and used to model BOLD responses for heart, stomach, and target conditions. Six motion parameters (three translations and three rotations) were included as nuisance regressors. Censoring was done at the regression step by removing volumes with either a Euclidean norm of the derivatives of the six motion parameters greater than 0.3 mm or greater than 10% outlier voxels, determined by 3dToutcount. Percent signal change during each condition was de ned as the estimated beta coe cient from single-subject analysis.

Demographic characteristics and clinical ratings
Independent sample t-tests examined differences between MDD and HC on age, sex, IPAQ exercise METminutes per week, BMI, and PROMIS depression. Chi^2 test was used to access sex differences between groups.

NEEV miRNA analysis
Statistical analyses on NEEV miR-93 were conducted in R. Scaled miR-93 data (counts per million) were log-transformed due to their non-Gaussian distributions determined by Shapiro-Wilks tests. Outliers were de ned as z = ±3 across subjects and set as missing. Independent sample t-tests were used to assess differences between MDD and HC, as well as between un-medicated and SSRI-medicated MDD subjects.
In addition, miR-9, a neuronal cell-speci c marker 56 was compared between NEEV and EV.

Relationship between NEEV miR-93 expression and in ammatory/metabolic markers
All in ammatory/metabolic markers (IL-1ra, TNF, IL-6, CRP, and leptin) were log-transformed due to their non-Gaussian distributions determined by Shapiro-Wilks tests. Outliers were de ned as z = ±3 across subjects and set as missing. Independent t-test was used to test the group difference on IL-1ra, IL-6, and TNF; relationships between NEEV miR-93 expression and IL-1ra, IL-6, and TNF within each group were tested by Pearson's correlations.
Even after log-transformation, the distributions for CRP and leptin were found to be non-Gaussian; therefore, Spearman's correlations were used to test their relationships to NEEV miR-93 expression within each group, and group differences on these two markers were tested using Mann-Whitney-Wilcoxon nonparametric tests. ANOVA tests were used to evaluate slope differences between MDD and HC groups.

Group-level fMRI imaging analysis
AFNI's 3dttest++ was used to assess the whole brain voxel-wise group by NEEV miR-93 interaction on BOLD activation of the interoception versus exteroception contrast. The group statistical map was corrected for multiple comparisons according to our previous neuroimaging approaches with this task (see Supplemental Methods). BOLD activation of the interoception versus exteroception contrast within clusters with signi cant group*miR-93 effects were extracted for follow-up analyses. Robust regression tested the slope of different relationships between NEEV miR-93 and BOLD for each signi cant cluster. False Discovery Rate correction for multiple comparisons was used across the resulting tests.

Demographics and clinical characteristics
The groups did not differ on age, sex, education, or employment status ( Table 1). The MDD group reported lower physical activity as well as higher BMI and PROMIS depression scores than the HC group. Figure 1 shows the results for EV and NEEV characterization following MISEV2018 guidelines 57 . Flow cytometry results (Fig. 1A) showed that NEEVs were positive for EV marker CD63-FITC and NEEV marker CD171-APC compared to EV-negative control, beads, or unstained samples. Western blot analysis ( Fig. 1B) showed that: (1) CD171 marker was enriched in NEEVs; (2) EV surface marker CD81 and EV internal marker Alix were present in EVs and NEEVs; and (3) EV negative marker Calnexin was not observed in EV or NEEV samples (note: the image displayed in Fig. 1B is from the same gel without cutting prior to antibody hybridizations; see Supplemental Figures S1-S4 for images of full-length gels). Transmission electron microscopy images of EVs and NEEVs (Fig. 1C) showed that the majority of EVs and captured NEEVs were in the small EV size range (Fig. 1D). The average concentration of NEEVs was approximately 1.98 x 10 10 particles per mL, which was approximately 30-fold higher than EV-depleted plasma (6.7 x 10 8 particles per mL). The approximate concentration of EV used for NEEV enrichment was 9.45 x 10 10 particles per mL.

NEEV miRNA results
The neuronal cell-speci c marker miR-9 was expressed at a 15-fold higher rate in NEEV than EV (Supplemental Figure S5). One outlier from the MDD group was excluded for miR-93 analysis. MDD exhibited signi cantly lower levels of NEEV miR-93 expression than HC (p = 0.037, d = 0.482) (Fig. 2), a difference that remained after controlling for BMI (p = 0.035). In addition, miR-93 expression did not differ between unmedicated and SSRI-medicated MDD individuals (p = 0.398).

Neuroimaging results
There were no signi cant group activation differences observed outside of the insular cortex, and thus, our analysis focused on clusters of observed insular activation. Speci cally, group differences were observed in the slope of the relationship between NEEV mR-93 and the interoception versus exteroception

Discussion
This study aimed to elucidate the molecular processes underlying previously described mid-insula dysfunction during interoceptive processing in depression using brain NEEV measurement, serum markers of in ammation and metabolism, and whole brain fMRI recording. There were three main ndings. Firstly, miR-93 expression in NEEVs was signi cantly diminished in individuals with Major Depressive Disorder (MDD) compared to Healthy Controls (HC). Secondly, a unique association emerged in MDD participants, where reduced miR-93 expression in NEEVs correlated with elevated serum concentrations of IL-1ra, IL-6, TNF, and leptin-establishing a connection between miR-93 expression in MDD and heightened in ammation. Lastly, in HC participants, but not in those with MDD, miR-93 expression in NEEVs exhibited a positive correlation with BOLD signals in the left and right dorsal midinsula during interoception, linking miR-93 regulation to adaptive interoceptive processing in healthy individuals. Taken together, while healthy individuals demonstrate increased responsiveness to stressinduced epigenetic regulation of insular function during interoceptive processing, MDD participants exhibit a failure to do so. This highlights the potential role of insu cient miR-93 signaling and its altered relationship with systems-level interoceptive processing in contributing to interoceptive processing abnormalities in MDD. The pathways unveiled in this study could offer novel therapeutic targets for rectifying interoceptive dysfunction among individuals suffering from depression.
MiR-93 expression was lower in individuals with MDD than those without MDD. To better understand the role of NEEV miR-93 in different neuronal processes, we performed a biological pathway analysis with miRWalk 58 by target mining the full mature miRNA, hsa-miR-93-5p, with miRBaseID. Several genes and biological pathways were identi ed after ltering with TargetScan, miRDB, and miRTarBase. Several differentially expressed genes were used during Gene Set Enrichment Analysis (GSEA), which identi ed 210 enriched genes, and 23 (out of 53) biological pathways that were signi cant, to include pathways centered on calcium ion transport, memory, and protein ubiquitination (See Supplemental Table S1). The pathways mentioned above are known to play a role in depression; for instance, the lack of ubiquitination of certain proteins [59][60][61] , memory disruption 62, 63 , and calcium ion signaling (linked to neuronal excitability and neurotransmitter release) 64, 65 , have all been linked to depression. These targets and more may be of interest or offer plausible explanations to the decreased interoceptive signaling found in depressed individuals, given that miR-93 in NEEV is attenuated.
We observed lower NEEV miR-93 associated with higher serum concentrations of in ammatory and metabolic markers, IL-1ra, IL-6, TNF, and leptin within MDD. Expanse literature shows that a subset of We did not observe an association between miR-93 expression and interoceptive signaling in the brains of individuals with MDD, but we did nd a relationship between NEEV miR-93 and higher interoceptionassociated insula activity in healthy individuals. This could suggest a homeostatic role of miR-93 from NEEV during intact interoception, although the precise nature of this relationship is unclear. Consistent with empirical and theoretical ndings implicating the role of insular activity in subjective interoceptive and emotional states [73][74][75] , our observation of associations between miR-93 and VIA BOLD signal in left and right dorsal mid-insula might be interpreted to suggest a mechanism whereby the tra cking or regulation of NEEV miR-93 activity is intact and involved in interoceptive processing in healthy individuals, but dysfunctional in depressed individuals. However, this is a speculative notion and warrants further study, particularly with respect to longitudinal assays of the relationship between affective states, NEEV miR-93 activity, and neural indicators of interoception in depression.
Interoception is a process allowing individuals to continuously sense and integrate numerous visceral, physiological signals including autonomic and nociceptive input, emotional stimuli, hunger signals, and sleep, which are then perceived by the brain during continuous feedback 76 . miR-93-5p regulates many diverse gene products that in uence a range of potentially associated processes ranging from in ammation to epigenetic modulation (See Supplemental Figure S6). It seems plausible that its expression is both a consequence of environmental exposure with long-term consequences, e.g. early life stress 77 , or subtle in ammatory processes 78 that have been implicated in the pathophysiology of depression. Chronic stress, which is a risk factor for MDD, has been proposed to lead to increased in ammation 77 , which in turn disrupt neural circuits involved in interoceptive processing 16 . Additionally, cytokines can affect the function of neurons and glia in the brain, leading to altered neural activity and connectivity 79 . Therefore, the association between lower miR-93 expression and higher serum concentrations of IL-1ra, IL-6, TNF, and leptin in individuals with MDD, might be interpreted to suggest that a low miR-93 expression level fails to regulate in ammatory cytokines in MDD. Again, further studies are needed to validate and extend this notion.
Strengths. We focused on evaluating how a neuronal process could be associated with interoceptive signaling using the innovative technique of NEEV isolation. For this, NEEV were isolated with the use of the transmembrane L1 cell adhesion molecule (L1CAM/CD171). A recent review article reported concern about using L1CAM/CD171 enriched EVs 80 , due the expression of L1CAM in other tissues of the body. We took careful steps to characterize and verify NEEVs following the MISEV2018 guidelines 57 , including: (1) ow cytometry of CD63 and CD171; (2) western blot analysis of EV surface markers, CD171 and CD81, EV internal marker -Alix, and the EV negative marker -calnexin; (3) EV and NEEV particle size and concentration measurements using the MRPS technology; (4) transmission electron microscopy imaging of EV and NEEV; (5) EV-depleted samples used as negative controls, and (6) the use of thrombin treatment for removal of brinogen from plasma, which potentially affects plasma EV separation and characterization. Additionally, data con rmed that the neuronal cell-speci c marker miR-9 was expressed at a much higher rate in NEEV than EV, supporting the enriched neuronal origin of NEEV 56 .
Limitations. While this study revealed new insight into the possible role of NEEV miR-93 in interoception, there are several limitations. First, more than half of MDD patients were taking SSRIs, raising the possibility of serotonergic in uences on the results. We believe this is unlikely as there was no difference on the VIA task or NEEV miR-93 expression between unmedicated MDD and SSRI-medicated individuals.
Future work may address this by repeating the study in unmedicated participants only. Second, other unmeasured factors could have potentially affected our results, such as other types of medication, genetics, diet, and socioeconomic status. Third, although we characterized the neuronal enrichment of NEEV, we were unable to further subdivide the NEEV populations to determine neuron type. Future studies should be undertaken to replicate these ndings, re ne the observed relationships to speci c neuronal subtypes, and longitudinally evaluate the degree to which NEEV signaling uctuates with interoceptive and affective changes, re ecting the varied emotional landscape of depression.

Conclusions
This study suggests that MDD is associated with lower NEEV miR-93 expression, which may lead to interoceptive processing dysfunctions through altered epigenetic modulation of insular function, whereas healthy individuals may be more reactive to stress-induced regulation of miR-93 expression during interoceptive processing. The combination of neuroimaging and brain-enriched extracellular vesicle approaches provides an exciting opportunity to discover novel cellular disease targets for depression.  Differential associations between neuronal-enriched extracellular vesicle miR-93 and serum IL-1ra, IL-6, TNF and leptin, in the MDD and HC groups. MDD = major depressive disorder. HC = health comparisons.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. T500NDEmiR93Supplement20230412.docx