Shifts in the immunoepigenomic landscape of monocytes in response to a diabetes-specific social support intervention: a pilot study among Native Hawaiian adults with diabetes

Clinical and immunological changes from participants enrolled in a diabetes-specific social support program

To address the health disparity of DM in NHs, a DM-SSP for NHs was previously developed and validated [35]. For this study, we recruited NH individuals enrolled in the 3-month DM-SSP intervention by our community partners. From 16 participants, we randomly chose 8 participants for our epigenetic study (Fig. 1). Demographic characteristics of study participants (n = 8) at baseline are shown in Table 1. Clinical data collected from participants at baseline and post-intervention are shown in Table 2. To ensure those chosen for our study were representative of the larger enrollment (n = 16), we compared clinical characteristics at baseline between our epigenetic study participants (n = 8) and the remaining participants from the DM-SSP enrollment (n = 8; Additional file 2: Table S1). To ensure homogeneity of monocyte populations for downstream analyses, samples enriched by magnetic cell separation were immunophenotyped to confirm the effectiveness of the enrichment of monocytes from PBMCs pre- and post-intervention, which exhibited robust enrichment (> 70% total monocytes with debris exclusion) at both timepoints (Mean ± Standard Deviation; Baseline = 80.3 ± 10.5, Post-Intervention = 75.7 ± 6.8). Glycemic status (i.e., HbA1c) was not significantly different between baseline and 3 months post-intervention (Baseline = 8.9 ± 1.3, Post-Intervention = 8.7 ± 2.4, Table 2). However, we observed significant, but clinically modest, changes in weight (Baseline = 224.9 ± 37.0 lbs., Post-Intervention = 220.5 ± 35.9, P < 0.05), BMI (Baseline = 36.2 ± 5.2 kg/m2, Post-Intervention = 35.5 ± 4.9, P < 0.05), and Diabetes Care Profile (Baseline = 2.7 ± 0.8, Post-Intervention = 4.0 ± 1.0, P < 0.01). Reductions in weight has been shown to have a significant impact on DM risk and management, delaying the onset of DM and improving glycemic control in those at-risk or known to have DM [36]. The Diabetes Care Profile, a validated survey instrument, used to assess the psychosocial factors related to diabetes care, such as diabetes self-management comprehension, has been associated with improved glycemic control [37]. The nominal changes to glycemic parameters (i.e., HbA1c) are not unexpected, given the short timeframe of the study, wherein improvements to the clinical features associated with diabetes may follow the cellular and molecular changes that are linked to long-term improvement in diabetic phenotypes, such as improved inflammation. To investigate this further, we sought to characterize the epigenetic modifications that may underlie improved monocyte inflammatory phenotypes.

Fig. 1figure 1

DM-SSP intervention-associated DNA methylation signatures in monocytes. a Differentially methylated loci (DML) in monocytes from participants at intervention timepoints, baseline (orange) and post-intervention (green), and among non-diabetic controls (red) identify distinct clusters of methylation patterns from unsupervised hierarchical clustering analysis (Manhattan distance, complete linkage method). Methylation values (β-value) are displayed as a range from low (0, blue) to high methylation (1, red). b Plot shows the expected (fuchsia) and observed (blue) CpG probe count for the DML at specific genomic regions (enhancer, promoter, gene body, 3’UTR, and intergenic) and the distribution around CpG Islands (CpG Island, N-shore, S-Shore, N-Shelf, S-Shelf, Open Sea). c Gene ontology analysis of DML enriched at the most significant biological processes indicated. d Differential DNA methylation between intervention timepoints (δβ-value =|Baseline β-value - post-intervention β-value|) at genes annotated to immune response-related functions. P  value of < 0.05, < 0.01, < 0.001, and < 0.0001 indicated by *, **, ***, and ****, respectively. Significance taken at P value < 0.05

Table 1 DM-SSP participants demographic characteristics (n = 8)Table 2 Clinical and immunological characteristics across DM-SSP intervention timepoints (n = 8)Distinct changes in monocyte DNA methylation following a diabetes-specific social support program intervention

To determine the extent to which genome-wide DNA methylation states in monocytes may be modified over the course of the DM-SSP intervention, we first isolated homogenous populations of total monocytes (CD14+CD16−/+) from each NH participant. To corroborate our immunophenotyping of magnetic bead-enriched monocytes, we compared monocyte-specific methylation patterns [38] with that of the corresponding methylation states of enriched monocytes from participants, resulting in a significant positive correlation at baseline (r = 0.85, P < 0.0001) and post-intervention (r = 0.85, P < 0.0001), indicating the sufficient homogeneity of the monocyte populations that we used for downstream DNA methylomic and transcriptomic analyses. Characterization of DM-SSP-associated DNA methylation patterns in monocytes between baseline and post-intervention was performed by filtering for CpG sites with absolute average differences in β-values between pre- and post-intervention timepoints at ≥ 0.15 β-units (δ of the β-value) after applying a resampling-based empirical Bayes approach on our dataset to exclude insignificant differences in DNA methylation [39]. This resulted in 1,061 differentially methylated loci (DML) that exhibited statistically significant and biologically relevant differences in DNA methylation. Unsupervised hierarchical clustering of the DML revealed a strong degree by which the methylation states in monocytes distinguished between baseline and post-intervention (Fig. 1a), indicating robust differential methylation of the 1,061 CpGs across all participants. These findings were consistent with previous reports supporting differential methylation in monocytes as a measure to identify and stratify study populations (e.g., case vs control) [40, 41]. Further, by incorporating DNA methylation data from monocytes of non-diabetic donors (clinical characteristics, Additional file 3: Table S2) at the same 1,061 CpGs into our hierarchical clustering analysis, we observed that the methylation states from non-diabetics clustered with DM-SSP participants post-intervention (Fig. 1a), suggesting a “normalization” of methylation levels to non-diabetic-like states after the intervention. Among DM-SSP-associated DML, we identified 13 hypermethylated (1.23%) and 1,048 hypomethylated (98.77%) loci post-intervention, indicating that a general hypomethylated state was associated with improved outcomes of the intervention. Furthermore, we observed significant mean differences in DNA methylation levels ranging from 15 to 33%, some of which were at CpGs within potential cis-regulatory regions. The maximum mean hypomethylated CpG locus post-intervention observed was 33% hypomethylated (in comparison with baseline) at one CpG locus within the 5’UTR of TRIM34. The maximum mean hypermethylated locus post-intervention observed was 21% hypermethylated (in comparison with baseline) at one CpG within the 3’UTR of FNBP1.

The genomic distribution of methylation associates with its distinct transcriptional regulatory functions. For instance, promoter methylation has been associated with transcriptional silencing [46], whereas gene body methylation is more nuanced and involved with alternative promoter usage and mRNA splicing [47, 48]. The localization of the DM-SSP-associated DML provided insight into the potential regulatory roles. We observed that the DML were significantly enriched at regulatory regions of the genome (Fig. 1b), including enhancers (Observed = 439, Expected = 224, P < 0.05) and gene bodies (Observed = 414, Expected = 359, P < 0.05); and significantly depleted in promoters (Observed = 315, Expected = 408, P < 0.05). In agreement with these observations, we found the distribution of the DML at CpG islands (Fig. 1b), regions dense in clusters of CpG content and typically located at gene promoters [21], to be de-enriched (Observed = 39, Expected = 328, P < 0.05). Further, there were significantly more CpGs than expected in open sea regions (Observed = 655, Expected = 385, P < 0.5), 5 kb or more from CpG islands, and both the north (N)-Shelf (Observed = 80, Expected = 54, P < 0.05) and south (S)-Shelf regions (Observed = 78, Expected = 49, P < 0.05) that flank CpG shores (≤ 2 kb from CpG islands) and extending outwards. These data suggested that the DM-SSP-associated DML were enriched at regions of the genome that may have cis-regulatory functions.

GO analysis, using Enrichr (https://amp.pharm.mssm.edu/Enrichr/) [51, 52], was applied to the DML to infer the potential cellular and biological processes. Our results revealed that the DML were enriched at genes annotated to biological processes related to DM and DM-related complications, including immune-related functions (e.g., cell migration, cytokine responses, etc.), metabolic processes (e.g., regulation of protein kinase B, regulation of PI3K, etc.), and cardiovascular pathways (e.g., regulation of cholesterol storage, regulation of arterial blood pressure, etc.) (Fig. 1c; full list in Additional file 4: Table S3). The enrichment of GO terms at cellular processes linked to DM and DM complications led us to suspect that the DML may contain a subset of CpGs that were localized at genes annotated to cardiometabolic diseases. Indeed, a subset of DML were located at genes associated with hypertension, myocardial infarction, stroke, insulin resistance, coronary artery disease, and DM (Additional file 4: Table S3). Finally, due to the contribution of inflammation in DM and its related cardiometabolic complications, we focused on a subset of 37 DML enriched at genes annotated to the immune response (Fig. 1d). Differential methylation analysis (δ β-Value =|Baseline β-Value - Post-intervention β-Value|) at this subset of CpGs revealed an absolute difference in methylation levels between pre- and post-intervention that ranged from 15 to 26% (Fig. 1d; Additional file 4: Table S3). An intragenic CpG (cg16429499) of NLRC3, a gene previously identified as having a role in regulating inflammation [53], exhibited the most robust differential methylation states between timepoints. Thirty-six of the 37 DML (97.2%) were characterized by DNA hypomethylation post-intervention compared to baseline (Additional file 4: Table S3). These results indicated DM-SSP-associated DML may functionally contribute to cellular processes relevant to DM (e.g., inflammation, insulin signaling pathways, etc.).

Diabetes-specific social support program-associated changes in monocyte gene expression may be epigenetically regulated

To examine whether DM-SSP-associated methylation differences may relate to transcription [54], we first identified differentially expressed genes (DEGs) in monocytes pre- and post-intervention in a subset of participants, which revealed 891 significant DEGs (PFDR < 0.05). Unsupervised hierarchical clustering of log-transformed gene expression data (Reads Per Million reads [RPM]) of the 891 DEGs revealed a strong degree by which gene expression partitioned both timepoints (Fig. 2a), consistent with a previous report examining differential expression in monocytes stratifying disease states [40]. GO pathway analysis of the DEGs revealed significant enrichment of genes relevant to the immune response, metabolic processes, and cardiovascular pathways (Additional file 5: Table S4). That these pathways were also revealed by the DML suggest these robust differences in gene expression may be epigenetically regulated. To explore this further, we integrated the gene-enriched DML (774 CpGs) with DEGs (891 genes) and observed 36 genes that overlapped both datasets (Fig. 2b). Twelve of the 36 DML-enriched genes were involved in the immune response, metabolic processes, or cardiovascular pathways, and cardiometabolic diseases (ADRB2, BACH2, BCL2, CD6, DOCK2, DUSP10, FCER2, GP1BA, ITGB1, LIPA, NCK1, and PCNXL2). To examine whether these DML-enriched genes may be epigenetically regulated, we performed correlation analyses between the methylation and expression states pre- and post-intervention. We observed significant associations between DNA methylation and gene expression at 9 out of the 12 DML-enriched genes, including ADRB2 (cg08370787, r = 0.61, P = 0.03, Fig. 2c), BCL2 (cg23756272, r = 0.77, P = 0.004, Fig. 2d), CD6 (cg01877352, r = 0.88, P = 0.0002, Fig. 2e), DOCK2 (cg00357551, r = − 0.84, P = 0.001, Fig. 2f), DUSP10 (cg19318653, r = − 0.59, P = 0.04, Fig. 2g), FCER2 (cg12387247, r = 0.85, P = 0.0004, Fig. 2h), LIPA (cg12555086, r = − 0.78, P = 0.003, Fig. 2i), NCK1 (cg00382999, r = 0.58, P = 0.05, Fig. 2j), and PCNXL2 (cg17894435, r = 0.79, P = 0.002, Fig. 2k). Among those with significant relationships between DNA methylation and gene expression, four CpGs were enriched at genes with known involvement in immune-related processes (i.e., cellular response to cytokine stimulus, regulation of JNK cascade/activity, acute inflammatory response, etc.): BCL2, CD6, DUSP10, and FCER2. These findings suggest these DML in epigenetically labile cis-regulatory regions involved in transcriptional regulation.

Fig. 2figure 2

DM-SSP intervention-associated differential gene expression overlaps DML. a Heatmap of Log2 (RPM) of the differentially expressed genes (DEGs) by unsupervised hierarchal clustering show two main clusters between baseline (orange) and post-intervention (green). RPM, reads per million reads. b Venn diagram of DML dataset at known genes (blue), DEGs (yellow), and overlap between CpGs-enriched at DEGs in both (purple). ck Plots display correlation between methylation (x-axis) and gene expression (y-axis) from baseline (orange) and post-intervention (green) for a subset of overlapping CpG-enriched genes (Fig. 3b) annotated to the immune response, metabolic processes, cardiovascular processes, and cardiometabolic diseases, including c ADRB2, d BCL2, e CD6, f DOCK2, g DUSP10, and h FCER2, i LIPA, j NCK1, and k PCNXL2. ADRB2: adrenoceptor beta 2; BCL2: BCL2 apoptosis regulator; CD6: cluster of differentiation 6; DOCK2: dedicator of cytokinesis 2; DUSP10: dual specificity phosphatase 10; FCER2: Fc fragment of IgE receptor II; LIPA: lipase A; NCK1: NCK adaptor protein 1; PCNXL2: pecanex-like protein 2. Significance taken at P value < 0.05

A diabetes-specific social support program associates with attenuated pro-inflammatory responses from monocytes

Given the enrichment of DM-SSP-associated differences in methylation and expression at genes involved in inflammation, we determined whether participation in DM-SSP led to changes in inflammatory states of monocytes. Thus, we performed monocyte intracellular cytokine staining (Mono-ICS), an ex vivo immunophenotyping assay used to determine monocyte inflammatory activity in response to the inflammatory stimuli lipopolysaccharide (LPS), in the same aliquot of monocytes used for molecular analyses. We selected 4 participants at baseline and 4 post-intervention (3 were matched participants at both timepoints), all of whose samples were included in our methylomic and transcriptomic profiling analyses. We first performed unsupervised hierarchical clustering analysis of the initial DM-SSP-associated DML (1,061 CpGs) in monocytes from this subset of participants, which robustly stratified individuals from both timepoints (Fig. 3a). Next, we measured inflammatory cytokine (IL-1β, IL-8, IL-6, and TNF-α) production from monocytes in resting conditions and stimulated with LPS. In an inflammatory stimuli-free condition (no-stim), participants displayed a low percentage of cytokine-specific producing monocytes (% of cytokine-specific+ monocytes/% of total monocytes) from both timepoints. Upon stimulation with LPS, we observed increased production of cytokine-producing monocytes at pre- and post-intervention relative to their respective stimuli-free conditions. However, comparing pre- to post-intervention timepoints under the LPS stimulated conditions, pre-intervention samples exhibited a significantly higher frequency of IL-1β (Baseline = 56.93 ± 7.96%, Post-Intervention = 30.33 ± 9.17%, P = 0.005, Fig. 3b) and IL-6 (Baseline = 26.98 ± 5.22%, Post-Intervention = 9.89 ± 2.06%, P = 0.001, Fig. 3c) producing monocytes than that of post-intervention. While these results are in agreement with previous reports suggesting that the beneficial effects of a diabetes self-management intervention include the reduction of inflammation and improved glucose tolerance [55], to our knowledge this attenuated monocyte-specific inflammatory response post-DM-SSP has never before been reported.

Fig. 3figure 3

Changes in monocyte inflammatory response between intervention timepoints determined by monocyte intracellular cytokine staining (Mono-ICS). a Unsupervised hierarchical cluster analysis (Manhattan distance, complete linkage method) of the DML from a subset of participants subjected to Mono-ICS. bc Figures represent the frequency of monocytes producing inflammatory cytokines at baseline (orange) and post-intervention (green) for b IL-1β, c IL-6 for two conditions: non-stimulated (No-Stim) and stimulation with inflammatory stimuli (LPS). Δ represents the difference in the frequency of cytokine-producing monocyte between LPS and No-Stim conditions. IL-1β, interleukin 1-beta; IL-6, interleukin 6; LPS, lipopolysaccharide. Bottom figures represent gating strategy employed for quantifying the frequency (%) of cytokine secreting monocytes from both treatments for b & c. P value for < 0.05, < 0.01, < 0.001, indicated by *, **, ***, respectively

We confirmed whether there was a heightened, or “hyper-responsive,” monocyte immune response to inflammatory stimuli pre-intervention that may have been attenuated post-intervention. We did this by comparing the difference in the frequency (%) of cytokine-producing monocytes between LPS-stimulated (LPS) and stimuli-free (no-stim) conditions (LPS - no-stim = Δ-value) separately at each intervention timepoint (Fig. 3b, c). Indeed, we observed a hyper-responsiveness to LPS challenge at pre- versus post-intervention for IL-1β (Baseline = 54.57 ± 7.13%, Post-Intervention = 20.10 ± 22.17%, P = 0.03, Fig. 3b) and IL-6 (Baseline = 26.54 ± 5.02%, Post-Intervention = 7.82 ± 3.55%, P = 0.001, Fig. 3c). That the overlapping DM-SSP-associated differences in DNA methylation and expression are at genes involved in the immune response, coupled with the attenuated inflammatory activity post-intervention, together suggest that DM-SSP facilitates epigenetic modulation that underlies monocyte function.

Epigenetic regulation of inflammatory genes associates with inflammatory monocytes

Previous data has shown that DNA methylation may facilitate pro-inflammatory responses to inflammatory stimuli [57]. As most, if not all, circulating monocytes would have turned over during the course of a 3-month intervention, the epigenetic, transcriptional, and functional differences we observed are likely a result of changes to the monocyte trajectory during differentiation. Thus, we focused on identifying a relationship between the immune response in monocytes and the monocyte methylation states of our previously observed immune-related DML that appear to be transcriptionally regulated (BCL2, CD6, DUSP10, and FCER2). From the same subset of participants used to assess monocyte inflammatory response, we compared Mono-ICS inflammatory response data (Δ % of cytokine-producing monocytes = LPS % of cytokine-producing monocytes - no stim % of cytokine-producing monocytes) to the methylation states of the DML at each timepoint. For an intragenic DML of BCL2 (cg23756272), monocyte methylation displayed a significant relationship with the frequency of cytokine-producing monocytes in response to LPS challenge for IL-6 (r = 0.73, P = 0.04, Fig. 4a). The association between methylation of an exon-localized DML of CD6 (cg01877352) was significantly correlated with IL-6+ monocytes (r = 0.80, P = 0.02, Fig. 4b). DUSP10, containing an intragenic DML (cg19318653), showed a significant positive association between DNA methylation and IL-6 (r = 0.83, P = 0.01, Fig. 4c) and TNF-α-secreting monocytes (r = 0.70, P = 0.05, Fig. 4d). We observed that the methylation state of the promoter-localized DML of FCER2 (cg12387247) was significantly positively correlated with both IL-1β+ (r = 0.72, P = 0.04, Fig. 4e) and IL-6+ (r = 0.76, P = 0.03, Fig. 4f) monocytes in response to LPS stimulation. Finally, we sought to investigate whether gene expression of each immune-related DML was associated with the cytokine-producing monocytes. Our results showed that the expression levels of only FCER2 was significantly associated with IL-1β+ monocytes (r = 0.70, P = 0.05, Fig. 4g). Together, these results revealed that the methylation and immune response states were dynamic between intervention timepoints, indicating that they were responsive to the DM-SSP intervention. These results suggest that the apparent modulation of DNA methylation states of genes associated with the inflammatory response of monocytes (i.e., BCL2, CD6, DUSP10, and FCER2) may be the result of a shift in the monocyte differentiation trajectory as the cell population replenishes over the course of the 3-month intervention.

Fig. 4figure 4

Association between monocyte immune responses and immune-related gene differential methylation. Plots represent correlation between the delta (Δ) frequency (%) of cytokine-producing monocytes (LPS-stimulated frequency of cytokine-producing monocytes [%] - no stimulation frequency of cytokine-producing monocytes [%]; y-axis) and af DNA methylation (β-value; x-axis; left) and g gene expression levels (RPM; x-axis; right) from baseline (orange circle) and post-intervention (green squares) samples for the subset of immune-related DML and DML-enriched DEG, including between a BCL2 & IL-6, b CD6 & IL-6, c DUSP10 & IL-6, d DUSP10 & TNF-α, e FCER2 & IL-β, f FCER2 & IL-6, g FCER2 & IL-1β. Significance taken at P value < 0.05

留言 (0)

沒有登入
gif