Vagal Stimulation Rescues HFpEF by Altering Cardiac Resident Macrophage Function

Abstract

Background:

We previously showed in a rat model of heart failure with preserved ejection fraction (HFpEF) that transcutaneous vagus nerve stimulation (tVNS) reduced cardiac fibrosis and inflammation. However, macrophage-mediated mechanisms through which tVNS rescues cardiac function remain poorly understood.

Methods:

We induced HFpEF in 8-week-old mice by a combination of a high-fat diet and L-NAME for 5 weeks followed by 4 weeks of tVNS or sham stimulation. At this time, we analyzed cardiac function by echocardiography and immune cell numbers by single cell RNA sequencing and flow cytometry.

Results:

Our data demonstrate that HFpEF mice exhibited diastolic dysfunction, left ventricular hypertrophy and fibrosis, consistent with HFpEF, and that tVNS significantly improved HFpEF severity. Analysis of merged single cell RNA sequencing data from control, HFpEF + sham, and HFpEF + tVNS mice showed that HFpEF was associated with accumulation of Spp1 expressing CCR2⁺ cardiac resident macrophages (CRM). Further, treatment with tVNS reduced the number of CCR2⁺ CRM and the expression of Spp1, while also inducing the expression of Igf1 in Timd4⁺/Lyve1⁺/Folr2⁺ (TLF⁺) and MHC2⁺ CRM. Global deletion of Spp1 or blockade of CCR2⁺ CRM recruitment improved HFpEF, whereas TLF⁺/MHC2⁺ specific deletion of Igf1 reversed the protective effect of tVNS on HFpEF. The benefits of tVNS were also abolished in the setting of disrupted acetylcholine (ACh)/α7 neuronal ACh receptor (α7nAChR) signaling, either via pharmacological inhibition of α7nAChR or choline acetyltransferase deletion in CD4⁺ T cells.

Conclusion:

Collectively, our data indicate that tVNS improves HFpEF by reducing Spp1 expressing CCR2⁺ CRM and inducing expression of pro-reparative Igf1 in TLF⁺/MHC2⁺ CRM. These effects are mediated through cholinergic signaling, highlighting a neuroimmune pathway in HFpEF.

Introduction

Heart failure with preserved ejection fraction (HFpEF) accounts for ~50% of heart failure cases, and this proportion is expected to increase along with the growing elderly population^{1, 2}. The 5-year mortality rate of this disease approaches 50% ³, yet effective therapies that have been shown to decrease morbidity or mortality remain limited^{4, 5}. While the pathogenic role of immune cells, particularly macrophages and T cells, in ischemic heart failure has been well characterized ^6–9, their contribution to HFpEF pathophysiology is poorly understood. This knowledge gap may partly explain the negative results of past HFpEF trials⁴. Recent animal and human studies suggest that inflammation plays a prominent role in the development of HFpEF¹⁰. In addition, it has recently been shown that the number of cardiac macrophages in the heart increases, both in animal models and in humans with HFpEF, suggesting that macrophages and their secreted proteins mediate cardiomyocyte hypertrophy and disease progression^10–14. However, the functional role of cardiac resident macrophages (CRM) in HFpEF development and progression remains unclear.

The vagus nerve modulates immune responses via the neuroimmune cholinergic anti-inflammatory pathway, which involves acetylcholine (ACh)-dependent activation of α7 nicotinic acetylcholine receptors (α7nAChR) on macrophages^15–17. Vagus nerve stimulation (VNS) exerts powerful anti-inflammatory and anti-fibrotic effects in both animals and humans by activating this pathway^15–18. Notably, transcutaneous VNS (tVNS) has emerged as a promising and highly translational treatment strategy for various cardiovascular conditions including atrial fibrillation, heart failure, and dysautonomia^19–21. Importantly, tVNS improves diastolic function, attenuates left ventricular (LV) inflammation and fibrosis, and suppresses proinflammatory gene expression compared to sham stimulation¹⁸, yet its mechanistic effects on CRM in HFpEF remain undefined.

In this manuscript, we used single cell RNA sequencing (scRNA-seq) to investigate the macrophage-mediated mechanism underlying tVNS-induced improvement in HFpEF. Our findings reveal that tVNS treatment reduces CCR2⁺ CRM number and their expression of Spp1, while promoting Igf1 expression in TLF⁺/MHC2⁺ CRM. Global deletion of Spp1 or blockade of CCR2⁺ CRM recruitment ameliorated HFpEF, whereas selective deletion of Igf1 in TLF⁺/MHC2⁺ CRM prevented tVNS-mediated HFpEF rescue. Moreover, the benefits of tVNS were abolished in the setting of disrupted Ach/α7nAChR signaling, either via pharmacological inhibition of α7nAChR or choline acetyltransferase (ChAT) deletion in CD4⁺ T cells. Collectively, these findings indicate that tVNS protects against HFpEF by reducing accumulation of pathogenic, Spp1-producing CCR2⁺ CRM while enhancing reparative Igf1 expression in TLF⁺/MHC2⁺ CRM through a neuroimmune cholinergic anti-inflammatory mechanism.

Methods

Data availability

Raw data for the scRNAseq experiments performed in this manuscript can be found in GEO using accession number GSE295382. Raw data from Lanzer et al. ³¹ was obtained using accession number GSE275031. All other data will be made available by either corresponding author upon reasonable request. Research materials are referenced in the Major Resources Table in Supplemental Material 1.

Experimental design

Wild type C57BL/6 CCR2^rfp/wt (B6.129(Cg)-Ccr2tm2.1Ifc/J) and CCR2^gfp/wt (B6(C)-Ccr2tm1.1Cln/J) mice were obtained from the Jackson Laboratory. Ms4a3^cre mice were the kind gift of Dr. Florent Ginhoux. All experimental procedures were approved by the University of Oklahoma Health Science Center Institutional Animal Care and Use Committee under protocol number 19–062.

HFpEF was induced in C57BL/6 CCR2-RFP mice that have the gene for red fluorescent protein (RFP) inserted into the endogenous CCR2 allele, serving as a reporter for cells actively expressing CCR2 ²². Eight-week-old C57BL/6 CCR2-RFP male mice were exposed to L-NAME, a nitric oxide synthase inhibitor (0.5g/L in drinking water) and high-fat diet (HFD; 60% calories from lard) for 5 weeks in order to induce HFpEF ²³. C57BL/6 CCR2-RFP mice fed a standard chow diet served as controls. We only used male mice because it was previously shown that female mice are protected from HFpEF development in this model ²⁴. At 13 weeks of age, echocardiography was performed to confirm the development of HFpEF, as evidenced of diastolic dysfunction (Figure 1A, red arrowhead) ²³. The animals were monitored twice daily for heart failure symptoms, including distress, reduced movement, difficulty in breathing, body edema, and cachexia. The development of heart failure was assessed using a scoring system that integrates these parameters (Supplemental Material 1)¹⁸. In brief, the heart failure score evaluates appearance, breathing, mobility, edema and body weight and assigns a score for each parameter (0 = normal, 1 = mildly abnormal, 2 = severely abnormal). The composite heart failure score is the sum of individual scores of all parameters. We also assessed exercise capacity using an exercise treadmill test, as described below. None of the mice were excluded because of lack of heart failure symptoms or absence of echocardiographic signs of HFpEF. The heart failure score was similar between groups.

Figure 1. Effect of tVNS treatment on cardiac phenotype.

A. Schematic representation of study design and timeline of events. Baseline assessment (red arrowhead). Endpoint assessment (blue arrowhead). B. Representative image of an anesthetized mouse undergoing transcutaneous vagus nerve stimulation (tVNS) with electrode clips on both ears. C. Representative electrocardiographic heart rate monitoring at baseline and during stimulation, showing prolongation of the cycle length during tVNS, indicating that vagal fibers were stimulated. Note the stimulation artifact (red arrows). D. Systolic blood pressure. n=18 for Control; n=17 for HFpEF+sham; n=17 for HFpEF+tVNS. E. Left ventricular (LV) ejection fraction. n=12 per group. F. Early diastolic mitral annulus Doppler velocity (e′). n=12 per group G. Ratio of the early mitral inflow Doppler velocity to the early diastolic mitral annulus velocity (E/e′). n=12 per group. H. Ratio of the early to late mitral inflow Doppler velocity (E/A). n=13 for Control; n=11 for HFpEF+sham; n=12 for HFpEF+tVNS. I. Representative examples of tissue Doppler echocardiography and mitral inflow Doppler echocardiography from animals in each group. J. Workload during treadmill exercise test. n=4 per group. K. Heart weight normalized to tibia length. n=17 per group L. Lung weight normalized to body weight. n=14 for Control; n=13 for HFpEF+sham; n=13 for HFpEF+tVNS. M. Representative examples of histological images from animals in each group, stained with Masson’s trichrome, showing fibrosis. N. Fibrosis area calculated as a percentage of the total LV area. n=11 per group.

After 5 weeks of L-NAME and HFD administration, mice were randomly assigned to either sham or tVNS treatment groups (Figure 1A, red arrowhead). A custom-made electrode was placed on the auricular concha area of each ear for tVNS (Figure 1B, Supplemental Material 1). Electrical stimulation was delivered daily for 20 minutes over a 4-week period beginning at 13 weeks of age, under 2% isoflurane inhalation anesthesia using a transcutaneous electrical nerve stimulation device (InTENSity Twin Stim; Current Solutions LLC, Austin, TX), with a frequency of 20 Hz, pulse duration of 200μs, and amplitude of 2mA, as in our previous study ¹⁸ (Figure 1C). Electrodes in the active tVNS group were positioned over the auricular concha region with the cathode inside and the anode outside ¹⁸, while in the sham group the electrodes were placed at the same area, but no stimulation was delivered.

Following the tVNS or sham treatment, at 17 weeks of age, the animals were assessed for blood pressure (BP), echocardiographic signs of diastolic dysfunction and exercise performance (Figure 1A, blue arrowhead). The animals were euthanized and the hearts were harvested for scRNA-seq analysis, flow cytometry and histology (Masson’s trichrome or picrosirius red staining to assess fibrosis).

Blood pressure measurement, echocardiography and exercise capacity

We measured blood pressure noninvasively using the tail-cuff method (Power Lab Data Acquisition System; ADInstruments, Bella Vista, NSW, Australia).

We performed echocardiography using a Vevo 3100 system (Visual Sonics, Toronto, ON) equipped with a 25MHz linear array transducer, under 2% isoflurane anesthesia. Briefly, 2D parasternal and apical 4-chamber LV images were obtained. LV ejection fraction (LVEF) from the parasternal short axis view, using M-mode was used as a measure of LV systolic function, in accordance with previous literature ¹⁸. Diastolic dysfunction was evaluated through pulse-wave Doppler recording of mitral inflow and mitral annulus tissue Doppler spectra obtained from the apical 4-chamber view. Parameters measured included the early diastolic mitral annulus velocity (e’), and the ratio of early mitral inflow Doppler velocity to early diastolic mitral annulus velocity (E/e’ ratio), which is a surrogate marker for LV filling pressures ²⁵. The echocardiographic analyses were performed in a blinded manner to ensure accuracy and eliminate bias.

Exercise performance assessment was performed according to an established protocol ²⁶. For these studies, mice were acclimated to a treadmill with a fixed 10% incline (Mice 5-Lane Touchscreen Treadmill, Panlab, Spain) once daily for three days. During acclimation, the treadmill was set to a speed of 10 m/min for 5 minutes. On the fourth day, exercise performance testing began with initial speed set at 10 m/min, which was incrementally increased by 2 m/min every 2 minutes until exhaustion or achievement of maximum speed of 40 m/min. Exhaustion was defined as failure to continue running for 10 seconds, despite gentle nudging with a tongue depressor. Total running time, maximum running speed and total running distance were recorded.

Quantification of fibrosis

LV tissue was collected at the endpoint, promptly fixed in 4% formalin, and embedded in paraffin. Paraffin blocks were cut into 5 um sections. Masson’s trichrome or picrosirius red staining was utilized to evaluate cardiac fibrosis. Images of the stained tissue were captured at x40 magnification, and the percentage of fibrosis was quantified as a percentage of the total stained LV tissue area using the Bioquant Osteo software (BIOQUANT Image Analysis Corporation, Nashville. TN), or ImageJ software ¹⁸. To avoid bias, the investigators conducting the measurements and data analysis were blinded to the group assignments.

Confocal imaging

Tissue sections (5 µm) were deparaffinized using xylene and then rehydrated through a graded serious of ethanol solutions (100%, 90%, 70%, and 50%) followed by distilled water. Antigen retrieval was conducted using the IHC-Tech Epitope Retrieval Steamer with a Tris-EDTA buffer for 40 minutes. The tissues were then permeabilized and blocked with 5% donkey serum (cat# 017-000-121; Jackson ImmunoResearch). After blocking, the sections were incubated overnight at 4°C with a primary antibody (1:100 dilution in 5% blocking buffer) targeting the Tdtomato antigen (cat# 600-401-379; Applied Biosystems, Thermo Fisher Scientific). The next day, the tissues were washed with Tris-buffered saline containing 0.1% Tween 20, followed by a 1-hour incubation at room temperature with an Alexa Fluor-594 secondary antibody (cat# 711-586-152; Jackson ImmunoResearch) diluted 1:500 in 5% blocking buffer. Nuclei were stained with DAPI (1µg/ml; Sigma #D9542). Finally, the samples were mounted (Epredia Shandon-Mount #1900331), and images were captured using a Leica SP8 confocal microscope at 40x (oil immersion) magnification. Secondary antibody-only and isotype controls (cat# 02–6102; Thermo Fisher Scientific) were assessed to confirm the specificity of the immunostaining.

Flow cytometry

Flow cytometry was performed on heart tissue using established methodology ^{27, 28}. Briefly, the hearts were minced with a razor blade and digested with a combination of collagenase I (3.6 mg/ml), DNase I (60 U/ml), and hyaluronidase (60 U/ml) for 1 hour at 37 ^oC. Red blood cells were lysed using ACK lysing buffer (Quality Biological cat#: 118-156-1001) and cells were resuspended in 1 ml PBS containing 1% BSA with Fc blocking solution for 30 minutes on ice. Approximately two million cells were stained at room temperature for 30 minutes with primary antibodies listed in Supplemental Material 1. In studies examining α7nAChR expression, we incubated cells with a non-conjugated rabbit anti-rat α7nAChR antibody (Alomone labs, Catalog #: ANC-007; 1:200 dilution) for 30 minutes in 1% BSA as done in previous studies²⁹. After 30 minutes of staining, cells were washed with 1% BSA and incubated with a goat anti rabbit Alexa647 antibody (Invitrogen, Catalog # A-21244; 1:500 dilution) for 30 minutes in 1% BSA. After washing with 1% BSA, cells were fixed in 2% PFA for 30 minutes, followed by washing with PBS. Cells were resuspended in PBS, run on an Aurora Cytek, and data analyzed using FlowJo version 10.9.0 software.

Flow cytometry for single cell experiments

After blocking, cells were spun at 1200 RPM (300g) and incubated with the antibodies listed in Supplemental Material 1 for 30 minutes at room temperature. Cells were spun, washed with 1% BSA, resuspended in 1% BSA, and sorted using a Becton–Dickenson FACSAriaII. For scRNA-seq studies, we sorted ~25,000 live, CD45+ cells into individual BSA coated tubes (see Figure 2 for gating strategy used for scRNAseq experiments). Cells were initially counted with trypan blue to ensure quality of cells after sorting. After reaching the core, cells were counted with a BioRad TC20 Automated Cell Counter (PN 1450102) and approximately 5,000 cells from each animal were subjected to bead emulsion and 10X genomics. This was done for N=2–3 animals from each experimental group (Control, HFpEF + sham, HFpEF + tVNS).

Figure 2. Subpopulations of cardiac immune cells identified using scRNA-seq and validation using flow cytometry.

A. Flow cytometry gating strategy used for single cell RNA sequencing experiments of CD45⁺ cells. B. UMAP showing the clustering of cells into 12 distinct subpopulations. C. Dotplot showing expression of key immune genes in each group of mice. D. Proportion of cell types in each experimental condition. E. UMAP demonstrating the clustering of CRM into four subpopulations. F. Dotplot showing expression of key CRM subset genes. G. Proportion of CRM subsets in each experimental condition. n=2 for Control; n=3 for HFpEF+sham; n=3 for HFpEF+tVNS. H. Gating strategy used to identify CRM subsets. I. Flow cytometry data showing the percentage of TLF⁺, MHC2⁺, and CCR2⁺ CRM across the experimental groups. n=15 for Control; n=13 for HFpEF+sham; n=14 for HFpEF+tVNS.

Taqman quantitative reverse transcription PCR

Heart tissues (20–30 mg) were homogenized in QIAzol lysis reagent using a bead beater (Biospec products), and total RNA was isolated using a commercial kit (cat#217004; miRNeasy Mini Kit; Qiagen) as per the kit procedure. Subsequently, about 1 µg of RNA was reverse transcribed to cDNA using the iScript gDNA Clear cDNA Synthesis kit (cat#172–5035; Bio-Rad). Afterward, TaqMan duplex qPCR was performed on a QuantStudio-5 Real-Time PCR system using the TaqMan Fast Advanced Master Mix (4444557; Applied Biosystems, Thermo Fisher Scientific) and Taqman probes under the following conditions: 95 °C for 40 sec; 50 cycles of 95 °C for 2 sec and 60°C for 30 sec. Each cDNA sample was examined in technical triplicates. The ribosomal gene 18S was used for normalization. The comparative ΔΔCT method was used to assess the relative gene expression and data are presented as fold change versus control. The probes for the genes Tnfa (Mm00443258_m1), Il6 (Mm00446190_m1), Il1b (Mm00434228_m1), Spp1 (Mm00436767_m1), Col3a1 (Mm00802300_m1), Col4a1 (Mm01210125_m1), Acta2 (Mm01546133_m1), Postn (Mm01284919_m1), Tgfb1 (Mm01178820_m1), and 18S rRNA (4319413E) were obtained from Applied Biosystems, Thermo Fisher Scientific.

10X genomics

Single cell libraries for 10x Chromium were generated from sorted cells according to the manual’s protocol. The sorted single cell suspension, 10x barcoded gel beads, and oil were loaded into a Chromium^™ Single Cell Chip G, and single cells were captured in nanoliter-scale oil droplets using the Chromium^™ Controller to form Gel Bead-In-EMulsions (GEMs). The experiments utilized the “Chromium^™ Single Cell 3’ GEM, Library & Gel Bead Kit v3.1”. Full-length cDNA libraries were created by incubating GEMs in a thermocycler. The GEMs with cDNAs were then disrupted, and all single cell cDNA libraries were combined, purified using DynaBeads MyOne^™ Silane beads (Fisher PN 37002D), and pre-amplified via PCR to produce enough material for sequencing library preparation. Sequencing libraries were prepared through the following steps: cDNA fragmentation, end repair and A-tailing, size selection with SPRIselect beads (Beckman Coulter, PN B23318), adaptor ligation, sample index PCR amplification, and a second size selection with SPRIselect beads. The final single cell libraries were sequenced on an Illumina NextSeq2000, aiming for at least 25,000 total reads per cell.

Single cell sequencing data processing

The Cellranger software (version 7.1.0) ‘mkfastq’ function was utilized to generate fastq files from the sequencer output. After creating the fastq files, the Cellranger ‘count’ function was used to align the raw sequencing reads to the reference genome using the STAR aligner. The ‘count’ function produced three output files (barcodes.tsv, features.tsv, matrix.mtx), which were imported into the R Seurat package (version 4.3.0) for downstream analysis (QC metrics-based selection and filtration of cells, normalization and scaling of data, and detection of highly variable genes) ³⁰. The Seurat vignette (https://satijalab.org/seurat/pbmc3k_tutorial.html) was followed to construct the Seurat data matrix object. Seurat objects from individual experiments were combined using the ‘merge’ function. Data reproducibility when harvested on individual dates was confirmed by integrating Seurat objects from each experiment using the “IntegrateData” function (dims = 1:30). We demontrated significant integration across all data sets prior to combining. Post-merging, low-quality cells were filtered out by retaining genes expressed in more than three cells and cells with at least 200 detected transcripts. Cells with mitochondrial gene content exceeding 10% or unique gene counts above 3,000 or below 200 were excluded. Data normalization was performed using Seurat’s ‘NormalizeData’ function, applying the LogNormalize method, which globally scales gene expression per cell to total gene expression, multiplies by a scale factor of 10,000, and log-transforms the result. Highly variable genes were detected using Seurat’s ‘FindVariableGenes’ function. Variations due to library size and mitochondrial gene percentage were regressed out using the ‘ScaleData’ function. Principal component analysis (PCA) was conducted on variable genes, with significant principal components determined via the ‘JackStraw’ function. The top 10 principal components were used as input for Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction, performed with Seurat’s ‘FindClusters’ and ‘DimPlot’ functions. Differentially expressed genes (DEGs) for each cell cluster, compared to all other clusters, were identified using the ‘FindAllMarkers’ function on normalized gene expression data. To identify DEGs between groups (control, HFpEF + sham, HFpEF + tVNS), we used the non-parametric Wilcoxon rank sum test and the “FindMarkers” function in Seurat. Finally, to compare gene expression profiles in different cell types, we combined our scRNAseq data with a publicly available dataset (GSE275031)³¹. We used R Package ‘Harmony’ version 1.2.3 ³² to correct for batch effects between datasets followed by downstream analysis of the data as described above.

Statistical analysis

Sample size was determined based on preliminary experiments assessing HFpEF phenotype and response to tVNS treatment, as well as prior studies¹⁸, using a significance alpha level of 0.05 and 80% power. Normality was assessed using the Shapiro-Wilk test and/or quantile-quantile (Q-Q) plots³³, particularly for small sample sizes (n<10), where visual inspection aided interpretation, acknowledging the limited power of formal normality testing in this context. Comparisons between two groups with normally distributed data were performed using unpaired two-tailed Student’s t-test, while non-normally distributed data were analyzed using Mann-Whitney U test. For comparisons among 3 groups, one-way ANOVA with Tukey’s post hoc correction or Kruskal-Wallis test with Dunn’s post hoc correction were used, as appropriate. When n<10, uncorrected post hoc tests (Fisher’s least significant differences or Dunn’s test) were used in order to maximize power, noting that these results should be interpreted with caution. Additionally, for very small sample sizes (n < 6 in any group), non-parametric tests were selected regardless of distribution. In cases of extremely small sample sizes (n < 3 in any group), no statistical testing was performed, and results are reported descriptively. Statistical significance was defined as p < 0.05. All analyses were conducted using GraphPad Prism version 9. Representative figures were selected to reflect average data. All experiments were performed using independent biological replicates. For qPCR technical replicates were used for each biological replicate (triplicate wells per cDNA sample). Additional statistical information, including exact sample sizes, is available in Supplemental Material 2 and 3.

Results

tVNS restores cardiac function in HFpEF mice

Following five weeks of L-NAME and HFD administration, C57BL/6 CCR2-RFP mice developed hallmark signs of heart failure, including edema, reduced mobility and tachypnea, as confirmed using a previously described scoring system that integrates these parameters ^{18, 34}, as well as a decline in exercise capacity (Supplemental Figure 1A, B). In addition, they exhibited diastolic dysfunction, as confirmed by echocardiography, while maintaining a normal LV ejection fraction (Supplemental Figure 1C–F).

Our lab previously demonstrated that tVNS significantly improves HFpEF phenotype in a high-salt diet rat model through anti-inflammatory and anti-fibrotic mechanisms¹⁸. To confirm these findings in our obese and hypertensive HFpEF mouse model, we administered tVNS daily for 20 minutes at the auricular concha region (20Hz, 2mA) for 4 additional weeks after inducing HFpEF (Figure 1A, B). An immediate heart rate reduction of 20 ± 8 bpm following tVNS confirmed effective vagal activation (Figure 1C). After 4 weeks of treatment, tVNS significantly attenuated systolic BP elevation compared to the HFpEF + sham group (Figure 1D). Echocardiographic markers of diastolic dysfunction, including e’ velocity and E/e’ ratio, were significantly improved in the HFpEF + tVNS group compared to the HFpEF + sham group, while LV ejection fraction remained unchanged (Figure 1E–I), indicating that tVNS improved myocardial relaxation and diastolic function. Additionally, tVNS improved exercise capacity, reduced heart weight normalized to tibia length, alleviated pulmonary congestion (as indicated by the reversal of the lung weight to body weight ratio), and decreased cardiac fibrosis (Figure 1J–N). Importantly, to confirm that these benefits were not due to BP reduction, we conducted an experiment using Olmesartan, an angiotensin II receptor blocker ¹⁸, which failed to replicate the protective effects of tVNS on cardiac diastolic function and heart weight, despite similar reduction in BP (Supplemental figure 2A–D). Overall, these results demonstrate that tVNS significantly improves HFpEF independent of changes in BP.

tVNS reduces CCR2⁺ CRM accumulation in HFpEF

To unbiasedly identify potential mechanisms through which tVNS mitigates the HFpEF phenotype, we performed scRNA-seq on immune cells isolated from hearts of control, HFpEF + sham and HFpEF + tVNS mice using the gating strategy outlined in Figure 2A. We focused our scRNA-seq analysis on immune cells due to their established role in acute and chronic cardiovascular disease ^{13, 27, 28, 35} , as well as the well-documented anti-inflammatory effects of tVNS ¹⁸; therefore, we hypothesized that alterations in immune cell populations may underlie the therapeutic benefit of tVNS on HFpEF. After removing low-quality cells (see methods), we used top differentially expressed genes to cluster and annotate 12 distinct immune cell populations, including B cells, T cells, natural killer cells, dendritic cells, neutrophils, monocytes, and cardiac resident macrophages (CRM; Figure 2B, C). Cluster abundance quantification is shown in Figure 2D.

We further studied CRM due to their established role in cardiac homeostasis and disease, where distinct subsets contribute to inflammation, fibrosis and remodeling ^{27, 28, 35, 36}. Our analysis identified 4 distinct clusters of CRM: Timd4⁺/Lyve1⁺/Folr2⁺ (TLF⁺) CRM, MHC2⁺ CRM, CCR2⁺ CRM, and IFNγ-responsive CRM (Figure 2E, F), consistent with previous reports ²⁸. Next, we examined CRM abundance across control, HFpEF + sham, and HFpEF + tVNS groups. The proportion of CCR2⁺ CRM was increased in HFpEF + sham mice compared to controls, whereas tVNS reversed this effect (Figure 2G). In contrast, HFpEF and tVNS had minimal impact on other CRM populations (Figure 2G), suggesting that the reduction of CCR2⁺ CRM accumulation may be central to the cardioprotective effects of tVNS on HFpEF.

To validate our scRNAseq results, we performed flow cytometry analysis on hearts from control, HFpEF + sham, and HFpEF + tVNS mice 4 weeks after tVNS initiation using the same experimental design as for the scRNA-seq analysis. The gating strategy for flow cytometry is shown in Figure 2H. Experiments were conducted in CCR2^rfp/wt reporter mice to accurately identify CRM actively expressing CCR2²². In agreement with our scRNA-seq data, flow cytometry confirmed that HFpEF increased CCR2⁺ CRM abundance, while tVNS reversed this effect (Figure 2I). In contrast, tVNS had minimal impact on other CRM subsets (Figure 2I). Collectively, these data demonstrate that CCR2⁺ CRM are elevated in HFpEF and that this effect is reversed by tVNS.

Loss of monocyte-derived CCR2⁺ CRM improves HFpEF

To investigate the mechanisms driving CCR2⁺ CRM accumulation, we induced HFpEF in fate-mapping reporter mice that were generated by crossing Ms4a3^cre Rosa stop^f/f TdT mice with Ccr2^gfp/wt reporter mice (Figure 3A). In these mice, the Ms4a3 promoter-driven Cre is selectively expressed in granulocyte-monocyte progenitors, including Ly6c^hi monocytes, allowing precise tracking of monocytes into the CRM niche³⁷. To assess the impact of CCR2 loss on CRM number, we generated fate-mapping mice with a functional CCR2^KO (Ms4a3^cre Rosa stop^f/f TdT Ccr2^gfp/gfp), in which both copies of the Ccr2 gene are replaced by a gfp allele. As a result, Ly6c^hi monocytes fail to egress from the bone marrow ³⁸, preventing the accumulation of monocytes and their progeny in peripheral tissues. Following 9 weeks of a HFpEF-inducing diet, flow cytometry confirmed that HFpEF increased total CCR2⁺ CRM abundance in Ms4a3^cre Rosa stop^f/f TdT Ccr2^gfp/wt mice, whereas this effect was blunted in CCR2^KO mice (Supplemental figure 3A–C). More specifically, HFpEF increased TdT⁺ CCR2⁺ CRM (expressed as a percentage of live single cells in the heart), but had no impact on TdT⁺ TLF⁺ or TdT⁺ MHC2⁺ CRM (Figure 3B). Loss of CCR2 significantly attenuated HFpEF-induced accumulation of TdT⁺ CCR2⁺ CRM, without affecting other CRM subsets (Figure 3B). Confocal microscopy further confirmed an increase in TdT⁺ CRM in HFpEF Ms4a3^cre Rosa stop^f/f TdT Ccr2^gfp/wt mice, while CCR2^KO mice exhibited fewer TdT⁺ CRM (Figure 3C). Collectively, these data demonstrate that CCR2⁺ CRM are monocyte-derived and that CCR2 loss prevents their accumulation in HFpEF.

Figure 3. Fate mapping data provide insight into the origin of CRM in HFpEF, as well as the impact of Ccr2 deficiency on the cardiac phenotype compared to wild-type controls.

A. Schematic representation of study design. B. Bar graphs present the proportion (plotted as percent of live single cells in the heart) of TdT⁺ (derived from circulating monocytes) TLF⁺, MHC2⁺, and CCR2⁺ CRM within each experimental group. n=5 per group. C. Representative confocal images showing TdT⁺ reporter cells in heart LV tissues. D. Early diastolic mitral annulus Doppler velocity (e′). n=6 per group E. Ratio of the early mitral inflow Doppler velocity to the early diastolic mitral annulus velocity (E/e′). n=6 per group. F. Representative examples of tissue Doppler echocardiography and mitral inflow Doppler echocardiography from CCR2 knockout (CCR2^gfp/gfp) mice compared to wild type (CCR2^gfp/wt). G. Heart weight normalized to tibia length. n=7 for HFpEF Ccr2^gfp/wt; n=8 for HFpEF Ccr2^gfp/gfp. H. Lung weight normalized to body weight. n=7 per group. I. Representative examples of histological images from animals in each group, stained with Masson’s trichrome, showing decreased fibrosis in CCR2^gfp/gfp mice compared to CCR2^gfp/wt mice. J. Fibrosis area calculated as a percentage of the total left ventricular (LV) area. n=6 per group. K. TaqMan qRT-PCR analysis of the expression of inflammatory genes (Tnfa, Il6, Il1b) n=6 for Conrol Ccr2^gfp/wt; n=6 for HFpEF Ccr2^gfp/wt; n=7 for HFpEF Ccr2^gfp/gfp. L. TaqMan qRT-PCR analysis of fibrotic genes (Spp1, Tgfb1, Col3a1, Col4a1, Acta2, Postn). n=6 for Conrol Ccr2^gfp/wt; n=6 for HFpEF Ccr2^gfp/wt; n=7 for HFpEF Ccr2^gfp/gfp.

Next, we assessed the cardiac phenotype of HFpEF CCR2^KO mice, which revealed improved diastolic function, as assessed by echocardiography, compared to HFpEF Ccr2^gfp/wt mice (Figure 3D–F). In line with these findings, CCR2^KO mice also exhibited reduced heart weight and lung congestion, while Picrosirius Red staining and quantification confirmed attenuated cardiac fibrosis (Figure 3G–J). Furthermore, pro-inflammatory (Tnfa, Il6, Il1b) and pro-fibrotic (Spp1, Tgfb1, Col3a1, Col4a1, Acta2, Postn) gene expression was significantly downregulated in CCR2^KO mice compared to Ccr2^gfp/wt mice (Figure 3K, L). The phenotypic rescue in CCR2^KO mice supports a pathogenic role of CCR2⁺ CRM in HFpEF.

To validate that CCR2⁺ CRM are the primary drivers for the pathological effects observed in our HFpEF mouse model, we pharmacologically blocked the colony-stimulating factor-1 (CSF1)/ CSF1 receptor (CSF1-R) signaling axis, which promotes self-proliferation of resident macrophages ^{39, 40}. For these experiments, HFpEF mice were randomized to vehicle vs. CSF1-R inhibitor (GW2580; 160mg/kg daily by oral gavage)⁴¹, followed by CRM and cardiac phenotype analysis nine weeks post-treatment. Flow cytometry confirmed that TLF⁺ and MHC2⁺ CRM were significantly reduced in the CSF1R inhibitor group, whereas CCR2⁺ CRM remained intact (Supplemental Figure 4A). Notably, depletion of TLF⁺ and MHC2⁺ CRM did not impact the cardiac phenotype (including diastolic dysfunction, heart weight and fibrosis) compared to vehicle-treated mice (Supplemental Figure 4B–G). Cumulatively, these data indicate that CCR2⁺ CRM promote HFpEF, whereas TLF⁺ and MHC2⁺ CRM have minimal impact on disease development.

CCR2⁺ CRM-derived Spp1 aggravates HFpEF

Our data suggest that tVNS ameliorates HFpEF by reducing the number of pathogenic CCR2⁺ CRM. To test the hypothesis that tVNS may also be restricting expression of pro-inflammatory cytokines in CCR2⁺ CRM, we analyzed DEGs in CCR2⁺ CRM from control, HFpEF + sham, and HFpEF + tVNS mice. An in-depth analysis of differentially expressed genes revealed that Spp1, which encodes osteopontin, a multifunctional extracellular matrix protein implicated in cardiac fibrosis, hypertrophy, and inflammation^{42, 43}, was elevated in CCR2⁺ CRM from HFpEF mice compared to controls, while tVNS reduced its expression (Figure 4A,B). To further assess the specificity of Spp1 expression across different cardiac cell types, we integrated our scRNAseq data with recent scRNAseq data from Lanzer et al³¹. and confirmed that Spp1 is predominantly expressed by CCR2⁺ CRM (Supplemental Figure 5A–C). To assess the functional role of Spp1, we induced HFpEF, as described above, in a global SPP1^KO mouse model (Figure 4C). Compared to SPP1^WT mice, SPP1^KO mice were protected from the HFpEF phenotype, displaying improved diastolic function, reduced hypertrophy and lung weight, and decreased cardiac fibrosis (Figure 4D–J). Furthermore, qRT-PCR analysis demonstrated that SPP1^KO mice exhibited reduced expression of pro-inflammatory (Tnfa, Il6, Il1b) and pro-fibrotic genes (Tgfb1, Col3a1, Col4a1, Acta2, Postn) compared to SPP1^WT mice (Figure 4K, L). Collectively, these data suggest that CCR2⁺ CRM-derived Spp1 promotes HFpEF, and that the protective effect of tVNS on HFpEF is mediated by both a reduction in CCR2⁺ CRM number and expression of Spp1.

Figure 4. Spp1 expression in CCR2⁺ CRM and the impact of global Spp1 deletion on cardiac phenotype in experimental mice.

A. Volcano plot showing genes that were differentially expressed in CCR2⁺ CRM across experimental groups. B. Violin plot showing expression of Spp1 in CCR2⁺ CRM. C. Schematic overview of the study design. D. Early diastolic mitral annulus Doppler velocity (e′). n=5 for Spp1^WT; n=10 for Spp1^KO. E. The ratio of early mitral inflow Doppler velocity to early diastolic mitral annulus velocity (E/e′). n=5 for Spp1^WT; n=11 for Spp1^KO. F. Representative images of mitral annulus tissue Doppler and mitral inflow Power Doppler echocardiography. G. Heart weight normalized to tibia length. n=5 for Spp1^WT; n=10 for Spp1^KO. H. Lung weight normalized to body weight. n=5 for Spp1^WT; n=9 for Spp1^KO. I. Representative histological sections stained with Picrosirius Red. J. Fibrosis area calculated as a percentage of the total LV area. n=8 for Spp1^WT; n=7 for Spp1^KO. K. TaqMan qRT-PCR analysis of inflammatory genes (Tnfa, Il6, Il1b) n=5 for Spp1^WT; n=7 for Spp1^KO. L. TaqMan qRT-PCR analysis of fibrotic genes (Spp1, Tgfb1, Col3a1, Col4a1, Acta2, Postn). n=4–5 for Spp1^WT; n=7–8 for Spp1^KO.

tVNS-induced Igf1 expression in TLF⁺/MHC2⁺ CRM supports HFpEF improvement

Previous data indicate that TLF⁺/MHC2⁺ CRM limit adverse remodeling following cardiac injury²⁸; this led us to investigate whether tVNS improves the HFpEF phenotype by enhancing TLF⁺/MHC2⁺ CRM function, in addition to reducing pathogenic Spp1-expressing CCR2⁺ CRM. To test this hypothesis, we analyzed DEGs in combined TLF⁺/MHC2⁺ CRM from HFpEF + sham and HFpEF + tVNS mice. Consistent with our previous findings that tVNS suppresses pro-inflammatory gene expression, we found that several pro-inflammatory genes, including Cxcl2, Nfkbiz, and Nlrp3, were downregulated by tVNS (Figure 5A). Remarkably, tVNS also upregulated multiple pro-reparative genes, such as Apoe and Igf1, in TLF⁺/MHC2⁺ CRM compared to HFpEF + sham mice (Figure 5A, B). Given its established role in promoting cardiac adaptation to hypertensive stress⁴⁴, we focused on Igf1.

Figure 5. Igf1 gene expression in TLF⁺ and MHC2⁺ CRM and the effects of macrophage-specific Igf1 deletion on cardiac phenotype in experimental mice.

A. Volcano plot showing differentially expressed genes in combined TLF⁺ and MHC2⁺ CRM after tVNS treatment. B. Violin plot showing Igf1 expression in TLF⁺ and MHC2⁺ CRM. C. Schematic overview of the study design. D. Early diastolic mitral annulus Doppler velocity (e′). n=11 for Igf1^WT; n=9 for Igf1^MKO. E. Ratio of early mitral inflow Doppler velocity to early diastolic mitral annulus velocity (E/e′). n=11 for Igf1^WT; n=9 for Igf1^MKO. F. Representative images of mitral annulus tissue Doppler and mitral inflow Doppler echocardiography G. Heart weight normalized to tibia length. n=11 for Igf1^WT; n=9 for Igf1^MKO. H. Lung weight normalized to body weight. n=10 for Igf1^WT; n=9 for Igf1^MKO. I. Representative histological sections stained with Picrosirius Red. J. Fibrosis area calculated as a percentage of the total LV area. n=8 for Igf1^WT; n=5 for Igf1^MKO. K. TaqMan qRT-PCR analysis of inflammatory genes (Tnfa, Il6, Il1b). n=13–14 for Igf1^WT; n=8–9 for Igf1^MKO. L. TaqMan qRT-PCR analysis of fibrotic genes (Spp1, Tgfb1, Col3a1, Col4a1, Acta2, Postn). n=13–14 for Igf1^WT; n=8–9 for Igf1^MKO.

To test the hypothesis that tVNS ameliorates HFpEF by inducing Igf1 expression, we crossed conditional Igf1^f/f mice with Cx3cr1cre^ERT2 mice, which express a tamoxifen-inducible Cre recombinase under the control of the Cx3cr1 promoter. Before conducting these experiments, we confirmed that Cre expression driven by the Cx3cr1 promoter was specific to TLF⁺/MHC2⁺ CRM by crossing Cx3cr1cre^ERT2 mice to Rosa stop^f/f TdT reporter mice (Supplemental Figure 6), in agreement with previous data ¹². Next, we delivered tamoxifen (6mg/kg, I.P.) for three consecutive days to 8-week-old Igf1^f/f (hereafter referred to as Igf1^WT mice) and Cx3cr1cre^ERT2 Igf1^f/f mice (hereafter referred to as macrophage-specific Igf1 knockout mice; Igf1^MKO). After a three-week recovery, Igf1^WT and Igf1^MKO mice were placed on a HFpEF diet at 11 weeks of age, which was maintained until the end of the study. At 16 weeks, a 4-week tVNS treatment was initiated (Figure 5C). At 20 weeks of age, HFpEF Igf1^MKO mice exhibited a significantly worsened phenotype compared to HFpEF Igf1^WT mice, characterized by worsened diastolic function, increased hypertrophy and lung weight, and exacerbated cardiac fibrosis (Figure 5D–J). In agreement with these findings, qRT-PCR analysis revealed that Igf1^MKO mice had elevated expression of pro-inflammatory (Tnfa, Il6, Il1b) and pro-fibrotic (Spp1, Tgfb1, Col3a1, Col4a1, Acta2, Postn) genes compared to Igf1^WT mice (Figure 5K, L). Overall, these data indicate that tVNS-promoted Igf1 expression in TLF^+/MHC2⁺ CRM contributes to HFpEF amelioration.

Cholinergic signaling mediates the protective effect of tVNS on HFpEF

The effect of tVNS on CRM, including suppression of pro-inflammatory genes (Spp1) and induction of pro-reparative genes (Igf1), suggests a potentially important mechanism in HFpEF amelioration. However, the precise signaling pathway through which tVNS exerts these effects on CRM remains unclear. To investigate how tVNS modulates CRM gene expression, we examined the role of α7 nicotinic acetylcholine receptor (α7nAChR) signaling and its interaction with acetylcholine (ACh)-producing ChAT⁺ CD4⁺ T cells, which have been shown to mediate the anti-inflammatory effect of the vagus nerve through the cholinergic anti-inflammatory pathway ⁴⁵. We thus hypothesized that tVNS regulates CRM function through α7nAChR/ACh signaling mediated by ChAT⁺ CD4⁺ T cells⁴⁵.

Prior to performing these studies, we first confirmed that CRM were the major cardiac cell type expressing the α7nAChR receptor via flow cytometry (Supplemental figure 7A–C). Surprisingly, we also found that α7nAChR expression was restricted to TLF⁺ and CCR2⁺ CRM, with MHC2⁺ CRM having virtually no α7nAChR expression (Supplemental figure 7A–C). To investigate the role of α7nAChR signaling in tVNS-mediated protection, we used methyllycaconitine (MLA), a selective α7nAChR antagonist. HFpEF mice were randomized into three groups: sham, tVNS, and tVNS + MLA (5mg/kg administered intraperitoneally 30 minutes before tVNS)¹⁸ (Figure 6A). Following four weeks of treatment, MLA administration abolished the beneficial effects of tVNS, as tVNS + MLA mice failed to exhibit the improvements observed with tVNS alone. Specifically, MLA-treated mice displayed worsened diastolic function, increased heart and lung weight, and exacerbated cardiac fibrosis, similar to HFpEF + sham group (Figure 6B–H). In agreement with these physiological findings, qRT-PCR analysis revealed that pro-inflammatory (Tnfa, Il6) and pro-fibrotic (Spp1, Tgfb1, Col3a1, Col4a1, Acta2, Postn) gene expression, which was significantly reduced by tVNS, remained elevated in the presence of MLA (Figure 6I, J). These results indicate that tVNS-mediated improvements in HFpEF require intact α7nAChR signaling.

Figure 6. Impact of MLA inhibition of α7nAChR and CD4⁺ T-Cell-specific deletion of ChAT on cardiac phenotype in tVNS mice.

A. Schematic representation of the study design for MLA inhibition of α7nAChR. B. Early diastolic mitral annulus Doppler velocity (e′). n=7 for HFpEF+sham; n=6 for HFpEF+tVNS; n=4 for HFpEF+tVS+MLA. C. The ratio of early mitral inflow Doppler velocity to early diastolic mitral annulus velocity (E/e′). n=7 for HFpEF+sham; n=6 for HFpEF+tVNS; n=4 for HFpEF+tVS+MLA. D. Representative images of mitral annulus tissue Doppler and mitral inflow Doppler echocardiography. E. Heart weight normalized to tibia length. n=9 for HFpEF+sham; n=6 for HFpEF+tVNS; n=4 for HFpEF+tVS+MLA. F. Lung weight normalized to body weight. n=9 for HFpEF+sham; n=6 for HFpEF+tVNS; n=5 for HFpEF+tVS+MLA. G. Representative histological sections stained with Masson’s trichrome. H. Fibrosis area calculated as a percentage of the total LV area. n=9 for HFpEF+sham; n=8 for HFpEF+tVNS; n=4 for HFpEF+tVS+MLA. I. TaqMan qRT-PCR analysis of inflammatory genes (Tnfa, Il6, Il1b). n=5 per group. J. TaqMan qRT-PCR analysis of fibrotic genes (Spp1, Tgfb1, Col3a1, Col4a1, Acta2, Postn). n=5 per group. K. Schematic representation of the study design of CD4⁺ T-Cell-specific deletion of ChAT gene. L. Systolic blood pressure measurements. n=5 for ChAT^KO; n=4 for ChAT^WT. M. Early diastolic mitral annulus Doppler velocity (e′). n=5 for ChAT^KO; n=4 for ChAT^WT. N. The ratio of early mitral inflow Doppler velocity to early diastolic mitral annulus velocity (E/e′). n=5 for ChAT^KO; n=4 for ChAT^WT. O. Representative images of mitral annulus tissue Doppler and mitral inflow Doppler echocardiography. P. Heart weight normalized to tibia length. n=5 for ChAT^KO; n=4 for ChAT^WT. Q. Lung weight normalized to body weight. n=5 for ChAT^KO; n=4 for ChAT^WT. R. Representative histological sections stained with Picrosirius Red. S. Fibrosis area calculated as a percentage of the total LV area. n=5 for ChAT^KO; n=4 for ChAT^WT.

Given that ChAT is the rate-limiting enzyme for ACh synthesis⁴⁵ and is predominantly expressed in CD4⁺ but not CD8⁺ T cells⁴⁶, we also sought to determine whether CD4⁺ T cell-derived ACh is essential for tVNS-mediated protection in HFpEF. To test this, we utilized a CD4⁺ T cell-specific ChAT knockout mouse model (CD4^cre ChAT^f/f; hereafter referred to as ChAT^KO), in which ChAT is constitutively deleted in CD4⁺ T cells ⁴⁶. HFpEF was induced in ChAt^WT and ChAT^KO mice, followed by four weeks of tVNS treatment (Figure 6K). ChAT^KO mice receiving tVNS exhibited significantly worsened BP, diastolic dysfunction, cardiac hypertrophy, lung congestion, and fibrosis compared to ChAT^WT mice receiving tVNS (Figure 6L–S), supporting a critical role for CD4+ T cell-derived ACh in mediating the protective effects of tVNS. Collectively, our results indicate that tVNS improves HFpEF via α7nAChR/ACh signaling, highlighting a key immunomodulatory mechanism underlying its cardioprotective effects.

Discussion

In this study, we demonstrate that tVNS improves HFpEF in a clinically relevant, obese and hypertensive mouse model²³. Using single cell RNA sequencing and genetic knockout mice, we found that the beneficial effect of tVNS on HFpEF was driven by a reduction in monocyte-derived CCR2⁺ CRM. Furthermore, tVNS downregulates Spp1 expression in these cells while simultaneously promoting Igf1 expression in TLF⁺/MHC2⁺ CRM, suggesting a shift towards a reparative macrophage phenotype. Mechanistically, these effects are mediated through cholinergic signaling, highlighting a neuroimmune pathway in HFpEF. Collectively, these findings suggest that targeting CRM through neuromodulation may represent a novel, non-invasive therapeutic strategy for HFpEF, consistent with our recent proof-of-concept data showing that tVNS treatment improved HFpEF-related outcomes in patients ⁴⁷.

Consistent with our results, HFD-related diastolic dysfunction is associated with proinflammatory macrophage accumulation, while macrophage depletion reverses this phenotype ¹⁴. Similarly, CCR2⁺ CRM play a detrimental role in ischemic cardiomyopathy and heart failure, with their loss improving cardiac function ^{27, 36, 48}. In this manuscript, we show for the first time that CCR2⁺ CRM are pathogenic in a clinically relevant mouse model of HFpEF ²³, as their reduction improved cardiac phenotype in our obese and hypertensive HFpEF model.

To explore the mechanism through which CCR2⁺ CRM may be promoting HFpEF, we analyzed their gene expression via scRNA-seq. This analysis demonstrated that Spp1, encoding the matricellular protein osteopontin, was elevated in CCR2⁺ CRM isolated from HFpEF mice (vs controls) and reduced by tVNS, suggesting a functional role in disease progression. Notably, the specificity of CCR2⁺ CRM for Spp1 expression compared to other cardiac cell types was further confirmed by scRNAseq analysis of integrated datasets. In support of our hypothesis, global deletion of Spp1 reduced HFpEF severity, aligning with recent reports implicating Spp1 in the progression of disease in mouse models of atrial fibrillation and heart failure^{49, 50}. These studies suggest that Spp1 promotes disease progression and fibrosis by binding to cognate receptors on fibroblasts, activating myofibroblasts, and driving extracellular matrix production and fibrosis. Our data suggest that CCR2⁺ CRM may be promoting HFpEF by secreting Spp1, promoting fibrotic remodeling.

Beyond reducing CCR2⁺ CRM numbers and Spp1 expression, we also showed that tVNS induced Igf1 expression in TLF⁺/MHC2⁺ CRMs, favoring a reparative macrophage phenotype. Prior studies have shown that CRM-expressed Igf1 is required for adaptive cardiomyocyte growth and protection against hypertension-induced heart failure⁵¹. While our data indicate that loss of TLF⁺/MHC2⁺ CRM (via CSF1R inhibition) did not impact HFpEF phenotype in the absence of tVNS, we found that TLF⁺/MHC2⁺ CRM-specific deletion of Igf1 abolished the protective effects of tVNS. This observation supports the role of Igf1 in mediating tVNS-driven cardiac repair. Given its established role in cardiac hypertrophy⁵¹, the worsened HFpEF phenotype in Igf1^MKO mice receiving tVNS suggests that Igf1 signaling is a critical mediator of the cardioprotective effects of tVNS neuromodulation.

The vagus nerve modulates immune responses via the cholinergic anti-inflammatory pathway, which involves acetylcholine production by CD4⁺ T cells and activation of α7nAchR on macrophages ^{15–17, 45, 46}. Importantly, tVNS for only short periods of time may be sufficient to induce a long-lasting anti-inflammatory and cardioprotective response ^{20, 47}. Our present findings extend these concepts, confirming that CRM are the primary α7nAChR-expressing cells in the heart, and demonstrating that tVNS for 30 minutes daily fails to rescue HFpEF mouse models with pharmacological inhibition of α7nAChR or ChAT deletion in CD4⁺ T cells. Our results also build upon previous evidence showing that tVNS improved diastolic function and LV fibrosis, and suppressed the expression of proinflammatory genes in rats with HFpEF, while this effect was attenuated by α7nAchR blockade ¹⁸. These results were corroborated by our recent pilot randomized clinical trial, which showed that tVNS for one hour daily, improved cardiac function, quality of life and reduced inflammatory cytokines in patients with HFpEF ⁴⁷. Importantly, improvement in cardiac function correlated with a decrease in inflammatory cytokines ⁴⁷. Collectively, our results provide the first evidence that an intact neuroimmune axis plays a central role in HFpEF pathophysiology and mediates the cardioprotective effects of tVNS.

Our findings suggest that tVNS protects against HFpEF by modulating CRM function through cholinergic signaling, suppressing pro-inflammatory pathways while enhancing reparative mechanisms. The reduction in CCR2⁺ CRM accumulation and downregulation of Spp1, along with upregulation of Igf1 in TLF⁺/MHC2⁺ CRMs, indicates that tVNS shifts the macrophage phenotype away from fibrosis-promoting and towards a tissue-reparative state (Figure 7). The dependence of these effects on α7nAChR signaling further supports the importance of the neuroimmune axis in HFpEF. Our results suggest that tVNS acts indirectly through the cholinergic anti-inflammatory pathway, involving the spleen⁴⁵. Consistent with this notion, neuromodulation of an anti-inflammatory neuroimmune pathway involving the splenic nerve, improved phenotype after cardiac injury⁵². Nonetheless, we cannot exclude that the protective effect of tVNS may also be mediated through direct actions on the heart. Notably, direct interaction of vagal efferent fibers with cholinergic myenteric neurons in close contact with muscularis resident α7nAChR-expressing macrophages, independent of splenic T cells, has been shown ⁵³. Moreover, amplification of non-neuronal cholinergic system in the heart has been shown to exert a beneficial immunomodulatory effect in mice following cardiac injury ⁵⁴. Together, our findings reveal a previously unrecognized neuroimmune mechanism in HFpEF, in which tVNS suppresses CCR2⁺ pro-inflammatory CRM activity and enhances Igf1-driven reparative macrophage function via α7nAChR/ACh signaling (Figure 7). This mechanistic insight positions tVNS as a novel and highly translational neuromodulatory strategy for targeting cardiac inflammation and fibrosis in HFpEF.

Figure 7. Pathogenic role of cardiac resident macrophages in heart failure with preserved ejection fraction (HFpEF) and their modulation by transcutaneous vagus nerve stimulation (tVNS).

CCR2⁺ cardiac resident macrophages (CRM) accumulate through recruitment and differentiation of circulating monocytes in HFpEF. Autonomic modulation with tVNS, mediated through the Ach/a7nAChR signaling pathway, reverses the accumulation of CCR2⁺ CRM and improves cardiac phenotype, similar to a CCR2⁺ knockout (CCR2^gfp/gfp) model. tVNS downregulates Spp1 expression in CCR2⁺ CRM, while simultaneously promoting Igf1 expression in TLF⁺/MHC2⁺ CRM, shifting the pro-inflammatory macrophage phenotype towards a reparative state.

Limitations

Our study has several limitations. First, animal models do not recapitulate all aspects of HFpEF, including the complexity of related comorbidities, such as aging and diabetes ⁴. However, for the purposes of this mechanistic study, we used an established, clinically relevant, obese and hypertensive (“two-hit”) mouse model of HFpEF ²³, recognizing that a variety of animal models of HFpEF exist, none of which is perfect ^{25, 55}. Second, assessment of cardiac function in our study was based on echocardiography, rather than on invasive pressure catheters placed retrogradely in the LV via arterial access ²⁵. Nonetheless, mitral inflow and tissue Doppler imaging to quantify diastolic dysfunction is an established method to quantify cardiac function ^{34, 56}. Third, we employed empirical parameters for tVNS, based on our previous experiments ¹⁸. Given that differential effects can be seen by manipulating the stimulation parameters ⁵⁷, we cannot exclude the possibility that optimization of stimulation parameters could have resulted in a more pronounced effect of tVNS ¹⁹. Fourth, we did not examine the effects of tVNS on non-neuronal cholinergic signaling. Therefore, we cannot exclude that the beneficial effects observed might be due to, at least in part, targeting non-neuronal cholinergic signaling. Notably, pyridostigmine-mediated enhancement of non-neuronal cholinergic signaling exerted a beneficial immunomodulatory effect and improved survival in mice following cardiac injury ⁵⁴. Finally, we acknowledge that the Spp1^KO model was global rather than macrophage-specific. Other cell types, including fibroblasts, cardiomyocytes and endothelial cells, may also contribute to Spp1 production. Future studies employing macrophage-targeted genetic models would help clarify the precise role of CRM-derived Spp1 in HFpEF.

Conclusion

In conclusion, this translational study demonstrates that tVNS rescues the HFpEF phenotype by reducing monocyte-derived CCR2⁺ CRM accumulation. Importantly, tVNS further alters the CRM inflammatory profile towards a reparative state, downregulating Spp1 in CCR2 CRM, and upregulating Igf1 in TLF⁺/MHC2⁺ CRM. Furthermore, our study establishes that tVNS exerts its protective effects through ACh/α7nAChR signaling, requiring CD4⁺ T cell-derived ACh for CRM modulation. These findings provide a mechanistic foundation for tVNS as a novel, non-invasive therapeutic strategy targeting the neuroimmune axis in HFpEF.

Novelty and significance.

What is known?

• Cardiac resident macrophages (CRM) play a central role in cardiovascular health and disease.

• Transcutaneous vagal stimulation (tVNS) improves the cardiac phenotype of heart failure with preserved ejection fraction (HFpEF) through its anti-inflammatory and anti-fibrotic effects.

What new information does this article contribute?

• tVNS rescues the HFpEF phenotype in an obese and hypertensive mouse model by reducing monocyte-derived CCR2⁺ CRM, while altering their inflammatory profile toward a pro-reparative state.

• These effects are mediated through the acetylcholine (ACh)/α7 nicotinic ACh receptor (α7nAChR) cholinergic signaling, highlighting a neuroimmune pathway in HFpEF.

• Collectively, these findings suggest that targeting CRM through neuromodulation may represent a novel, non-invasive therapeutic strategy for HFpEF.

CRM are macrophages that reside in the heart prior to injury, and are the most abundant immune cells in the heart. tVNS is a translational, non-invasive intervention that has been shown to exert anti-inflammatory and anti-fibrotic effects in the heart, improving the HFpEF phenotype. However, our mechanistic insights into this process remain limited. Our study demonstrates that tVNS selectively reduces the monocyte-derived CCR2⁺ CRM, while having minimal impact on the abundance of other CRM subtypes. Furthermore, tVNS shifts the inflammatory profile of CRM toward a pro-reparative state, by reducing osteopontin (Spp1) and increasing Igf1 expression. In order for tVNS to excert its effects, an intact ACh/α7nAChR signaling axis is required. That involves – but is likely not limited to – ACh production from ChAT⁺ T cells and functional a7nAchR expression on CRM. Collectively, our findings support the notion that the neuroimmune axis holds a central role in HFpEF and underscore the potential of tVNS as a neuromodulatory treatment strategy to counteract HFpEF-related cardiac remodeling.

Non-standard Abbreviations and Acronyms:

ACh

Acetylcholine

ACK

Ammonium-Chloride-Potassium

Acta2

Alpha-smooth muscle actin

α7nAChR

α7 Neuronal acetylcholine receptor

ANOVA

Analysis of variance

BP

Blood pressure

BSA

Bovine Serum Albumin

CCR2

C-C chemokine receptor type 2

ChAT

Choline acetyltransferase

Col3a1

Collagen 3 alpha 1

Col4a1

Collagen 4 alpha 1

CRM

Cardiac resident macrophages

CSF1

Colony-stimulating factor-1

CSF1-R

CSF1 receptor

CSF1-Ri

CSF1 receptor inhibitor

DAPI

4′,6-diamidino-2-phenylindole

DEGs

Differentially expressed genes

E/e’ ratio

Ratio of early mitral inflow Doppler velocity to early diastolic mitral annulus velocity

e’

Early diastolic mitral annulus velocity

GEMs

Gel Bead-In-EMulsions

GEO

Gene Expression Omnibus

HFD

High-fat diet

HFpEF

Heart failure with preserved ejection fraction

Igf1

Insulin-like growth factor 1

Il1b

Interleukin 1β

Il6

Interleukin 6

L-NAME

L-N^G-nitro arginine methyl ester

LV

Left ventricular

LVEF

Left ventricular ejection fraction

MLA

Methyllycaconitine

PCA

Principal component analysis

PBS

Phosphate Buffered Saline

PFA

Paraformaldehyde

Postn

Periostin

RFP

Red fluorescent protein

RPM

Revolutions per minute

scRNA-seq

Single cell RNA sequencing

Spp1

Osteopontin

Tgfb1

Transforming Growth Factor Beta 1

TLF⁺

Timd4^+/Lyve1⁺/Folr2⁺ (macrophage subpopulation characterized by these markers

Tnfa

Tumor necrosis factor alpha

tVNS

Transcutaneous vagus nerve stimulation

UMAP

Uniform Manifold Approximation and Projection

VNS

Vagus nerve stimulation