Immune profiling of murine cardiac leukocytes identifies triggering receptor expressed on myeloid cells 2 as a novel mediator of hypertensive heart failure

Abstract

Aims

Heart failure with preserved ejection fraction (HFpEF) is characterized by diastolic dysfunction, microvascular dysfunction, and myocardial fibrosis with recent evidence implicating the immune system in orchestrating cardiac remodelling.

Methods and results

Here, we show the mouse model of deoxycorticosterone acetate (DOCA)-salt hypertension induces key elements of HFpEF, including diastolic dysfunction, exercise intolerance, and pulmonary congestion in the setting of preserved ejection fraction. A modified single-cell sequencing approach, cellular indexing of transcriptomes and epitopes by sequencing, of cardiac immune cells reveals an altered abundance and transcriptional signature in multiple cell types, most notably cardiac macrophages. The DOCA-salt model results in differential expression of several known and novel genes in cardiac macrophages, including up-regulation of Trem2, which has been recently implicated in obesity and atherosclerosis. The role of Trem2 in hypertensive heart failure, however, is unknown. We found that mice with genetic deletion of Trem2 exhibit increased cardiac hypertrophy, diastolic dysfunction, renal injury, and decreased cardiac capillary density after DOCA-salt treatment compared to wild-type controls. Moreover, Trem2-deficient macrophages have impaired expression of pro-angiogenic gene programmes and increased expression of pro-inflammatory cytokines. Furthermore, we found that plasma levels of soluble TREM2 are elevated in DOCA-salt treated mice and humans with heart failure.

Conclusions

Together, our data provide an atlas of immunological alterations that can lead to improved diagnostic and therapeutic strategies for HFpEF. We provide our dataset in an easy to explore and freely accessible web application making it a useful resource for the community. Finally, our results suggest a novel cardioprotective role for Trem2 in hypertensive heart failure.

Graphical Abstract

Graphical Abstract.

Time for primary review: 34 days See the editorial comment for this article ‘Triggering receptor expressed on myeloid cells 2 restricts cardiac inflammation and hypertrophy in hypertensive mice’, by M. Delgobo and S. Frantz, https://doi.org/10.1093/cvr/cvad132.

1. Introduction

Heart failure with preserved ejection fraction (HFpEF) is a syndrome of symptoms and signs of heart failure in the setting of left ventricular hypertrophy, diastolic dysfunction, and normal or near-normal ejection fraction.^1,2 HFpEF is a common and highly morbid condition with few approved therapies. Therefore, better understanding of the pathophysiology of HFpEF and development of new therapeutics is a critical unmet need.

Preclinical HFpEF models utilize common comorbidities to induce diastolic dysfunction and heart failure in rodents, including hypertension, obesity, and ageing.^3,4 Hypertension is the most common co-morbidity in HFpEF,^5–7 and mounting evidence implicates immune activation in hypertensive end-organ damage.⁸ Deoxycorticosterone acetate (DOCA)-salt is a hypertension model and has been used to study HFpEF in rats.^3,9–11 In mice, DOCA-salt treatment causes mild hypertension and diastolic dysfunction but has not been extensively characterized as a model of HFpEF.^3,12–14 DOCA-salt induced hypertension has also been shown to have an inflammatory component, as depletion of macrophages or blocking monocyte recruitment is associated with protection from hypertension-induced end-organ damage.^15–17

Increasing evidence suggests HFpEF is associated with a pro-inflammatory state in humans. Pro-inflammatory markers such as interleukin (IL)-6, tumour necrosis factor-alpha, and C-reactive protein are elevated in HFpEF patients during acute decompensation.¹⁸ At the tissue level, endomyocardial biopsy specimens from patients with HFpEF show fibrosis and increased macrophage abundance.¹⁹ Furthermore, high circulating leucocyte counts associate with adverse outcomes in HFpEF.²⁰ Animal models have shown a causal role for immune cell subsets in cardiac dysfunction due to HFpEF-relevant comorbidities.^21–24 Cardiac macrophages contribute to diastolic dysfunction through secretion of pro-fibrotic cytokines in a mouse HFpEF model,²⁵ and blockade of macrophage recruitment prevents diastolic dysfunction in a pressure overload model.²⁶ In addition to macrophages, most major immune cell populations are present in the heart, and a recent study revealed activation of a broad range of immune cell types in a pressure-overload model.²⁷ Adaptive immune cells such as CD4⁺ T helper cells and cytotoxic CD8+ T cells have also been implicated in hypertension and its associated end-organ damage.⁸ However, the extent of immune activation in a hypertensive HFpEF model has not been fully characterized.

Here, we show that the DOCA-salt model in mice recapitulates key elements of human HFpEF including diastolic dysfunction, exercise intolerance, and pulmonary congestion. Using cellular indexing of transcriptomes and epitopes by sequencing (CITE-Seq), we identified that DOCA-salt treatment altered abundance and/or transcriptional programmes in multiple cardiac immune cell types, most notably macrophages. Within cardiac macrophages, we identified several novel genes that were up-regulated in hearts of DOCA-salt treated mice including triggering receptor expressed on myeloid cells 2 (Trem2). Trem2 is a transmembrane receptor whose expression is restricted to monocytes, macrophages, osteoclasts, and microglia. Trem2 regulates cell activation and phagocytosis^28,29 and interacts with a host of potential ligands, many of which are increased during tissue damage.³⁰ Although the role of Trem2 has been most extensively studied in microglia in Alzheimer’s disease, Trem2-expressing macrophages have been identified in adipose tissue and aorta in the context of obesity and atherosclerosis.^31,32 Recent studies have found an increase in Trem2 expressing macrophages after myocardial infarction.^33,34 In addition, a recent study found that Trem2 is important for scavenging dysfunctional mitochondria in sepsis-induced cardiomyopathy.³⁵ Currently, the role of Trem2 in hypertensive HFpEF is unknown.

Using mice with genetic deletion of Trem2, we found that loss of Trem2 leads to exacerbated cardiac hypertrophy and renal dysfunction in DOCA-salt treated mice, as well as loss of pro-angiogenic transcriptional programmes and increased pro-inflammatory cytokine production in isolated macrophages. Moreover, we found that humans with HFpEF have increased plasma levels of soluble TREM2 (sTREM2) compared to healthy controls. Together, our results provide evidence for a novel protective role of Trem2 in cardiac remodelling. Furthermore, our CITE-Seq results in healthy and HFpEF hearts provide an atlas for identification of additional targets that may lead to improved diagnostic and therapeutic strategies for HFpEF.

2. Methods

2.1. Mice

Wild type C57BL/6J (stock no. 000664) and Trem2^−/− (stock no. 027197) mice on a C57BL/6J background were obtained from Jackson Laboratory. For all studies, age-matched male mice were used between 10 and 12 weeks of age. All mice were housed in a pathogen-free environment with 12 h light and dark cycles. All procedures conform to the guidelines from the NIH Guide for the Care and Use of Laboratory Animals.

2.2. DOCA-salt model

To induce DOCA-salt hypertension, uninephrectomy was performed and a DOCA pellet (100 mg; Innovative Research of America, Sarasota, FL, USA) was implanted subcutaneously under anaesthesia. Mice in the control group received a sham procedure where the kidney was externalized but not removed. Anaesthesia was induced with ketamine/xylazine (90–120 mg/kg + 10 mg/kg; 1:1 volume) through intraperitoneal injection. In addition, for the DOCA-salt treatment group, the drinking water was supplemented with 1% NaCl for 21 days after uninephrectomy. Mice were euthanized by CO₂ inhalation at the conclusion of experiments.

2.3. BP measurement

BP was measured non-invasively using tail cuff plethysmography (Hatteras Instruments, North Carolina, USA) twice a week. Prior to baseline measurements, mice were trained for 2 days to acclimate them to the procedure. On the day of recording, the mice were restrained and allowed to acclimate for 5 min prior to recording. A set of 10 measurements was taken and then discarded. Two additional sets of 10 measurements were recorded and averaged for each day. Two days of tail cuff were performed each week to obtain a weekly average for each mouse.

2.4. Urine collection and albuminuria measurements

Urine was collected in metabolic cages overnight during a 16–18 h collection period. Urine albumin and creatine were measured using Albuwell M and Creatinine Companion (Ethos Biosciences, Logan Township, NJ, USA), respectively. Albumin values were then normalized to creatinine values.

2.5. Exercise testing

Mice were acclimatized to a single-lane treadmill for 10 m at 10 m/min the day prior to testing. Mice were tested by beginning at 10 m/min and increasing the speed by 4 m/min every 3 min until exhaustion. Exhaustion is defined as the mouse sitting on the shock pad for 5 consecutive seconds. Time to exhaustion and total running distance were recorded.

2.6. Lung wet-to-dry ratio

Lungs were excised immediately after euthanasia and prior to perfusion. Lungs were weighed prior to drying at 65 °C for 1 h in an oven. Lungs were then weighed again an hour later to confirm complete drying.

2.7. Echocardiography

Echocardiography was performed in conscious mice with the VisualSonics Vevo2100 platform and MS-400 transducer. Images were acquired using 2-D B-mode and M-mode. Left ventricular dimensions were measured from M-mode in parasternal short axis and used to calculate echocardiography parameters using VevoLab. The echocardiographer was blinded to animal groups and genotypes during acquisition. Two independent persons analysed the echocardiography data in a blinded fashion and their values were averaged.

2.8. Invasive haemodynamic measurements

To perform haemodynamic measures, mice were anaesthetized with 2–3% isoflurane and orotracheally intubated with a 22-gauge catheter. Animals were mechanically ventilated with vapourized isoflurane general anaesthesia. After positioning the mice supine on a heated operating table, an abdominal incision was made. Cautery was used to incise the diaphragm and expose the heart. A 1.4 French Mikro-tip catheter was inserted into the left ventricle to obtain pressure and volume measurements which were continuously recorded with a Millar MPVS-300 unit coupled to a Powerlab 8-SP analogue-to-digital converter acquired at 1000 Hz. Mice were then sacrificed and data were analysed in LabChart using the PV Loop analysis tool.

2.9. Flow cytometry

Single-cell suspensions of left ventricles were prepared. Left ventricles were isolated from perfused hearts and further rinsed in PBS. Left ventricles were mechanically minced using fine scissors and then enzymatically digested in RPMI 1640 media containing Liberase TM (0.12 mg/mL) for 30 min at 37 degrees. After digestion, suspensions were filtered through a 40 µm cell strainer. For viability staining, Live/Dead Fixable Violet Dead Cell Stain (Life Technologies) was used and the following surface antibodies CD45-BV421, CD64-PE/Dazzle594, CD11b-BV510, I-A/-E-BV711, Ly6C-BV785, CD45-BV650, CD31-AF594, PDPN-BV421, and ICAM1-APC Fire750. All antibodies are from Biolegend (San Diego, CA, USA). Flow cytometry was performed on BD FACSCanto II or Cytek Aurora systems and analysed using FlowJo software. Gates were set for positive antibody staining according to fluorescence minus one control.

2.10. CD11b + cell isolation

For isolation of cardiac CD11b + cells, a single-cell suspension was prepared as described above. Cells were then incubated with CD11b + microbeads (Miltenyi Biotec, 130–049-601, Germany) according to manufacturer’s instructions. Cells were then washed and passed through LS columns (Miltenyi) twice for positive selection. The resulting positive fraction of cells was then spun down and lysed for downstream RNA isolation.

2.11. Histology

Mice were perfused with 10% neutral buffered formalin. Hearts were embedded in paraffin and cut into 5 µm cross sections. For collagen staining, slides were stained with Masson’s Trichrome, scanned using a Leica SCN400 slide scanner, and quantified in ImageJ from at least three non-serial sections per mouse. Quantification of LV fibrosis was determined by total area positive determined by colour threshold divided by total tissue area. To quantify cardiomyocyte cross-sectional area, H&E scanned images were analysed in ImageJ where rounded myocytes were visually identified and measured. For each mouse, a minimum of 125 myocytes were measured in a blinded fashion from at least four different locations and the median taken for statistical analysis between groups. To assess capillary density, CD31 staining was performed by the Vanderbilt Translational Pathology Shared Resource Core. For each mouse, six to eight images were processed and diaminobenzidine staining quantified through colour deconvolution and threshold analysis in ImageJ. The threshold value was obtained by averaging the automatic threshold from six different mice.

2.12. RNAscope

RNAscope was performed according to the manufacturer’s instructions using the following probes and reagents: Trem2-C1 (catalogue no. 404111), Cd68-C2 (catalogue no. 316611-C2), and Multiplex v2 kit (catalogue no. 323100). Briefly, tissue was fixed in 10% neutral-buffered formalin at room temperature for 24–48 h and then embedded and sectioned. Slides were deparaffinized and dried prior to quenching endogenous peroxidase activity. Target retrieval was performed in ACDBio (Newark, CA, USA) RNAscope 1 × Target Retrieval Reagent at 99–102 °C for 15 min. We applied RNAscope Protease Plus for 15 min at room temperature and proceeded to run the RNAscope assay with TSA Plus Cy3 or Cy5 (Akoya Biosciences, Marlborough, MA, USA) reagents for fluorescent detection. Afterward, tissue was exposed to DAPI for 30 s, mounted in ProLong Gold and dried overnight. Images were acquired on an Olympus IX81 fluorescent microscope. Post-processing was applied using ImageJ and quantification of RNAscope puncta was performed in CellProfiler using default parameters. A minimum of five images were quantified per cardiac section to obtain a median value for each mouse.

2.13. Bone marrow-derived macrophages

Bone marrow was harvested and BMDMs were differentiated for 6–7 days in RPMI (Roswell Park Memorial Institute) with L-glutamine and 10% serum containing media supplemented with 10 ng/mL M-CSF (Macrophage colony stimulating factor, Peprotech, Cranbury, NJ, USA) prior to plating for experiments. BMDMs were plated at equal densities and treated with 5 ng/mL LPS (eBioscience, San Diengo, CA, USA) for 24 h for activation experiments.

2.14. mRNA measurement

mRNA was isolated using TRIzol reagent (ThermoFisher, Waltham, MA, USA) from flash frozen left ventricle samples according to the manufacturers’ instructions after tissue homogenization. mRNA was isolated from BMDMs using the Qiagen RNeasy Mini Kit. Reverse transcription was performed using the iScript cDNA synthesis kit (Bio-Rad). Quantitative PCR was performed using either TaqMan universal PCR master mix (ThermoFisher) or SYBR Green PCR master mix (Applied Biosystems, Waltham, MA, USA) and primers listed in the Supplementary material online. Samples were run in duplicate using Gapdh primers as internal control for tissues and Rplp0 as control for BMDMs. Relative mRNA expression was calculated using the 2^−(ΔΔCt) method.

2.15. Single-cell sample processing

Immune cells were enriched from single-cell suspensions isolated from left ventricles using CD45 microbeads and MACS magnetic separators (Miltenyi Biotec). Cells were stained for viability with LIVE/DEAD violet fixable dead cell staining kit (Life Technologies, Waltham, MA, USA), an AlexaFluor488-conjugated CD45 antibody for sorting (Biolegend) and Total-SeqC oligo-conjugated antibodies towards CD3, CD4, CD8a, CD19, CD11b, CD11c, CD64, NK1.1, TCRyd, Ly6G, PD-1, CD44, Galectin-3, CD80, CD39, TIGIT, and CCR7 (Biolegend) according to manufacturer’s instructions. Total-SeqC Hashtag 1–4 (Biolegend) were also included in the staining cocktail to identify each biological replicate. Viable immune cells (CD45+) were sorted on a FACSAria III system. Cells were then loaded onto the 10 × Genomics Chromium system for droplet-based encapsulation. Separate libraries were prepared for feature barcodes, 5′-enriched VDJ sequences, and mRNA library. Sequencing was performed on the Illumina NovaSeq6000.

2.16. Single-cell data analysis

Data were processed to generate expression matrices using the 10 × CellRanger (version 3.1.0) pipeline. Reads were aligned using mm10–3.0.0 transcriptome. Single-cell data were analysed using the Seurat package version 4 in R.³⁶ Standard pre-processing steps for quality control were performed. Cells expressing fewer than 200 representing low-quality cells and cells expressing more than 3000 genes, which are likely doublets, were excluded. Additionally, cells with greater than 20% of reads being mitochondrial indicative of a dying cell were excluded. Genes detected in less than five cells per sample were excluded. Samples were demultiplexed within the Seurat package. Cells categorized as doublets (two hashing antibodies) or negative (no hashing antibody) were excluded from downstream analyses. Remaining high-quality singlets were processed according to recommended Seurat pipeline using the default global-scaling normalization method, identification of variable features, and scaling of data. Variable features were used to perform principal components analysis for both the RNA and antibody data. Clustering was performed in Seurat using the FindMultiModalNeighbors() and FindClusters() functions to construct a weighted nearest-neighbours UMAP visualization based on both the RNA transcriptome and the protein level antibody data. Subclustering was performed by identifying a cell type(s) of interest and re-running the pipeline to identify variable genes, perform principal component analysis and cluster the cells. Graphs were created using the additional packages ggplot2, ggrepel, and dplyr in R.

2.17. Pathway enrichment analysis

Pathway analysis was performed using Metascape’s default parameters using significantly differentially expressed genes with adjusted P value less than 0.05 as the input sorted by P-value.⁷⁸

2.18. Bulk RNA-sequencing

Available data for comparison of WT and Trem2-deficient BMDMs were downloaded from GEO accession number GSE98563. Only normal M-CSF concentrations without LPS treatment samples were included (GSM2599743-46, GSM2599759-62). Differential expression analysis was performed using default parameters in DESeq2.⁷⁹ Z-scores of normalized counts were then plotted in Prism.

2.19. VEGF-A measurements

VEGF-A was measured in the supernatant of BMDMs using the mouse VEGF Quantikine ELISA kit (R&D Systems, Cat no. MMV00).

2.20. sTREM2 measurements in human plasma

Plasma was obtained from healthy individuals at an outpatient clinical research site and from acutely hospitalized patients within 24 h of admission for acute heart failure (IRB191468). The investigation conforms to the principles outlined in the Declaration of Helsinki. Inclusion criteria for HFpEF subjects included: age > 35 years, able to provide informed consent, physician confirmed diagnosis of HFpEF with corroborating evidence of current or prior left ventricular ejection fraction > 50%, current or prior symptomatic heart failure (New York Heart Association class II or greater), and current or prior evidence of elevated left ventricular filling pressures by any of the following: LV end diastolic pressure (EDP) or pulmonary capillary wedge pressure (PCWP) >12 mmHg, brain natriuretic peptide (BNP) >100 pg/mL, echocardiographic E/e′ > 15, chest X-ray showing volume overload, or recent or planned administration of IV diuretics. Inclusion criteria for healthy subjects included age >35 years, able to provide informed consent, no known significant cardiovascular disease, including but not limited to coronary heart disease, heart failure, valvular stenosis or moderate valvular regurgitation, arrhythmias, pericardial disease, stroke, and/or peripheral arterial disease. Exclusion criteria for all subjects included active unstable angina or acute coronary syndrome, heart failure due to pericardial disease or infiltrative disease (e.g. amyloid, sarcoid), heart failure due to right ventricular failure and/or pulmonary arterial hypertension alone, estimated glomerular filtration rate < 15 mL/min/1.73 m² and/or on dialysis, active pregnancy, prior organ transplant, recent use of systemic steroids or other immunosuppressant, chronic daily non-steroidal anti-inflammatory drugs (NSAIDs), presence of active infection, autoimmune or rheumatologic disorders, and known diagnosis of HIV. Patients were screened by a physician prior to informed consent and sample collection. Peripheral blood was centrifuged to collect plasma which was then stored at −80 degrees. sTREM2 levels were measured using Human TREM2 ELISA kit (abcam, ab224881, Walham, MA, USA) according to manufacturer’s instructions at 1:25 dilution.

2.21. sTREM2 measurements in mice

sTREM2 measurements were made in plasma isolated from mice upon sacrifice and was adapted from Zhong et al.⁸⁰ Briefly, the ELISA was performed by coating a 96-well plate with a TREM2 capture antibody (R&D Systems, MAB17291–199, Minneapolis, MN, USA). The plate was blocked with 1% BSA in PBS, washed three times, and then incubated with samples overnight at 4 °. Recombinant mouse Trem2 (R&D Systems, 1729-T2) was reconstituted according to manufacturer’s instructions and used to create a standard curve. A mouse Trem2 biotinylated antibody (R&D Systems, BAF1729) was used in conjunction with streptavidin-HRP for detection. Plasma from Trem2 deficient mice showed no signal in our mouse ELISAs, confirming specificity of the assay.

2.22. Single-cell ShinyApp

A web-based ShinyApp was created in R through modification of previously published code⁸¹ based on the R package ShinyCell⁸² to facilitate distribution and further interrogation of our dataset. The data can be accessed at the following url: https://madhurlab.shinyapps.io/MouseHFpEF/.

2.23. Statistics

All data are presented as mean ± standard error of the mean (SEM) or as a violin plot. Data were analysed in GraphPad Prism using parametric or non-parametric tests as appropriate. For comparisons of multiple groups, one-way or two-way analysis of variance (ANOVA) was used with post-tests to obtain multiple comparison corrected P values. Single-cell sequencing data were analysed using a Wilcoxon rank sum test with default parameters in Seurat, including multiple comparison testing corrections. Specific tests stated in legends. P-values < 0.05 were considered significant and significance for all figures determined as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

2.24. Study approval

Studies involving animals were performed in accordance with protocols approved by the IACUC of Vanderbilt University Medical Center. Studies involving humans were in accordance with protocols approved by the IRB (#191468) of Vanderbilt University Medical Center.

3. Results

DOCA-salt recapitulates key features of human HFpEF. To determine whether the DOCA-salt model is an acceptable HFpEF model, we subjected 10–12-week-old male C57BL/6J wild-type (WT) mice to unilateral nephrectomy, implantation of a DOCA pellet, and 1% salt supplementation in the drinking water for 3 weeks. Sham mice received a sham procedure and normal drinking water. As expected, DOCA-salt treated mice exhibit increased blood pressure (BP) (see Supplementary material online, Figure S1). In addition, DOCA-salt treated mice develop cardiac hypertrophy as evidenced by increased heart weight normalized to tibia length (Figure 1A) and glomerular damage with a significant increase in urinary albumin to creatinine ratio (Figure 1B). To evaluate cardiac function, we performed echocardiography, invasive haemodynamics, and exercise testing. On echocardiography, DOCA-salt treated mice display a preserved ejection fraction with no significant change in fractional shortening or heart rate (Figure 1C). As previously demonstrated,¹² invasive haemodynamics revealed an increase in left ventricular end-diastolic pressure and the relaxation constant Tau in DOCA-salt treated mice, consistent with diastolic dysfunction (Figure 1D). When challenged with a treadmill stress test, DOCA-salt treated mice display a shorter time to exhaustion (Figure 1E). In agreement with elevated left-sided cardiac pressures, DOCA-salt treated mice also display greater wet-to-dry lung weight ratio, a measure of pulmonary congestion (Figure 1F). Histology reveals increased cardiac fibrosis in hearts from DOCA-salt treated mice, particularly in the perivascular regions (Figure 1G). Quantification of cardiac capillary density was performed by staining for the endothelial cell marker CD31 which revealed a significant decrease in CD31 + area in DOCA-salt treated mice (see Supplementary material online, Figure S2). Additionally, DOCA-salt treated mice have elevated gene expression of the heart failure markers atrial natriuretic peptide (Nppa) and brain natriuretic peptide (Nppb) in left ventricular tissue (Figure 1H). We also observe increases in left ventricular gene expression of cytokines Il1b, Il6, and Il10 and no significant change in Tnf or Tgfb1 (Figure 1H). Together, these data suggest the DOCA-salt mouse model recapitulates key features of human HFpEF in mice.

Figure 1.

DOCA-salt model recapitulates key features of HFpEF. (A) Heart weight normalized to tibia length (HW/TL, n = 10–15). (B) Urine albumin to creatinine ratios (n = 6–10). (C) Echocardiographic measures of heart rate (HR), ejection fraction (EF), and fractional shortening (FS) after 3 weeks in sham controls and DOCA-salt treated animals (n = 6–7). (D) Invasive haemodynamic measures of the diastolic relaxation constant Tau (τ) and end-diastolic pressures (EDP, n = 5–8). (E) Exhaustion time during treadmill exercise stress test (n = 7–8). (F) Ratio of lung wet weight to dry weight indicative of pulmonary congestion (n = 7–8). (G) Representative images of Masson’s trichrome staining demonstrating collagen deposition and quantification of positive staining as total left ventricle (LV) fibrosis per total tissue area per section (n = 7). (H) Relative gene expression in left ventricles of sham and DOCA-salt treated mice normalized to Gapdh (n = 5–9). Data are expressed as mean ± SEM and analysed by Mann–Whitney U test. *P < 0.05, **P < 0.01, ***P < 0.001.

CITE-seq of immune cells in healthy and HFpEF murine hearts reveals broad population changes. To deeply phenotype the cardiac immune landscape in healthy and HFpEF murine hearts, we performed CITE-seq, a multimodal single-cell RNA-sequencing protocol which enables simultaneous identification of surface proteins using oligonucleotide tagged antibodies to extracellular antigens and single-cell messenger RNA (mRNA) sequencing. Immune cells were isolated from saline perfused left ventricles of four sham and four DOCA-salt treated mice using magnetic selection and stained with hashtags and extracellular antibodies. Biological replicates were pooled and sorted using fluorescence activated cell sorting to obtain live CD45 + cells and loaded onto the 10× chromium controller (Figure 2A). A total of 7600 and 4359 cells were sequenced from hearts of DOCA-salt and sham treated mice, respectively, and analysed using unbiased computational tools in Seurat (Figure 2B). Weighted nearest neighbours dimensional reduction, using both the RNA transcriptome and protein tags,³⁶ and population annotation was performed using cluster marking genes and protein expression of lineage markers (see Supplementary material online, Tables S1 and S2; Figure 2C and D). Cell hashing allowed for separation of individual mice within the same experimental group (see Supplementary material online, Figure S3). We observed concordance between levels of RNA transcripts and the corresponding antibody directed towards its protein product (see Supplementary material online, Figure S3). Almost all major immune lineages were identified in our dataset. Compared to sham controls, we observed a significant increase in percent of neutrophils, γδ T cells, monocytes, dendritic cells, and CD8+ T cells with a decrease in frequency of B cell clusters in hearts of DOCA-salt treated mice relative to sham controls (Figure 2E). Thus, unbiased cardiac immunophenotyping identifies a wide range of immune cells and reveals changes in subset abundance associated with hypertensive remodelling. For further study, we provide our dataset in an open access, interactive web portal (https://madhurlab.shinyapps.io/MouseHFpEF/), making it a useful resource for the community.

Figure 2.

CITE-seq of immune cells in healthy and HFpEF hearts reveals broad population changes. (A) Schematic created with Biorender representing hashing and staining of cardiac leucocytes from four sham and four DOCA-salt treated mice. (B) Weighted nearest neighbours uniform manifold approximation and projection (UMAP) plot based on RNA transcriptomes and protein markers displaying clusters of immune cells sequenced from all samples. (C) RNA expression of lineage marking genes supporting cluster identification. (D) Protein expression of antibody panel used in CITE-seq analysis. (E) Log-fold difference of immune cell subsets as percent of CD45 + immune cells using scProp. Clusters significantly different based on a false discovery rate < 0.05 are highlighted in red and denoted by dashed line.

T-cell abundance and transcriptional programmes are altered in HFpEF. T-cell-derived cytokines contribute to mouse models of hypertension and associated end-organ damage.⁸ To investigate potential changes in T-cell phenotypes, we performed clustering of CD3+ T cells (CD4, CD8, and γδ T cells) to identify novel alterations in cardiac T cells in HFpEF. Five different T-cell clusters were identified (Figure 3A and B). The frequencies of γδ T cells and Cxcr3-expressing T cells were increased in DOCA-salt treated mice with a concurrent decrease in CD4 cells lacking markers of naive cells such as Tcf7 and Igfbp4³⁷ (Figure 3C; see Supplementary material online, Table S3). Expression of Cxcr3 has been shown to mark interferon gamma (IFN-γ) producing T cells,³⁸ which promote hypertension.^39–41 γδ T cells were characterized by expression of IL-17A related genes, including the transcription factors Rorc and Sox13, and cytokine receptors Il1r1, Il17re and Il23r (Figure 3D). Differential expression analysis of cardiac T cells revealed modest changes with relatively few genes being globally altered by DOCA-salt treatment (Figure 3E; see Supplementary material online, Table S4). We did, however, observe a significant increase in Tmem173 in most T-cell subsets (Figure 3F), which encodes STING1 and drives type 1 interferon production. Together, these data suggest T-cell cytokine production, particularly IL-17A, IFNγ, and type I interferons, is increased in hearts of HFpEF mice.

Figure 3.

Altered T-cell abundance and transcriptional landscape in HFpEF. (A) UMAP representation of cardiac T-cell subsets. (B) RNA expression of markers differentiating T-cell populations. (C) Log-fold difference of T-cell subsets as a percent of CD3 + cells using scProp. Sub-clusters significantly different based on a false discovery rate < 0.05 are highlighted in red. (D) RNA expression of selected T-cell polarization and activation related genes by cluster. (E) Volcano plot of differentially expressed genes in all T-cell subsets highlighting genes up-regulated in the DOCA-salt treated group (red) vs. genes enriched in the Sham group (blue). Horizontal dotted lines represent adjusted P value of 0.05, and vertical dotted lines represent average log₂-fold change of 0.25. (F) Violin plot of Tmem173 expression split by cluster and experimental condition.

Myeloid cell abundance and transcriptional programmes are altered in HFpEF. We then interrogated the myeloid compartment to better phenotype subclusters of myeloid cells. We identified six macrophage subclusters, three monocytes subclusters, one dendritic cell cluster, and one neutrophil cluster (Figure 4A). All macrophages were characterized by high expression of CD64 protein expression (Figure 4B). Consistent with prior reports,^42,43 a resident macrophage population was identified expressing Timd4, Lyve1, and Folr2 (TLF + macs) (Figure 4B; see Supplementary material online, Table S5). We observed a cluster of macrophages with high expression of the chemokine receptor CCR2 and the pro-fibrotic mediator osteopontin (Spp1), similar to monocyte-derived macrophage populations described by others.^44,45 We also observed a population with similar transcriptional profiles as previously described major histocompatibility complex (MHC)-II^hi macrophages⁴⁶ that lacked expression of defining markers for TLF + and CCR2 + macrophages, which we term Lilra5 + macrophages (see Supplementary material online, Figure S4). This population is characterized by expression of Lilra5, Cd14, and St3Gal6 as well as higher levels of some but not all MHC-II genes (H2-Aa, H2-Eb1). Finally, we observed macrophages actively expressing cell cycle markers (Ccnb2) as well as two populations characterized by high and low expression of Mrc1, which encodes the alternative macrophage marker CD206 (Cd206hi and Cd206lo macs).

Figure 4.

Altered myeloid cell abundance and transcriptional landscape in HFpEF. (A) UMAP representation of myeloid cell subsets along with a UMAP of cardiac macrophages split by experimental condition. (B) RNA expression of markers differentiating myeloid populations. (C) Log-fold difference of myeloid cell subsets as a percent of CD11b + cells using scProp. Sub-clusters significantly different based on a false discovery rate < 0.05 are highlighted in red and denoted by dashed line. (D) Volcano plot of differentially expressed genes in all macrophage subsets highlighting genes up-regulated in the DOCA-salt treated group (red) vs. genes enriched in the Sham group (blue). Horizontal dotted lines represent adjusted P value of 0.05, and vertical dotted lines represent average log₂-fold change of 0.25. (E) Up-regulated pathways in cardiac macrophages from DOCA-salt treated mice. (F) Overlaid expression of selected genes onto UMAP representation of macrophage clusters seen in (A), split by experimental condition.

We then quantified the relative abundance of myeloid populations and observed an increased frequency of neutrophils, CCR2 + macrophages, cycling macrophages, and Ly6C + monocytes and a decreased frequency of Lilra5 + macrophages and non-classical monocytes in DOCA-salt treated mice compared to Sham controls (Figure 4C). Differential gene expression of all cardiac macrophages (Figure 4D; see Supplementary material online, Table S6) showed increases in genes known to associate with cardiac remodelling (Ccr2, Lgals3, Spp1, Apoe) and potentially novel genes (Trem2, Lgals1). Pathway analysis of up-regulated genes yielded significant enrichment of terms associated with inflammation, cell activation, angiogenesis, and prostaglandin synthesis (Figure 4E). Some genes (Trem2, Lgals1) were globally up-regulated across multiple macrophage subsets, while others such as Ccr2 showed some cluster specificity (Figure 4F). In monocytes, we observed up-regulation of genes related to both activation and immunoregulation (Ly6c2, Pirb, Wfdc17, Chil3) (see Supplementary material online, Figure S5). Intriguingly, Chil3 expression has been tied to immunoregulatory monocytes,⁴⁷ and Wfdc17 has been suggested as a marker of monocyte derived suppressor cells and also monocyte activation.^48,49 We mined available single-cell sequencing data⁵⁰ to determine if similar transcriptional programmes were activated in acute cardiac injury. Indeed, we found up-regulation of monocyte and macrophage Trem2, Ccr2, Spp1, Lgals1, Lgals3 expression post-myocardial infarction (see Supplementary material online, Figure S6), suggesting potential shared mechanisms in distinct forms of cardiac remodelling.

Using flow cytometry in an independent cohort of sham and DOCA-salt treated mice, we observed a significant shift in cardiac macrophage populations by flow cytometry with loss of MHC-II^high macrophages and expansion of MHC-II^low macrophages (see Supplementary material online, Figure S7). MHC-II^low macrophages have been reported to associate with remodelling^51,52 and are consistent with decreased macrophage expression of MHC-II genes (H2-Eb1, H2-Aa) in our CITE-seq data (Figure 4D). We were also able to reproduce significant increases in absolute numbers of neutrophils and Ly6C + monocytes via flow cytometry in independent mice to support the validity of the CITE-seq dataset (see Supplementary material online, Figure S7). Together these data suggest shifts not only in abundance of myeloid cell subsets but also global shifts in cell state across multiple subsets.

Cardiac macrophage Trem2 is increased after DOCA-salt treatment. Intriguingly, the transcriptional response of cardiac macrophages is similar to the gene signature described in ‘disease associated microglia’, with up-regulation of common genes (Apoe, Spp1, Ctsd, Ctsl, Trem2).⁵³ This signature is dependent in part on Trem2, which led us to hypothesize that Trem2 signaling might drive macrophage activation in our model. We hypothesized that loss of Trem2 signaling would prevent up-regulation of this disease associated gene signature and impact cardiovascular function in our model, thereby determining if this gene signature is beneficial or deleterious for the heart. Of note, Trem2 expression was not restricted to a specific subset but rather was seen in multiple subsets including recruited CCR2 + macrophages, resident TLF + macrophages, and to a lesser degree in Ly6C + monocytes (see Supplementary material online, Figure S8). We validated a significant increase in Trem2 gene expression in left ventricles in an independent cohort of DOCA-salt treated mice (Figure 5A). We also observed in an increase in plasma sTrem2 (Figure 5B), which has been reported to correlate with Trem2 + macrophage infiltration.⁵⁴ To validate macrophage localization of Trem2, we performed RNAscope using Trem2 and Cd68 probes. We found increased total Trem2 signal per region and on a per macrophage basis (Figure 5C). Importantly, Trem2 was not detected outside of Cd68 expressing cells, consistent with an expression restricted to macrophages in the heart. In a large available human single-cell database, Trem2 expression is also restricted to myeloid cells and only detected in macrophages within the heart.⁵⁵

Figure 5.

Cardiac macrophage Trem2 is increased after DOCA-salt treatment. (A) Left ventricle gene expression of Trem2 in Sham and DOCA-salt treated mice. (B) Plasma sTrem2 levels measured by ELISA in sham and DOCA-salt treated mice. (C) RNAscope images of cardiac sections using probes targeted towards Cd68 (macrophage marker) and Trem2 (left). Quantification of median Trem2 signal per high-powered field and median Trem2 signal per cardiac macrophage using CellProfiler (right). A total of five to eight images were analysed per mouse to obtain median values. Scale bar 25 microns. Inset enlarged for visualization. Data are expressed as mean ± SEM and analysed by Mann–Whitney U test. *P < 0.05, **P < 0.01.

Loss of Trem2 exacerbates cardiac hypertrophy and renal injury in DOCA-salt treated mice. As Trem2 is important for transcriptional responses to tissue damage,⁵⁶ we hypothesized that loss of Trem2 would alter cardiac remodelling in response to DOCA-salt. To test a causal role for Trem2 in HFpEF, we subjected Trem2^+/+ and Trem2^−/− mice to DOCA-salt treatment. We performed urine collection, exercise testing, and echocardiography at baseline and after HFpEF development. Trem2^−/− mice display a concentric hypertrophic response to DOCA-salt treatment without evidence of wall thinning or dilation (Figure 6A). BP was measured by tail cuff plethysmography and did not differ significantly between Trem2^+/+ and Trem2^−/− mice at baseline or in later stages of DOCA-salt treatment, but Trem2^−/− mice had an exaggerated hypertensive response to DOCA-salt treatment at week one (Figure 6B). Trem2-deficient animals exhibit exacerbated cardiac hypertrophy as measured by heart weight to body weight ratios and myocyte cross sectional area (Figure 6C and D). No difference was observed in heart weight to body weight of untreated animals (Figure 6C). Invasive haemodynamic measurements of left ventricular function revealed impaired diastolic function in Trem2^−/− DOCA-salt treated mice compared to Trem2^+/+ DOCA-salt treated controls, as evidenced by an increase in the diastolic relaxation constant Tau (Figure 6E). Trem2^−/− mice also exhibit a greater degree of glomerular injury, as indicated by increased albumin to creatinine ratio after DOCA-salt treatment (Figure 6F). A time course revealed albuminuria was well established by week two of DOCA-salt treatment, which was higher in Trem2 deficient animals (see Supplementary material online, Figure S9). We did not observe changes in cardiac fibrosis, exercise ability, or pulmonary congestion (see Supplementary material online, Figure S10). We found that female mice lacking Trem2 also exhibit significantly increased cardiac hypertrophy and diastolic dysfunction (see Supplementary material online, Figure S11). Similar to Trem2^+/+ mice, Trem2^−/− mice exhibit no change in ejection fraction or fractional shortening after DOCA-salt treatment (see Supplementary material online, Figure S12).

Figure 6.

Loss of Trem2 exacerbates cardiac hypertrophy and renal injury in DOCA-salt treated mice. (A) Representative H&E staining of Trem2^+/+ and Trem2^−/− mice after DOCA-salt treatment showing concentric hypertrophic response, scale bar 500 microns. (B) Systolic BP as measured by tail cuff plethysmography (n = 11–12 per group). (C) Heart weight (HW) normalized to body weight (BW) of Trem2^+/+ and Trem2^−/− mice after DOCA-salt treatment and in age-matched unmanipulated mice (untreated). Data from three independent cohorts of mice (n = 6–9). (D) Violin plot of cross-sectional area (CSA) of cardiomyocytes after sham or DOCA-salt treatment. Each dot represents an individual measurement, and each violin represents a mouse from two independent cohorts. The median value for each mouse is quantified on the right (n = 3–6). (E) Invasive haemodynamic measures of the diastolic relaxation constant Tau (τ) (n = 3–4). (F) Urine albumin to creatinine ratios after a 16-h overnight urine collection (n = 6–10). Data are expressed as mean ± SEM and analysed by two-way ANOVA with repeated measures and Sidak’s multiple comparison tests (B), two-way ANOVA with Tukey’s post-test (C, D, F), Mann–Whitney U test (E). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Loss of Trem2 increases pro-inflammatory cytokine production and decreases pro-angiogenic gene expression in macrophages. Consistent with prior reports,⁵⁷ osteopontin and galectin-3 were both reduced in Trem2^−/− bone marrow-derived macrophages (BMDMs) with no observed change in galectin-1 expression (Figure 7A). Given that Il6 and Il10 are elevated in DOCA-salt treated hearts, we examined macrophage expression of Il6 and Il10 levels in resting macrophages and after lipopolysaccharide (LPS) activation. Loss of Trem2 leads to higher expression of Il6 and Il10 after LPS activation, in line with previous studies⁵⁸ (Figure 7B). To further evaluate why Trem2 deficiency leads to exacerbated end-organ damage after DOCA-salt treatment, we mined existing bulk RNA-sequencing data from Trem2^+/+ and Trem2^−/− BMDMs⁵⁷ and performed pathway analysis. Pathways down-regulated in Trem2^−/− macrophages include collagen degradation, regulation of cytokine production, enzyme-linked receptor protein signaling, response to wounding, and heart development as top terms (Figure 7C). Many of these are parent terms related to vascular biology such as ‘angiogenesis’, ‘blood vessel development’, ‘regulation of vascular permeability’, and ‘cardiac cell development’ (see Supplementary material online, Table S7). Enrichment of these terms was driven by decreased expression of genes such as Vegfa, Pdgfa, Pdgfb, Mmp12, and Mmp13 in Trem2^−/− macrophages (Figure 7D). We validated decreased expression of Vegfa, Mmp12, and Mmp13 in Trem2-deficient BMDMs by quantitative reverse transcriptase-polymerase chain reaction qRT–PCR (Figure 7E). We also validated that Trem2^−/− macrophages secrete less vascular endothelial growth factor-A (VEGF-A) in the supernatant by enzyme-linked immunosorbent assay (ELISA) (Figure 7F). To examine the translatability of these findings to our in vivo model, we isolated CD11b + myeloid cells from hearts of Trem2^+/+ and Trem2^−/− DOCA-salt treated mice and performed qRT–PCR analysis. Similar to our in vitro LPS stimulation experiment, we found an increase in Il6 and Il10 expression in Trem2^−/− cardiac myeloid cells after DOCA-salt treatment along with a decrease in Vegfa and Spp1 expression (Figure 7G). Furthermore, we measured gene expression from whole LV tissue from Trem2^+/+ and Trem2^−/− DOCA-salt treated mice and confirmed an increase in Il6 and Il10 expression (Figure 7H). Given the decrease in angiogenesis related gene transcripts in Trem2^−/− macrophages and the known presence of coronary vessel rarefaction in human HFpEF,⁵⁹ we hypothesized that loss of Trem2 would result in decreased cardiac capillary density after DOCA-salt treatment. Indeed, we observe a decrease in CD31 positivity, which labels endothelial cells, in cardiac sections of Trem2^−/− mice compared to Trem2^+/+ controls following DOCA-salt treatment (Figure 8A). Using flow cytometry as a second method of endothelial cell quantification, we found that DOCA-salt treatment and Trem2 deficiency decreases absolute numbers of cardiac vascular endothelial cells lacking podoplanin expression (PDPN)(CD45-CD31 + PDPN-) (Figure 8B). In addition, DOCA-salt treatment and loss of Trem2 up-regulated intracellular adhesion molecule 1 (ICAM1) expression on cardiac vascular endothelial cells, consistent with endothelial inflammation (see Supplementary material online, Figure S13).

Figure 7.

Loss of Trem2 alters macrophage gene expression and cytokine production. (A) Relative expression of Lgals1 (galectin-1), Lgals3 (galectin-3), and Spp1 (osteopontin) in Trem2^+/+ and Trem2^−/− BMDMs. Each point is a biological replicate (n = 4). (B) BMDMs were differentiated and stimulated with LPS (M1) or not (M0) for 24 h and then harvested. Relative expression of Il6 and Il10 normalized to Rplp0. Each point is a biological replicate (n = 4). (C) Enrichment analysis of genes down-regulated in Trem2^−/− BMDMs compared to Trem2^+/+. (D) Z-score values of selected genes from bulk RNA-sequencing of Trem2^+/+ (WT) and Trem2^−/− (KO) BMDMs. (E) Relative expression of Vegfa, Mmp12, and Mmp13 normalized to Rplp0 in BMDMs. Points represent biological replicates (n = 4). (F) VEGF-A concentration in supernatants of BMDMs after 48 h of plating as determined by ELISA. Each point is a biological replicate from two independent experiments (n = 6–8). (G) Gene expression in CD11b + fraction of cardiac cells normalized to Gapdh in Trem2^+/+ and Trem2^−/− mice after DOCA-salt treatment (n = 4–5). (H) Left ventricular tissue gene expression normalized to Gapdh in Trem2^+/+ and Trem2^−/− mice after DOCA-salt treatment (n = 7–9). Data are expressed as mean ± SEM and analysed by two-way ANOVA with Tukey’s post-test (B), Student’s t-test (G: Il10), or Mann–Whitney U test (A, E, F, G, H).*P < 0.05, **P < 0.01, ***P < 0.001.

Figure 8.

Loss of Trem2 is associated with decreased cardiac capillary density after DOCA-salt treatment. (A) Representative images of immunohistochemistry for CD31 as a measure of capillary density and quantification of CD31 positivity as percent tissue area. A total of six to eight images with myocytes in cross-section were averaged per animal to obtain a single value per mouse (n = 6–7). Scale bar is 50 microns. (B) Representative gating strategy for quantification of vascular endothelial cells (VECs, CD45-CD31 + PDPN-) based on CD31 and PDPN expression with absolute numbers of cells in Trem2^+/+ (black circles) and Trem2^−/− (red squares) male mice after sham of DOCA-salt treatment from two cohorts (n = 4–6). Data are expressed as mean ± SEM and analysed by Student’s t-test. *P < 0.05.

sTREM2 is elevated in human heart failure. To test whether Trem2 might play a role in human heart failure, we collected plasma from patients admitted to Vanderbilt University Medical Center for a heart failure exacerbation and from non-heart failure controls (see Supplementary material online, Table S8). Patients admitted with HFpEF all had significant elevation of NT-proBNP (N-terminal prohormone of brain natriuretic peptide) upon admission and required treatment with intravenous diuretics. Exclusion criteria included evidence of acute coronary syndrome, active infection, autoimmune disease, and malignancy. Samples were collected within 24 h of admission. sTREM2 is produced by proteolytic cleavage after ligand-receptor binding and has been shown to correlate with tissue TREM2⁺ macrophages.⁵⁴ Furthermore, we showed that plasma sTREM2 was elevated in the DOCA-salt mouse model of heart failure (Figure 5B). Non-heart failure controls were matched for age, sex, and body mass index (BMI). Compared to non-heart failure controls, patients with HFpEF were more likely to have coronary artery disease, diabetes, and hypertension (see Supplementary material online, Table S8). Patients with HFpEF had increased levels of sTREM2 compared to non-heart failure controls (Figure 9). HFpEF risk factors such as BMI, age, and sex did not correlate with plasma sTREM2 levels, and heart failure status remained significantly associated with sTREM2 levels in a multivariable regression when adjusted for these factors (see Supplementary material online, Table S9). Furthermore, including the additional demographic variables of diabetes, CAD, and SBP in the multivariable regression did not alter the results, as only heart failure status remains significantly associated with sTREM2 levels (see Supplementary material online, Table S9).

Figure 9.

sTREM2 is elevated in human heart failure. Plasma samples were collected from control individuals without heart failure (n = 9) and from individuals with heart failure (n = 11) within 24 h of an admission for acute heart failure exacerbation and an underlying diagnosis of HFpEF. Levels of sTREM2 were measured by ELISA. Note that the high value in the HFpEF group did not meet outlier criteria by ROUT testing and removal does not change the results. Data are expressed as box and whisker plots showing median and range along with individual points and analysed by Mann–Whitney U test. ***P < 0.001.

4. Discussion

Inflammation is a common underlying contributor to the risk factors associated with HFpEF. Here, we used an unbiased method of immune cell profiling to elucidate pathways relevant to cardiac remodelling in a mouse model of hypertensive HFpEF driven by hypertension and identify myeloid Trem2 as a modulator of cardiovascular relevant gene programmes. Though there is increasing recognition that inflammation plays a role in chronic diseases, we are just beginning to unravel some of the mechanisms underlying immune cell activation in cardiovascular disease. Our study extends a growing body of literature supporting a critical role for macrophages in cardiac remodelling and provides new avenues of investigation for future studies.

HFpEF is a clinically heterogeneous syndrome but a major risk factor is hypertension.⁶⁰ Because not all patients with HFpEF have identical risk factors, having a multitude of available preclinical models is necessary to further our understanding of its pathophysiology. DOCA-salt treatment is similar to the SAUNA (salty drinking water/unilateral nephrectomy/aldosterone) model previously described with differences in the delivery and form of exogenous mineralocorticoid.^25,61 Our results are concordant with and extend findings previously published in the SAUNA model with increases in cardiac IL-10 production and distinct alterations in myeloid cells.²⁵ Advantages of the DOCA-salt model include its rapidity and reproducibility. The up-regulation of clinically relevant heart failure markers such as atrial and brain natriuretic peptides, osteopontin, and galectin-3 provide further support for the translational relevance of the DOCA-salt model.⁶² We chose the three week time point for our studies in an attempt to model a chronic disease state with evidence of multi-organ dysfunction. This is likely to reflect the human disease state when individuals are identified and diagnosed with heart failure. However, earlier timepoints may reflect other pathophysiological processes of interest such as the inciting inflammation during the initial disease onset.

We provide our dataset in a searchable and freely accessible web application (https://madhurlab.shinyapps.io/MouseHFpEF/), making it a useful resource for the community for further interrogation. To our knowledge, this is the first unbiased immune phenotyping of a hypertensive HFpEF model. We found broad changes in the cardiac immune landscape beyond those explored in detail here. One limitation of our single-cell dataset is that while hearts were perfused, we did not utilize in vivo antibody labelling to exclude leucocytes retained in capillaries. However, the majority of our study focuses on macrophages, which are generally absent from the circulation. We selected Trem2 as one intriguing candidate to investigate, but we also observed increases in frequencies of cardiac neutrophils, monocytes, CD8+ T cells, and γδ T cells, which merit further investigation. Both γδ T cells⁶³ and one of their characteristic cytokines, IL-17A,⁶⁴ are known contributors to mouse models of hypertension, and γδ T cells associate with hypertension and cardiac strain in humans.⁶⁵ We also describe a population of macrophages expressing the matricellular protein osteopontin and the chemokine receptor CCR2 which resemble those found in other disease contexts.^{25, 45, 66} Recent studies on cardiac macrophages frame their discussion based on embryonic-derived resident macrophages (Timd4 + Lvye1 + Folr2+) and recruited BMDMs (Ccr2+); however, we and others have observed heterogeneity in macrophage cell subsets beyond this dichotomy.^{42–44,67,68} In particular, Dick et al. described a third population termed MHC-II^hi macrophages based on higher expression of H2-Eb1 and H2-Aa,⁴⁶ which has been recapitulated by others.³⁴ However, the MHC-II^hi subset of cardiac macrophages had similar or less protein expression of MHC-II compared to CCR2 + macrophages by flow cytometry.⁴⁶ Thus, we termed the analogous subset of macrophages in our dataset Lilra5 + macrophages, which was also a top cluster defining marker in the Dick et al. study.

We observed changes in both cluster specific genes (Spp1, Ccr2) and changes in genes that were expressed in multiple clusters (Trem2), highlighting the utility of both cell-type and cluster specific analyses when performing deep phenotyping methods such as CITE-seq.

Taken together, our data reveal a novel cardioprotective role for Trem2 in heart failure. Trem2 was chosen for further downstream study given the similarity of macrophage gene expression changes observed in our model to previously described ‘disease associated microglia states’ in the context of Alzheimer’s Disease. This led us to believe that Trem2 signaling might control macrophage gene programmes relevant to cardiac remodelling. There are few studies describing the role of Trem2 in the cardiovascular system; however, the limited literature is consistent with a protective effect of Trem2. Intriguingly, loss of Trem2 led to increased inflammation and infarct size in a stroke model,⁶⁹ and a study of hepatic injury showed de-differentiation of endothelial cells after injury in Trem2 deficient mice.⁷⁰ Our study suggests novel antihypertrophic and pro-angiogenic roles for Trem2 in the heart. We show that loss of Trem2 leads to exacerbated cardiac hypertrophy, increases in local Il6 and Il10 expression, decreased cardiac capillary density, and increased cardiac endothelial cell activation. In addition, although loss of Trem2 resulted in decreased expression of matrix metalloproteinases (MMP) MMP12 and MMP13 from BMDMs, we did not see an effect of TREM2 deficiency on cardiac fibrosis in our study. A possible explanation for this is substrate specificity of MMP12 and MMP13. MMP12 lacks collagenase activity but cleaves elastin which we did not quantify. MMP13 expression preferentially cleaves type II collagens which are present in cartilage and bone but not abundant in the heart. The target of MMP13 in the hearts of DOCA-salt treated mice is unclear and is a topic for future investigation. Therefore, Trem2 deficiency may not affect the major collagens in the heart, and subsequently not display differences by Masson’s trichrome. Macrophage derived MMPs may play important roles in altering other aspects of the extracellular matrix or inflammatory signaling, as MMP12 has been implicated in inflammation resolution by terminating complement activation and dampening neutrophil influx.⁷¹ We also found that loss of Trem2 led to decreased Vegfa expression both in BMDMs and cardiac myeloid cells, but we were unable to ascribe Vegfa expression to a particular subset of macrophages due to poor expression in the single-cell dataset. In line with work done in microglia, we see that loss of Trem2 leads to decreased galectin-3 and osteopontin expression in macrophages⁵⁶ and extend those findings to additional cardiovascular-relevant genes. Recent work described an increase in cardiac Trem2^high macrophages in a model of myocardial infarction and demonstrated a blunted decline in systolic function when mice were injected with a gelatin hydrogel containing sTrem2 compared to controls.³³ Although these results support a protective role for sTrem2, it is unclear whether this represents increased or decreased Trem2 signaling since sTrem2 may act as decoy receptor to reduce Trem2 signaling. In contrast, our results using genetic deletion of Trem2 definitively show protective effects of Trem2 signalling in a hypertensive model of diastolic dysfunction and are in line with other inflammatory disease contexts.^54,56,69,72 We also report that loss of Trem2 is deleterious for hypertension-induced glomerular injury in the DOCA-salt model. A recent study using a cardiac pressure overload model implicated a heart-brain-kidney axis in which sympathetic tone activated local renal cells to secrete cytokines that affect cardiac macrophage populations and the cardiac adaptive response.⁷³ Further study into the role of Trem2 in renal macrophages and renal function is warranted to better dissect the connection between renal and cardiac function in diverse models.

Loss of Trem2 leads to multiple changes in macrophage function and activation. For example, Trem2 plays a critical role in macrophage phagocytosis,⁷⁴ which was not examined in this study. In addition, we find that Trem2 is expressed on both macrophages with markers of residency (Timd4, Lyve1, Folr2) as well as recruited macrophages (Ccr2), but whether Trem2 deficiency impacts both recruited and resident macrophages similarly requires additional investigation. Further studies are also needed to parse out whether Trem2 deficient mice have alterations in clearance of apoptotic cells in the heart or other relevant organs. This process, known as efferocytosis, has been implicated in settings of acute cardiac injury but remains relatively unexplored in chronic cardiac stress.^75–77 One limitation of this study is that it used a whole-body knockout of Trem2. The restricted expression of Trem2 in myeloid cells and microglia limits this concern, but we cannot exclude a role for Trem2 outside of the heart in our model. A major gap in knowledge that challenges further advances in Trem2 biology is the uncertainty surrounding Trem2 ligands and the lack of commercially available Trem2 agonists. More work is needed to develop tools to identify relevant disease- and tissue-specific ligands and manipulate Trem2 signaling in vivo.

We found that plasma sTREM2 is elevated in both the DOCA-salt mouse model and in patients with acute HFpEF exacerbation. Our human study is limited in numbers, and thus a larger study is required to more definitively determine how risk factors might impact sTREM2 levels and its possible utility as a prognostic biomarker. Few data exist on whether sTREM2 represents a decoy receptor that might limit cell surface Trem2 signaling or whether it might have distinct biological functions. Regardless, sTREM2 is elevated in other contexts of increased Trem2 expression or signaling.^33,54 Whether elevations in circulating sTREM2 represent increased cardiac Trem2 expression in humans will require tissue samples, which are difficult to obtain from patients with HFpEF. However, our mouse data, which mirrors the human data, suggest that increased circulating sTREM2 does indeed correlate with increased cardiac TREM2 expression. Taken together, our data suggest a novel role for Trem2 signaling in human HFpEF.

In summary, we utilized an unbiased immunophenotyping method to identify known and novel alterations in cardiac immune cells in a hypertensive mouse model of HFpEF. We chose Trem2 as a proof-of-principle candidate to test whether this approach can identify novel mediators of cardiac remodelling. We propose a framework in which loss of Trem2 alters macrophage gene expression programmes in a detrimental fashion, leading to exacerbated hypertrophy and decreased capillary density in response to cardiovascular stress. We also show that Trem2 is relevant to human HFpEF, suggesting that modulation of Trem2 signaling might be of therapeutic potential to alter the course of hypertensive HFpEF and end-organ damage.

Supplementary material

Supplementary material is available at Cardiovascular Research online.

Authors’ contributions

M.S.M. conceived the study. M.S.M. and C.D.S. planned the study and designed experiments. C.D.S. carried out the experiments and performed data analysis with help from D.J.F. and J.W.W. A.C.D. aided in experimental design. V.A., N.L.F, and A.R.H. assisted with collection and analysis of invasive haemodynamics data. D.D.D. and D.K.G. helped recruit and obtain human plasma samples from heart failure patients. M.A.C. and A.H.H. designed and helped implement the data portal. C.D.S. and M.S.M. wrote the manuscript with input from other authors.

 

Translational perspective.

Here, we demonstrate that the deoxycorticosterone acetate-salt model is a useful preclinical model of heart failure with preserved ejection fraction (HFpEF) that recapitulates key elements of human disease. Using a novel single-cell sequencing technology to profile cardiac immune cells, we found alterations in both innate and adaptive components of the immune system. Myeloid Trem2 was selected as a proof-of-concept target. We show that Trem2 plays a cardioprotective role in HFpEF. Furthermore, we found elevations of plasma soluble Trem2 levels in patients with HFpEF, suggestive of increased Trem2 signaling. Thus, preclinical models may identify relevant pathways of immune cell activation in human HFpEF, and enhancing Trem2 signaling may be a novel therapeutic strategy for the cardiac dysfunction in HFpEF.