Selective loss of resident macrophage derived Insulin-like Growth Factor-1 abolishes adaptive cardiac growth to stress

Summary:

Hypertension affects one-third of the world’s population, leading to cardiac dysfunction that is modulated by resident and recruited immune cells. Cardiomyocyte growth and increased cardiac mass are essential to withstand hypertensive stress, however whether immune cells are involved in this compensatory cardioprotective process is unclear. In normotensive animals, single-cell transcriptomics of fate-mapped self-renewing cardiac resident macrophages (RMs) revealed transcriptionally diverse cell states with a core repertoire of reparative gene programs, including high expression of Insulin-like Growth Factor-1 (Igf1). Hypertension drove selective in situ proliferation and transcriptional activation of some cardiac RM states, directly correlating with increased cardiomyocyte growth. During hypertension, inducible ablation of RMs, or selective deletion of RM-derived Igf1 prevented adaptive cardiomyocyte growth and cardiac mass failed to increase, which led to cardiac dysfunction. Single-cell transcriptomics identified a conserved IGF1-expressing macrophage subpopulation in human cardiomyopathy. Here, we defined the absolute requirement of RM-produced IGF-1 in cardiac adaptation to hypertension.

eTOC blurb:

Hypertensive stress requires cardiac muscle growth to maintain organ function. Zaman et al. reveal that the ability of the heart to adapt to hypertension through cardiomyocyte growth is entirely dependent on local IGF-1 produced by resident cardiac macrophages.

Graphical Abstract

Introduction:

Hypertension affects nearly 30% of the world’s population, causing profound morbidity and mortality (Drazner, 2011). Adult cardiomyocytes lack proliferative capacity. Thus, in response to hypertension, cardiomyocytes (and the myocardium itself) grow in size to minimize wall stress and maintain contractile function despite the increased workload. The left ventricle thickens as a result, and cardiac function remains preserved due to this adaptive remodeling process. Adaptive cardiomyocyte growth is of critical importance and a central compensatory process to hypertensive stress in humans. Over time, adverse remodeling in the form of fibrosis gradually develops and stiffens the myocardium. While poorly understood, a subset of patients develop impaired cardiac contractile function and progress to heart failure (Drazner, 2011).

A variety of factors and complex multicellular interactions govern this process, with well-characterized contributions from cardiomyocytes, fibroblasts, and endothelial cells. Macrophages are numerically the most abundant immune cell in the heart. Non-specific depletion of all macrophages in the setting of hypertension worsens cardiac function but improves fibrosis – suggesting they may possess both protective and pathologic functions, either as a single homogenous subset that possess multiple roles, or as heterogeneous subsets that each possess unique functions (Kain et al., 2016; Liao et al., 2018; Zandbergen et al., 2009). Cardiac macrophages can be categorized broadly into self-renewing, resident macrophages or monocyte-dependent macrophages (Emanuel et al., 2014). Monocyte-dependent macrophages have been shown to promote tissue damage and fibrosis in hypertension (Ishibashi et al., 2004; Kuwahara et al., 2004). However, little is known about self-renewing resident macrophages in hypertensive heart disease as prior studies have lacked the genetic tools required to track or target this population specifically.

Cardiac macrophages have been traditionally seen as a homogenous group of alternatively activated (“M2”-like) macrophages. However, recent observations suggest that rather than being a single population, they are composed of at least three subsets based on single-cell RNA sequencing, each possessing different lifecycles (Dick et al., 2019; Hulsmans et al., 2017). Most cardiac macrophages exist in an essentially closed system in adult animals; they are free of adult monocyte input and replenish almost exclusively through proliferative self-renewal (Dick et al., 2019). Cardiac macrophages that express the chemokine receptor CCR2 are monocyte derived in adult animals, and gradually replaced by circulating monocytes, while subsets that lack CCR2 expression but co-express TIMD4, LYVE1 and FOLR2 have virtually no monocyte input (Dick et al., 2019). Our focus here is on self-renewing cardiac resident macrophages (RMs) in adult animals, as defined by their independence from circulating blood monocytes. Cardiac RMs are exquisitely tied to the maintenance of tissue homeostasis and confer protection following myocardial infarction (Bajpai et al., 2019; Dick et al., 2019; Leid et al., 2016; Nicolás-Ávila et al., 2020). Loss of cardiac RMs during ischemic injury leads to increased, pathological cardiomyocyte hypertrophy (Dick et al., 2019).

Here, we investigate self-renewing cardiac RMs using a genetic fate-mapping approach to define their role in hypertensive heart disease. We generated a high-density single-cell transcriptional map of cardiac RMs during hypertensive stress and found that cardiac RMs were organized into large parental clusters, each made up of smaller, yet unique transcriptional states. Importantly, all cell states were pre-established in naïve mice prior to the onset of hypertension, suggesting cardiac RMs were poised to respond to a variety of challenges. During hypertensive stress, we observed a single-cell landscape that reinforced intrinsic core reparative potential of cardiac RMs, such as high expression of Insulin-like Growth Factor-1 (Igf1), combined with an induction of generalized and state-specific transcriptional programs in acute and chronic stress. Genetic ablation of RMs, or specific deletion of Igf1 within RMs demonstrated an absolute requirement of RM-produced IGF-1 in the compensatory growth of cardiomyocytes during hypertension, allowing the myocardium to withstand hypertensive stress. This study reveals highlights an entirely cardiac RM driven cardiac adaptation process to hypertension, the most prevalent chronic disease in humans.

Results:

Hypertensive stress drives numerical expansion of self-renewing resident cardiac macrophages.

We utilized the Cx3cr1^creERT2; Rosa26^Td (termed RM^Td) system to precisely fate-map RMs in the heart. Administration of tamoxifen chow to 3-week-old mice (for 7 days) labels all circulating monocytes and cardiac CD45⁺CD11b⁺CD64⁺ macrophages with expression of Td in this mouse line (Figure 1A–B) (Dick et al., 2019). Monocyte-dependent macrophages were replenished with unlabeled progenitors leaving only self-renewing cardiac RMs selectively labelled with Td (~71% Td⁺ macrophages, 4 weeks post-tamoxifen discontinuation, Figure 1B). Using flow cytometry, we observed that the majority of cardiac RMs lacked CCR2 expression, consistent with their bone marrow monocyte independence and expressed intermediate to high cell surface marker TIMD4 (Figure 1C).

Figure 1. Self-renewing resident cardiac macrophages expand numerically during hypertensive stress.

(A) Diagram of tamoxifen administration and AngII infusion in Cx3cr1^creERT2; Rosa26^Td (RM^Td) mice.

(B) Representative flow cytometric analyses of blood CD11b⁺ CD115⁺ monocytes and cardiac CD11b⁺ CD64⁺ macrophages at 4 weeks and 8 weeks of age, as indicated.

(C) Representative flow cytometric analyses of cardiac CD11b⁺ CD64⁺ macrophages gated on Td⁺ or Td⁻ cells, as indicated.

(D-F) Mice were infused with AngII for 28 days (D28 AngII) and compared to mice that underwent sham surgery (Sham).

(D) Quantification of aortic systolic and diastolic blood pressure from Millar catheterization (top), left ventricular ejection fraction (LVEF) from echocardiography, and heart weight to initial body weight ratio (bottom); n=8–15, pooled from two experiments.

(E) Representative immunofluorescence images and quantification of cardiomyocyte cross-sectional area from Wheat Germ Agglutinin staining, scale bar = 25 μm; n=6–12, pooled from N=2. (F) Representative histochemical images and quantification of fibrotic area in cardiac tissue from Sirius Red staining, scale bar = 50 μm; n=8–9, pooled from two experiments.

(G-I) RM^Td mice were infused with AngII for 4, 7 and 28 days (D4, D7, and D28 AngII).

(G) Quantification of infiltrating Ly6C^hi CD11b⁺ monocytes and cardiac CD11b⁺ CD64⁺ macrophages with flow cytometry; n=5–6, pooled from two experiments.

(H) Representative flow cytometry analyses (top) and quantification (bottom) of cardiac CD11b⁺ CD64⁺ macrophages gated on Td⁺ or Td⁻ cells, as indicated; n=4–6, pooled from two experiments.

(I) Representative immunofluorescence images and quantification of Td⁺ cells per 20× field of view (FOV) in cardiac tissue sections from RM^Td mice as indicated, scale bar = 50 μm; n=4, pooled from two experiments.

AngII: angiotensin II. *p < 0.1; ***p < 0.001; ****p < 0.0001. Data is shown as box and whisker plots (D-G, I) or as mean ± SEM (H). Analysis was performed using the unpaired t-test (D-F), and one-way ANOVA with multiple comparisons vs. Sham (G-I). See also Figure S1.

We induced hypertensive stress in RM^Td mice with the infusion of angiotensin II (AngII), which is an established, clinically relevant model of hypertension (Figure 1A). The AngII pathway has been effectively targeted by multiple, clinically approved therapies for not only hypertension, but also the management of heart failure – the latter showing direct benefits on mortality (Chamsi-Pasha et al., 2014; Dandona et al., 2007; Flather et al., 2000; Heran et al., 2012; Liu et al., 2002; Opie and Sack, 2001; Takemoto et al., 1997). We assessed various stages of disease by comparing steady state (Sham) mice to acute and sub-acute hypertension at days 4 and 7 AngII (D4 and D7 AngII) and chronic hypertension at day 28 AngII (D28 AngII) (Figure 1A). 28 days of AngII infusion models hallmarks of pathophysiology and adaptation in chronic hypertension, such as blood pressure elevation, mildly depressed cardiac function (as measured by left ventricular ejection fraction) and increase in cardiac mass (Figure 1D). Cardiac tissue growth correlated with increased cardiomyocyte area in hypertensive mice (Figure 1E). Sustained hemodynamic stress within the heart led to development of peri-vascular fibrosis at D28 AngII (Figure 1F).

In acute hypertensive stress, we observed a robust infiltration of Ly6C^hi monocytes and numerical expansion of cardiac macrophages (CD45⁺CD64⁺CD11b⁺) at D4 AngII, both of which subsided, but remained higher when compared to baseline at D28 AngII (Figure 1G). Using the RM^Td mice, we next assessed the contribution of monocyte-dependent macrophages vs. self-renewing cardiac RMs to macrophage expansion in hypertensive stress. Within cardiac tissue, hypertensive stress induced a transient increase in monocytes at D4 AngII, with gradual differentiation into Td⁻CCR2⁺ macrophages followed by downregulation of CCR2, as has been previously observed (Figures 1H–I; S1A) (Epelman et al., 2014; Liao et al., 2018). Total Td⁺ cardiac RMs numerically expanded over time, both in absolute numbers per heart and normalized to cardiac mass, which robustly increased over 28 days of AngII infusion (Figure 1H – flow cytometry, Figure 1I - immunofluorescence). Therefore, we conclude that fate-mapped self-renewing cardiac RMs increase numerically during hypertensive challenge, directly correlating with increasing cardiac mass and cardiomyocyte area.

Single-cell RNA sequencing reveals differentially activated resident cardiac macrophages during acute and chronic hypertensive stress.

We utilized single-cell RNA sequencing (scRNA-seq) to create a high dimensional map of cardiac RMs at steady state, and during acute and chronic hypertensive stress. We performed two independent experiments: The first experiment consisted of steady state, acute and chronic hypertension, with four RM^Td mice (50% male and 50% female) pooled and sequenced together in each condition. In the second experiment, we focused on steady state and acute hypertension, and performed cell hashing to individually label each biological replicate (n=3 per condition, 2 males, 1 female) with hashtag barcoded antibodies (Figure S2A) (Stoeckius et al., 2018). The barcoded antibody signal was demultiplexed to assign cells back to their replicate of origin.

We began our analysis using a low-resolution approach to broadly define the major cardiac RM clusters (Figure 2), understanding these subsets may be composed of smaller groupings of more numerous transcriptional states (see Figure 3 analyses). Total Td⁺ cardiac RMs were sorted from RM^Td at steady state (Sham), D4 AngII (acute hypertension) and D28 AngII (chronic hypertension) (Figure S2B). In addition, a small number of cardiac-infiltrating monocytes (CD64⁻CD11b⁺Ly6-C^Int-hi) were also sorted from each group as a comparator (Figure S2B). We employed graph-based clustering and UMAP (Uniform Manifold Approximation and Projection) visualization of the combined analysis of all groups (Figure S2C; 16,714 cells). Clusters were annotated on the basis of canonical macrophage (Fcgr1, C1qa, C1qb, C1qc, Mafb), monocyte (Plac8, Ly6c2, Ace, Chil3) and proliferation (Mki67, Top2a) genes (Figure S2D–E, Table S1).

Figure 2. Hypertensive stress induces the selective increase of fate-mapped TIMD4^hiMHC-II^lo resident cardiac macrophages.

Sorted resident cardiac macrophages (CD45⁺DAPI⁻CD11b⁺CD64⁺Td⁺) from Cx3cr1^creERT2; Rosa26^Td (RM^Td) mice following sham surgery (Sham; steady state), 4 days (D4 AngII; acute hypertension) and 28 days of AngII infusion (D28 AngII; chronic hypertension) were processed individually for scRNA-seq (10x Genomics). Two independent experiments were performed, each using 3–4 pooled animals per sample. Experiment 1 consist of Sham, D4 AngII and D28 AngII, and Experiment 2 consisted of Sham and D4 AngII.

(A) UMAP projection of the combined analysis of Sham, D4 AngII and D28 AngII visualized together.

(B) Feature plots illustrating expression of subset-defining genes.

(C) Quantification of unique or shared differentially expressed genes (DEGs) of each cluster, with selected genes highlighted (min.pct = 0.2, logFC threshold = 0.2, adjusted p-value <0.05).

(D) UMAP projection of Sham, D4 AngII and D28 AngII visualized separately (left). Quantification of the relative abundance of each cluster in each condition (right).

(E) The proportion of resident cardiac macrophages (all subsets combined) that are proliferating in Sham, D4 AngII and D28 AngII as assessed by scRNA-seq. Sham and D4 AngII represent the average of Experiment 1 and Experiment 2, shown as mean ± SEM.

(F) Representative flow cytometric analyses of cardiac Td⁺CD11b⁺CD64⁺ macrophages, as indicated from Sham, D4 AngII, D7 AngII and D28 AngII; green = TIMD4^hiMHC-II^lo; blue = TIMD4^hiMHC-II^hi; orange = TIMD4^loMHC-II^hi (left). Quantification of the abundance of each subset (right); n=4–8, pooled from two experiments.

(G) The proportion of cardiac Td⁺CD11b⁺CD64⁺ macrophages in each subset that are proliferating (BrdU⁺) across conditions; n=4–8, pooled from two experiments.

(H) Quantification of each Td⁺CD11b⁺CD64⁺ subset as absolute amount per heart (left) and normalized to per mg cardiac tissue (right) across conditions; n=4–8, pooled from two experiments.

Representative immunofluorescence images (I) and quantification (J) of LYVE-1⁺Td⁺ cells (indicated with white arrowheads) per 40× field of view (FOV) in cardiac tissue sections from RM^Td mice as indicated, scale bar = 50 μm; n=3.

AngII: angiotensin II; DEGs: differentially expressed genes; *** p-value<0.001, **** p-value<0.0001. Data is shown as mean ± SEM. Analysis was performed on pooled data in A-E and using two-way ANOVA with multiple comparisons vs. Sham (E-G). See also Figure S2.

Figure 3. High resolution analysis of resident cardiac macrophages reveals nine transcriptional states and maintenance of reparative programs in acute and chronic hypertensive stress.

(A) High resolution clustering identifies nine resident cardiac macrophage states (S1–9) in Sham, D4 AngII and D28 AngII split by condition (UMAP projection on left). The relative abundance of each cell state (S1–9) across conditions (right).

(B) Heatmap depicting the top 30 DEGs for each cell state (logFC threshold = 0.2, min.pct = 0.2, adjusted p-value <0.05). Colour bar denotes the high-resolution state (top; S1–9), low-resolution cluster (C1–3; middle), and condition (bottom). The number of DEGs in each cell state is shown (bottom).

(C) Pathway enrichment analysis (gProfiler, GO Biological Processes) using DEGs for each cell state (S1–9).

(D) DEGs (upregulated and downregulated combined) at D4 AngII (pink bars) or D28 AngII (blue bars) relative to Sham for each cell state (S1–9).

(E) Percentage of DEGs that were unique to D4 or D28 AngII (relative to Sham) or shared in both conditions for selected cell states.

(F) Percentage of DEGs that were shared or unique to each cell state at D4 or D28 AngII (relative to Sham), as indicated. Genes were considered shared if they were either upregulated at both D4 and D28 AngII or downregulated at both D4 and D28 AngII.

(G) Pathway enrichment analysis (gProfiler, GO Biological Processes, KEGG) using DEGs upregulated at D4 AngII relative to Sham for selected cell states.

(H) Pathway enrichment analysis (gProfiler, GO Biological Processes, KEGG) using DEGs upregulated at D28 AngII relative to Sham for selected cell states.

(I) DEGs of total resident cardiac macrophages relative to monocytes were computed for Sham, D4 AngII and D28 AngII for both experiments separately. The overlap of these DEGs were used for pathway enrichment analysis (gProfiler, GO Biological Processes).

(J) Feature plots of the expression of growth factors that are constitutively expressed in resident cardiac macrophages at steady state (Sham) and during hypertensive stress (D4/D28 AngII).

AngII: angiotensin II. DEGs: differentially expressed genes. data is from two pool experiments, merged for analysis. See also Figure S3.

We focused our analysis on 15,139 pooled cardiac RMs (Figure 2A) to generate a low-resolution map to assess major changes in cardiac RM transcriptional organization over the course of hypertensive stress. We found 3 cardiac RM clusters, and one proliferating cluster. Both Clusters-1 and 2 expressed high Timd4, Folr2 and Lyve1, with Cluster-1 lowly expressing antigen presentation genes (H2-Aa, H2-Ab1) compared to Cluster-2 (Figures 2B, S2F). Cluster-3 lacked Timd4, Folr2 and Lyve1, but had increased expression of antigen presentation genes and inflammatory cytokines and chemokines (Il1b, Tnf, Ccl4; Figure 2B, Table S2). Transcriptionally, Cluster-1 and Cluster-3 were distinctive and possessed 131 and 79 differentially expressed genes (DEGs), respectively. Cluster-2 was an intermediate population with majority of DEGs overlapping with either Cluster-1 or Cluster-3 (Figures 2C, S2F). These three clusters were detected reproducibly with similar frequencies across conditions in each biological replicate (Figure S2H–I). The highest proportion of proliferating cardiac RMs was observed at D4 AngII (Figure 2E), suggesting changes in composition may be linked with in situ proliferation.

We next assessed whether these cardiac RM clusters could be resolved in fate-mapped Td⁺ RMs with flow cytometry using cell-surface expression of markers TIMD4 and MHC-II. Indeed, 3 populations were detected, that paralleled our single-cell analysis: TIMD4^hiMHC-II^lo, TIMD4^hiMHC-II^hi and TIMD4^loMHC-II^hi RMs (Figure 2F). All three cardiac RM populations behaved differently during hypertensive stress. TIMD4^hiMHC-II^lo RMs proliferated robustly during acute hypertensive challenge and increased in frequency and number over time; TIMD4^hiMHC-II^hi RMs also proliferated, but numerically contracted during hypertensive stress while TIMD4^loMHC-II^hi RMs did not proliferate and numerically contracted (Figure 2F–H). We further validated the expansion of TIMD4^hiMHC-II^lo RMs with immunofluorescence staining for LYVE1, a surrogate marker for TIMD4, and observed a specific increase in Td⁺LYVE1⁺ cells at D28 AngII (Figure 2I–J). We concluded that our single-cell Cluster-1 corresponded to TIMD4^hiMHC-II^lo RMs, Cluster-2 to TIMD4^hiMHC-II^hi RMs and Cluster-3 to TIMD4^loMHC-II^hi RMs. The numerical expansion of cardiac RMs was exclusively driven by the increase of TIMD4^hiMHC-II^lo RMs, with each cardiac RM cluster possessing unique lifecycle properties during hypertensive challenge.

High resolution analysis of resident cardiac macrophages identifies nine transcriptional states with unique functions.

How to assess the transcriptional response to stress within closely related cells is an open question in macrophage biology. Initially, we defined the presence of three broad cardiac RM clusters (Figure 2). We wondered whether cardiac RMs could assume more numerous transcriptional states, as has been shown for microglia (Hammond et al., 2019; Li et al., 2019). The potential advantage with higher resolution clustering is that numerically smaller, but transcriptionally more diverse populations may be discovered. Here, we use the term “states” to refer to high resolution clustering, where multiple individual transcriptional states are derived from parent RM clusters defined in Figure 2.

We performed a higher resolution analysis of cardiac RM clusters 1–3 in steady state and hypertensive stress. Our analysis of the pooled data revealed nine transcriptional states differing by an average of 96 DEGs (range 15–225 DEGs). Each state was pre-established prior to hypertensive stress, with slight perturbations in composition over time (Figure 3A–B, Table S3). Cardiac RM states, particularly those that are rare in frequency, may be transient. Therefore, we assessed the reproducibility of RM composition at high resolution. We observed the presence of each cell state in both experiments, as well as in each biological replicate with similar relative frequencies (Figure S3A). Transcriptionally, there was an average overlap of 49% (Sham) and 37% (D4 AngII) in cell state defining DEGs between experiments 1 and 2 (Figure S3B). Within each state, biological replicates were transcriptionally similar and had negligible DEGs compared to each other (Figure S3C). We assessed DEGs in cardiac RMs using three approaches: an exploratory approach (permissive cut-offs), a conventional approach (Seurat default cut-offs), and a stringent approach (restrictive to the top DEGs; Table S3). In the stringent strategy, the most abundant RM state (S1) retains no DEGs relative to the other states, hampering our ability to understand the profile for the largest RM group, which adopts a small unique transcriptional signature using less restrictive strategies (Table S3). Here, we studied RM states using the exploratory approach in order to investigate heterogeneity within very closely related cells. Our exploratory approach preserves 70% of the DEGs from the conventional strategy (Table S3), and was reproducible in two independent experiments (Figure S3).

These numerically smaller transcriptional cell states were enriched in distinct pathways (Figure 3C). Cluster-1 consisted of 4 cell states (S1-S4); S1 (expressing Igf1, Igfbp4 and Trem2), and enriched in several pathways related to insulin-like growth factor 1 receptor (IGF-1R) signaling and regulation of glial apoptotic processes (Figure 3C); S2 (expressed Ccl6, Ccl9 and Anxa2) had predicted pathways mediating wound healing, tissue regeneration, angiogenesis, supramolecular fiber organization and muscle structure development – highlighting the potential relevance of this cell state adaptive cardiac remodeling (Figure 3C–D). Cell states S5-S6 corresponded to Cluster-2, which were largely enriched in antigen presentation and T cell activation pathways (Figure 3C). S7-S9 fell within Cluster-3 and were functionally diverse. S7 was enriched in dendritic cell differentiation and S9 was enriched in contractile filament assembly, while S8 was enriched in the growth of cardiac muscle tissue and muscle stretch pathways (Figure 3C). Of note, S4 distinctly expressed numerous interferon-stimulated genes (ISGs; Ifit1, Ifit2, Ifit3b), suggesting this subset received tonic interferon signaling, as described in dendritic cells (Schaupp et al., 2020). These granular data suggested a deeper heterogeneity within all clusters present in steady state, acute and chronic hypertension that may be related to a division of labor.

Resident cardiac macrophage states respond differentially to hypertensive stress and are constitutively enriched in tissue reparative pathways.

We have highlighted that cardiac RM cell states possess different transcriptional signatures, with several states enriched in functions relevant for adaptive cardiac remodeling (Figure 3C). We questioned whether upon hypertensive challenge, cardiac RMs alter gene expression programs within individual cell states or maintain pre-programmed functions sufficient to exert a biological effect. We observed the greatest transcriptional change in S1, S2, S3, S5 and S7, suggesting that these cell states were the most responsive to hypertensive stress, while S4 and S6 were transcriptionally quiescent (Figures 3D, S3D, Table S4). Hypertension elicited a temporally tailored response, with unique DEGs in acute or chronic hypertensive stress (Figure 3E).

The majority of genes induced in either D4 or D28 AngII infusion were unique to a single state (Figure 3F). In acute hypertension (D4 AngII), the induced genes expressed were mapped to pathways involved in response to lipopolysaccharide (S1), autophagy (S2) and muscle tissue development (S8), suggesting acquisition of additional functions (Figure 3G). We also observed several pathways that were shared across states, such as cell chemotaxis, IL-17 signaling and TNF signaling. These pathways mapped to a combination of genes either selectively induced by individual states or commonly upregulated by several states. For example, while Ccl8 and Cxcl2 were recurring genes induced across states involved in cell chemotaxis. The majority of pathways induced in chronic hypertension (day 28 AngII) were shared between states and related to apoptosis and protein folding (Figure 3H). The numerically dominant state (S1), appeared to have the greatest number of unique pathways induced during chronic hypertension, including response to muscle stretch, which together revealed both unique and overlapping functional responses across states.

Next, we explored whether core cardiac RM transcriptional programs exist at steady state and during hypertensive stress. We observed that cardiac RMs, irrespective of cluster, state, experiment, or condition, were enriched in reparative pathways related to wound healing, angiogenesis and vasculature development (Figure 3I). These pathways consistently mapped to several key growth factors, including Igf1, Pf4 and Ang, which were constitutively expressed in all cardiac RMs at steady state and hypertensive stress (Figure 3J). Thus, despite large scale structural and functional changes to the heart during hypertension, no additional cardiac RM transcriptional clusters or more granular transcriptional states developed. Rather, core tissue reparative programs were maintained, while in parallel, hypertension induced a variety of state-specific and shared pathways.

Resident cardiac macrophages are required for cardiomyocyte growth and preservation of cardiac function during hypertensive stress.

In order to define the functional role of cardiac RMs in hypertension, we utilized a diphtheria toxin (DT)-based system to deplete fate-mapped RMs as previously described (Dick et al., 2019). We crossed Cx3cr1^creERT2; Rosa26^Td (RM^Td) with Rosa26^DTR to generate Cx3cr1^creERT2; Rosa26^Td/DTR (RM^Td-DTR) mice (Figure 4A) (Parkhurst et al., 2013), which allows tracking and depletion of RMs upon DT administration. Administration of DT once every 5 days decreased Td⁺ cardiac RMs (~70%), and all of the cardiac RM subsets remained depleted at D7 AngII (Figures 4B, S4A–B). The residual RM populations remaining may reflect an intrinsic difference in labelling efficiency between the Rosa26-Td and Rosa26-DTR alleles with the cre recombinase upon tamoxifen administration. The Rosa26-Td insertion contains a reinforced (CAG) promoter while the Rosa26-DTR insertion does not.

Figure 4. Inducible depletion of tissue resident macrophages inhibits adaptive cardiomyocyte growth and leads to cardiac dysfunction during chronic hypertensive stress.

Cx3cr1^creERT2; Rosa26^Td/DTR (RM^Td-DTR) mice or RM^Td controls were infused with AngII for 7 or 28 days (D7 AngII or D28 AngII) concurrently with DT administration in all groups.

(A) Diagram of tamoxifen, AngII and DT treatment in RM^Td-DTR mice.

(B) Representative flow cytometric analyses and quantification of cardiac Td⁺CD11b⁺CD64⁺ macrophages from D7 AngII with DT treatment; n=4.

(C-G) At D28 AngII with DT treatment, cardiac function and disease pathology were assessed.

(C) Quantification of body weight loss shown as mean ± SEM; asterisks indicate statistical significance of each RM^Td-DTR group against respective RM^Td controls; n=3–6.

(D) Indices of cardiac growth; heart weight to initial body weight ratio, left ventricle (LV) anterior wall thickness (measured by echocardiography) and LV mass (approximated with echocardiography); n=3–6.

(E) Representative immunofluorescence images and quantification of cardiomyocyte cross-sectional area from Wheat Germ Agglutinin staining, scale bar = 50 μm; n=3–6.

(F) Representative histochemical images and quantification of fibrotic area in cardiac tissue from Sirius Red staining, scale bar = 100 μm; n=3–6.

(G) Echocardiographic assessment of cardiac function and remodeling; LV ejection fraction (LVEF), LV end systolic volume (ESV), LV end diastolic volume (EDV), LV internal diameter end diastole (LVIDd); n=3–6.

AngII: angiotensin II; DT: diphtheria toxin; LV: left ventricle; * p-value<0.1 ** p-value<0.01 *** p-value<0.001, **** p-value<0.0001. Data is shown as box and whisker plots unless otherwise indicated. Analysis was performed using two-way ANOVA with multiple comparisons. See also Figure S4.

RM^Td-DTR mice (or RM^Td controls) were infused AngII for 28 days, during which time DT was injected every 5 days in all groups. While we did not observe any mortality, hypertensive RM^Td-DTR mice lost body weight compared to RM-Td mice (Figure 4C). We assessed whether hypertensive RM^Td-DTR mice underwent adaptive cardiac growth during hypertensive stress compared to RM^Td controls. Increases in heart weight to initial body weight ratio, left ventricular (LV) anterior wall thickness, and LV mass from hypertensive challenge were absent in hypertensive RM^Td-DTR mice (Figure 4D). Accordingly, the marked hypertension-induced increase in cardiomyocyte cross-sectional area (~3-fold) was completely absent in RM^Td-DTR mice (Figure 4E). We confirmed this phenotype to be independent of blood pressure in RM^Td-DTR hypertensive mice as we detected no differences in blood pressure elevation compared to RM^Td controls given AngII (Figure S4C). Moreover, loss of cardiac RMs correlated with increased fibrosis in the perivascular and interstitial regions of the myocardium in hypertensive RM^Td-DTR mice (Figure 4E).

To determine the effects of RM depletion on cardiac function, we performed echocardiography in chronically hypertensive mice. We observed RM^Td-DTR mice had lower LV ejection fraction (LVEF), with increased LV end diastolic volume, end systolic volume and LV internal diameter in diastole compared to hypertensive RM^Td controls also receiving DT (Figure 4G). Together, these data demonstrated that selective reduction of RM abundance led to a complete absence of cardiomyocyte growth in response to hypertensive stress, combined with adverse cardiac chamber dilation and severe cardiac dysfunction.

Importantly, depletion of RMs in steady state (absence of hypertension) did not affect cardiac function, chamber size or tissue fibrosis (Figure 4F–G). We also observed no change in the proinflammatory cytokines Il1b, Tnfa, Il6 and Ccl2 in cardiac tissue post-RM depletion (Figure S4D). During RM-depletion, we did not detect any infiltration of Ly6C^hi monocytes and neutrophils (Figure S4E) at steady state, although we did observe an increase in Td⁻ macrophage subsets, which were composed of cell populations that were not depleted and CCR2⁺ macrophages; which could have resulted from local in situ proliferation or marginal monocyte recruitment we did not detect (Figure S4F–H). Monocytes have been previously reported to infiltrate cardiac tissue following macrophage depletion to repopulate a depleted macrophage niche (Epelman et al., 2014). Therefore, we conclude that obvious tissue myocardial inflammation and cardiac injury was not triggered by RM depletion in normotensive mice.

Specific deletion of IGF-1 from resident macrophages abolishes adaptive cardiomyocyte growth during chronic hypertensive stress.

We next sought to identify how cardiac RMs specifically influence adaptive cardiomyocyte growth. The RM^Td-DTR model depletes CX3CR1 expressing cells labeled at the time of tamoxifen administration systemically, affecting the resident macrophage pool in several organs, and causing weight loss. Moreover, as cardiac RMs possess numerous functions (all of which would be ablated with RM depletion), we utilized a more selective approach. Cardiac RMs are enriched in several growth factors that may facilitate this process (Figure 3J). We chose to focus on cardiac RM derived IGF-1, which has been previously shown to direct vascular development and pruning in the neonatal heart in vitro (Leid et al., 2016). Indeed, the growing neonatal heart and adult hypertensive heart both require contextual organ growth.

Cardiac RMs were highly enriched for Igf1 relative to monocytes (Figure 5A), and we detected Igf1 expression in all cell states (Figure S5A). The cell state that comprised the majority of cardiac RMs (S1, Figure 3A) had higher Igf1 expression relative to other states (Figure S5A). Igf1 expression was maintained in all cardiac RMs through acute and chronic hypertensive stress and we further observed downregulation of Igfbp4, a negative regulator of IGF-1 bioactivity, at D4 AngII (Figure S5B) (Duan and Clemmons, 1998). Hypertension caused an increase in cardiac IGF-1 protein, and that increase was completely abrogated with cardiac RM depletion (Figure 5B). Together, this suggests that cardiac RMs are an important source of local IGF-1 in the heart at steady state and hypertensive stress elicits a cardiac RM response that includes production of IGF-1 protein.

Figure 5. Specific deletion of IGF-1 from tissue resident macrophages abolishes adaptive cardiomyocyte growth and leads to cardiac dysfunction during chronic hypertensive stress.

(A) Violin plot showing mRNA expression of Igf1 in resident cardiac macrophages and monocytes. (B) Quantification of IGF-1 protein in cardiac tissue (left; n=8–12, pooled from two experiments) and serum (right; n=3–6) through ELISA in RM^Td and RM^Td-DTR mice at D28 AngII with diphtheria toxin treatment, shown as mean ± SEM.

(C-G) Cx3cr1^creERT2; Igf1^fl/fl (RM^ΔIgf1) mice were infused with AngII for 28 days (D28 AngII) and compared to control Igf1^fl/fl (Igf1 flox) mice.

(C) Diagram of tamoxifen and AngII administration in RM^ΔIgf1 mice.

(D) Indices of cardiac growth; left ventricular (LV) anterior wall thickness (measured by echocardiography) and LV mass (approximated with echocardiography); n=6–12, pooled from two experiments.

(E) Representative immunofluorescence images and quantification of cardiomyocyte cross-sectional area from Wheat Germ Agglutinin staining, scale bar = 50 μm; n=6–12, pooled from two experiments.

(F) Echocardiographic assessment of cardiac function and remodeling; LV ejection fraction (LVEF), LV end systolic volume (ESV), LV end diastolic volume (EDV), LV internal diameter end diastole (LVIDd); n=6–12, pooled from two experiments.

(G) Quantification of fibrotic area in cardiac tissue from Sirius Red staining; n=3–6.

AngII: angiotensin II; LV: left ventricle; * p-value<0.1 ** p-value<0.01 *** p-value<0.001, **** p-value<0.0001. Data is shown as box and whisker plots unless otherwise indicated. Analysis was performed using Wilcoxon Rank Sum test (A) or two-way ANOVA with multiple comparisons (D-G). See also Figure S5.

To study the role of cardiac RM-derived IGF-1, we created Cx3cr1^creERT2; Igf1^fl/fl (RM^ΔIgf1) mice in which IGF-1 derived from RMs was conditionally deleted upon tamoxifen administration (Figure 5C) (Stratikopoulos et al., 2008). We validated the model by detecting decreased IGF-1 protein in the hearts of normotensive and hypertensive RM^ΔIgf1 mice (Figure S5C), similar to the loss of cardiac IGF-1 when RMs were depleted (Figure 5B). With immunofluorescence, we further localized this loss to a specific decrease in IGF-1⁺LYVE1⁺ and IGF-1⁺CCR2⁻ macrophages during steady state and hypertension in RM^ΔIgf1 mice (Figure S5D), and not in IGF-1⁺CCR2⁺ cardiac macrophages (Figure S5E). We did observe a small increase in IGF-1 protein in CD68⁻ non-macrophages in RM^ΔIgf1 mice (Figure S5E). These data demonstrated that cardiac RMs were responsible for ~20% of IGF-1 protein in the heart during steady state, and all the inducible cardiac IGF-1 protein during hypertension. However, hypertension did not influence circulating IGF-1.

We assessed adaptive cardiomyocyte growth in hypertensive RM^ΔIgf1 mice. Left ventricular anterior wall thickness and mass were reduced in hypertensive RM^ΔIgf1 mice compared to hypertensive control (Igf1 flox) littermates (Figure 5D). Cardiomyocytes in RM^ΔIgf1 did not grow during hypertensive stress and remained the identical size when compared to sham mice (Figure 5E). During hypertension, we observed decreased cardiac systolic function and adverse left ventricular chamber dilation in RM^ΔIgf1 mice (Figure 5F), but in constrast to RM-depletion, did not observe exacerbation of cardiac fibrosis (Figure 5G). Importantly, systemic weight loss observed in RM^Td-DTR mice was not seen in RM^ΔIgf1 mice (Figure S5F). Thus, cardiac RM-derived IGF-1 is required for adaptive cardiomyocyte growth, allowing the myocardium to withstand hypertensive stress.

As a control, we examined cardiac macrophage subset distribution in RM^ΔIgf1 mice. We did not detect changes in CCR2⁺ cardiac macrophages in sham or hypertensive RM^ΔIgf1 mice, which suggests that the lack of cardiomyocyte growth and decreased cardiac function during hypertension was not due to enrichment of inflammatory cardiac macrophages, but rather a direct effect of IGF-1 deletion in cardiac RMs (Figure S5G). We also observed reduced LYVE1⁺MHC-II⁻ cardiac macrophages at steady state and during hypertension in RM^ΔIgf1 mice, suggesting that cardiac RM-derived IGF-1 affects RM survival through either an autocrine feedback loop or indirectly through growth factors produced by other cell types (Figure S5G).

An IGF1^hi resident cardiac macrophage population is conserved in human heart failure.

We next assessed whether a similar cardiac macrophage subpopulation is present in the human heart. The first sample was from the right ventricle of a 5-month-old neonate undergoing corrective surgery (normal cardiac function, however the right ventricle was pressure loaded), and the second was from a patient with hypertrophic cardiomyopathy and end-stage heart failure at the time of heart transplantation. We performed scRNA-seq on sorted CD45⁺ cardiac immune cells for each sample individually and pooled for analysis (4760 cells after quality control; Figure S6A–B, Table S5). We focused our analyses on macrophages and monocytes (expressing C1QC and FCN1, respectively; Figure 6A) and excluded dendritic cells and lymphocytes. We observed two groups of monocytes (expressing FCN1), which consisted of classical monocytes (expressing CD14, SELL) and non-classical monocytes (expressing FCGR3A), as well as three populations of macrophages (expressing C1QC; Figure 6A–C).

Figure 6. IGF1-expressing cardiac macrophage subset is conserved in human heart failure.

Single-cell RNA sequencing (10x Genomics) was performed on human cardiac tissue from two patients: 5-month-old neonate undergoing corrective surgery (with normal cardiac function) and adult human end-stage heart failure from non-ischemic origins.

(A) UMAP dimensionality reduction revealed two clusters of monocytes and three clusters of macrophages with distinct gene expression profiles.

(B) UMAP projection split by patient sample (left) and the frequency of each cluster present in each sample (right).

(C) Heatmap depicts the top 30 DEGs in each cluster (min.pct = 0.2, logFC threshold = 0.2, adjusted p value <0.05).

(D) Pathway enrichment analysis (gProfiler, GO Biological Processes) using DEGs for each macrophage cluster (MF 1–3).

(E) A murine resident cardiac macrophage gene signature was generated by computing the top DEGs in resident cardiac macrophages relative to monocytes (Figure S2). The machine-learning algorithm Garnett was used to classify cells in both human datasets individually using the gene signature generated from murine resident cardiac macrophages. Heatmap shows the percentage of cells from each cluster in each sample that were classified as resident cardiac macrophages due to similar gene expression profiles.

(F) Feature plots depict the gene expression of LYVE1 and IGF1. Outlined region highlights the cluster which specifically and distinctly expresses these genes, and were the most enriched in the murine resident cardiac macrophage signature (E). MF: macrophage. See also Figure S6.

Closer analysis of the three macrophage subsets revealed that MF-1 and MF-2 had negligible expression of the monocytic marker FCN1, suggesting a non-monocytic origin (Figure S6C). MF-1 was enriched in pathways related to antigen presentation and T cell proliferation, while MF-2 was enriched in endocytosis and heart development (Figure 6D). MF-3, uniquely present in adult heart failure, was enriched in cell activation, response to stress and type I interferon signaling, and expressed FCN1, suggesting a monocyte origin (Figure 6D).

We generated a murine cardiac RM gene signature using the Garnett machine-learning algorithm (Pliner et al., 2019). We applied this mouse “Garnett classifier” to the neonatal and adult macrophage and monocyte subpopulations. We found that MF-2 was the most transcriptionally similar to murine cardiac RMs (Figure 6E). MF-2 expressed a similar core gene set as murine cardiac RMs, including LYVE1, IGF1, F13A1, FOLR2, CD163, GAS6 and MRC1 (Figure 6F and S6C; Supplementary Table S5) suggesting this subset is similar between species. Together, these data reveal the conservation of a subset of IGF1-expressing cardiac RM subpopulation across species.

Discussion:

Cardiac adaptation to injury occurs in both neonatal and adult animals, with divergent responses. For example, neonatal cardiomyocytes proliferate robustly after injury in a process driven by neonatal cardiac macrophages, but adult cardiomyocytes lack this proliferative capacity (Lavine et al., 2014). Instead, the only mechanism by which adult cardiomyocytes augment cardiac function in the setting of hypertension is by increasing cell size and contractile machinery through adaptive cardiomyocyte growth. Other forms of cardiac remodeling, such as myocardial fibrosis and chamber dilatation, also occur during hypertensive stress, however these are pathologic in nature. Cardiomyocyte growth can, in late phases of hypertensive stress, itself become pathologic and lead to impaired cardiac function (Drazner, 2011). Importantly, immune cells can influence the cardiac remodeling process during hypertension (Fujiu et al., 2017; Hulsmans et al., 2018; Ishibashi et al., 2004; Liao et al., 2018). Given that hypertension is perhaps the most widespread chronic disease in humans, it is essential to dissect pathological from protective cardiac remodeling pathways.

RMs are heterogeneous not only in transcriptional patterns, but also in life cycle. In this study, we focused on those cardiac RMs that possess self-renewal potential without adult monocyte input and excluded transient populations. We have demonstrated that cardiac RMs increase numerically during acute and chronic hypertension, however, in parallel, there is also notable recruitment of monocytes. Prior studies have focused on monocyte infiltration that promotes increased oxidative stress and cardiac fibrosis (Ishibashi et al., 2004; Wenzel et al., 2011). Yet, the precise role of cardiac RMs has not been assessed. Here, utilizing a combination of genetic fate-mapping and scRNA-seq, we mapped the behaviour of numerous functionally diverse cardiac RM cell states. We showed that these cardiac RM cell states were pre-established before injury, yet differentially responded to hypertensive stress in a fashion that correlated with a reparative response. Moreover, specific depletion of RMs and RM-derived IGF-1 had the identical effect of preventing adaptive cardiomyocyte growth, which led to cardiac dysfunction, and adverse ventricular chamber dilation during hypertensive stress.

One question that typically arises when analyzing single-cell data sets is whether different “cell-states” or parental “clusters” from which those cell states were derived are discrete cell populations or a single population with a spectrum of activated states. Here, we posited that both processes are involved, and by separating transcriptionally coarse clusters versus higher resolution cell states, some clarity emerges. For example, two parental clusters that expressed Timd4 (TIMD4^hi) proliferated upon hypertensive challenge, with one population increasing numerically, while the other decreased numerically, suggesting coordinated proliferation but differential survival or differentiation. Conversely, Timd4^lo populations did not proliferate, and reduced in frequency, indicating three distinct behaviours correlated with three cardiac RM clusters. The acute proliferative burst we observed in some cardiac RMs during hypertensive stress coincided with the tissue infiltration of monocytes, suggesting a potential association of the two phenomena. However, we have previously shown that during acute hypertension, monocyte deficient mice exhibit similar cardiac macrophage proliferation and population frequency compared to wild type controls (Epelman et al., 2014). Thus, recruited cells do not appear to influence cardiac RM proliferation in response to AngII infusion.

Our analyses of cardiac RMs revealed unique transcriptional cell states that each possessed a specialized repertoire of functions poised to respond to given stimuli, combined with many overlapping functions and gene expression profiles. Thus, regardless of how they are defined, individual cardiac RM subpopulations appear to have the inherent ability to rapidly engage in antigen processing pathways, antiviral pathways, and reparative activities upon challenge. During hypertensive stress, we examined both the cell-state specific responses, and the combined responses of all cardiac RM subpopulations together – akin to a bulk transcriptomic analysis, thus defining shared pathways across RMs, and potentially unique pathways within individual states. A number of unique cell state-specific transcriptional programs were triggered in some states (particularly those of higher abundance such as S1-S3), while some cell states were transcriptionally quiescent. This suggests inherent differences between cell states and an unequal division of labour during hypertension that requires further exploration.

Perhaps functionally more pertinent, we observed both maintenance of established reparative programs, and induction of additional reparative programs. For example, cell state S1 (Timd4^hi) was numerically the largest population with the highest basal expression of Igf1. Upon hypertensive challenge, S1 increased in frequency and expressed genes that mapped to pathways shared with other responsive states, thus revealing a generalized cardiac RM response to hypertension. We still do not understand if one cell state can transform into another during stress, and if selective proliferation or death affect composition. Further work developing tools that track with individual states will help answer these more nuanced questions.

Some of the transcriptional responses (or quiescence) may be influenced by the physical niche in which the cell types reside. TIMD4⁺LYVE1⁺ cardiac macrophages are found near vasculature, while those subsets expressing high MHC-II are found near nerves, suggesting spatial regulation may be involved (Chakarov et al., 2019; Dick et al., 2019; Lavine et al., 2014; Leid et al., 2016). In our study, we observed cardiac RMs in close proximity to cardiomyocytes. Recent studies have shown that direct cardiomyocyte-macrophage crosstalk, where cardiac RMs take up damaged mitochondria released by cardiomyocytes, and separately, focal adhesion complexes between cardiac RMs and cardiomyocytes, activate cardiac RMs in response to muscle stretch to secrete growth factors, including IGF-1 (Nicolás-Ávila et al., 2020; Wong et al., 2021).

We wondered why, with so much available cardiac and circulating IGF-1, only RM-derived cardiac IGF-1 (~20% of total IGF-1 in the heart) drives cardiomyocyte growth during hypertension. Liver-specific IGF-1 deletion leads to a ~75% reduction of circulating IGF-1 without affecting post-natal growth in mice, highlighting the non-redundant role of local IGF-1 to developmental growth (Sjögren et al., 1999; Yakar et al., 1999). We demonstrated that cardiac RMs were the sole source of AngII inducible IGF-1 protein within the myocardium. RM-independent local IGF-1 was not able to support cardiomyocyte growth, suggesting either local production of IGF-1 or IGF-1 binding proteins could regulate bioactivity. Indeed, we detected downregulation of Igfbp4, a negative regulator of IGF-1 bioactivity (Duan and Clemmons, 1998) by cardiac RMs during acute hypertensive stress. IGF-1 acts directly on cardiomyocytes to induce cardiomyocyte growth (Cittadini et al., 1996; Ito et al., 1993) and induces proliferation of endothelial cells (Lin et al., 2017). Vascular growth is essential to supply the growing myocardium with oxygen and nutrients. Given the localization of cardiac RMs between cardiomyocytes and endothelium, cardiac RM-derived IGF-1 may act on both cell types in concert to promote growth.

We used scRNA-seq to identify a cardiac RM subset that is transcriptionally analogous to mouse cardiac RMs, defined by its unique enrichment in LYVE1, FOLR2, IGF1 and several other cardiac RM markers. The samples included a neonate with a pressure loaded right ventricle due to congenital heart disease, and an adult with end-stage hypertrophic cardiomyopathy. Despite differences in age and degree of cardiac stress, the IGF1-expressing macrophage subset was conserved in both subjects. In our study, human MF-3 was present only in the adult heart failure sample, and expressed both monocytic and macrophage genes, suggested peripheral recruitment. Our characterization of neonatal cardiac RMs resembles the recently reported human heart cell atlas, where LYVE1⁺ macrophages and antigen-presenting macrophages (LYVE1⁻FOLR2⁻) observed in that study (Litviňuková et al., 2020) are similar to MF-1 and MF-2 in our study.

Cardiac RMs are not required for all forms of cardiomyocyte growth. Following myocardial infarction, depletion of RMs leads to increased pathological cardiac hypertrophy in the peri-infarct and remote non-infarct regions (Dick et al., 2019). These observations highlight the very specific role cardiac RMs, and RM-derived IGF-1 play in adaptive cardiomyocyte growth during hypertensive stress, a process that is not generalizable to all forms of cardiac injury. Together, these data reveal the essential role of cardiac RM derived IGF-1 in mediating adaptive cardiac growth during hypertension.

Limitations of Study:

The Cx3cr1^creERT2 resident macrophage targeting system utilized is not selective for cardiac macrophages, as it affects self-renewing CX3CR1⁺ tissue-resident macrophage populations in other organs. Our scRNA-seq reproducibly detected nine cardiac RM states in mice, however, our data provides a mere snapshot. A current limitation with granular transcriptomic analyses is inability to infer biological significance of low abundance transcriptional states detected in a dynamic setting. These states may be unstable, and their gene expression profiles may be transient.

STAR METHODS

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Slava Epelman (slava.epelman@uhn.ca).

Materials availability

This study did not generate new unique reagents. All reagents and mouse lines used to generate experimental animal models for this study are commercially available as of the date of publication.

Data and Code Availability

All single-cell RNA sequencing data generated in this study have been deposited in the Gene Expression Omnibus (GEO) public database under the accession number GEO: GSE179343. Additional microscopy and flow cytometry data beyond what is reported in this paper will be shared by the lead contact upon request.

R scripts for processing of single-cell RNA sequencing data are available on GitHub: https://github.com/HomairaH/EpelmanLab_ResidentCardiacMacrophages_Hypertension. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Animals

All mice used in this study were purchased from The Jackson Laboratory and were bred in our animal facility before use. Stock numbers are listed in the key resources table. All mice were bred and housed in a pathogen-free environment at the University Health Network Animal Facility and were fed Teklad global soy protein-free extruded diet (Envigo) unless otherwise stated. 10–12-week-old age-matched littermate controls were used for all experiments, with sample sizes of n=4–6 in experimental groups. All experimental procedures were approved by the Animal Care Committee of the Toronto General Research Institute and were performed according to the guidelines of the Canadian Council on Animal Care.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Monoclonal anti-mouse CD45, clone 30-F11	Biolegend	Cat# 103132; RRID: AB_893340
Monoclonal anti-mouse CD64, clone X54-5/7.1	Biolegend	Cat# 139323; RRID: AB_2629778
Monoclonal anti-mouse/human CD11b, clone M1/70	Biolegend	Cat# 101224; RRID: AB_755986
Monoclonal anti-mouse TIMD4, clone RMT4-54	Biolegend	Cat# 130009, RRID: AB_2565718
Monoclonal anti-mouse CCR2, clone 475301	R&D Systems	Cat# FAB5538A; RRID: AB_10645617
Monoclonal anti-mouse I-Ab, clone M5/114.15.2	eBiosciences	Cat# 56-5321-80
Monoclonal anti-mouse BrdU, clone 3D4	BD Biosciences	Cat# 560810; RRID: AB_2033930
Monoclonal anti-mouse CD115, clone AFS98	eBiosciences	Cat# 17-1152-80
Monoclonal anti-mouse Ly6C, clone HK1.4	Biolegend	Cat# 128006; RRID: AB_1186135
Monoclonal anti-mouse Ly6G, clone 1A8	Biolegend	Cat# 127623; RRID: AB_10645331
Monoclonal anti-mouse CD3, clone 17A2	Biolegend	Cat# 100204; RRID: AB_312661
Monoclonal anti-mouse B220, clone RA3-6B2	Biolegend	Cat# 103205; RRID: AB_312990
Monoclonal anti-mouse NK1.1, clone PK136	Biolegend	Cat# 108706;RRID: AB_313393
Monoclonal anti-mouse CD68, clone FA-11	Biolegend	Cat# 137001, RRID: AB_2044003
Monoclonal anti-mouse LYVE1, clone ALY7	eBiosciences	Cat# 13-0443-82
Monoclonal anti-mouse MHC-II, clone M5/114	Biolegend	Cat# 107601; RRID: AB_313316
Monoclonal anti-mouse CCR2, clone 475303	R&D Systems	Cat# MAB55382
Polyclonal anti-mouse IGF1	ThermoFisher	Cat# PA527207; RRID: AB_2544683
Monoclonal anti-human CD45, clone 2D1	Biolegend	Cat# 368504; RRID: AB_2566352
Biological samples
Human cardiac tissue samples	University Health Network Biobank, Toronto General Hospital Research Institute	N/A
Chemicals, peptides, and recombinant proteins
Tamoxifen chow	Envigo	Cat# TD 130856
Angiotensin II	Bachem	Cat# H-1705
Diphtheria toxin	Sigma	Cat# D0564
Collagenase I	Sigma	Cat# C0130
DNase I	Sigma	Cat# D4513
Hyaluronidase	Sigma	Cat# H3506
BrdU	Sigma	Cat# B5002
Fast Green/Sirius Red	Chondrex	Cat# 9046
Wheat Germ Agglutinin	ThermoFisher	Cat# W11261
SlowFade Diamond Antifade mountant with DAPI	Invitrogen	Cat# S36973
OCT medium	Sakura Finetek	Cat# 4583
TRIzol Reagent	ThermoFisher	Cat#: 15596026
LightCycler 480 SYBR Green I Master Mix	Roche	Cat# 4707516001
Anti-mouse CD45 magnetic beads	Miltenyi Biotec	Cat# 130-052-301
Critical commercial assays
Cytofix/Cytoperm	BD Biosciences	Cat# 554714
TSA Fluorescein kit	PerkinElmer	Cat#SAT705A001EA
RNeasy Micro Kit	QIAGEN	Cat# 74004
iScript cDNA synthesis kit	Bio-Rad	Cat# 1708841
Mouse anti-IGF-1 ELISA kit	R&D Biosystems	Cat# MG100
Cell hashing protocol	Stoeckius et al., 2018	N/A
Chromium Single cell 3′ Reagents Kit	10x Genomics	N/A
Deposited data
Raw and processed single-cell RNA sequencing data	This research article	GEO: GSE179343
Experimental models: Organisms/strains
Mouse: B6.129P2(Cg)-Cx3cr1tm2.1(cre/ERT2)Litt/WganJ	The Jackson Laboratory; Parkhurst et al., 2013	JAX: 021160
Mouse: B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J	The Jackson Laboratory	JAX: 007914
Mouse: C57BL/6-Gt(ROSA)26Sortm1(HBEGF)Awai/J	The Jackson Laboratory	JAX: 007900
Mouse: FVB.129(B6)-Igf1tm1Dlr/J	The Jackson Laboratory; Stratikopoulos et al., 2008	JAX: 012663
Oligonucleotides
Primers for qPCR, see Table S6	NCBI primer design	N/A
Software and algorithms
FlowJo v.10.7.1	FlowJo, LLC	https://www.flowjo.com/; RRID: SCR_008520
LabChart Pro	AD Instruments
ImageJ software	Schneider et al., 2012	https://imagej.nih.gov/ij/
Zen Pro software	Zeiss	N/A
HALO software	Indica Labs	N/A
Prism software	GraphPad	N/A
Cell Ranger	10x Genomics	N/A
STAR 2.7.9a RNA-seq aligner		N/A
MULTI-seq demultiplexing algorithm	McGinnis et al., 2019	N/A
Seurat v.3.1	Stuart et al., 2019, Stuart and Satija, 2019	https://github.com/satijalab/seurat
R 3.6.1		N/A
SCTransform	Hafemeister and Satija, 2019	https://github.com/ChristophH/sctransform
gProfiler		https://biit.cs.ut.ee/gprofiler/gost/
Harmony	Korsunsky et al., 2019	https://github.com/immunogenomics/harmony
Garnett	Pliner et al., 2019	https://cole-trapnell-lab.github.io/garnett/
Other
Osmotic minipumps	Alzet	Cat# 1004D, 1007D, 2004D
VisualSonics Vevo 2100 System	VisualSonics	N/A
Millar pressure-transducing catheter	Millar Instruments	N/A
LSR Fortessa	BD Biosciences	N/A
Axioimager Microscope	Zeiss	N/A
LSM700 Confocal Microscope	Advanced Optical Microscopy Facility, Princess Margaret Cancer Research Tower	N/A
AutoMACS	Miltenyi Biotec	N/A
Aria Fusion	BD Biosciences	N/A
HiSeq2500 sequencer	N/A	N/A

Human Subjects

Cardiac tissue samples were obtained from the University Health Network Biobank at the Toronto General Hospital Research Institute, Toronto, Ontario. The first sample was obtained from the left ventricle at the time of heart transplant of a 46-year-old male patient with hypertrophic cardiomyopathy. The second sample was obtained from the right ventricle of a 5-month-old male patient with congenital heart disease (Teratology of Fallot) undergoing corrective surgery. Informed consent was obtained from all patients and/or parents/legal guardians. The protocol was approved by institutional research ethics boards.

METHOD DETAILS

Experimental Mouse Models

Fate-mapping.

To induce recombination in Cx3cr1^creERT2 mice, tamoxifen was administered via tamoxifen-containing chow (Envigo) for 7 days at the age of 3 weeks at the time of weaning.

Angiotensin II infusion.

Mice were anesthetized with 1% isoflurane, back was shaved and osmotic mini pumps (Alzet) containing either saline or angiotensin II (AngII) (Bachem, 2.0 mg–1kg⁻¹day⁻¹) were implanted. The incisions were closed with silk sutures.

Resident macrophage depletion.

Diphtheria toxin (Sigma) was diluted in 200 μl PBS and administered to mice i.p. For long term experiments, 250 ng diphtheria toxin was given every 5 days for a period of 28 days to all mice in the experimental cohort.

Tissue isolation and cell-surface staining for flow cytometry

Sample preparation.

Mice were euthanized by CO₂ inhalation and tissue was processed as described previously (Dick et al., 2019). Hearts were perfused with 20 ml of cold PBS. Hearts were chopped finely and digested, while being shaken, for 45 min at 37 °C in DMEM containing collagenase I (450 U ml⁻¹), DNase I (60 U ml⁻¹) and hyaluronidase (60 U ml⁻¹) enzymes (all from Sigma). The digested material was filtered through 40 μM filters and pelleted by centrifugation (400g for 5 min at 4°C) in Hank’s balanced salt solution (HBSS) supplemented with 2% bovine serum + 0.2% BSA. Red blood cells were lysed in ACK lysis buffer (Invitrogen) for 5 min at room temperature and resuspended in FACS buffer (PBS containing 2% bovine serum and 1 mM EDTA). Blood was collected in syringes containing 50 μl heparin, red blood cells were lysed, and pellets were resuspended as described above. Single-cell suspensions were then labeled for cell-surface markers using antibodies listed below. Cells were labeled in 50 μl FACS buffer with 0.2 μl antibody per sample, except for the antibody to CCR2 (5 μl), for 30 min at 4 °C. Cells were washed and resuspended in 400 μl FACS buffer to be analyzed by flow cytometry (BD Fortessa BYGRV).

BrdU incorporation.

For proliferation experiments, 2 mg of BrdU (Sigma) was injected i.p. at 2 h before organ harvest as described previously (Epelman et al., 2014). To detect intracellular BrdU, the BD Bioscience Cytofix/Cytoperm protocol was used. Cells were fixed after cell surface labeling, washed and resuspended with permeabilization buffer overnight. DNA was digested for 1 h with DNase (Sigma), cells were labeled with Fc receptor–blocking reagent for 15 min and were then labeled with an antibody to BrdU for 30 min.

Flow cytometry gating strategy, antibodies, and quantification.

After gating on CD45⁺ cells, doublets were excluded, and live cells were analyzed using forward scatter (FSC) and side scatter live-dead exclusion. Single cell analysis software (FlowJo) was used to analyze cell cytometric data. All positive gated events were normalized to organ weights of individual mice to obtain cells/mg as indicated.

Gating strategy.

Blood monocytes were identified as CD115⁺CD11b⁺ cells. A separate gating strategy was utilized for tissue infiltrating monocytes, as the CD115 molecule is internalized from cell surface during tissue digestion at 37°C. Therefore, Ly6C⁺ monocytes in the myocardium were identified as CD45⁺CD11b⁺CD64^lowLy6C⁺ cells. Cardiac macrophages were identified as CD45⁺CD11b⁺CD64⁺ and were further parsed by CCR2, MHC-II, TIMD4, and Td.

Echocardiography

Mouse echocardiography was performed using the VisualSonics Vevo 2100 System using 1% isoflurane anesthetic as described previously (Dick et al., 2019). Temperature was held constant at 37°C and heart rate held between 500 and 600 beats min⁻¹. Two-dimensional B-mode images were obtained in the long- and short-axis views. VevoLab program was used for Simpson’s volume measurements. For ejection fraction calculation, left ventricle tracing was done in triplicate. Measurements were made in a blinded fashion. For end diastolic volume calculations and anterior wall diameter measurements were done in both diastole and systole, in triplicate.

Millar catherization

Hemodynamics measurements were performed on mice anesthetized with 1% isoflurane. The right carotid artery was dissected and cannulated with a micromanometer 1.4F Millar pressure-transducing catheter (Millar Instruments). Heart rate and aortic blood pressure were recorded. Data were collected and analyzed using LabChart Pro (AD Instruments).

Immunohistochemistry and histology

Hearts were cut cross-sectionally and fixed overnight in 10% buffered formalin. Hearts were then paraffin embedded and sectioned at two cutting levels at 8-μm thickness. After deparaffinization and rehydration steps, sectioned were stained with Sirius Red/Fast Green (Chondrex) for fibrosis, or Wheat Germ Agglutinin (1:1000 dilution, ThermoFisher) and mounted with SlowFade Diamond Antifade Mountant with DAPI (ThermoFisher) for cardiomyocyte hypertrophy. Slides were imaged using the Zeiss Axio Imager microscope at 20×. Percentage fibrosis was quantified with ImageJ software (Schneider et al., 2012). Wheat Germ Agglutinin images were processed using Zen software. All cardiomyocytes in cross-section were counted and measured per field of view. All measurements were made in a blinded fashion.

Immunofluorescence

Hearts were isolated at reported times after AngII infusion and cut cross-sectionally. They were fixed in 4% paraformaldehyde overnight and followed by a 30% sucrose gradient. Tissue was embedded in OCT medium (Sakura Finetek) and flash frozen in isopentane suspended in liquid nitrogen. Tissue was sectioned in 8-μm slices at two cutting levels. Sections were blocked for 1 h in a 50:50 solution of permeabilization block (5% bovine serum, 0.01% Triton X in PBS) and normal serum block (Biolegend). Sections were washed with PBS and the primary antibody was diluted 1:100 in FACS buffer and added overnight at 4°C in a hydrated chamber. Sections were washed with PBS and were labeled with secondary antibodies diluted 1:250 in PBS for 1 h at 4°C. For LYVE1 staining the TSA Fluorescein kit was used (PerkinElmer). Sections were mounted with SlowFade Diamond Antifade Mountant with DAPI (ThermoFisher) and imaged on the LSM700 confocal microscope at 20× or 40× with immersion oil (AOMF, Advanced Optical Microscopy Facility, Princess Margaret Cancer Research Tower, Toronto, Canada). Whole slide scans were analyzed using HALO software (Indica Labs) for colocalization. A section without the primary antibody, but labeled with the secondary IgG1 antibody, was used to control for background.

RNA isolation and qPCR

RNA was isolated from homogenized tissue using Trizol Reagent (Ambion) and the RNeasy Micro Kit (QIAGEN) as per manufacturer’s instructions. Samples were digested with DNase (QIAGEN) prior to reverse transcription using the iScript cDNA synthesis kit (Bio-Rad) using 800ng of RNA. The resultant cDNA was analyzed by qPCR using LightCycler^® 480 SYBR Green I Master Mix in a Roche LightCycler (Roche). RNA measurements were normalized to GAPDH for each individual mouse.

IGF-1 ELISA

Hearts were flushed with PBS and flash frozen. Following tissue homogenization in PBS, samples underwent 3 freeze-thaw cycles and supernatant was collected after centrifugation (5000g for 5 min at 4 °C). Mouse anti-IGF-1 ELISA kit (R&D Biosystems) was used on the supernatants.

Single-cell RNA sequencing

Mouse tissue.

Resident cardiac macrophages (CD45⁺DAPI⁻CD11b⁺CD64⁺Td⁺) were sorted from Cx3cr1^creERT2; Rosa26^Td mice of three conditions: sham, day 4 and day 28 post AngII infusion. Monocytes (CD64⁻CD11b⁺Ly6-C^lo-hi) were also spiked in at a 1:100 ratio and sequenced together. In Experiment 1, 2 males and 2 females were pooled for each condition (sham, day 4 and day 28 post AngII infusion). In Experiment 2, we performed cell hashing as previously described (Stoeckius et al., 2018). We first individually labelled each biological replicate using unique hashtag barcoded antibodies (2 males and 1 female each for sham and day 4 AngII infusion). Following cell staining and cell sorting, replicates in each condition were pooled and sequenced together (10x Genomics).

Cell sorting.

Mouse hearts were isolated and enzymatically digested as described above, with the inclusion of 1 mM flavopiridol in the digestion buffer. Digestions were stopped after 30 min and cells were processed into a single-cell suspension on ice. Cells were labeled with CD45 magnetic beads (Miltenyi Biotec) and subsequently fluorescently tagged antibodies to CD45, CD64, CD11b, B220, CD3, Ly6G, NK1.1, Ly6C and MHC-II. Hematopoietic cells were positively enriched using the AutoMACS instrument (Miltenyi Biotec). Live resident macrophages (DAPI⁻CD45⁺CD64⁺CD11b⁺Td⁺) and monocytes (DAPI⁻CD11b⁺Ly6C^lo-hi) were sorted on the Aria Fusion (BD Bioscience) under low pressure into DMEM containing 50% bovine serum for 10x Genomics single-cell RNA sequencing.

Human tissue.

Cardiac tissue samples were obtained from the University Health Network Biobank at the Toronto General Hospital Research Institute, Toronto, Ontario. Informed consent was obtained from all patients and/or parents/legal guardians. The protocol was approved by both institutional research ethics boards. Resected tissue was immediately placed into sterile, cold DMEM media supplemented with 50% fetal bovine serum. Human cardiac tissue was processed similar to mouse tissue and stained with anti-CD45 antibody. Cell sorting was performed using Aria II instrumentation (BD Bioscience). Total live immune cells were sorted by gating on DAPI⁻CD45⁺ cells. Samples were sorted directly into DMEM + 50% bovine serum for single-cell RNA sequencing submission.

Library preparation and sequencing

Single cell suspensions were prepared as indicated in the Chromium Single Cell 3’ Reagents Kits User Guide (v3 Chemistry). Samples were loaded onto the v3 10x Chromium for the generation of sequencing libraries and processing in accordance with the methods described by 10x Genomics. Human cardiac tissue samples (CD45⁺ immune cells) were similarly processed; the first sample (46-year-old) was processed using v2 Chemistry and the second sample (5-month-old) using v3 Chemistry. Cell Ranger (10x Genomics) was used to pre-process sequenced cells and to generate the expression matrices. Raw base call (BCL) files from HiSeq2500 sequencer were demultiplexed into FASTQ files. Reads were aligned using STAR and filtered, followed by barcode and UMI counting to generate the feature-barcode matrices. In the second single-cell RNA sequencing experiment, count matrices were generated from the hashtag antibody libraries and the antibody signal was demultiplexed to assign cells to their replicate of origin using the MULTI-seq algorithm, as previously described (McGinnis et al., 2019). Cells were assigned either as singlets for a particular hashtag, as negative if there was not sufficient signal for any hashtag antibody, or as doublets if the cell had sufficient antibody signal for more than one hashtag. Cells that were classified as doublets were removed from subsequent analyses.

Pre-processing and quality control

The package Seurat (v3.1) was used for all scRNA-seq analyses using R 3.61 (Stuart et al., 2019; Stuart and Satija, 2019). Genes not detected in a minimum of three cells were removed. Low-quality cells with less than 200 expressed genes were excluded. Cells expressing a high number of genes (>4000–6000) were considered as putative doublets or multiplets and were removed. Dead or lysed cells were removed by exclusion of cells with a high percentage of transcripts mapping to mitochondrial genes (>20–30%). Lastly, we removed cells expressing a high percentage of transcripts mapping to dissociation-associated genes (DAGs; 20–30%) as we considered these cells to be stressed or dying, as previously reported (O’Flanagan et al., 2019). For each of these parameters, outlier cells were removed in relation to the cell distribution within the given dataset. Sham, day 4 AngII and day 28 AngII datasets were first pre-processed and filtered individually, and then merged for all subsequent analyses using the “merge” function in Seurat.

Normalization, dimensionality reduction, clustering and cell annotation

To remove technical variation while preserving biological variation, data was normalized using SCTransform as implemented in Seurat v3.1 (Hafemeister and Satija, 2019). SCTransform uses regularized negative binomial regression to normalize the data, find variable features and scale the data. Highly variable features were selected (3000 by default) using the variance-stabilizing transformation (vst) method in SCTransform. Mitochondrial gene percentage and the number of counts (nCount_RNA) were regressed out. Dimensionality reduction was performed using principal component analysis (PCA) and the most statistically significant PCs were chosen for subsequent clustering as determined by examination of the standard deviation of each principal component depicted on an elbow plot. Graph-based clustering was performed using the FindNeighbors and FindClusters functions. Non-linear dimensionality reduction and visualization was performed using Uniform Manifold Approximation and Projection (UMAP). In order to assess the finer division of resident cardiac macrophages, we assessed each parent cluster separately and increased resolution (Cluster-1 resolution 0.3, Cluster-2 resolution 0.3, and Cluster-3 resolution 0.2). Clusters were identified and annotated based on differential gene expression testing using the Wilcoxon Rank Sum Test. In particular, we used the following parameters in the FindAllMarkers function: min.pct: 0.2; logFC threshold: 0.2; adjusted p-value < 0.05. Cluster-defining genes were found in this manner for all heatmaps and venn diagrams. In heatmaps, the top 30 differentially expressed genes of each cluster were shown, and each cluster was downsampled to 50 cells for visualization. The FindMarkers function was used to find differentially expressed genes of two clusters relative to each other, such as those used for pathway enrichment analyses

Differential gene expression across conditions

To compare resident cardiac macrophage gene expression across conditions (sham, day 4 and day 28), we used the “subset” function to separate each resident cardiac macrophage state and computed differentially expressed genes of day 4 relative to sham, or day 28 relative to sham, including both upregulated and downregulated genes (FindMarkers function; min.pct: 0.2; logFC threshold: 0.2; adjusted p-value < 0.05). These genes were used for subsequent pathway enrichment analyses across conditions, where indicated.

Pathway enrichment analysis

gProfiler (https://biit.cs.ut.ee/gprofiler/gost) was used to measure over-representation of our defined gene list against the Gene Ontology (GO) database (http://www.geneontology.org). We focused on enriched biological processes of GO (BP, data version released in 2020) and reported the enrichment score (−log₁₀ of the adjusted p-value) for each pathway.

Harmony batch correction

We integrated data using Harmony to correct for batch effects in experiment 1 and 2 of the mouse scRNA-seq and the two human scRNA-seq samples (Korsunsky et al., 2019). Harmony adjusts PCA embeddings, and these harmony embeddings were used to re-cluster the data.

Garnett

Garnett is an R package for the automated classification of cell types in scRNA-seq data (https://cole-trapnell-lab.github.io/garnett/) (Pliner et al., 2019).The machine learning algorithm first trains a classifier in order to attain a gene expression profile for each cell type. The check_markers() function and the murine database is used to validate that the chosen cell type markers have low ambiguity for the respective cell type. The train_cell_classifier() function was used to train our murine resident cardiac macrophage classifier based on the expression profile of the provided gene signature (upregulated genes relative to monocytes). The predictive model was built using the default parameter of 500 unknown type cells as the outgroup and a minimum of 8 representative cells were required to include the cell type in the predictive model. Cells in the human datasets were tested using the classify_cells() function using the default parameters.

QUANTIFICATION AND STATISTICAL ANALYSIS

In all experiments, data are presented as mean ± S.E.M unless otherwise indicated. Statistical tests were selected based on relevant assumptions about data distribution and variance. Student’s t test with Welch correction (two-tailed) was used for statistical analysis of differences between two groups. One-way ANOVA was used for statistical analysis of differences between more than 2 groups. Two-way ANOVA with multiple comparisons was used for statistical analysis of differences between more than 2 groups with 2 different independent variables. Significant differences were defined at p < 0.05. All statistical analyses were performed using GraphPad Prism software. Sample sizes were chosen according to standard guidelines. Number of animals is indicated as “n”. Statistical details for graphs can be found in respective figure legends.

Supplementary Material

MMC3

Table S2: Differentially expressed genes of three resident cardiac macrophage clusters. Related to Figure 2.

MMC4

Table S3: Differentially expressed genes of nine transcriptional cell states of resident cardiac macrophages using three strategies. Related to Figure 3.

MMC6

Table S5: Differentially expressed genes in total immune cells and macrophage clusters in human neonatal and adult heart failure samples. Related to Figure 6 & S6.

MMC7

Table S6: qPCR primers to assess changes in inflammatory cytokine production following cardiac RM depletion. Related to Figure 4 & S4.

MMC2

Table S1: Differentially expressed genes in resident cardiac macrophages and monocytes. Related to Figure 2 & S2.

MMC5

Table S4: Differentially expressed genes in resident cardiac macrophage cell states across conditions. Related to Figure 3.

Highlights:

• Cardiac resident macrophage (RM) subsets respond differentially to hypertension

• Cardiac RM-derived IGF-1 drives compensatory cardiac muscle growth to hypertension

• Loss of cardiac RM-derived IGF-1 during hypertension leads to heart failure

• An IGF1-expressing cardiac macrophage subset is conserved in human heart failure

Inclusion and diversity:

We worked to ensure sex balance in the selection of non-human subjects. One or more of the authors of this paper self-identifies as an underrepresented ethnic minority in science. One or more of the authors of this paper self-identifies as a member of the LGBTQ+ community.