As previously reported (13), the number of T cells in the heart increased after TAC ( Figure 1A and Supplementary Figures 1A–D ). Our preliminary data suggest that most CD8⁺T cells in the heart before TAC showed CD44⁺CD62L^high memory phenotype, and CD44⁺CD62L^low effector fraction slightly increased after TAC (data not shown). To determine the effects of T cells on heart failure, we performed TAC on T cell-deficient CD3ϵ^-/- mice as a model for heart failure. T cell deficiency can affect macrophage functions. Echocardiography showed an increase in left ventricular posterior wall thickness at diastole (LVPWd) 2 to 4 weeks after TAC, followed by a delayed increase in left ventricular diastolic diameter (LVDd) ( Figures 1B, C ). Left ventricular ejection fraction (LVEF) decreased 2 weeks after TAC and continued to decline ( Figure 1D ). In contrast, although CD3ϵ^-/- mice showed a rapid increase in LVPWd 2 weeks after TAC, LVDd was smaller in CD3ϵ^-/- mice than in WT mice, and LVEF was maintained at high levels in CD3ϵ^-/- mice compared with WT mice ( Figures 1B–D ), indicating that cardiac dilatation was suppressed and cardiac output was effectively maintained in CD3ϵ^-/- mice. Eight weeks after TAC, LVDd was strongly suppressed in CD3ϵ^-/- mice, and heart size was smaller in CD3ϵ^-/- mice than in WT mice ( Figures 1C, E ). Furthermore, CD3ϵ^-/- mice had lower mRNA levels of Nppb, which encodes brain natriuretic peptide (BNP), a marker of heart failure, as well as lower levels of Col1a1, a fibrosis marker, compared to WT mice ( Figure 1F ). These data indicate that in the genetic absence of T cells protects the heart from heart failure induced by pressure overload.

However, we noticed that the LVPWd was greater in CD3ϵ^-/- mice than in WT mice 2 weeks after TAC, indicating that the myocardium had thickened by the absence of T cells ( Figure 1B ). Histological analysis confirmed the cardiac hypertrophy is caused by cardiomyocyte enlargement in CD3ϵ^-/- mice ( Figure 1G ). Indeed, heart weight and wall thickness were consistently larger in CD3ϵ^-/- mice than in WT mice; this difference was especially marked 2 weeks after TAC ( Figures 1B, H ). The tendency of higher HW/BW in sham CD3ϵ^-/- mice is probably due to smaller BW of CD3ϵ^-/- mice (sham WT 24.56 ± 1.45g, sham CD3ϵ^-/- 22.46 ± 1.56g), though there was no difference in BW after TAC (TAC WT 24.50 ± 1.22g, TAC CD3ϵ^-/- 24.93 ± 1.48g). The hearts of CD3ϵ^-/- mice enlarged rapidly 2 weeks after TAC but did not enlarge progressively and did not show severe fibrosis 6 to 8weeks after TAC ( Figures 1F, I ). Figure 1I showed a representative viewing of the heart. We confirmed the there is no difference in collagen at mRNA levels before TAC (data not shown). This implies that heart hypertrophy does not always lead to heart failure; accordingly, we hypothesized that the cardiac hypertrophy seen in the CD3ϵ^-/- mice in the early stages after TAC is cardioprotective.

To confirm the presence of the cardioprotective hypertrophy in the early stages after TAC, we performed RNAseq analysis on isolated cardiomyocytes 2 weeks after TAC treatment ( Figure 2A ). Sarcomeric cardiac β-myosin heavy chain gene (Myh7) has been shown to be associated with myocardial hypertrophy (16, 17). Natriuretic Peptide A (Nppa) has secreted by cardiac stretch stimulation, and collagen type I alpha 1 Chain (Col1a1) is a marker of subsequent fibrosis (13). The expression levels of Myh7, Col1a1, and Nppa were higher in CD3ϵ^-/- mice than in WT mice ( Figure 2A ), and we confirmed that TAC induced Myh7 more intensively in CD3ϵ^-/- mice than in WT mice ( Figure 2B ).

The cardioprotective cytokine receptors Egfr, Osmr, Pdgfra, Igf1r, Il1rl1 (receptors for EGF, OSM, PDGF, IGF-1, and IL-33, respectively) and macrophage colony stimulation factor (Csf1) were also higher in CD3ϵ^-/- mice than in WT mice ( Figure 2A ). These data further confirmed that cardioprotective hypertrophy occurs in CD3ϵ^-/- mice at an early stage after TAC.

Next, to determine which CD4⁺ or CD8⁺T cells are involved in cardioprotective hypertrophy, we depleted these cells using CD4- or CD8-specific antibodies. Antibodies to CD4 and CD8 were administered every three days for 2 weeks after TAC and heart weight and Myh7 mRNA levels were measured ( Figures 2C, D and Supplementary Figures 2A, B ). Anti-CD4 antibody alone had little effect, while anti-CD8 antibody or a combination of anti-CD4 and anti-CD8 antibodies increased heart weight and promoted Myh7 expression as observed in CD3ϵ^-/- mice. CD8⁺T cell depletion decreased the expression of Nppb, a heart failure marker (13), 6 weeks after TAC ( Figure 2E ).

We determined whether antigen recognition of CD8⁺T cells is required for cardioprotective hypertrophy through experiments with OT-I mice, in which most CD8⁺T cells recognize the OVA peptide/class I MHC complex. We transferred CD8⁺T cells isolated from the spleens of WT (WT-CD8⁺T) or OT-I (OT-I-CD8⁺T) mice into CD3ϵ^-/- mice, then performed TAC. One week after TAC, CD8⁺T cell infiltration into the heart was significantly higher in mice transferred with WT-CD8⁺T cells than in those transferred with OT-I-CD8⁺T cells ( Figure 2F Supplementary Figures 2C, D ). Heart weight at 2 weeks post-TAC was lower in mice transferred with WT-CD8⁺T cells than in those transferred with OT-I-CD8⁺T cells ( Figure 2G ). Histological analysis supports that WT-CD8⁺T cell transfer suppress cardiac hypertrophy, but not in OT-I-CD8⁺T transfer group ( Figure 2H ). Since OT-I-CD8⁺T cells and WT-CD8⁺T cells might infiltrate and proliferate in the heart differently, this experiment suggests a potential importance of antigen recognition of CD8⁺T cells.

These results indicate that the absence of CD8⁺T cells enhanced early cardioprotective hypertrophy induced by pressure overload.

Next, to define the mechanism underlying CD8⁺T cell depletion-mediated cardioprotective hypertrophy, we performed scRNAseq analysis of cardiomyocytes isolated from mice 2 weeks after TAC using Langendorff perfusion (16). We have previously shown that, during early hypertrophy, cardiomyocytes activate mitochondrial metabolism and protein synthesis genes, and that their expression levels are correlated with cell size (16). In the present study, principal component analysis (UMAP) revealed dramatic differences among WT mice, CD8⁺T cell-depleted mice, and CD3ϵ^-/- mice in gene expression in cardiomyocytes after TAC ( Figure 3A ). Expressions of the hypertrophy marker Myh7 and the tissue repair-related genes Igf1r and Csf1 were most evident in CD8⁺T cell-depleted mice and CD3ϵ^-/- mice compared with WT mice ( Figure 3B ), which is consistent with total RNAseq data ( Figure 2A ).

We used the Random Forests machine learning algorithm to extract modules that were significant for cell classification, and identified modules M6 and M12 as the gene modules that were most drastically different between WT mice and CD8⁺T cell-depleted or CD3ϵ^-/- mice ( Figures 3C, D ). M6 was enriched for genes involved in the oxidation-reduction process, the TCA cycle, and fatty acid beta-oxidation, while M12 was enriched for genes involved in translation, ATP metabolism, and muscle contraction. The representative genes that form these modules are shown in Figures 3E, F . Cardiomyocytes of CD8⁺T cell-depleted mice and CD3ϵ^-/- mice showed lower M6 expression and higher M12 expression compared with those of WT mice. Pseudo-time analysis is a powerful tool for the transcriptional dynamics during the trajectory of cardiomyocyte remodeling as reported previously (16). The pseudo-time calculated by trajectory analysis confirmed that the M6/12 genes and cardioprotective cytokine receptor genes were sorted in pseud-time order ( Figures 3G, H ). Decrease in M6 genes suggest reduced oxidative stress and increase of M12 genes suggests anabolism and preservation of muscle functions. The resulting metabolic and translational signature in cardiomyocytes may be responsible for the observed myocardial protection and subsequent alleviation of heart failure.

Macrophages have been shown to play essential roles in the pathology of TAC. For example, bone marrow-derived CD72^hi macrophages and CD68^hi inflammatory macrophages have been shown to promote cardiac injury (18, 19), while an increased M2-type macrophage count protects against heart injury (20). T-cell depletion promotes the expression of Csf1, a macrophage colony stimulating factor (M-CSF) that usually promotes M2 type macrophage differentiation (21) ( Figures 2A , 3B ). Therefore, we hypothesized that the presence of CD8⁺T cells modifies macrophage status which leads to cardioprotective hypertrophy.

To verify this hypothesis, we first performed FCM analysis of cardiac inflammatory cells 1 week after TAC, when cardiac hypertrophy was not very evident. At this time point, TAC increased the number of CD11b⁺ cells and F4/80⁺ macrophages as well as Ly6c^low cardiac tissue-resident macrophages, and CD3ϵ^-/- mice showed much higher numbers of these cells compared to WT mice ( Figure 4A and Supplementary Figures 1E, F ). A similar tendency was observed in CD8⁺T cell-depleted mice, but not in CD4⁺T cell-depleted mice. The deletion of CD8⁺T cells had no effect on the number of CD4⁺T cells ( Figure 4B ). The number of resident and infiltrated macrophages did not change in the spleens of WT, CD8⁺T cell-depleted and CD3ϵ^-/- mice after TAC ( Figure 4C ), suggesting that T cell deficiency increased monocytes and macrophages in the heart in a tissue-dependent manner.

To analyze the differences in the monocyte and macrophage fractions in detail, we performed scRNAseq analysis of the CD45⁺ mononuclear cells found in the heart. We identified macrophage/monocyte by the expression of Csf1r, Lyz1, and Lyz2. Among them, cluster 4 and 6 were identified as monocytes with low expression of Adgre1 (F4/80) and Cd68. In particular, cluster 6 has high expression of Cd300e.　Cluster1, 3, and 5 were identified as CCR2^- macrophage and others are CCR2⁺ macrophages (cluster 2, 7, 9, 10, 12, and 17) ( Figures 4D, F and Supplementary Table 2 ). At steady-state levels, the majority of monocytes/macrophages in the heart were Ly6c^lowCCR2^- resident macrophages (clusters 1, 3, and 5) in both WT and CD3ϵ^-/- mice ( Figures 4D, E, G ). One week after TAC, these Ly6c^lowCCR2^- clusters drastically decreased while Ly6c^low CCR2⁺ clusters 2 and 7 increased ( Figures 4D, E, G ). It is notable that TAC increased clusters 4 and 6, which represent CD11b⁺F4/80^- monocytes, in both CD3ϵ^-/- mice and CD8⁺T cell-depleted mice compared with WT mice ( Figure 4G ). Cluster 17, containing Ly6c^lowCCR2⁺ macrophages, was specifically increased in CD3ϵ^-/- mice and strongly expressed Arg1, a cardioprotective cytokine (9) ( Figure 4F ). Cluster 5, which also contained Ly6c^lowCCR2⁺ macrophages, was increased in CD8⁺T cell-depleted TAC mice, but not in CD3ϵ^-/- mice, compared with WT TAC mice; these cells strongly expressed Cd163, a M2 type macrophage marker ( Figure 4F ). Although there are some differences in the subclusters of macrophages between CD8⁺T cell-depleted mice and CD3ϵ^-/- mice, the overall picture suggests the depletion of CD8⁺T cells facilitated the infiltration of monocytes into the heart and the expansion of Ly6c^lowCCR2⁺ macrophages.

Ly6c^highCCR2⁺ cells are usually bone marrow-derived inflammatory macrophages, but here they showed both M1- and M2-like phenotypes expressing Nos1, Trl2/4, Arg1, and Msr1, probably because their seven days of residence in the heart modulated their phenotypes, in keeping with our previous finding that Ly6c^highCCR2⁺ cells can convert from inflammatory cells into repairing M2-like cells expressing Msr1 and IGF-1 in the brain from day 3 to day 6 after ischemic brain injury (22). Because clusters 4, and 9 also highly expressed Msr1 ( Figure 4F ), these data suggest that a lack of CD8⁺T cells increases the numbers of both resident and infiltrated macrophages and monocytes and also promotes phenotypic conversion into M2-type like cells.

To further investigate the difference in monocyte/macrophage differentiation between WT, CD3ϵ^-/-, and CD8⁺T cell-depleted TAC mice, we performed RNA velocity analysis against monocyte/macrophage subclusters ( Figures 5A, B ). In sham mice, the direction of differentiation of resident macrophages was similar among WT and CD3ϵ^-/- mice. In TAC mice, the direction of Ly6c^hi monocytes (clusters 0, 7, and 12) was similar among WT and CD3ϵ^-/- mice ( Figure 5C ). There seems to be two types of monocyte-like cells (clusters 3 and 5), which are probably a different population from the peripheral blood monocyte (cluster 7: Ly6C^high, 12: Ly6C^low). There was a direction from cluster 3 to cluster 5 and 0 or cluster 10, with a particularly strong direction to cluster 5 and 0 in WT TAC mice. Whereas, in CD3ϵ^-/- TAC mice, there was strong direction from cluster 5 to cluster 3 and 10. In CD3ϵ-/- and CD8⁺T cell-depleted TAC mice, the direction from cluster 3 to cluster 4 was also observed. Especially in CD8⁺T cell-depleted TAC mice, there was a strong supply flow from cluster 3 to resident macrophages (cluster 1, 2, 4, and 8). In addition, our velocity analysis indicated that CD3ϵ^-/- TAC mice-specific cluster 13 originated from cluster 10 ( Figure 5C ). These results showed a pathway from cluster 3 and 5 (CCR2⁺monocyte-like cells) to cluster 10 and 13 (resident macrophages) and this flow was facilitated in the absence of CD8⁺T cells.

Next, we compared the characteristics of cardiac monocytes/macrophages whose numbers were increased by TAC. As shown in Figures 6A, B , clusters expressing Csf1, Il10, Ptgs2, Areg, and Msr1 were increased in CD3ϵ^-/- mice and CD8⁺T cell-depleted mice compared with WT mice, suggesting that the macrophages and monocytes whose numbers were increased in CD3ϵ^-/- mice and CD8⁺T cell-depleted mice had M2-type like characters. We also observed that the monocytes and macrophages whose numbers were increased in CD3ϵ^-/- mice and CD8⁺T cell-depleted mice expressed higher levels of tissue-repairing factors such as Igf1, Areg, and Osm (9, 23, 24); Igf1 was upregulated in monocyte/macrophage clusters 2, 8, 10, and 13 ( Figure 6C , left panel), Osm was particularly upregulated in monocyte/macrophage cluster 0, 6, 7, and 13 ( Figure 6C , right panel), and Areg was diffusely expressed in the clusters of Ly6c^low ( Figure 6C , middle panel). Upregulation of these genes in mononuclear cells depleted of T cells and B cells was confirmed through conventional RT-PCR analysis ( Figure 6D ).

From RNA velocity analysis and expression pattern, we propose that the absence of CD8⁺T cells creates a niche for monocytes and allows some of these cells to proliferate and differentiate after TAC into Ly6c^lowCCR2⁺ macrophages expressing tissue-repairing growth factors that cause cardioprotective hypertrophy.

C57BL/6N mice were purchased from CLEA JAPAN. Cd3ϵ^-/- and OT-I-TCR transgenic mice have been described previously (31). All mice had a C57BL/6 genetic background. Male mice, aged 8 weeks and weighing 20–25 g, were used under cohousing conditions in specific pathogen-free facilities. All animal experiments were approved by the Animal Research Committee and Ethics Committee of Keio University School of Medicine.

Mice underwent TAC to induce heart failure or a sham operation (32). In TAC, the transverse aorta was constricted with a 7-0 nylon suture across a 27-gauge blunt needle, which was removed after constriction. Transthoracic echocardiography was performed with a Vevo 2100 instrument (VisualSonics) equipped with an MS-400 imaging transducer. Mice were anesthetized during the echocardiographic examination, and heart rate was maintained at 500 to 550 bpm in all mice. M-mode echocardiographic images were obtained from the longitudinal axis to measure left ventricular size and function.

Each mouse was deeply anesthetized and intracardially perfused with 40 mL of ice-cold PBS to exclude blood cells. The heart was dissected, minced with fine scissors, and then enzymatically digested with a cocktail of type II collagenase (Worthington), elastase (Worthington), and DNase I (Roche Diagnostics) for one hour at 37°C with gentle agitation as described elsewhere (33). After digestion, the tissue was triturated and passed through a 70-µm cell strainer. Leukocyte-enriched fractions were isolated through 40% to 65% Percoll (GE Healthcare) density gradient centrifugation as described elsewhere. Cells were removed from the interface and washed with RPMI-1640 supplemented with 2% FBS cell culture medium for further analysis. To prevent the nonspecific binding of antibodies to Fc receptors, isolated cells were first incubated with anti-CD16/32 antibody (2.4G2) at 4°C for five minutes. FCM data were acquired with a FACSCanto II (BD Biosciences) and analyzed with FlowJo 9.9.3 software (FlowJo). All antibodies and reagents are described in Supplementary Table 1 .

The heart tissue was fixed in paraformaldehyde overnight, and 5-µm tissue sections were cut and stained with Masson’s trichrome staining or Wheat Germ Agglutinin (WGA)-Rhodamine conjugate (Vector) and DAPI. Measurements and analysis were performed using a BZ-X800 fluorescence microscope and accompanying software (Keyence).

Cardiomyocytes were isolated using Langendorff perfusion from the left ventricular free wall 2 weeks after TAC or sham operation. Echocardiography was used to assess whether the heart was appropriately exposed to pressure overload. Mice whose hearts had not been appropriately exposed to pressure overload were excluded from single-cardiomyocyte RNA-seq analysis. Enzymatic dissociation using Langendorff perfusion was performed as described previously (16). To prevent hypercontraction, the cardiomyocytes were resuspended in medium (NaCl 130 mM, KCl 5.4 mM, MgCl2 0.5 mM, NaH2PO4 0.33 mM, D-glucose 22 mM, HEPES 25 mM, FBS 0.2%, pH 7.4) containing a low concentration of calcium (0.1 mM). For bulk RNA-seq of myocardium, cardiomyocytes were lysed and mRNA was extracted using RNeasy Micro Kit (QIAGEN) as described (34). The analysis was performed by GENEWIZ. For scRNAseq, rod-shaped live cardiomyocytes were collected immediately after isolation from two mice with a 0.2–2-µL pipette (sample volume, 0.5 µL) and incubated in lysis buffer. Single-cell cDNA libraries were generated using the Smart-seq2 protocol and the efficiency of reverse transcription was assessed by examining the cycle threshold (Ct) values of control genes (Tnnt2) from quantitative real-time polymerase chain reaction (q-PCR). The following primer sets were used for qPCR: Tnnt2 mRNA forward, TCCTGGCAGAGAGGAGGAAG; Tnnt2 mRNA reverse, TGCAGGTCGAACTTCTCAGC. A Ct value of 25 was set as the threshold. The libraries were sequenced on a HiSeq 2500. RefSeq transcripts (coding and non-coding) were downloaded from the UCSC Genome Browser (http://genome.ucsc.edu). Reads were mapped to the mouse genome (mm9) with the parameters “-g 1 -p 8 mm9—no-coverage-search” using TopHat (35). FPKM was calculated with reads mapped to the nuclear genome (36). We removed the low-quality cardiomyocytes with less than 3,000 detected genes for the downstream analysis. Analysis of the single-cell transcriptome profiles was performed using the R statistical language and environment (16). For weighted co-expression network analysis, all genes expressed at an FPKM value of ≥5 in at least 5 cardiomyocytes were used to construct a signed network using the WGCNA R package (37). Gene ontology analysis was performed using DAVID (38). For pseudotime analysis, Slingshot was used to detect trajectories (39).

One week after TAC surgery, heart CD45⁺ cells were isolated using an autoMACS Pro Separator by positive selection after Percoll enrichment. To obtain single-cell RNA-seq data, we used an BD Rhapsody system (BD Biosciences) and TAS-Seq as previously reported (40). A single-cell suspension of ~20,000 cells was captured from all isolated cells, without selection, on an array of >200,000 microwells through a limited dilution approach. Beads with oligonucleotide barcodes were added to saturation so that one bead was paired with each cell in a microwell. After exposure to cell lysis buffer, polyadenylated RNA molecules hybridized to the beads. Beads were retrieved into a single tube for reverse transcription and Exonuclease treatment. cDNA amplification was performed by using TAS-Seq approach. Sequencing libraries were prepared by using NEBNext Ultra II FS library prep kit for Illumina (New England Biolabs). The sequencing analysis were performed by ImmunoGeneTeqs, Inc (Chiba, Japan) by a Novaseq 6000 sequencer (Illumina, San Diego, CA, USA) and a Novaseq S4 200 cycles v1.5 kit (Illumina).

Pair-end fastq files (R1: cell barcode reads, R2: RNA reads) of TAS-Seq data were processed as a previous report. For assignment of each tag to each cell barcode, read counts of each tag in each valid cell barcode, which is defined by the cDNA matrix, were extracted from the tag/cell barcode expression matrix. Unassigned cell barcodes were labeled as “not-detected” cells. Then, a sum of the total read counts of each tag was normalized to the minimum sum count of each tag, and log₂ fold-change between first most tag counts and second most tag counts within each cell barcode. Each cell barcode was ranked by the fold-change ascending order, and the top 3.28% cells were identified as doublets, which were theoretically detectable doublets calculated by the Poisson’s distribution based on the number of loaded cells, total Rhapsody well number, and the number of tags used. Finally, remained cell barcodes were assigned to the first most counted tags.

Clustering of single cells of each dataset was performed by using Seurat v2.3.4 (41) in R 3.6.3. Seurat object for each dataset was created by using CreateSeuratObject function (min.cells=5, min.genes=500). Cells of which mitochondrial gene proportion over 0.4 was filtered out by FilterCells function in Seurat v2.3.4. On the mitochondrial gene proportion vs nGene plot, the nGene low and mitochondrial gene high dead cells are mainly mitochondrial gene proportion above 0.4. The population between 0.25 and 0.4 was also minor, thus we filtered out mitochondrial gene proportion over 0.4. The expression data was normalized by normalizeData function (scale.factor = 1,000,000), and scaled with ScaleData function in Seurat v2.3.4. Read counts of each cell within each dataset were regressing out as a confounding factor within ScaleData function. Highly-variable genes of each dataset were identified using FindVariableGenes function in Seurat v2.3.4. with the following parameters: mean.function = ExpMean, dispersion.function = LogVMR, x.low.cutoff = 0.1, x.high.cutoff = Inf, y.cutoff = 0.5. Then, principal component analysis (PCA) against identified highly-variable genes and projection of PCA onto entire data was performed using RunPCA (number of calculated PCs were 100) and ProjectPCA functions in Seurat v2.3.4. Enrichment of each PC was calculated using JackStraw function (num.replicate = 100), and PCs that were significantly enriched statistically (p ≤ 0.05) were selected for clustering and dimensional reduction analyses. Cell clustering was performed using FindClusters function (resolution = 1.5) in Seurat v2.3.4 against the significant PCs, and dimensional reduction was performed using python wrapper of Fast interpolation-based t-stochastic neighbor embedding (FIt-SNE) (42) v1.2.1 (perplexity = 100, df = 0.9, random_seed=42, max_iter=1000, and all the other parameters were set as defaults) through reticulate package (https://github.com/rstudio/reticulate) in R 3.6.3. Statistically significant marker genes of each identified cluster were identified using parallelized FindMarkers function in Seurat v2.3.4 (test.use=“wilcox”, only.pos=TRUE, min.pct=0.1, logfc.threshold=0.25, adjusted p (Bonferroni correction) ≤ 0.05). Then, each identified cluster was manually annotated by their marker genes that were previously reported as cell subset-defining marker genes, and the lineage marker double-positive cells were annotated as doublets. Next, we further sub-clustered monocytes/macrophages using Seurat v2.3.4 by the similar workflow of whole-cell data, and incorporate their annotation into Seurat object of whole-cell data (resolution = 2, perplexity=50, df = 0.9). Cell subset annotations and their compositions were visualized in 2D FIt-SNE space and stacking plot, respectively. All of the identified marker genes are shown in Supplementary Table 2 .

For RNA velocity analysis, TAS-Seq data cDNA reads were mapped to reference genome (build GRCm38_101) by using HISAT2-2.2.1 (43) by the following parameters: -q -p 6 –rna-strandness F –very-sensitive –seed 656565 –reorder –omit-sec-seq –mm. For HISAT2 index build, Corresponded ensembl gtf file was filtered to retain protein-coded RNA, long non-coding RNA, and T cell chain/immunoglobulin chain annotations according to 10X Genomics’s method (https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/advanced/references#mkgtf). Then, cell barcode information of each read were added to the HISAT2-mapped BAM files, and associated gene annotations were assigned by using featureCounts v2.0.2 (44) by following parameters: -T 2 -Q 0 -s 1 -t gene -g gene_name –primary -M -O –largestOverlap –fraction -R BAM. In featureCounts analysis, “gene” annotation was used for capturing un-spliced RNA information for RNA velocity analysis, and primary annotations were kept. Then, resulted BAM file was split by valid cell barcodes by using nim 1.0.6 and hts-nim v0.2.23. Resulted BAM files were processed into loom files by using velocyto run with -c and -U option, and the loom files were concatenated by loompy.combine function (45). RNA velocity estimation of sub-clustered monocytes/macrophages was performed by using scVelo (46) with reticulate package in R 3.6.3. Briefly, gene filtering and normalization was performed by filter_and_normalize function (min_shared_counts=as.integer (30), min_shared_cells=5), and moments were calculated pp.moments function (n_pcs=32, n_neighbors=30). The number of principal components was determined by the significantly-enriched principal components identified by Jackstraw function of Seurat analysis. RNA velocity was estimated by tl.velocity function with dynamical mode and visualized by pl.velocity_embedding_stream function in 2D FIt-SNE space.

mRNA was extracted from renal tissues using an RNAiso kit (TaKaRa Bio). Total RNA was reverse-transcribed to cDNA using a high-capacity cDNA reverse transcription kit (Applied Biosystems). qPCR was performed on the cDNA samples using EvaGreen (Bio-Rad). The relative quantification value was expressed as 2^−ΔCt, in which ΔCt is the difference between the mean Ct value of triplicate measurements and the endogenous Hprt1 control. CD11b⁺ cells (CD45⁺CD11b⁺CD3^-CD19^-Ly6g^-) were sorted on a BD Aria III for qPCR analysis (47).

For CD4⁺ or CD8⁺T cell depletion, mice received an i.p. injection of 100 µg anti-CD4 antibody (GK1.5) or 400 μg anti-CD8 antibody (3.115) on day 0 and every three days until 2 weeks after TAC. For transfer experiments, CD8⁺T cells were isolated from the spleens of WT and OT-I mice using an autoMACS Pro Separator by negative selection (B220^-CD4^-CD11b^-CD11c^-CD19^-CD105^-Gr1^-NK1.1^-TCRγδ^-Ter119^-). The CD8⁺T cells (1 × 10⁶ cells) were transferred into CD3ϵ^-/- mice five days prior to the TAC operation.

Data are expressed as means ± SEM. Statistical significance was determined by one-way ANOVA followed by post hoc Tukey’s multiple-comparisons tests to analyze differences among three or more groups, and by unpaired Student’s t-test to analyze differences between two groups. p < 0.05 was considered to indicate a significant difference. Statistical analysis was performed using Prism 8.0 (GraphPad Software).