Male C57BL/6J mice aged 8 weeks were purchased from Shulaibao Biotechnology Co., Ltd. (Wuhan, China) and housed in a specific pathogen-free facility in Animal Center of Tongji Medical College, Huazhong University of Science and Technology. All animal studies were performed according to the principles of the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Animal Care and Use Committee of Union Hospital, Huazhong University of Science and Technology, China.

Mice ischemia models were performed according to the protocol we previously described (24). Briefly, after being anesthetized by intraperitoneal injection (1% sodium pentobarbital solution, 50mg/kg), the mice were intubated and connected to a ventilator. Then, the heart was exposed to find the left anterior descending coronary artery (LAD) under a stereo microscope, and the LAD was then ligated at about 1-2mm below the left atrial appendage with a 6-0 suture needle. For the MI model, the anterior wall of the left ventricle below the ligation line turned pale, and the ligation line was not released, while for the MIRI model, the ligation wire was released after 30 minutes, and the ventricular wall below the ligation point gradually returned to pink. The silk thread was passed under the LAD but without ligation for sham models. Finally, the thorax was closed and the mice fed for 3 days. We gave 100ul of buprenorphine (0.1mg/mL, subcutaneous injection) to relieve the pain of the mice when they returned to autonomic respiration after the operation, and additional doses were given every 4-6 hours for 24 days after surgery. In addition, ibuprofen (0.2 mg/mL) was added to drinking water 2 days before surgery and ≤3 days after surgery.

Three days after operation, the mice were sacrificed by cervical dislocation, and the hearts were quickly removed after the blood in the ventricular chambers was washed using PBS. Then, the infarct area in the left ventricle of the heart was chopped up and digested with collagenase B (1mg/ml) according to the product instruction. RPMI 1640 medium supplemented with 10% FBS was used to stop the enzyme digestion reaction. Then, the cardiac mononuclear cell suspension was obtained by density gradient centrifugation with Percoll solution.

The cardiac mononuclear cell suspension was separated into EP tubes with 5 × 10⁶ cells per tube; one sample from each group was selected to sort ILCs and each sample consisted of pools of ten mice. Then, FITC-labeled Mouse Lineage Antibody Cocktail, APC/Cy7 labeled anti-mouse CD45 antibody and PE-labeled anti-mouse CD127 antibody were added and incubated with the cells for 30 minutes at 4°C in the dark. ILCs were sorted on a flow cytometer (LE-SH800) and 7-AAD was added to label the dead cells.

The ILC cells obtained by flow sorting were subjected to single-cell RNA sequencing on the 10X Genomics DNBSEQ platform. The prepared ILC suspension was then partitioned into Gel Beads in Emulsions (GEMs) in the automated Chromium Controller, and then mRNAs were reverse- transcribed into cDNAs. A reaction system was configured for breaking GEMs; after reacting at the suitable temperature for a fixed period of time, the products were purified. The PCR reaction system was prepared and the reaction time was suitable for a certain temperature. After that, the cDNAs were subjected to fragmentation, end repair, “A” base addition and adaptor ligation. The PCR reaction system was configured. After reacting at the suitable temperature for a fixed period of time, amplification was processed via PCR. The circularization program was then performed to produce single-stranded cyclized products and digested DNA molecules that were uncyclized and linear. After that, rolling cycle amplification was used to replicate the single-stranded circle DNA molecules and generated DNA nanoballs (DNBs) which contained multiple copies of DNA. The high-intensity DNA nanochip technique was then used to load enough quality DNBs and combinatorial Probe-Anchor Synthesis was used for sequencing. The data generated by sequencing were collected in a FASTQ format and then aligned with the Mouse Genome (mm10) using the CellRanger pipeline (v5.0.1) (25). The data were then analyzed using Rstudio.

Cardiac mononuclear cell suspension of MI/MIRI/sham mice was used to verify ILC subsets (ILCdc, ILC2a, ILC1, ILCt). Each group contained three samples, and 3-5 C57BL/6J mice were mixed into one sample. Cardiac mononuclear cells were obtained as described above. ILCdc was identified by the fluorescently labeled antibody Lin^-CD45⁺CD127⁺CD74⁺, ILC2a was identified by the fluorescently labeled antibodies Lin^-CD45⁺CD127⁺ST2⁺, ILC1 was identified by the fluorescently labeled antibodies Lin^-CD45⁺CD127⁺NKP46⁺ and ILCt were identified by the fluorescently labeled antibodies Lin^-CD45⁺CD127⁺CD2⁺. ILC subsets were sorted by the flow cell sorter Becklo-MoFlo-XDP, and 200 to 500 cells per sample were collected for subsequent qRT-PCR. In addition, we set FMO (Fluorescence Minus One) as control to improve the accuracy of the flow cytometry gating strategy when detecting cardiac mononuclear cells.

Based on RT-PCR amplification methods, the sorted ILC subsets were reverse- transcribed to obtain cDNA. cDNAs were synthesized with a Single Cell Sequence Specific Amplification Kit (Vazyme, Nanjing, China). First, the amplification primers of different genes were mixed to make an Assay Pool (the final concentration of each primer was 0.1 μM), and then the reaction system was configured in the Nuclease-free centrifuge tube as follows: 2x Reaction Mix 2.5ul, 0.1 uM Assay Pool 0.5ul, RT/Taq enzyme 0.1ul, Nuclease-free ddH20 up to 5.0 μl. Single-cell ILC subsets were added to the reaction system, and then the following reaction was performed on the PCR instrument: 50° for 60 min and 1 cycle, 95° for 3 min and 1 cycle, 95° for 15 s, 60° for 15 min and 14 cycles. At the end of the reaction, 20 μl of Nuclease-free ddH2O (1:5 dilution) was added to each tube, centrifuged at 3000 rpm (1000 × g) for 2 min, and subsequent qPCR reactions were performed immediately.

The qRT-PCR reactions were carried out in the Bio-Rad CFX96 Real-time PCR Detection System, using AceQ^® qPCR SYBR Green Master Mix (Vazyme, Nanjing, China). The program was set as a two-step method, 95° for 10 s, 60° for 30 s and 40 cycles. The gene expression results were analyzed using the 2^−ΔΔCT method, and GAPDH was used as an endogenous control for mRNA expression.

SAS was used for statistical analysis. For flow cytometry data analysis, two-tailed unpaired T-test was used to compare the differences between two groups, and one-way ANOVA was used to compare the differences between multiple groups. The data were expressed as mean ± sd, n = 3. P<0.05 indicated statistical significance.

To elucidate the cellular constitutions of ILCs in mouse heart, we performed scRNA-seq using 10X technology ( Figure 1A ). Live heart ILCs (Lin^-CD45⁺CD127⁺) from mice in the sham, MI and MIRI models were sorted by a fluorescence-activated cell sorter (FACS) ( Figures S1A–C ). The quality control of scRNA-seq is shown in Figures S1D, E . It is worth noting that the number of doublets identified in Figure S1F was very low and these cells were subsequently removed. Then, to further remove low-quality cells, cells were screened by setting the thresholds as the expression of Cd127 (Il7r) and Cd45 (Ptprc) and no expression of Cd3e, Cd3d, Cd4, Cd8a, Cd19, Cd11c(Itgax), Cd11b (Itgam), Cma1, Ly6c1, Adgre1 (F4/80) or Pdgfra. Finally, 1365, 962 and 495 cells were screened from the MI, MIRI and sham groups, respectively.

Unbiased clustering of the cells identified five clusters based on uniform manifold approximation and projection (UMAP) analyses ( Figure 1B ). We then annotated the clusters by using the FindAllMarker function to calculate marker genes for each cluster by using scType as a reference, and in combination with the ILC- specific marker genes reported in the literatures. Finally, five clusters, namely ILC1, ILC2a, ILC2b, ILCdc and ILCt, were obtained ( Figure 1B ). As shown in Figure 1C , Ptprc and Il7r were present in every cluster; at the same time, Cd4, Cd3d, Cd3e and Cd8a in T lymphoid cells, Cd19 in B cells, Itgam in myeloid cells, Itgax in dendritic cells, Ly6c1 in monocyte cells, Adgre1 in macrophages, Cma1 in mast cells and Pdgfra in fibroblasts were not expressed. Inhibitor of DNA-binding protein 2 (Id2), which is a crucial transcriptional regulator for ILC development, was present in every cluster. These results confirmed the successful purification of ILC subsets. Interestingly, the expression levels for typical ILC3 markers, including Rorc, Il17a and Il22, were very low. In clusters ILC2a and ILC2b, the expression levels of Klrg1, Gata3, Rora, Il5 and Areg were high. In addition, ILC2b highly expressed Stmn1, Birc5 and Pclaf, indicating that this cluster may be different from classical ILC2s. In cluster ILC1, Klrc1 and Ifng were highly expressed when compared to other clusters. Meanwhile, the expressions of other ILC1 markers, such as Tbx21and Tnf, were also high. DC markers, including Cd74, Cd83, H2-Ab1, H2-Aa and H2-Eb1, were highly expressed in the ILCdc cluster, indicating the potential DC-like roles this cluster may play. T-cell surface glycoprotein CD3 gamma chain (Cd3g) and lymphoid enhancer-binding factor 1 (Lef1), which are important during T cell development and activation, were highly expressed in the ILCt cluster.

In order to identify gene expression patterns in the ILCs subgroups, particularly for some novel genes, a heatmap using the top five genes in each cluster was performed as demonstrated in Figure 1D . All the differentially expressed genes (DEGs) are presented in Table S1 . In addition, the expression of representative marker genes in the cell populations is shown in Figure 1E . We also checked the expression of ILCreg- related genes, such as Id3, Il1b and Bcl2a1b, and found that they were highly expressed in ILCdc ( Figures S1G–I ). Moreover, proliferation-related gene expression shown by feature plots, such as Mki67, Pcna and Top2a, were highly expressed in the ILC2b cluster ( Figures S1J–L ).

To investigate the differential changes of ILC subsets in the ischemic myocardium of mice, we visualized them with UMAP plots ( Figure 2A ) and calculated the number and proportion of ILC subsets in the mouse MI/MIRI/Sham group ( Figure S2A ). In general, cells from every group were present in every cluster. It ‘s worth noting that the ILCdc and ILC2a clusters accounted for a large number and percentage, meaning that they are important ILC subclusters.

Next, to predict hub genes and construct gene regulatory networks in each ILC cluster, hdWGCNA was used. As shown in Figure 2B , ten major gene modules and the related top ten genes in each module were identified ( Figure S2B ). Then, after comparing with DEGs in ILCs ( Table S1 ), these gene modules were clustered into ILC clusters as shown in Figure 1B according to the genetic similarities ( Figure S2C ). GO-Biological Process and Molecular Function were then performed to predict ILC-related signal pathways ( Figures S2D, E ). ILC2a was involved in cytoplasmic translation, cotranslational protein targeting to membrane and SRP-dependent cotranslational protein targeting to membrane. Furthermore, ILC2a may have functions in large ribosomal subunit rRNA binding, ubiquitin-protein transferase inhibitor and ubiquitin ligase inhibitor activity. ILCdc was involved in the regulation of neutrophil activation and played a role in IgG binding, superoxide-generating NAD(P)H oxidase activity and oxidoreductase activity. ILC2b was involved in nuclear cell cycle DNA replication, double-strand break repair via break-induced replication and mitotic DNA replication. Additionally, ILC2b may play a role in DNA replication origin binding, single-stranded DNA helicase activity and 3’−5’ DNA helicase activity. The hub genes in each module are shown in Figure 2C . We specifically highlighted our hdWGCNA analysis for ILCdc, ILC2a and ILC2b. CD74 was predicted to be a crucial gene for ILCdc. Finally, the gene regulatory networks are demonstrated in Figure S2F . In general, the hub genes obtained here had high correlation with the aforementioned DEGs.

In order to explore the potential biological functions of ILC subsets, we used the “GSVA” R package for pathway enrichment analysis ( Figure 2D ). GSVA analysis showed that the ILC subgroups were involved in 186 signaling pathways ( Table S2 ). ILC2a was mainly involved in O-glycan biosynthesis, ABC transporters, vascular smooth muscle contraction, Notch signaling pathway, glycerolipid metabolism and inositol phosphate metabolism. ILC2b was mainly involved in base excision repair, nucleotide excision repair, mismatch repair, DNA replication, RNA polymerase, RNA degradation, basal transcription factors, spliceosome, cell cycle, homologous recombination, purine metabolism and pyrimidine metabolism. ILC1 was mainly involved in the regulation of autophagy, natural-killer-cell-mediated cytotoxicity, adipocytokine signaling pathway and basal transcription factors. Additionally, ILCdc was involved in antigen processing and presentation, chemokine signaling pathways, Toll -like receptor signaling pathway, Jak STAT signaling pathway, MAPK signal pathway, PPAR signal pathway, leukocyte transendothelial migration, apoptosis, renin angiotensin system, cardiac muscle contraction, dilated cardiomyopathy and hypertrophic cardiomyopathy. ILCt was mainly involved in T -cell receptor signaling pathway, valine leucine and isoleucine biosynthesis, RIG1-like receptor signaling pathway, adipocytokine signaling pathway, WNT signaling pathway, ECM receptor interaction and regulation of autophagy. Through GSVA analysis, we can understand the different potential biological functions of ILC clusters better.

In order to explore the gene expression of ILCs subgroups in different disease conditions, we used Monocle2 software to conduct a pseudotime analysis. As showed in Figures 3A–C , the five ILC subclusters were ordered along a trajectory, and different states with branching points were identified for the sham, MIRI and MI groups. For ILCs at rest conditions ( Figure 3A ), ILC2a and ILC2b were found in the early stage of the trajectory, and ILCdc was found in the later stage of the trajectory; meanwhile, ILC1 and ILCt were also found in the later stage but not in the same branch as ILCdc. Interesting, the ILC clusters in ischemic heart showed different trajectories. ILCdc was found in the early stage of the trajectory for both the MIRI and MI group, and ILC2a and ILC2b were found within the both two branches of the later stage of the trajectory; meanwhile, ILC1 and ILCt were within the same branch of the later stage ( Figures 3B, C ). Notable, there were two branch points for MI but only one point for both the sham and MIRI group. Marker gene expressions of different ILC clusters were further demonstrated and confirmed the cell distributions along the pseudotime trajectory ( Figure 3D , Figures S3A, B ). For example, ILCdc showed a rising trend in the sham group and a downward trajectory in the MI and MIRI groups, but ILC2a and ILC2b showed an opposite pattern.

Further analysis identified several gene cohorts that changed during the pseudotime transition ( Figure 3E , Figures S3C, D ). Notably, the expression levels of ILC2s-related genes, such as Il5 and Areg, decreased, while ILCdc-related genes, such as CD74, H2-Aa and Cst3, increased during pseudotime. Gene expression patterns were also identified between the two transition states ( Figure 3F , Figures S3E, F ). Analysis of branch point one for the sham group discovered the upregulation of ILCdc-related genes such as CD74, H2-Aa and Cst3 at cell fate 1 ( Figure S2G ). Upregulation of ILC1-related genes, such as Xcl1 and Gzmb, was revealed at cell fate 2. A similar analysis for the MIRI and MI groups was performed and is demonstrated in Figures S3G, H . However, interestingly, the expressions of ILCdc-related genes, such as CD74, H2-Aa and Cst3, were high at the pre-branch point and then decreased in both cell fate 1 and cell fate 2 in the MI group.

To further assess the mechanisms underlying the heterogeneity of ILCs infiltrating the ischemic myocardium, we used SCENIC software to predict and analyze the transcription factors that regulate ILC subsets. SCENIC allowed the identification of TF-centered gene co-expression networks. As expected, a set of TFs implicated in heart ILCs exhibited different expression patterns among the sham, MI and MIRI groups and also among different ILC clusters. We identified previously known transcription factors that regulate ILC subsets, and also identified some novel transcription factors involved in regulating the fates of ILCs subsets.

TFs with high regulon specificity scores for every cluster in different groups a re demonstrated in Figure 4A and Figures S4A, B . Tbx21 was predominantly expressed in ILC1s in all three groups in our results as a previous well-known TF for ILC1s ( Figure 4B ). Furthermore, our results revealed upregulated expression of Irf8 in the sham group and increased activity of Nfatc1 in the MI group for ILC1. For the ILCt cluster, Irf1 was the TF with high regulon specificity score in both the sham and MI groups; meanwhile, Lef1 was the special TF in the MIRI group. Interestingly, for the ILC2b cluster, Tfdb1 was the potential TF in the sham group, but there was no special TF for the MI and MIRI groups.

As shown in Figure 4A and Figures S4A, B , ILC2a and ILCdc had more specific regulons than other clusters, indicating the higher cell activities they may have. ILC2a showed high regulon activity of GATA3 in all three groups as expected, and additional increased TF regulon signatures included Rora/Nfkb/Ets/Fosb/Junb/Jund for the sham group, Hif/Crem/Fosl2 for the MI group and Ets1/Crem for the MIRI group. Furthermore, ILCdc had high regulon activity of Rel/Irf5/Tcf4/Bcl11a/Bach1/Spib, Etv6/Batf3/Hmgn3/Rel/Foxp1/Irf8 and Irf5/Relb/Batf3 for the sham, MI and MIRI groups, respectively. We specifically highlighted our SCENIC analysis for ILCdc and found that Relb was predominantly expressed in the ILCdc cluster ( Figure 4B ). Additionally, we also displayed the heatmaps of different TFs expressions in ILC clusters for different groups in Figure 4C and Figures S4C, D .

ILCs have been found to play roles though communications with other cells in several diseases. In order to study the interaction relationships among the ILC subclusters, we used the CellCall toolkit to uncover the communication networks in the sham, MIRI and MI groups. As shown in Figure 5A and Figures S5A, B , ILCdc was the prominent receiver of signals from other cells in the sham, MIRI and MI groups, suggesting the importance of ILCdc in ischemic heart. Hence, we focused on ILCdc in the following analysis. Pathway activity analysis showed that the intercellular signal from ILC2b to ILCdc was mainly enriched in the TNF signaling pathway in the sham group, JAK-STAT signaling pathway in the MIRI group, and necroptosis in the MI group ( Figure 5B , Figures S5C, D ). In addition, the intercellular signal from ILC2a to ILCdc was mainly enriched in the JAK-STAT signal pathway for the MIRI and MI groups. Further ligand–receptor interaction heatmaps demonstrated that ILC2b interacted with ILCdc through Vegfb-Kdr, Vegfc-Kdr, Flt3l-Flt3, Tnfsf11-Tnfrsf11a and Ccl4-Ccr1 in the sham group, and through Csf2-Csf2ra/Csf2rb2, Il5-Csf2ra/Csf2rb2, Lif-Il6st, Il4-Il13ra1/Il4ra/Il21r, Il6-Il6st, Il13-Il13ra1/Il2rg/Il4ra and Tnfsf11-Tnfrsf11a in the MIRI and MI groups ( Figure 5C , Figures S5E, F ). Additionally, ILC2a interacted with ILCdc through Flt3l-Flt3 and Tnfsf11-Tnfrsf11a in the sham group, but through Csf2-Csf2ra/Csf2rb2, Il5-Csf2ra/Csf2rb2 and Il13-Il13ra1/Il2rg/Il4ra in the MIRI and MI groups.

Then, Sankey diagrams were used to demonstrate the ligand–receptor–transcription factor regulatory axis (L-R-TF) among ILCs in the heart. For example, in the MI group, ILC2a may secrete Il5, which is received by Csf2ra and Csf2rb2 in ILCdc, and then transcription factor Stat4 is activated (Il5-Csf2ra/Csf2rb2-Stat4) ( Figure 5D ). It is worth noting that this conclusion was predicted by cell communication “cellcall”, which needs to be further verified by experiments. Tnfsf11-Tnfrsf1la-Relb, Il4-Il13ra1/Il2rg/Il4ra/Il21r-Stat4, Il13-Il13ra1/Il2rg/Il4ra-Stat1 and Csf2-Csf2rb2/Csf2ra-Stat1 were the other L-R-TF axis for ILC2a and ILCdc in the MI group. The L-R-TF axis between ILC2a and ILCdc in the sham and MIRI groups are also shown in Figures S5G, H . Interestingly, the messages derived from ILCdc only sent to ILCdc itself and ILC2b ( Figure 5A , Figures S5A, B ). Sankey diagrams for ILCdc-ILC2b and ILCdc-ILCdc L-R-TF axis are shown in Figures S5I, J . The ligands for ILCdc-ILC2b connection were TNF and Tgfb1. In addition, ILCdc may have an autocrine effect on itself via several ligands, such as CXCL16. As shown in Figures S5K, L , ILCdc specially expressed CXCL16 and moderately expressed Tgfb1. Furthermore, to demonstrate the regulation of downstream genes by intracellular transcription factors for ILCdc, we used GSEA to display the set of genes downstream of transcription factor regulation ( Figure 5E ). Finally, the cellular communication pattern among ILCdc and other ILCs in MI is shown in Figure 5F .

To analyze the heterogeneity of ILCdc further, ILCdc was subjected to unsupervised clustering by using UMAP. Five subclusters were revealed with unique signature genes, named ILCdc clusters 0-4 ( Figure 6A ). ILCdc cluster 0 and cluster 1 were found to be the predominant clusters. In addition, GSVA showed the pathways that ILCdc clusters may be involved in ( Figure 6B ). Next, the unique signature genes of each ILCdc cluster were demonstrated in Figure 6C . Fscn1, CD63, Gyg, Ccl22 and Zmynd15 were the top five genes for ILCdc cluster 0, and Cfp, Napsa, Ifitm1, Ifitm2 and Ifitm3 were the top five genes for ILCdc cluster 1. Pseudotime trajectory analysis of ILCdc clusters was then performed to explore the development stages of ILCdc subgroups in the MI group. As shown in Figure 6D , ILCdc cluster 2 was found to be in the earliest stage. In addition, ILCdc clusters 0 and 1 were found to be in the late stage but in different cell fate directions. Further analysis demonstrated the gene cohorts that changed during the pseudotime transition for the ILCdc cluster in the MI group ( Figure S6A ). Gene expression patterns were also identified between the two transition states ( Figure S6B ). Analysis of branch point one for the MI group discovered the downregulation of ILCdc cluster 0 genes, such as CD63 and Gyg. Upregulation of ILCdc cluster 1-related genes, such as Ifitm1 and Ifitm3, were revealed at cell state 2.

To better reveal the transcriptional characteristics of ILCdc, several analysis methods were used here. Firstly, the volcano plot in Figure 6E shows the DEGs for ILCdc in the sham and MI groups (whole DEGs are provided in Table S3 ). The expression of Gyg, H2-Q7, Map4k4, Serpinb6b, Spint2, Calm1, Wnk1 and Cd63 were more highly expressed in the MI group, while that of Mgl2, Ms4a4c, Ifi27l2a and Ifitm3 were more highly expressed in the sham group for ILCdc. Interestingly, under the same threshold, there were no DEGs between the sham and MIRI groups. Then, KEGG enrichment analysis was performed using the DEGs acquired previously. Figure 6F shows that ILCdc was involved in oxidative phosphorylation and apoptosis in the MI group.

Next, we compared the gene expression patterns of ILCdc between the sham and MI groups by using pseudotime trajectory analysis. As shown in Figure S6C , ILCdc in the sham groups was predicted to be in the earlier stage; meanwhile, ILCdc in the MI group was predicted to be in the later stage with two different cell fates. Gene cohorts that changed during the pseudotime transition and gene expression patterns between the two transition states are also shown in Figures S6D, E .

In a similar manner, we explored the transcriptional features of ILC2a. As shown in Figure S7A , ILC2a was divided into four subclusters, named ILC2a clusters 0-3. The cell number and proportion of ILC2a subclusters are shown in Figures S7B, C , indicating that ILC2a clusters 0-2 shared a similar expression level. The top five genes for ILC2a cluster 2 were Ramp3, Gata3, Isy1, Il5 and Il13, suggesting that this subcluster was the classical ILC2 ( Figure S7D ). Furthermore, the top five genes for ILC2a cluster 0 were Pkm, Ly6a, AA467197, Nme1 and Mdh1. Additionally, for ILC2a cluster 1, the top five genes were Zfp36, Fos, CD69, Atf3 and Hspa1b. The pseudotime analysis suggested that ILC2a clusters 0 and 3 were in the earliest stage, while clusters 0 and 2 were in the middle stage and cluster 1 was in the later stage ( Figure S7E ). Gene cohorts that changed during the pseudotime transition and gene expression patterns between the two transition states were also identified in Figures S7F, G .

To further verify the existence of ILC subsets, we used flow cytometry to identify ILCdc (Lin^-CD45⁺CD127⁺CD74⁺), ILC2a (Lin^-CD45⁺CD127⁺ST2⁺), ILC1(Lin^-CD45⁺CD127⁺NKP46⁺) and ILCt (Lin^-CD45⁺CD127⁺CD2⁺). Heart and spleen mononuclear cells in the MI/MIRI/sham groups were isolated and FACS was performed. Here, we plot the flow cytometric gating strategy for ILC and list the histograms of fluorescence channels for single antibodies versus blank controls separately ( Figures 7A, B ). In addition, we set FMO (Fluorescence Minus One) as control to improve the accuracy of the flow- cytometry gating strategy when detecting ILCdc ( Figure S8B ), and the presence of ILCdc was confirmed in both the heart and spleen ( Figures S8C, S10 ). In the heart tissue, compared with the sham group, the ILCdc in the MI group decreased significantly (P=0.004) ( Figures 7C–E ). When compared with the sham group, there was no significant difference in the MIRI group (P=0.3575) ( Figures S9C–E ).

For other ILCs clusters, when compared with sham group, the proportion of ILC2a decreased significantly both in the MI group (P<0.0001, Figures 7C–E ) and in the MIRI group (P=0.0009, Figures S9C–E ).

And the proportion of ILC1 in the MI group decreased significantly (P=0.0248) ( Figures 7C–E ) but showed no significant difference in the MIRI group (P=0.50) ( Figures S9C–E ). As for ILCt, the proportion of ILCt in the MI group decreased significantly (P=0.017) ( Figures 7C–E ) when compared with the sham group, and there was no significant difference in the MIRI group (P=0.49) ( Figures S9C–E ).

In the spleen, ILCdc was decreased significantly in the MI group when compared with the sham group (p=0.037), but there was no significant difference between the MIRI group and sham group (p=0.96) ( Figures S10C, D ).

To further validate the marker genes of ILC subsets in cardiac mononuclear cells, we sorted ILC subsets (ILCdc, ILC2a, ILC1, ILCt) using flow cytometry and performed qRT-PCR. The results showed that the expressions of Relb, Ciita, IL6ra, IL4ra, Ifngr2 and IL21r in ILCdc were significantly higher in the MI group than those in the sham group ( Figure 8A ). The expressions of IL5 and IL13 in ILC2a were significantly lower in the MI group than those in the sham group ( Figure 8B ). However, the expression of IL4 and Ifng in ILC2a was not significantly different in the MI group when compared with the sham group ( Figure 8B ). The expression of Ifng, Tnfsf11 and Tnf in ILC1 showed an upward trend in the MI group, but there was no statistically significant difference when compared with the sham group ( Figure 8C ). The expression of Ifng, Tnfsf11, and Tnf in ILCt was not significantly different between the MI and sham groups ( Figure 8D ).