Identification of cardiac fibroblast subpopulation F-Act
To illustrate the cellular characteristics of fibroblasts after MI, we performed a scRNA-seq of the cardiac interstitial cell population in the ischemic myocardium using the 10X Genomics Chromium platform at the 7th and 14th day post-sham or post-MI surgery. The transcriptional profile of 13441 cells was captured after quality control filtering. We performed unbiased clustering on an aggregate of cells using the Seurat R package9 to identify cells with distinct lineage identities and transcriptional states, with cell populations visualized in uniform manifold approximation projection (UAMP) dimensionality reduction plots.
The cardiac interstitial cells were represented by a total of 9 distinct cell lineages (Fig. 1a), with the most significant proportion of fibroblasts (Col1a1+Pdgfrα+, 6118, 45.5%), followed by endothelial cells (Kdr+Pecam1+, 4371, 32,5%) and macrophages (Cd68+Itgam+, 1898, 14.1%). For further clustering analysis of fibroblasts, we observed 11 subpopulations (Fig. 1a). F-SH was characterized by high expression of Sca1, a stem/progenitor cell marker enriched in the uninjured adult mice with multilineage differentiation and self-renewal potential9, 10. F-MFC represents matrifibrocytes, which were characterized by high expression of Comp and mainly present in the myocardial scars and play a functional importance role in maintaining the integrity of the mature scar11, 12, 13. Myofibroblasts, termed as F-Myo, increased significantly after 7 days post-MI and showed strong upregulation of numerous collagen genes (Col1a1, Col3a1, and Col5a1), as well as Postn and Acta2 at a high level (Fig. 1b), indicative of activate state and contractile phenotype14, 15, 16. Interestingly, in addition to F-Myo, another Postn-positive fibroblast subpopulation was termed as F-Act and distinguished from F-Myo by low Acta2 expression (Fig. 1b)10. Gene ontology (GO) enrichment analysis in the categories of biological process (BP), molecular functions (MF), and cellular component (CC) revealed that the upregulated genes in F-Act were enriched in cell adhesion, extracellular matrix binding, and fibronectin binding (Supplementary Fig. 1a), indicating that F-Act was also activated and participated in collagen deposition.
To clarify the temporal-spatial distribution of F-Act and F-Myo, we performed the immunofluorescence (IF) staining for periostin and αSMA in the heart tissue at different time points after MI (Fig. 1c and Supplementary Fig. 1b, c). Periostin-positive fibroblasts (stained green) were mainly distributed in the infarct and peri-infarct zone but barely in the remote zone and persisted until 28 days post-MI. Among them, αSMA-positive (stained red) F-Myo was enriched in the infarct zone, which appeared on the 7th day, peaked on the 14th day post-MI, and gradually decreased17. At the same time, αSMA-negative F-Act was observed in the peri-infarct zone from 7 days post-MI and maintained until 28 days, indicating that F-Act was persistently activated after scar maturation and might be a critical etiological factor in abnormal termination of fibrosis after MI.
To further examine the quantitative dynamic changes of F-Act after MI, Pdgfrα reporter mice (Pdgfra-CRE:tdTomato) were generated to trace cardiac interstitial fibroblasts. The proportion of the F-Act in cardiac fibroblasts at different time points after MI was detected by flow cytometry (Fig. 1d). In the hearts of uninjured adult mice, pdgfrα-positive cardiac fibroblasts account for approximately 5–10%. In the injured myocardium, the percentage of pdgfrα-positive fibroblasts, mainly periostin-positive fibroblasts, increased significantly after MI, peaked at about 50% on the 14th day post-MI, and then gradually decreased, indicating that a large number of fibroblasts were activated and proliferated for tissue repair and became quiescent or underwent apoptosis after scar formation18, 19. F-Myo predominated in the first two weeks after MI and gradually dropped to the baseline level on the 28th day post-MI. In comparison, F-Act progressively increased from the 7th day to at least the 28th day post-MI. F-Act began to surpass F-Myo to become the dominant fibroblast subpopulation on the 21st day post-MI.
Taken together, we observed a subpopulation of cardiac fibroblasts termed F-Act that was persistently activated in the peri-infarct zone until scar maturation and became the dominant subpopulation on the 21st-day post-MI.
Screening for the characteristic surface antigen of F-Act
Considering the pathological role of the F-Act, we aimed to employ the CAR-T cell immune therapy to remove it precisely. Previous studies have used FAP as a fibroblast antigen for CAR-T cells6, 20. However, UAMP plots show that Fap is widely expressed in all fibroblast subpopulations, especially in the F-Myo and F-MFC (Supplementary Fig. 2a). The continuation of FAP as a target for CAR-T cells inevitably harms other beneficial fibroblast subpopulations that form and stabilize myocardial scars, so exploring the F-Act-specific antigen markers is of great significance.
To identify surface proteins expressed specifically in the F-Act, we compared all genes encoding membrane proteins and highly expressed in the F-Act from scRNA-seq data, referring to the Human Protein Atlas (HPA) (Fig. 2a). Excluding genes those highly expressed in F-Myo or F-MFC (e.g., Pempa1 and Pi16) and those indiscriminately expressed in all fibroblast subpopulations (e.g., Pmp22, Mtch1, Tspo), Cd248, Tmem100, and Ackr3 were picked out. However, given that Ackr3 was also expressed in monocytes/macrophages and endothelial (Supplementary Fig. 2b) and Tmem100 was functionally expressed in the lung(Supplementary Fig. 2c, d)21, we finally focused on Cd248, which is highly expressed only in fibroblast subpopulations F-Act and F-SH and almost not expressed in other major cardiac interstitial cells (Fig. 2b).
CD248 is a type 1 transmembrane glycoprotein found on the plasma membrane of activated mesenchymal cells. CD248 functions only during embryonic development in the physiological state and is not expressed or found at deficient levels in adult tissues22. Alignment of the mouse and human amino acid sequences showed an overall 77.5% identity23. Protein and RNA expression overview from HPA confirmed that CD248 was expressed mainly in the duodenum, adipose, small intestine, and moderately in the heart muscle (Supplementary Fig. 2e, f). The scRNA-seq of human heart muscle from HPA showed that Cd248 was expressed mainly in fibroblasts, a small amount in smooth muscle cells, and barely in cardiomyocytes (Fig. 2c). As the evidence of the sequencing IF staining of healthy mice’s liver, spleen, lung, and kidney showed that CD248 was expressed only in small amounts around vessels and was mainly colocalized with activated fibroblasts (periostin+) and smooth muscle cells (αSMA+) but not with endothelial cells (CD31+), neutrophils (Ly6G+), and macrophages (CD68+) (Supplementary Fig. 3).
Identification of CD248 as the target antigen for CAR-T cells against F-Act
To demonstrate the specificity of CD248 as a characteristic antigen of the F-Act, we performed IF staining of mouse hearts at different time points after MI (Fig. 3a and Supplementary Fig. 4a). In the infarct and peri-infarct zone, CD248-positive (stained red) cells fused with periostin-positive (stained green) fibroblasts, and the proportion of double-positive cells continued to increase until 28 days after MI. At the same time, CD248-positive cells were rarely present in the remote zone of undamaged myocardium. Western blot analysis of the left ventricular and ventricular septal myocardium also showed that the protein expression of CD248 gradually increased over time after MI (Supplementary Fig. 4b). On the 14th day post-MI, CD248-positive cells coincided barely with αSMA-positive F-Myo and very little with COMP-positive F-MFC (Fig. 3b), demonstrating that CD248 was expressed in F-Act specifically and targeting CD248 to clear F-Act would not damage F-Myo and F-MFC.
Considering the high homology of mouse and human CD248 amino acid sequences, we performed IF staining of the heart sample collected from the patient diagnosed with MI prior to heart transplantation. (Fig. 3c and Supplementary Fig. 4c). Consistent with the mouse MI model, CD248-positive cells were enriched in the infarct and peri-infarct zone and colocalized with periostin-positive fibroblasts. Still, they did not coincide with αSMA-positive F-Myo or COMP-positive F-MFC. Interestingly, a small amount of CD248-positive and periostin-positive F-Act was found in the remote zone of undamaged myocardium, suggesting that some fibroblasts may migrate from the infarct and peri-infarct zone to the uninjured remote zone.
Furthermore, we performed a flow cytometry assay on Pdgfra-CRE:tdTomato mouse hearts (Fig. 3d). After MI, the proportion of CD248-positive cells in all living cells and pdgfrα-positive fibroblasts gradually increased until the scar matured. From the 7th day post-MI, pdgfrα-positive fibroblasts accounted for more than 80% of CD248-positive cells, proving that CD248-positive cells were mainly fibroblasts. The dynamic trend of the number of CD248-positive cells in the process of MI was consistent with that of periostin-positive and αSMA-negative cells described above in Fig. 1d. Moreover, after 7 days post-MI, the proportion of periostin-positive cells in CD248-positive cells remained above 60%, indicating that most CD248-positive cells were activated.
In summary, the expression property of CD248 made it possible to be an effective and safe target antigen for the CAR-T cells against F-Act.
Construction and functional verification of the BBIR-T Cell system target CD248
Given that CD248 was expressed in some abdominal organs, we drew on previously reported BBIR-T cell strategy to optimize CAR-T cell therapy in order to reduce the off-target effect8. BBIR-T cell used dcAv as the extracellular T-cell domain, which could recognize and bind a variety of biotinylated antigen-specific molecules effectively linked to an intracellular T cell signalling domain (Fig. 4a). We planned to label F-Act by intramyocardial injection of BF immediately after the MI surgery and remove F-Act by intravenous injection of BBIR-T cells after the infarct scar matured. This therapy used biotinylated antigen-specific molecules as intermediaries to connect effect cells and target cells, which could limit the effect site of BBIR-T cells to the heart.
Figure 4b shows the gene constructs of BBIR containing dcAv fused to the human CD3ζ cytosolic domain in combination with the CD28 and CD137 costimulatory modules. While the independent anti-CD19 CAR containing anti-CD19 Fc fragment fused with CD3ζ and CD137 was constructed as a control. HEK-293T cells were transfected by transfer plasmids pdcAv.BBIR and panti-CD19.CAR containing green fluorescent protein (GFP) and stained with biotin labelled by fluorescein AF594. BBIR-293T cells could express avidin and exogenous CD3ζ and recognize and bind biotin (Supplementary Fig. 5a-c).
Human peripheral blood-derived T cells were used for in vitro functional verification, and mouse spleen-derived T cells were used for in vivo treatment. Flow cytometry showed that CD3-positive T cells accounted for more than 90% of all viable cells 5 days after extraction, of which CD4-positive and CD8-positive cells accounted for approximately 50% and 40%, respectively (Supplementary Fig. 5d). The transfer plasmids were packaged into lentiviruses to transfect T cells to prepare BBIR-T cells and anti-CD19 CAR-T cells. Flow cytometry showed that more than 70% lentivirus transfection efficiency could be achieved, and more than 70% of BBIR-T cells could recognize and bind biotin (Fig. 4c). Western blotting was used to detect the expression of exogenous CD3ζ in BBIR-T cells and anti-CD19 CAR-T cells (Fig. 4d).
To alleviate the immunogenicity of the whole antibodies24, anti-CD248 F(ab’)2 fragments were prepared through pepsin digestion and ultrafiltration purification (Supplementary Fig. 5e, f), which could be detected by anti-mouse IgG F(ab’) secondary antibody (Supplementary Fig. 5g). Labelling anti-CD248 F(ab')2 with biotin enables it to be recognized by BBIR-T cells via a biotin-avidin bridge. Western blotting of CD248 using BF as the primary antibody demonstrated the specificity of this engineered antibody (Fig. 4e). Mouse heart IF showed that BF could recognize and bind CD248-positive cells, consistent with the results in Fig. 3, where labelled cells were mainly located in the infarct and peri-infarct zone, co-expressing periostin but not αSMA or COMP (Fig. 4e).
In vitro killing effect of BBIR-T cell therapy on CD248-positive fibroblasts
To verify the effectiveness of the BBIR-T cell therapy, we prepared target cells—the mouse embryonic fibroblast line NIH-3T3, which was previously reported to express CD24823. Overexpression (OE) or knockout (KO) of Cd248 in NIH-3T3 cells was performed by lentivirus transfection and puromycin screening (Supplementary Fig. 6a-c). Target cells were labelled with BF and cocultured with effector cells—BBIR-T cells. The effector cell to target cell ratio (E:T ratio) gradient was set from 2:1 to 20:1. Cytokine secretion in the supernatant was detected 24h after coculture by enzyme-linked immunosorbent assay (ELISA). The concentrations of IFN-γ and IL-2 in NIH-3T3 cells and Cd248-OE NIH 3T3 cells increased with increasing E:T ratio, indicating that BBIR-T cells were successfully activated (Fig. 5a, b). Cytotoxicity was detected by cell counting kit-8 (CCK-8) and lactic dehydrogenase (LDH) release experiments. When the E:T ratio reached 10:1, BBIR-T cells killed more than 80% of CD248-positive cells (Fig. 5c, d), so subsequent experiments used a fixed E:T ratio of 10:1. On the other hand, the BBIR-T system had no noticeable killing effect on Cd248-KO NIH 3T3 cells and did not lead to the activation of BBIR-T cells.
To demonstrate the safety of the BBIR-T cell therapy, we prepared different effector cells, including activated T cells, BBIR-T cells, and anti-CD19 CAR-T cells. The cell experiments were divided into 5 groups: BF + Control, BF + Activated T, BBIR-T, BF + BBIR-T, and BF + anti-CD19 CAR-T. Most of the target cells were labelled with BF except the BBIR-T group. The target cells in the BF + Control group were not cocultured with effector cells. At the same time, the target cells in the other group were cocultured with different effector cells for 24h. By detecting the secretion of IFN-γ and IL-2 in the supernatant and target-cell cytotoxicity (Fig. 5e-h and Supplementary Fig. 6d-k), it was found that only in the BF + BBIR-T group, the T cells could be activated and clear NIH-3T3 cells as well as Cd248-OE NIH-3T3 cells, but had no effect on Cd248-KO NIH-3T3 cells. The above results showed that three conditions must be met simultaneously for the BBIR-T cell therapy: the target cell expressed CD248, the target cells were labelled by BF, and the effector cells were present as BBIR-T cells.
Three-times injection of BBIR-T cells improved the cardiac function of MI mice
To verify BF's feasibility in labelling CD248-positive cells in vivo, C57BL/6J WT mouse was modelled for MI and BF was injected intramyocardial immediately after MI surgery. After 14 days, the mouse heart was obtained for IF of biotin and periostin. As shown in Supplementary Fig. 7a, biotin-positive cells could still be observed after 14 days of intramyocardial injection, and most of them were colocalized with periostin-positive cells.
To explore the injection regimen and the effect of the BBIR-T cell therapy in clearing F-Act in vivo, as shown in Fig. 6a, we performed MI surgery on Pdgfra-CRE:tdTomato mice and injected intramyocardial with BF into the infarcted myocardium immediately after the MI surgery. From the 14th day post-MI, 106 BBIR-T cells were injected into the tail vein every 3 days, and 5 mouse hearts were collected in batches 3 days after administration for flow cytometry detection (Fig. 6b). The results showed that the proportion of CD248-positive cells gradually decreased as the number of administrations increased. When BBIR-T cells were injected three times, CD248-positive cells decreased by approximately 2/3 and were not further reduced after the 4th injection of BBIR-T cells. Therefore, follow-up treatment was administered with three intravenous injections.
The following animal experiments were designed to further validate the BBIR-T cell therapy's efficacy, as shown in Fig. 6c, d. The Pdgfra-CRE:tdTomato mice were divided into 6 groups. The sham group only opened the chest without ligating the left anterior descending (LAD); the remaining 5 groups underwent LAD ligation. Except for the sham and MI + BBIR group, the other groups were injected intramyocardially with BF immediately after MI. On the 14th, 17th, and 20th days post-MI, except for the sham and MI + BBIR groups, the other groups were injected intravenously with different effect cells. Ultrasound cardiogram (UCG) was performed before MI surgery and 7, 14, and 28 days after MI to evaluate cardiac systolic function and left ventricular movement. Mice were sacrificed after LV catheterization testing 28 days post-MI. Although mouse deaths in the MI + BF + BBIR-T group decreased after implementing the BBIR-T cell therapy, the Kaplan-Meier survival curve showed no statistical significance in 1-month survival between different MI groups (Supplementary Fig. 7b).
Echocardiography showed that LV ejection fraction (LVEF) and LV fraction shortening (LVFS) gradually decreased with the prolongation of time after MI, suggesting a decline in LV systolic function, and only the MI + BF + BBIR-T group can delay the deterioration of cardiac function (Fig. 6e). On the 28th day after MI, the LV anterior wall (LVAW) of the MI + BF + BBIR-T group was thicker than the other groups in both the systolic and diastolic phases. At the same time, there was no significant difference in the LV posterior wall (LVPW) (Supplementary Fig. 7c). The LV volume was smaller in the MI + BF + BBIR-T group than in the other groups, suggesting that the BBIR-T cell therapy delayed cardiac deterioration primarily by delaying ventricular remodelling. The results of LV catheterization further confirm the conclusions of echocardiography (Fig. 6f). Compared with the sham group, after 28 days of MI, the LVP and LV dP/dt maximum decreased, and the LVEDP) and -dP/dt minimum increased, implying impaired LV systolic and diastolic function. Among the MI groups, the LVEDP and LV -dP/dt minimum in the MI + BF + BBIR-T group was lower than in other groups. At the same time, the LV dP/dt maximum was higher, indicating that the BBIR-T cell therapy not only delayed the deterioration of LV systolic function but also improved LV diastolic function.
The BBIR-T cell therapy could inhibit fibrosis expansion and improve cardiac remodelling but not aggravate inflammation response and tissue damage
Flow cytometry of cardiac pdgfrα-positive fibroblasts in the infarct and peri-infarct zone showed that only combined pre-labelled target cells and infusion of BBIR-T cells could clear F-Act. In contrast, unlabelled target cells or infusion of other effect cells had no significant effect on the number of the CD248-positive cells (Fig. 7a), further demonstrating the specificity of the BBIR-T cell therapy. The effect of clearance of F-Act on ventricular remodelling and cardiac fibrosis was evaluated by pathological analysis (Fig. 7b). Hematoxylin-eosin (HE) staining and Sirius red staining of the mouse hearts on the 28th day post-MI showed that the left ventricle had been severely remodelled, the ventricular wall was thinned, and the heart was enlarged. The thickness of the infarcted ventricular wall in the MI + BF + BBIR-T group was significantly thicker than in the other MI groups. The infarct zone (IZ) in the pathology was defined as the area of ventricular wall thinning. The peri-infarct zone (PIZ) was the area around the infarct zone that remained fibrotic, which was calculated by total fibrotic area minus the IZ area (Supplementary Fig. 7d).25 Semi-quantifying the ratio of IZ size and the PIZ size to the LV area found that the BBIR-T cell therapy could not only reduce the infarct size but also inhibit the expansion of the peri-infarct zone. The pathological results showed that targeted clearance of F-Act could effectively inhibit fibrosis expansion and ventricular remodelling.
Brain natriuretic peptide (BNP) is mainly synthesized and secreted by ventricular myocytes. It reflects the compensatory function of the heart, and its level is proportional to the degree of heart failure26. Serum BNP level on the 28th-day post-MI was detected by ELISA (Fig. 7c), which in the MI + BF + BBIR-T group was lower than other MI groups. Cardiomyocyte hypertrophy and pulmonary edema are important pathologic manifestations of ventricular remodelling. We measured the ratio of heart weight (HW) to body weight (BW) ratio, HW to tibia length (TL) ratio, and lung wet weight to dry weight ratio (Wet/Dry) (Fig. 7d-f) and found that the BBIR-T cell therapy could mitigate cardiomyocyte hypertrophy and pulmonary edema. In the undamaged myocardium of the remote zone, wheat germ agglutinin (WGA) staining, which labels cardiomyocyte membranes, showed that the cross-sectional area (CSA) of cardiomyocytes in the MI + BF + BBIR-T group was smaller than that of other MI groups. All these results further demonstrated that the BBIR-T cell therapy effectively alleviated the extent of heart failure.
In vivo, activation of traditional CAR-T cells leads to the sustained release of a large number of cytokines, known as cytokine release syndrome (CRS)27, which increases the risk of myocarditis. In comparison, activation of BBIR-T cells is limited by the number of small molecules BF and does not activate indefinitely. Using Luminex X200 to detect the level of inflammatory cytokines in mouse myocardial tissue and peripheral blood 24 hours after injecting effector cells (Fig. 8a, b), it was found that the BBIR-T cell therapy did not aggravate the release of inflammatory cytokines. HE staining of the liver, spleen, lung, and kidney showed no significant tissue damage, reflecting that the BBIR-T cell therapy did not cause damage to the off-target organs (Fig. 8c). The above results demonstrate the safety of the BBIR-T cell therapy.
Together, the BBIR-T cell therapy could effectively clear CD248-positive F-Act. Moreover, the clearance of F-Act after scar maturation could significantly inhibit fibrosis expansion and cardiomyocyte hypertrophy, thereby improving ventricular remodelling and delaying the deterioration of cardiac function.