Cardiac fibroblasts regulate the development of heart failure via Htra3-TGF-β-IGFBP7 axis

Tissue fibrosis and organ dysfunction are hallmarks of age-related diseases including heart failure, but it remains elusive whether there is a common pathway to induce both events. Through single-cell RNA-seq, spatial transcriptomics, and genetic perturbation, we elucidate that high-temperature requirement A serine peptidase 3 (Htra3) is a critical regulator of cardiac fibrosis and heart failure by maintaining the identity of quiescent cardiac fibroblasts through degrading transforming growth factor-β (TGF-β). Pressure overload downregulates expression of Htra3 in cardiac fibroblasts and activated TGF-β signaling, which induces not only cardiac fibrosis but also heart failure through DNA damage accumulation and secretory phenotype induction in failing cardiomyocytes. Overexpression of Htra3 in the heart inhibits TGF-β signaling and ameliorates cardiac dysfunction after pressure overload. Htra3-regulated induction of spatio-temporal cardiac fibrosis and cardiomyocyte secretory phenotype are observed specifically in infarct regions after myocardial infarction. Integrative analyses of single-cardiomyocyte transcriptome and plasma proteome in human reveal that IGFBP7, which is a cytokine downstream of TGF-β and secreted from failing cardiomyocytes, is the most predictable marker of advanced heart failure. These findings highlight the roles of cardiac fibroblasts in regulating cardiomyocyte homeostasis and cardiac fibrosis through the Htra3-TGF-β-IGFBP7 pathway, which would be a therapeutic target for heart failure.

The study examines the role of the serine peptidase Htra3 in cardiac homeostasis and in heart failure. The authors show that Htra3 is predominantly expressed in cardiac fibroblasts and that global germline loss of Htra3 causes a hypertrophic cardiomyopathy that is further accentuated in response to pressure overload induced through TAC. The effects of Htra3 are mediated through TGFbeta degradation: suppression of endogenous Htra3 following pressure overload induces an exaggerated TGF-beta response.
General comment: The study deals with an interesting and novel concept. The data are generally of high quality. The main problems relate to the exclusive reliance on a global KO line (and the absence of fibroblastspecific approaches), and the overinterpretation of the findings to support an effect of fibroblasts on cardiomyocyte senescence. The following concerns need to be addressed: Major comments: 1.The main weakness of the study is the use of a global Htra3 KO line to suggest fibroblast-specific effects. The authors attempt to circumvent this by showing single cell data suggesting that Htra3 is predominantly expressed in fibroblasts. However, considering that the protein is expressed at high levels in the myocardium and in other muscle tissues (compared to other organs), expression at the protein level in cardiomyocytes (and in other muscle cells) is likely and may be involved in regulation of cell size. Conclusions regarding the fibroblast-specific effects of Htra3 require fibroblast-specific loss-of-function approaches. The authors should document the predominant expression of Htra3 in fibroblasts vs cardiomyocytes at the protein level, and acknowledge the limitation related to the lack of cell-specific loss-of-function approaches.
2.Htra3 KO mice need to be characterized. Mice with global loss of Htra3 had significant baseline hypertrophy. What is the basis for this abnormality? Was systemic blood pressure affected? Were there other effects on cardiomyocytes? Where there any effects in other organs? If Htra3 is expressed in all fibroblasts and inhibits TGF-beta signaling, abnormalities in other organs may be prominent.
3.The cardiomyopathic response in Htra3 KOs needs to be characterized. Is this progressive? Does it result in heart failure and mortality in the absence of injury? 4.The authors suggest that the phenotype in Htra3KOs is driven by overactive TGF-beta. Does TGF-b neutralization abolish the hypertrophic response at baseline? D r a f t O n l y 5.How do the authors explain the absence of baseline fibrosis in mice lacking Htra3?
6.The criteria used by the authors to suggest cardiomyocyte "senescence" are problematic. They seem to define senescence as practically any activation of oxidative responses , or induction of certain inflammatory genes. DNA damage is not documented. The authors need to tone down statements regarding cell senescence and simply refer to their observations, limiting the use of speculative statements.
7.There is a tendency to overinterpret. The title epitomizes this problem: "Cardiac fibroblasts are critically involved in the development of heart failure by inducing cardiomyocyte senescence". However, fibroblast-specific approaches are lacking and cardiomyocyte senescence is not demonstrated (unless one uses the term senescence to describe any injury). Please revise the title and conclusions, interpreting the findings.
8.Data on the fibroblast-specific expression of Htra3 need to be strengthened. Please provide comparative analysis of levels in different myocardial cell types. Was Htra3 expressed in all fibroblast subsets? The images documenting Htra3 expression in Pdgfra+ cells need to be improved. It is impossible to appreciate the colocalization.

Minor:
Htra3 was identified as a molecule located "at the center of the network". Please explain what that means.
Results: "To investigate molecular interactions leading to the induction of senescent failing cardiomyocytes". Why did the authors assume that cardiomyocytes become "senescent" following pressure overload? The study investigates effects of pressure overload not cell senescence.
The abstract requires minor revisions and clarifications: a. please explain the rationale for the focus on Htra3, b. "governing the identity", perhaps a more specific term would be preferable.
Reviewer #2 (Remarks to the Author): By using single-cell RNA-seq, spatial transcriptomics, and genetic modyfications the authors showed that high-temperature requirement A serine peptidase 3 (Htra3) is a critical regulator of cardiac fibrosis and cardiomyocyte senescence. Htra3, specifically expressed in cardiac fibroblasts keeps them quiescent and prevents TGF-β degradation. TGF-β is known inducer of so called secondary senescence, but the authors showed that Htra3-TGF-β-IGFBP7 pathway is responsible for cardiac D r a f t O n l y fibroblast activation and cardiomyocyte senescence. They provided evidence that cardiomyocyte senescence is DNA damage dependent governing the identity of quiescent cardiac fibroblasts through degrading TGF-β.
I have some comments concerning this issue: 1. It cannot be excluded that in KO Htra3 deficient mice after intense proliferation and collagen production also cardiac fibroblasts undergo senescence and in this state non-degraded TGF-β induces secondary senescence in cardiomyocytes. Moreover SASP of senescent fibroblasts can be fibrogenic (doi.org/10.1038/ncomms14532). Can the authors exclude senescence of fibroblasts in their model of heart failure? 2. DNA damage-dependent senescence of cardiomyocytes could be better documented (Fig.4) By genetic perturbation in several models and with scRNAseq and spatial transcriptomics the authors identify fibroblast expressed Htra3 as a critical regulator and its downregulation via activated TGF-signalling of cardiac fibrosis and cardiomyocyte senescence through DNA damage. The Htra3-TGF--IGFBP7 was explored in mouse pressure overload, AMI and knockout models. Further, it was confirmed in human failing heart and demonstrated that the cytokine IGFBP7 is secreted from senescent cardiomyocytes and is a putative therapeutic target in heart failure.
Seminal for scientific results is detailed description of methods so they and results can be critically scrutinized. Here is a list on what might be improved.
D r a f t O n l y 1. Ethics: How was heart samples for human controls obtained. What was the time delay between death and tissue isolation? How does it compare to patient samples? It is not trivial how live human cardiomyocytes can be obtained from a healthy donor.
2. Line 501: How were live human cardiomyocytes isolated and processed? It is not described in ref 1. In Ext data fig 2b UMAP of human cells are shown for a number of cell types. Where these isolated and processed as for the human cardiomyocytes?
3. Mouse cardiomyocytes were isolated by Langendorf perfusion. To what extent are they representative for the whole heart? What was the yield? Did you consider using snRNAseq to get RNA for all cell types in the same analysis? 4. Cluster analysis: It seems that both WGCNA, Seurat 3 and Seurat 4 were used for cluster analysis. Why did you use different methods? What were the conditions used in the different analyses? What was the results of quality analysis? When you merged clusters analysed for different samples, did you correct for batch effects? 5. Spatial transcriptomics: Which method did you use for the projection of scRNAseq clusters on spatial transcriptomics maps? Which scRNAseq cell type classification have been used? Is it from fig 1 where classification was done with WGCNA, another method of classification. If so, has this procedure been validated? Or is it de novo scRNAseq from cells in the infarct zone as indicated in the text? In that case, the methods, results and analysis of the scRNA is not given in results and methods sections.
6. Fig 6: You show spatial transcriptomics w 5 clusters. Cl3 seem to be mainly characteristics of blood. You show projections of scRNAseq cell types. How was this done? What does the proportions mean? For example does a proportion of 0.4 mean that 40% of the cells were of that type? If so, the number of cells in cl2 is much more than possible 100%. 7. Number of detected genes: In the different experiments you used different cutoff for the genes to be analysed. What principle did you use to decide on the cutoff? 8. Figure 1: g: Coexpression network of cardiac fibroblasts. As I understand this is the basic analysis to decide on exploration of Htra-3. Why did you decide on only HTRA-3 and not the other main genes in the figure? h: Top genes correlated with the fibroblast module. This method is only superficially described in the methods section and there is no xls file on the genes related to this module or any other. 9. The only two xls files provided are computed tables and the only on basic analysis ligand-receptor pairs between modules. This table is perplexing: For example NPPA: Ligand module 0, receptor module 2. Where 0 may be interpreted as no cell type and 2 as M-2 endothelial cell.
What does zero mean? NPPA have no ligand module? Npr2 no receptor module?
10. It is difficult to get an overview of the 44 modules and what they stand for. How were they discriminated and what genes were systematically identified?
D r a f t O n l y 11. Line 309 It is stated that IGFBP7 had higher diagnostic power than NT-proBNP. The statistical difference between the ROC curves for this conclusion is not given. D r a f t O n l y

Responses to the reviewers' comments
We thank all reviewers for their insightful comments regarding our manuscript. In view of the reviewers" suggestions, we have performed additional analyses and revised the manuscript. The sentences that were revised according to the reviewers" comments are highlighted in yellow in the main text.

Response
We conducted qRT-PCR to verify the specific expression of Htra3 in the heart among major organs (Extended Data Fig. 1j), which was consistent with the previous report (ref. 27). Next, scRNA-seq analysis revealed the expression of Htra3 was only seen in cardiac fibroblasts among various cell-types in the heart (Fig. 1i Data Fig. 1k). Based on these results, Htra3 is considered to be predominantly expressed in cardiac fibroblasts. We also mentioned the limitation of the lack of cell-specific loss-of-function approaches in the Discussion section in our revised manuscript.

Response
We checked the blood pressure, but there was no significant difference in blood pressure between WT and Htra3 KO mice (Extended Data Fig. 2b). For organs other than the heart, we could not find any difference in their size and tissue histology (HE staining) (Extended Data Fig. 2d). As we mentioned before, the expression of Htra3 is predominantly seen in cardiac fibroblasts, which may explain the normal phenotype of other organs. We also compared transcriptomic profiles from WT and Htra3 KO mice after sham operation. In cardiomyocytes from Htra3 KO mice, genes involved in oxidative-reduction process were down-regulated and genes involved in protein synthesis were up-regulated without pressure overload (Extended Data Fig. 4e), which may lead to cardiomyocyte hypertrophy in Htra3 KO mice (Fig. 2c).
3. The cardiomyopathic response in Htra3 KOs needs to be characterized. Is this progressive? Does it result in heart failure and mortality in the absence of injury?

Response
We compared the natural history of cardiac function between WT and Htra3 KO mice.
As shown in Extended Data Fig. 2c, we could not observe any differences in LVDd/Ds and FS between WT and Htra3 KO mice with aging up to 60 weeks old.

Response
We performed TGF-β neutralization in Htra3 KO mice without any additional surgery.
Since ventricular hypertrophy has already been observed in Htra3 KO mice of 8 weeks old, we injected anti-TGF-β antibodies from 4 weeks old, showing that TGF-β neutralization successfully inhibited the ventricular hypertrophy in Htra3 KO mice (Extended Data Fig. 3h).
D r a f t O n l y 5. How do the authors explain the absence of baseline fibrosis in mice lacking Htra3?

Response
As shown in Fig. 2d, we quantified the percentage of fibrosis area. Htra3 KO mice showed significantly more interstitial fibrosis than WT mice even without TAC surgery.
6. The criteria used by the authors to suggest cardiomyocyte "senescence" are problematic. They seem to define senescence as practically any activation of oxidative responses, or induction of certain inflammatory genes. DNA damage is not documented.
The authors need to tone down statements regarding cell senescence and simply refer to their observations, limiting the use of speculative statements.

Response
We agree with the reviewer"s comment. The sentences and words revised according the suggestion were highlighted in yellow in the revised manuscript.
7. There is a tendency to overinterpret. The title epitomizes this problem: "Cardiac fibroblasts are critically involved in the development of heart failure by inducing cardiomyocyte senescence". However, fibroblast-specific approaches are lacking and cardiomyocyte senescence is not demonstrated (unless one uses the term senescence to describe any injury). Please revise the title and conclusions, interpreting the findings.

Response
We again agree with the reviewer"s comment. According to the suggestion, we revised the title, results, and conclusions, based on the experimental findings.

Response
We showed the fibroblast-specific expression of Htra3 in the heart using violin plot (Extended Data Fig. 1b). After myocardial infarction, the fibroblast population was divided into 4 clusters (Extended Data Fig. 6c). Fibroblast clusters 2 (FB2) and 4 (FB4), which were increased in the early phase after myocardial infarction, showed lower expression levels of Htra3 (Extended Data Fig. 6d-f), which is consistent with the D r a f t O n l y finding that expression of Htra3 in cardiac fibroblasts was down-regulated after pressure overload to the heart (Fig. 2e) or mechanical stretch (Fig. 2g).
We sincerely apologize for the confusing description regarding the images documenting Htra3 expression in the original paper. Since it was difficult to perform immunostaining and in situ hybridization in the same tissue slide, we used paired mirror cardiac sections to perform immunostaining of Pdgfr-α and in situ hybridization of Htra3.
Therefore, same cells would appear in both sections. In Fig. 1j, Arrows indicate the colocalization of Pdgfr-α (immunostaining) and Htra3 (in situ hybridization) in the same cells appeared in both sections.

Minor:
Htra3 was identified as a molecule located "at the center of the network". Please explain what that means.

Response
Since we conducted weighted gene coexpression network analysis to derive the fibroblast network, the central location of Htra3 in Fig. 1g means that its expression was strongly correlated with the expression of other cardiac fibroblast module genes. By calculating the correlation coefficient of each gene expression with the fibroblast module expression, we also found that Htra3 expression was strongly correlated with expression of the cardiac fibroblast module (Fig. 1h), suggesting that Htra3 defines the identity of cardiac fibroblasts.
Results: "To investigate molecular interactions leading to the induction of senescent failing cardiomyocytes". Why did the authors assume that cardiomyocytes become "senescent" following pressure overload? The study investigates effects of pressure overload not cell senescence.

Response
As the reviewer states, we investigated the effects of pressure overload and revealed the appearance of failing cardiomyocytes with DNA damage and secretory phenotype.
We revised the manuscript according to the suggestions. The abstract requires minor revisions and clarifications: a. please explain the rationale for the focus on Htra3, "governing the identity", perhaps a more specific term would be preferable.

Response
Thank you for your suggestion. We conducted single-cell RNA-seq of cardiac fibroblasts isolated from WT and Htra3 KO mice to show that Htra3 basically inhibits TGF-β signalling and that pressure overload and Htra3 deletion synergistically activates TGF-β signalling, leading to activation of fibroblasts. Therefore, we changed the sentence from "governing the identity of quiescent cardiac fibroblasts" to "maintaining the identity of quiescent cardiac fibroblasts".
Reviewer #2 (Remarks to the Author):  Fig. 4g, h). DNA damage in cardiac fibroblasts may also induce fibroblast senescence, consistent with our single-cell RNA-seq finding that cardiac fibroblasts of Htra3 KO mice expressed genes involved in senescence-associated secretory phenotype (e.g., Igfbp7 and Tgfb3) (Fig. 3h, i, j). (Fig.4). Double positive (γH2AX, p21) nuclei are hardly visible. What about quantitative analysis? The levels of p53 and/or ATM would be desirable.

Response
We sincerely apologize for the confusing presentation in the Fig. 4j.  Data Fig. 4h). We further showed that expression of DNA damage-related genes, such as Cdkn1a and Trp53, was up-regulated in cardiomyocytes of Htra3 KO mice after TAC surgery (Fig. 4k).

Response
Although there is no valid data about the number of cells in the murine whole heart, we isolated more than 10 5 cells from one mouse heart, which was equivalent to the previous reports (such as Nature 497, 249-253 (2013) and Nature 582, 271-276 (2020)). Besides, we also confirmed that there were no clumps of cardiac tissue after isolation of cardiomyocytes through Langendorf perfusion. Actually, we performed single-nucleus RNA-seq of the heart, but did not quantitatively detect many genes; therefore, we obtained single-cell expression profiles of cardiomyocytes by the Smart-seq2 full-length method and non-cardiomyocytes by droplet-based Chromium Controller (10x Genomics). Response Since single-cell RNA-seq through full-length cDNA synthesis by Smart-seq2 quantitatively detected many genes, we applied weighted gene co-expression network analysis to detect gene modules and use their expression levels for clustering analysis.
The other single-cell RNA-seq data were analyzed using Seurat with the "FindIntegrationAnchors" function to correct batch effects. Seurat V4 was applied only for spatial transcriptomics data analysis. We added these explanations in the Methods section.

Response
We apologize for the sufficient description in methodology. To predict the proportion of cell types in each spot, predict.score was calculated by using "FindTransferAnchors" and "TransferData" functions in Seurat (Fig. 6f). Dot size represents the average of predict.score of each spot in each cluster. The scRNA-seq data of cells isolated from mice after sham or MI operation (Extended Data Fig. 6) were used as reference data.
We added these descriptions in the Methods section.

D r a f t O n l y
As the reviewer pointed out, we also consider that since heart sections used for spatial transcriptomics analysis contained red blood cells, cluster 3, which was characteristic for genes related with red blood cells, was generated. However, since we carefully performed perfusion by PBS before and during cell isolation for single-cell RNA-seq, we did not detect red blood cells in our single-cell RNA-seq analysis. As we answered to the previous comment, we used scRNA-seq data of cells isolated from mice after sham or MI operation as the reference, and re-analyzed to predict cell types in each spot and derive the proportion of cell types in each cluster (Fig. 6f).

Number of detected genes: In the different experiments you used different cutoff for
the genes to be analysed. What principle did you use to decide on the cutoff?

Response
In single-cell RNA-seq through full-length cDNA library synthesis by Smart-seq2, we calculated detected genes for each cell and generated histogram to set the cutoffs for genes to be analyzed (Extended Data Fig. 1c, 3a, 4a, and 7a). We added this explanation in the Methods section.

Response
We calculated the correlation coefficient between each gene expression with fibroblast module expression in Fig. 1h. We also generated Supplementary Table 3 as original matrix data for Fig. 1h. We added these explanations in the Methods section. In Fig. 1h Fig. 7d). Genes involved in extracellular matrix organization and TGF-beta receptor signaling pathway were enriched in M44 (Fig. 7k). Expression dynamics of representative genes in M44 along with pseudotime are shown in Fig. 7k, l. 11. Line 309 It is stated that IGFBP7 had higher diagnostic power than NT-proBNP. The statistical difference between the ROC curves for this conclusion is not given.

Response
The statistical difference between the ROC curves in Fig. 7q was P-value = 0.1063. We  Fig. 5d.
Finally, the authors would like to thank the editor and reviewers again for these valuable comments and suggestions.

Response
We deeply appreciate the Reviewer's comments.
For constructing the LR interaction network map in the heart, we used weighted co-expression network analysis to generate co-expression gene modules and annotate cell-type-specific gene modules by using cell-type-specific gene expression profiles, which were obtained by UMAP projection of single-cell data. For weighted co-expression network analysis, all genes expressed at an FPKM value of ≥10 in at least one of the samples were used to construct a signed network using the WGCNA R package, which was also used for cell type annotation in Supplementary Data 1 (The ligand and receptor interaction pairs in the heart). The soft power threshold was analyzed with the "pickSoftThreshold" function and applied to construct a signed network and calculate the module eigengene expression using the "blockwiseModules" function. Modules with <30 genes were merged with their closest larger neighbouring module. To visualize the weighted co-expression networks, Cytoscape (version 3.7.2) with "edge-weighted force-directed" was used. We described these explanations in the Methods section D r a f t O n l y (page 15 line 29).
In single-cell RNA-seq analysis of human cardiomyocytes, we performed Random forest and overlap analysis to identify co-expression gene modules M1, M2, and M44 as significantly involved in cell classification and conserved between human and mouse ( Supplementary Fig. 7b-d). We did not perform ligand receptor interaction network analysis using human cardiomyocytes in the manuscript. By calculating the overlap between gene modules detected from single-cardiomyocyte RNA-seq in human and mice, we found that M44, which corresponded to murine cardiomyocyte M9 ( Supplementary Fig. 7d), was enriched with genes involved in extracellular matrix organization and activated specifically in trajectory 2 ( Fig. 7k-m). We described these explanations in the main text (page 10 line 5 and page 10 line 19).
We also added the description about Pathway, GO, and co-expression analysis in the Methods section in our revised manuscript (page 16 line 29).
For cell-type classification of single-cell analysis in Fig. 1i (mouse) or Supplementary (mouse) and Supplementary Fig. 2f (human).
Since we didn't know where line 550-553 in the reviewer's manuscript corresponded to in our manuscript, we revisited the manuscript thoroughly and entirely, including the Methods section. The revised sentences are highlighted in yellow in the manuscript.
Finally, the authors would like to thank the editor and reviewers again for these valuable comments and suggestions.
D r a f t O n l y