Lrat germline knockout mice (Lrat^−/−) were purchased from Jackson Laboratories (strain #018866) (20). Lrat^−/− mice were cross-bred with Stra6 germline knockout mice (Stra6^−/−) to generate mice with germline deletion of both Lrat and Stra6 (Lrat^−/− x Stra6^−/−; DKO) (21). Mice were on a pure C57/Bl6J genetic background, and genotyping was performed as previously described (20, 21). Wildtype (WT) and Stra6^−/− mice were fed a standard chow diet (C1000, Altromin, Lage, Germany). Lrat^−/− and DKO mice received a vitamin A-deficient diet (C1016, Altromin, Lage, Germany). Dietary treatment was initiated at 4–6 weeks of age and continued until tissue harvest. MI was induced after 4 weeks of dietary treatment at the age of 8–10 weeks. Animals were housed in a 14 h light/10 h dark cycle with ad libitum access to food and water. Studies were performed in male mice. For tissue harvest, mice were euthanized by cervical dislocation under isoflurane anesthesia. All experiments were performed in accordance with the ARRIVE guidelines and with protocols approved by local state authorities (Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit; protocol numbers: 23/00360 and 24/00719), which conform to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health.

Total RNA from the infarct border zone was isolated using the NucleoSpin RNA kit (Macherey-Nagel, Düren, Germany) and cDNA synthesis was performed (MaximaTM H Minus First Strand cDNA Synthesis kit, Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturers’ instructions. Quantitative RT-PCR analysis was performed using the Luna® Universal qPCR Master Mix (New England Biolabs, Ipswich, MA, USA) and a LightCycler® 96 PCR system (Roche, Mannheim, Germany) (22). Primers are listed in Supplementary Table S1.

Retinoids were extracted from liver tissue. All-trans-retinal [lower limit of quantification (LLOQ): 0.992 pmol/sample] and all-trans-retinol levels (LLOQ: 3.4 pmol/sample) were determined as previously described (14).

MI was induced by left anterior descending coronary artery (LAD) ligation. Mice were pre-treated with metamizole (1A Pharma, Holzkirchen, Germany) dissolved in drinking water (500 mg/kg) the day before surgery. Before surgery, mice were treated by intraperitoneal injection of butorphanol (2 mg/kg; Cp-Pharma, Burgdorf, Germany) and subcutaneous injection of carprofen (5 mg/kg; Cp-Pharma, Burgdorf, Germany). Anesthesia was induced with 2%–4% isoflurane dissolved in oxygen. After oral intubation, mice were mechanically ventilated (MiniVent Type 683, Harvard Apparatus, Holliston, MA, USA), and anesthesia was maintained with 1%–4% isoflurane dissolved in oxygen. Mice were placed on a heating pad (37°C), and an eye care solution was applied to prevent corneal injury. Local anesthesia (a combination of 0.5% lidocaine and 0.25% bupivacaine) was applied before a left thoracotomy in the fifth intercostal space. After opening the pericardium, the LAD was ligated using a 6-0 Prolene suture (Ethicon, Norderstedt, Germany). MI was evident from LV discoloration. The thorax was closed, and mice recovered at 32°C. Animals received subcutaneous injections of carprofen (5 mg/kg) for an additional 3 days post-surgery. Sham-operated mice were subjected to similar surgery, except that no ligature around the LAD was placed.

Transthoracic echocardiography was performed under isoflurane anesthesia (induced at a concentration of 5% and sustained with 1% isoflurane). Mice were placed in the supine position on a heating pad (37°C). An eye care solution was applied to prevent corneal injury. Two-dimensional B-mode images were acquired, and endocardial silhouettes were traced manually. Ejection fraction was determined in long-axis projection using the VevoStrain software (VisualSonics Inc.). LV end-diastolic area (LVEDA) and LV end-systolic area (LVESA) were determined in long-axis parasternal projections. Fractional area change (FAC) was calculated as [(LVEDA—LVESA)/LVEDA] * 100 (14, 22).

Paraffin sections of cardiac tissue were prepared, stained with hematoxylin and eosin (H&E) or Picrosirius red (PSR) staining solutions, and quantified as previously described (14). As previously reported, hearts were embedded in OCT, cut into 10-µm-thick sections, and stained with PSR staining solution for transverse sections (23).

Protein extraction from the infarct border zone, immunoblotting, and densitometric analysis were performed as previously described (22). Antibodies used for immunoblotting are listed in Supplementary Table S2.

Total RNA from the infarct border zone was isolated using the NucleoSpin RNA kit (Macherey-Nagel, Düren, Germany). Quality/integrity control of total RNA was performed using a 2100 Bioanalyzer System (Agilent Technologies, Waldbronn, Germany) (22). The RNA sequencing library was generated from 500 ng of total RNA using the NEBNext® Ultra^TM II Directional RNA Library Prep Kit for Illumina® (New England BioLabs, Frankfurt, Germany) with the NEBNext® Poly(A) mRNA Magnetic Isolation Module according to the manufacturer's instructions. RNA libraries were sequenced using a NovaSeq 6000 system and the NovaSeq 6000 S1 Reagent Kit (100 cycles, paired end run) with an average of 3 * 10⁷ reads per sample (Illumina, San Diego, CA, USA). Quality reports were generated for each FASTQ file using the FASTQC tool (https://www.bioinformatics.babraham.ac.uk/projects/fastqc). Raw FASTQ files were trimmed on base call quality and sequencing adapter contamination using fastq-mcf (https://expressionanalysis.github.io/ea-utils/). Reads shorter than 15 bp were removed from FASTQ files. Next, trimmed reads were aligned to the murine reference genome using the open-source short-read aligner STAR with settings according to the log file (24). Data analysis was performed with the statistical programming language R (v4.1.1) (25). Feature counts were determined using the R package “Rsubread” (v2.2.6), and transcript annotation was performed using the R package “bioMaRt” (v2.44.4) (26, 27). The R package “DESeq2” (v1.32.0) was used to evaluate differential gene expression (28–31). To ensure robust expression analysis, only transcripts with a count of at least 10 in about 40% of total samples were used for further processing. A total of two libraries were removed from further analysis based on outlier properties in principal component analysis (PCA) and Cook's distance to assess sample-level variance. Transcripts involved in retinoid signaling and metabolism were identified based on previous publications and their enrichment in the top 500 genes contributing to PC2 was determined by Fisher's Exact Test (32–40). The R package clusterProfiler was applied to the top 500 genes contributing to PC2 and gene ontology (GO) term enrichment analysis was performed. The identified GO terms were filtered to extract GO terms containing keywords related to retinoid biology, which identified the following pathways: “retinoic acid metabolic process”, “retinoic acid biosynthetic process”, and “retinoid metabolic process”. Functional analysis was performed using the Ingenuity Pathway Analysis (IPA) tool (Qiagen, Germantown, MD, USA; https://digitalinsights.qiagen.com/products-overview/discovery-insights-portfolio/analysis-and-visualization/qiagen-ipa). Transcripts were classified based on KEGG pathways [Wnt signaling pathway: mmu04310, cytoskeleton in muscle cells: mmu04820, extracellular matrix (ECM)-receptor interaction: mmu04512; https://www.genome.jp/kegg]. Gene set enrichment analysis (GSEA) was performed with software version 4.1.0 (41–44). Cumulative distribution analysis was performed with R. P-values are reported for two-sided Kolmogorov–Smirnov tests (cutoff for significance: p < 0.05). Spline interpolation was used for cumulative distribution function (ECDF) plots for visual clarity.

Data are expressed as mean ± SEM. Two-way ANOVA was performed to analyze differences after Stra6^−/− by Lrat^−/− in non-operated groups, after Stra6^−/− by VitAD in Sham-operated groups at time points investigated, and after Stra6^−/− by VitAD in MI-operated groups at the 3-day time point, each followed by Holm-Šídák post hoc analysis. T-Tests were performed to analyze differences after MI relative to Sham, which had the same Stra6 expression and vitamin A availability. One-way ANOVA was performed to analyze differences between MI-operated groups at the 2-week and 4-week time points, followed by Holm-Šídák post hoc analysis. Survival was analyzed using a log-rank test and adjusted for multiple testing using the Bonferroni method. Statistical analyses were performed using GraphPad Prism software version 8.0 (GraphPad Software, San Diego, CA, USA) and for RNA sequencing analysis as described above. For all analyses, a p-value of <0.05 was considered significantly different.

RT-PCR analysis of LV tissue confirmed deletion of Lrat in Lrat^−/− and DKO mice, Stra6 in Stra6^−/− and DKO mice, and both in DKO mice (Figures 1A,B). Liver retinoid levels reflect whole body vitamin A status (45, 46). Lrat germline deletion (Lrat^−/−) decreases hepatic retinoid levels and impairs tissue retinoid levels in the absence of dietary vitamin A (15, 20, 47, 48). To test the hypothesis that Stra6 expression attenuates ischemic injury-induced HF following MI by vitamin A-dependent mechanisms, we subjected Lrat^−/− and DKO mice to a vitamin A-deficient diet starting at 4- to 6 weeks of age (groups indicated as “VitAD” and “Stra6^−/− × VitAD” respectively). WT and Stra6^−/−-mice were fed with a vitamin A-sufficient diet (groups indicated as “WT” and “Stra6^−/−” respectively). MI was induced after 4 weeks of dietary treatment and was continued until tissue harvest. Sham-operated mice with the same genotype and dietary treatment served as controls (Figure 1C). All-trans-retinal and all-trans-retinol levels were nearly absent in livers from the VitAD and Stra6^−/− × VitAD groups (all-trans-retinal [pmol/mg tissue]: WT 0.394 ± 0.071, Stra6^−/− 0.458 ± 0.041, VitAD 0.002 ± 0.000, Stra6^−/− × VitAD 0.002 ± 0.000; all-trans-retinol [pmol/mg tissue]: WT 92.3 ± 10.8, Stra6^−/− 104.1 ± 13.3, VitAD not detectable, Stra6^−/− × VitAD not detectable; Figures 1D,E). These data confirm VitAD independent of Stra6 expression. Tissue RA levels are regulated by CYP26 hydroxylases, including CYP26A1. RA mediate CPY260 enzyme expression in a feedback regulation loop and CPY26 levels correlate with vitamin A availability (49–51). Cyp26a1 mRNA expression trended to decrease in the VitAD-Sham relative to the WT-Sham group (−61.5%, p = 0.059; Supplementary Figure S1), which suggests decreased cardiac RA levels in the VitAD group.

Following ischemic injury, we observed increased mortality in Stra6^−/− mice, with the majority of mice dying by 4 days post-MI. Mortality was not increased in the other groups investigated relative to Sham-operated mice (Figure 2A). Autopsies revealed cardiac rupture in all Stra6^−/− mice post-MI as diagnosed by the presence of a blood clot in the chest cavity and around the heart. Interestingly, transthoracic echocardiography revealed HF independent of vitamin A status and Stra6 expression as indicated by a similar increase in LVEDA and a similar decrease in ejection fraction 3 days post-MI (Figures 2B–D; Supplementary Table S3). LVEDA similarly increased, and ejection fraction similarly decreased in MI-operated groups 4 weeks post-surgery (Figures 2E,F; Supplementary Table S3). Heart weights normalized to tibia length (Figure 2G; Supplementary Table S4), and mRNA expression of HF markers (Figures 2H,I) increased 3 days post-MI relative to Sham-operated groups independent of vitamin A status and Stra6 expression. Similarly, HF marker mRNA expression was increased 4 weeks post-MI relative to Sham-operated groups independent vitamin A status and Stra6 expression (Supplementary Figure S2). Quantification of fibrotic tissue showed no difference between groups at the 3-day time point (Figures 2J,K). The non-receptor tyrosine kinase janus kinase 2 (JAK2) is recruited to STRA6, which activates signal transducer and activator of transcription 5 (STAT5) (18). I/R activates STAT5A, and Stat5a-deficient hearts cannot be preconditioned (17). STAT5 activates the PI-3 kinase (PI3K)/Akt signaling cascade, which is cardioprotective under conditions of ischemic preconditioning (IPC) (17, 19). Similarly, STRA6 activates the JAK2/STAT3 signaling module and STAT3-mediated signaling is cardioprotective following I/R (52, 53). We hypothesized that Stra6 expression is cardioprotective following ischemic injury by preserving STAT3 and STAT5 signaling. Interestingly, STAT3, but not STAT5/Akt signaling was activated following ischemic injury independent of Stra6 expression and vitamin A availability 3 days post-surgery (Supplementary Figure S3). Together, these data reveal that Stra6 deletion does not impact cardiac structure and contractile function under basal conditions. Moreover, VitAD attenuates increased mortality in Stra6^−/− mice following ischemic injury independent of STAT3 and STAT5/Akt signaling.

To delineate the mechanisms contributing to cardiac rupture in Stra6-deficient heart post-MI and the reversal under VitAD conditions, we performed differential gene expression analysis by RNA sequencing using tissue from the infarct border zone. A total of 19,023 transcripts were considered for analysis (Supplementary Table S5). Principal component analysis (PCA) revealed a predominant effect of surgery performed on gene expression (Figure 3A), which is supported by a heatmap presenting the top 100 genes with the greatest variance across groups (Figure 3B). Using a list of previously identified genes that play a key role in retinoid signaling and metabolism (Supplementary Table S6), we identified that retinoid-related transcripts are significantly enriched in the top 500 genes contributing to PC2 (p = 0.0338, odds ratio = 11.91; Supplementary Table S7; Supplementary Figure S4A), but not in the top 500 PC1-contributing genes. Gene ontology enrichment analysis using PC2 loading genes identified significantly enriched terms related to retinoid biology, including “retinoic acid metabolic process”, “retinoic acid biosynthetic process”, and “retinoid metabolic process” (Supplementary Figure S4B). These results suggest that PC2 captures transcriptional variation related to retinoid acids.

We explored differences in gene expression using pairwise comparisons of groups. We investigated 12 comparisons representing differences in surgery performed (comparisons I–IV), in Stra6 expression (comparisons V–VIII), and vitamin A availability (comparisons IX to XII; Figures 3C–N). A total of 6,301 transcripts was induced and 5,922 were repressed in WT-MI vs. WT-Sham (comparison I), 6,400 induced/6,051 decreased in Stra6^−/−-MI vs. Stra6^−/−-Sham (comparison II), 6,673 induced/6,403 decreased in VitAD-MI vs. VitAD-Sham (comparison III), and 6,575 induced/6,150 decreased in Stra6^−/− × VitAD-MI vs. Stra6^−/− × VitAD-Sham (comparison IV, Figures 3C–F, cutoff: FDR < 0.01). In contrast, the expression of a relatively small number of genes was altered for comparisons V–XII after applying the same cutoff (Figures 3G–N). These data support our PCA and indicate a predominant effect of surgery performed relative to Stra6 expression and vitamin A availability on gene expression. Moreover, ischemic injury mediates the expression of retinoid signaling and metabolism genes (Supplementary Figure S5; Supplementary Table S6). Cumulative distribution analysis showed decreased expression of cytoskeleton and ECM-receptor interaction genes in Sham-operated groups following Stra6 deletion and vitamin A deficiency, which was reversed by the combination of both (Supplementary Figure S6).

Next, we focused on differentially expressed genes between MI-operated groups (comparisons VI, VIII, X, and XII; Figure 4A; Supplementary Table S8). We investigated comparison VI (Stra6^−/−-MI vs. WT-MI) to identify transcripts by which Stra6 deletion results in cardiac rupture post-MI and comparison XII (Stra6^−/− × VitAD-MI vs. Stra6^−/−-MI) to delineate transcripts by which VitAD attenuates this effect. Interestingly, IPA identified decreased wingless/int-1 protein (Wnt) ligand biogenesis and trafficking as the top canonical pathway for comparison VI and increased interferon alpha/beta signaling for comparison XII (Figures 4B,C; cutoff: |log₂ fold change| > 0.6 and p < 0.05). We next investigated transcripts that were altered in comparisons VI and XII in the opposite direction (gene sets indicated as ABCD, ABD, ACD, and AD in Figure 4A) and identified a total of 105 transcripts (Figures 4D,E; Supplementary Table S8; cutoff: |log₂ fold change| > 0.6 and p < 0.05). Of note, Wnt ligand biogenesis and trafficking was also the top canonical pathway for the identified 105 inversely regulated transcripts (Figure 4D). Moreover, GSEA identified decreased expression of Wnt signaling genes for comparison VI (cutoff: |log₂ fold change| > 0.6 and p < 0.05; Figure 4F), which supports the results of the IPA performed (Figure 4B). The oppositely expressed Wnt signaling transcripts between comparisons VI and XII comprise Wnt5b, Wnt9b, and secreted frizzled-related protein (Sfrp5), which have been associated with the pathogenesis of HF (Figure 4G) (54–58). Results of the RNA sequencing experiment were confirmed by RT-PCR analysis (Figure 4H). Together, these data suggest that VitAD attenuates cardiac rupture in Stra6-deficient hearts following ischemic injury, at least in part, by the opposite expression of Wnt5b, Wnt9b, and Sfrp5 and reversing the decrease in Wnt signaling.