Cardiac ISL1-interacting Protein (CIP) (14) or Muscle-enriched A-type Lamin-Interacting Protein (MLIP) (18) has been previously identified as a striated muscle-enriched gene. In the heart, CIP is dominantly expressed in cardiomyocyte cells. However, transcriptional factor(s) that control the expression of CIP in cardiomyocytes have not yet been investigated. Genome-wide binding sites for multiple cardiac transcriptional factors, including GATA4, Tbx5, Nkx2-5, Mef2A, and SRF, have been carefully investigated (19). Binding sites for GATA4, Tbx5, and Nkx2-5 were detected in the promotor region of CIP (Figure 1A). Knock-down of GATA4 alone or both GATA4 and mef2a decreased the expression of CIP in cardiac cells (Supplementary Figure 1), indicating GATA4 is a key transcriptional factor for the expression of CIP. To determine whether CIP is directly regulated by GATA4, we generated luciferase reporters controlled by a series of CIP promotor sequence variants, each of which contains the GATA4-binding sequences detected by previous CHIP-seq assay (19). In transient cell-based luciferase assays, all CIP-Luc reporter variants were responsive to GATA4 transactivation. The 554bp CIP-Luc reporter construct, containing a proximal GATA4 binding sequence, was the most responsive one in the assay, indicating a functional GATA4 enhancer located in this binding sequence (Figure 1B). Four putative GATA4 binding motifs, which are evolutionarily conserved, were found in 554 bp CIP promotor fragment (Figure 1C). To determine which of the binding motifs were functional, we generated three mutant variants of the 554 bp CIP-Luc reporter that contained mutant versions of motifs 1 and 2, motif 3, and motif 4, respectively. The results of the luciferase assay revealed that mutations in motif 3 and motif 4, but not motif 1 and 2, decreased the responsivity of 554 bp CIP-Luc reporter to GATA4 transactivation (Figure 1D), suggesting motif 3 and motif 4 are functional GATA4 binding motifs in CIP promotor. Therefore, our data demonstrated that GATA4 is a key transcriptional factor regulating the expression of CIP in cardiomyocytes.

In order to further investigate the relevance of CIP in human cardiac diseases, we collected a total of 194 RNA-seq datasets based on analyses of human heart tissue from the NCBI public database, including 53 non-failing heart samples (NF), 28 hypertrophic cardiomyopathy samples (HCM), 40 ischemic cardiomyopathy samples (ICM), and 73 dilated cardiomyopathy samples (DCM). After normalizing the data from different batches, we found that expression of CIP was significantly decreased in all groups of diseased hearts, while cardiac disease markers, NPPA and NPPB, were significantly upregulated (Figure 2A). Furthermore, the downregulation of CIP in human DCM hearts was further confirmed by Western blotting analysis (Figure 2B). Next, we examined the data to detect genes co-expressed with CIP to define a potential CIP-based gene regulation network. Human genes were divided into multiple modules based on the Spearman correlation coefficient score of genes, then the module containing CIP was further subjected to KEGG pathway analysis. Pathways related to “Parkinson disease,” “oxidative phosphorylation,” “Huntington disease,” and “metabolic pathways” were on the top of the list (Figure 2C). Given that we focused on cardiac disease in this study, our data suggest that CIP was involved in the regulation of OXPHOS pathway. OXPHOS-related genes with the strongest correlation coefficient, including ATP5PF, SDHB, SDHA, etc., are illustrated in a Circos plot (Figure 2D). Together, our data from this genome-wide analyses of human diseased hearts suggest CIP plays an important role in human heart disease by regulating oxidative phosphorylation in cardiomyocyte.

In a previous study, we reported that overexpression of CIP in the heart before cardiac stress inhibited the stress-induced cardiac mal-remodeling (15). In order to test the therapeutic potential of CIP in interfering the progression of adverse cardiac remodeling, we performed transverse aortic constriction (TAC) surgery, which induced ventricular pressure overload, on an inducible cardiac-specific CIP overexpressing mouse model (Rosa26-CIP-flox; Myh6-MerCreMer, and CIP-OE mice). Mice were administrated Tamoxifen to induce the cardiac-specific overexpression of CIP 2 weeks after TAC surgery (Supplementary Figure 2). The establishment of cardiac remodeling was confirmed at 2 weeks after TAC as indicated by the echocardiography examination of the left ventricular posterior wall (LVPW; Figures 3A,B and Table 1). Cardiac parameters were measured at 2 and 8 weeks post-surgery by echocardiography (Table 1). Compared to the sham-operated group, the ventricular wall became thicker and the ventricular chamber was dilated in control mice when the stress was prolonged (Figures 3A,C). As expected, their cardiac function reflected by fraction shortening (FS) decreased significantly, indicating disease progression to heart failure (Figures 3A,D). In contrast, overexpression of CIP in the heart during stress significantly repressed the thickening of the ventricular wall and chamber dilation. More importantly, cardiac function was preserved by CIP overexpression during the prolonged stress (Figures 3A–D).

Mice were sacrificed at 8 weeks post-surgery for cardiac tissue collection. Consistent with echocardiographic data, significant adverse cardiac remodeling was induced by pressure overload in control mice, which was indicated by the ratio of ventricular weigh vs. body weight (Vw vs. Bw) and histological examination (Figures 4A,B). Overexpression of CIP in the heart resulted in a smaller heart under pressure overload. Consistently, cardiomyocyte size, indicated by its cross area determined by Wheat Germ Agglutinin (WGA) staining, was significantly larger in the control group after TAC but became smaller when CIP was overexpressed (Figure 4C). Heart failure is often accompanied with increased cardiac fibrosis. Sirius red/Fast green staining showed less cardiac fibrosis in CIP-OE hearts under cardiac stress compared to the control group (Figure 4D). We further examined the molecular markers for cardiac disease and fibrosis, including NPPA, NPPB, ACTA1, and FBN1. The induction of these marker gene expression by cardiac stress was significantly repressed by the CIP overexpression (Figure 4E). Together, all these data demonstrated that cardiac overexpression of CIP after the pathological remodeling has been initiated is able to inhibit disease progression and protect the heart from failure.

In order to investigate the potential mechanism of the protective effect of CIP in cardiac remodeling, we carried out unbiased transcriptome profiling with hearts of CIP-OE mice at 8 weeks post-surgery by performing RNA-seq. In total, 444 genes, including CIP, were significantly up-regulated, while 792 gene were significantly down-regulated in CIP-OE hearts (fold change > 1.5, p < 0.05; Figure 5A). A hierarchical clustering heatmap revealed that the dysregulation of these genes was induced by TAC surgery in the control group but rescued by the overexpression of CIP in stressed hearts (Figure 5B). In order to further characterize this subset of genes, 1,236 dysregulated genes were subjected to GO term analysis. Consistent with what we had found in human diseased hearts, genes related to “Mitochondrion” and “Oxidative phosphorylation” were enriched in the down-regulated genes (Figure 5C). To gain more information of the gene regulation network, we further searched for upstream regulators of these dysregulated genes using Ingenuity Pathway Analysis (IPA). Several regulators, including IGF1R (p = 7.27e-8; Figure 5D), Rictor (a core component of mTORC2, p = 6.25e-14; Figure 5D) and TGFB1 (p = 4.76e-7; Supplementary Figure 3), were showed in the top list. It is worth noting that the IGF1R was reported to modulate the activity of FoxO1 through regulating AKT, which is consistent with our previous report that CIP regulated pathological cardiac remodeling through the FoxO1/CnA signaling cascade (15). Dysregulation of downstream targets for IGF1R from IPA analysis, including TGFB1, BAX, CEBRA, CEBRB, COX6A2, COX8B, DES, NDUFA1, and NDUFB9, as well as several downstream targets for Rictor and TGFB1 were further confirmed by qRT-PCR (Figure 5E). All these data suggested that the inhibitory function of CIP on the progression of adverse cardiac remodeling was mediated, at least in part, by IGF and mTORC2 signaling pathways.

Left ventricular (LV) tissues were collected from patients with end-stage heart failure during heart transplantation performed in the First Affiliated Hospital, Sun Yat-sen University. In brief, diseased hearts were removed at the time of transplantation and LV tissue was subsequently dissected and snap-frozen. We used LV samples from not implanted healthy hearts to serve as controls (Supplementary Table 1). All the procedures followed the protocol approved by the First Affiliated Hospital, Sun Yat-sen University, Guangzhou, China.

Cardiac ISL1-interacting Protein-KI-flox mice were generated in a previous study (15). CIP-KI-flox mice, which have a Rosa-CIP allele (the stop codon is present and floxed) were bred with aMHC–Mer–Cre–Mer mice to obtain CIP-OE (CIP-KI-flox; aMHC–Mer–Cre–Mer) mice. Tamoxifen was administrated through intraperitoneal injection to activate the expression of Cre recombinase and the excision of the stop codon for the ectopic expression of CIP transgene in the heart in CIP-OE mice. CIP-KI-flox littermates were used as controls.

Echocardiographic measurements were performed on mice using a Visual Sonics Vevo^® 2100 Imaging System (Visual Sonics, Toronto, ON, Canada) with a 40 MHz MicroScan transducer (model MS-550D). Mice were anesthetized with isoflurane (2.5% isoflurane for induction and 0.1% for maintenance). Heart rate and LV dimensions, including diastolic and systolic wall thicknesses, LV end-diastolic and end-systolic chamber dimensions were measured from 2-D short-axis under M-mode tracings at the level of the papillary muscle. LV mass and functional parameters such as percentage of fractional shortening (FS%) and ejection fraction (EF%) were calculated using the above primary measurements and accompanying software.

Mice were anesthetized with isoflurane (3–4% isoflurane for induction, 1–2% isoflurane for maintenance). The chest was shaved and cleaned with alcohol. A suture was placed around the front upper incisors and pulled taut so that the neck was slightly extended. The tongue was retracted and held with forceps, and a 20-G catheter was inserted into the trachea. The catheter was then attached to the mouse ventilator via a Y-shaped connector. Ventilation was performed with a tidal volume of 220–240 μl for a 25–30 g mouse and a respiratory rate of 130–140 breaths per min. 100% oxygen was provided to the inflow of the ventilator. The chest was opened through a left 2nd intercostal thoracotomy. The 26-G needle without its sharp tip was put on the ascending aorta. They were tightly ligated together using 7-0 Nylon suture (Ethicon, Edinburgh, Scotland) at the position between brachiocephalic artery and left common carotid artery, and the 26-G needle was removed immediately after ligation. In the sham operation, no ligation was performed. Isoflurane was stopped, and the lungs were slightly overinflated to assist in removal of air in the pleural cavity. Dissected intercostal space and chest skin were closed using 6-0 silk suture (Ethicon, Edinburgh, Scotland). All manipulations were performed by an operator without knowledge of genotype.

Mouse heart tissues were dissected from the animals, rinsed with PBS and fixed in 4% paraformaldehyde (pH 8.0) overnight. After dehydration through a series of ethanol baths, samples were embedded in paraffin wax according to standard laboratory procedures. Sections of 5 μm were stained with haematoxylin and eosin (H&E) for routine histological examination with light microscope. For Sirius red/fast green collagen staining sections were fixed with pre-warmed Bouins’ solution, 55°C for 1 h then washed in running water. Sections were stained in 0.1% fast green solution for 10 mins then washed with 1% Acetic acid for 2 mins. After rinsing in tape water, sections were stained in 0.1% Sirius resolution for 30 mins. After staining, sections were dehydrated and cleared with Xylene. The images were examined with light scope and quantified with ImageJ software.

Total RNAs were isolated using Trizol Reagent (Life Technologies, Carlsbad, CA, United States) from cells and tissue samples. For Quantitative RT-PCR, 2.0 μg RNA samples were reverse-transcribed to cDNA by using random hexamers and MMLV reverse transcriptase (Life Technologies) in 20 μl reaction system. In each analysis, 0.1 μl cDNA pool was used for quantitative PCR. The relative expression of interested genes is normalized to the expression of ACTB or PRKG1. Primers used in this study were listed (Supplementary Table 2). For Western blot analyses, tissue homogenate were cleared by 10,000 × g centrifugation for 10 min. Samples were subsequently analyzed by SDS/PAGE and transferred to PVDF membranes that were incubated with 5% non-fat dry milk in TBST and Anti-CIP (1:2,000, 21st Century Biochemical, Marlboro, MA, United States) or Anti-GAPDH (1:5,000, Proteintech, Rosemont, IL, United States) overnight at 4°C and then washed three times with TBST buffer before adding IgG secondary antibody. Specific protein bands were visualized through chemiluminescent detection.

HEK293T cells were cultured in DMEM supplemented with 10% FBS in a 5% CO² atmosphere at 37°C. Wildtype, mutant or truncated CIP promoter sequences were cloned into multiple cloning sites of the pGL3-Basic vector (Promega, Madison, WI, United States) to generate CIP-Luc reporters used in this study. The indicated combinations of CIP-Luc reporter, pRL Renilla reporter (internal control) and GATA4 construct were transfected into HEK293T cells with PEI reagents. Forty Eight hours after transfection, cell extracts were prepared and luciferase activity was determined. For dual-luciferase assay, normalized firefly luciferase expression from triplicate samples in 12-well plates relative to renilla luciferase expression was calculated.

Total RNAs from mouse heart were used to perform RNA-seq in BGI Genomics (Wuhan, China). RNA-seq reads were mapped to mouse genome mm10 by STAR and reads counts were calculated with FeatureCounts. Expression analysis was run in RStudio. DESeq2 was employed to perform statistical analysis of differential gene expression. An adjusted P value of 0.05 were used as cutoff to identify differentially regulated genes. Volcano plot were performed with the ggplot2 library. Hierarchical clustering heatmap was made with the pheatmap library. The raw data of RNA-seq in this study were deposited in Gene Expression Omnibus (GEO) database of the National Center for Biotechnology Information (NCBI) (Accession: GSE194149).

Human diseased heart RNA-seq data, including GSE57344, GSE71613, GSE116250, GSE46224, GSE108157, GSE55296, GSE120836, and GSE130036, were downloaded from NCBI database. Gene-level quantification were calculated by featureCounts-v1.6.3. To perform strand-specific reads counting, the strand type (non-strand, stranded, reversely stranded) of each sample was inferred from sorted bam file using infer_expriment.py (3.0.0). Then we provided featureCounts with strand type information to calculate read counts of every gene in each sample and merged the quantification results together to make an expression matrix for differential gene expression analysis. Differential gene expression analysis was performed using DESeq2-1.24.0. The design matrix in DESeq2 model was written as “∼series + gender + phenotype” to adjust the differences between data series and gender. Only differential expressed genes with FDR < 0.05 and log2FoldChange >0.25 identified by DESeq2 were kept. Then, we applied the classic weighted correlation network analysis (WGCNA) algorithm for co-expression analysis. The R implementation of WGCNA (version: 1.68) was used in our study.

Values are reported as means ± SEM unless indicated otherwise. Statistical significance was determined with ANOVA. For multiple group comparisons, a post hoc Tukey’s test was performed when ANOVA reached significance. Values of P < 0.05 were considered statistically significant.