TREATMENT AND PREVENTION OF DISEASE MEDIATED BY WWP2
20220251569 · 2022-08-11
Assignee
Inventors
Cpc classification
A61K31/519
HUMAN NECESSITIES
A61P29/00
HUMAN NECESSITIES
C12Y603/02019
CHEMISTRY; METALLURGY
International classification
C12N15/113
CHEMISTRY; METALLURGY
A61K31/519
HUMAN NECESSITIES
A61K31/55
HUMAN NECESSITIES
Abstract
Methods of treating and preventing fibrosis and pathological inflammation through inhibition of WWP2 are disclosed, as well as agents for use in such methods.
Claims
1. A WWP2 inhibitor for use in a method of treating or preventing fibrosis.
2. Use of a WWP2 inhibitor in the manufacture of a medicament for use in a method of treating or preventing fibrosis.
3. A method of treating or preventing fibrosis, comprising administering a therapeutically or prophylactically effective amount of a WWP2 inhibitor to a subject.
4. The WWP2 inhibitor for use according to claim 1, the use according to claim 2, or the method according to claim 3, wherein the fibrosis is of the heart, kidney, liver, lung, skeletal muscle, blood vessels, eye, skin, pancreas, bowel, small intestine, large intestine, colon, brain, and/or bone marrow.
5. A WWP2 inhibitor for use in a method of treating or preventing pathological inflammation.
6. Use of a WWP2 inhibitor in the manufacture of a medicament for use in a method of treating or pathological inflammation.
7. A method of treating or pathological inflammation, comprising administering a therapeutically or prophylactically effective amount of a WWP2 inhibitor to a subject.
8. The WWP2 inhibitor for use according to claim 5, the use according to claim 6, or the method according to claim 7, wherein the pathological inflammation is associated with a chronic infection, a cancer, an autoimmune disease, a degenerative disease or an allergic disease.
9. The WWP2 inhibitor for use, the use or the method according to any one of claims 1 to 8, wherein the WWP2 inhibitor is an inhibitor of a WWP2 isoform comprising the C2 domain, optionally wherein the WWP2 inhibitor is an inhibitor of WWP2-FL and/or WWP2-N.
10. The agent, use or method according to any one of claims 1 to 9, wherein the WWP2 inhibitor is a WWP2-binding molecule, a WWP2 target-binding molecule, or a molecule capable of reducing expression of WWP2.
11. The agent, use or method according to any one of claims 1 to 10, wherein the WWP2 inhibitor is a small molecule.
12. The agent, use or method according to any one of claims 1 to 11, wherein the method comprises administering the agent to a subject in which expression and/or activity of WWP2 is upregulated.
Description
BRIEF DESCRIPTION OF THE FIGURES
[0306] Embodiments and experiments illustrating the principles of the invention will now be discussed with reference to the accompanying figures.
[0307]
[0308]
[0309]
[0310]
[0311]
[0312]
[0313]
[0314]
[0315]
[0316]
[0317]
[0318]
[0319]
[0320]
[0321]
[0322]
[0323]
[0324]
[0325]
[0326]
[0327]
[0328]
[0329]
[0330]
[0331]
[0332]
EXAMPLES
[0333] In the following Examples, the inventors describe the identification of WWP2 as a positive regulator of a pro-fibrotic gene network common to a range of chronic diseases. They show that increased WWP2 expression in the human diseased heart leads to increased expression of this gene network of extracellular matrix proteins and pro-fibrotic genes (including collagens, MMPs, cytokines, etc.), and that the positive association between expression of WWP2 and pro-fibrotic gene expression exists in tissues of various organs in the context of chronic disease.
[0334] The inventors further demonstrate that WWP2 regulates the TGFβ-induced transcriptional response of pro-fibrotic genes, and that transgenic mice lacking the N-terminal region of WWP2 also show a reduction in the complement and coagulation cascade, which is responsible for the activation of immune cells during the fibrogenic response.
[0335] The inventors further demonstrate that inhibition of WWP2 expression reduces macrophage accumulation and expression of proinflammatory factors by macrophages in a fibroinflammatory disease setting.
[0336] The inventors further demonstrate that inhibition of WWP2 expression reduces mortality and tissue fibrosis in a model of lung fibrosis, and that inhibition of WWP2 expression reduces tissue fibrosis in a model of kidney fibrosis, that WWP2 expression is upregulated in renal tissue of human subjects with chronic kidney disease, and that inhibition of WWP2 expression abrogates TGFβ-induced upregulation of ACTA2 expression in fibroblasts.
[0337] The inventors further demonstrate that inhibition of WWP2 using a small molecule inhibitor targeting the WWP2 N-terminal isoform (WWP2-N) comprising the C2 domain abrogates TGFβ-induced upregulation of ACTA2 expression in fibroblasts, confirming that WWP2 inhibitors are useful for the treatment/prevention of fibroinflammatory disease.
Example 1: Materials and Methods
[0338] 1.1 Description of Rat Segregating Population and Human Cohorts (RNA-seg, Phenotypic and Genetic Data) and Data Processing
[0339] Rat Recombinant Inbred (RI) Strains Data
[0340] Rat RI Left Ventricle (LV) Expression Data
[0341] This expression dataset consists of the heart LV RNA-sequencing in the 30 RI strains. This data was published in [3]. TopHat 1.2.0 [4] was used to map the reads to the BN reference genome RGSC 3.4 (known splice junctions in the Ensembl reference database [5] were supplied, de novo splice junction detection was also enabled). After mapping the reads to the reference genome, Cufflinks 1.0.2 was used to assemble the aligned RNA-seq reads into transcripts by using the Ensembl annotation of the rat transcriptome. Cufflinks is able to reconstruct the set of transcripts that “explains” the reads observed in our RNA-seq experiment, in our case, by using a transcriptome reference annotation. The underlying reason for using a reference-based transcriptome assembly approach is that, comparative transcriptome assembly analyses have shown that, overall, reference-based transcriptome assembly approaches have a better performance than de novo approaches, in both sensitivity and the ability to discover splicing patterns [6]. Cufflinks measures transcripts abundances in Fragments Per Kilobase of transcript per Million mapped reads (FPKM). One of the assumptions of reference-based transcriptome assembly approaches is that all the isoforms of all the genes are known. However, transcriptomic reference sets are still incomplete, more specially in the case of the rat transcriptome [6]. Hence, the transcriptomic assembly is not very accurate in rat. Therefore, all the RNA-seq expression data analyses in the rat have been carried out by summarising expression at the gene level.
[0342] Steps followed for gene level FPKM computation and filtering of the data: 1) Replace the values of all the isoforms with isoform quantification status not OK by missing values. The isoform quantification status is part of the output from Cufflinks and it measures the successful deconvolution of each isoform. 2) Remove all isoforms with more than 10 missing values across all samples. 3) Impute all the missing values by using R function pca from the pcaMethods R package 1.56.0 [7]. The initial number of isoforms was 39,492. After removing those with more than 10 missing values across all samples, the number of isoforms got reduced to 39,181. 4) Sum up all the FPKM values of all the isoforms of each gene to get the FPKM value at the gene level (Number of genes: 29,469). 5) Filter out genes that did not have FPKM>1 in at least a 5% of the samples (in at least 2 samples). 6) Replace Os in the data by the minimum number in the data set higher than 0. Afterwards, Log2 transform the data. By following these steps, we ended up with a filtered data set of 12,061.
[0343] The inventors also inspected the presence of unmeasured confounding factors (i.e. potential sources of expression variation that are not being measured in the study and they may be affecting gene expression) [8]. We computed the variability explained by each of the PCs of the transcriptomic data. We found out that the 1st PC was explaining a large proportion of the variance present in the data (38%) and it was significantly correlated to the library concentration of the Bioanalyser, (a library quantification measure, Spearman's ranked correlation=0.70, p-value=1.5×10.sup.−5). To account for possible confounding effects introduced by the differences in library concentration, we adjusted the gene expression data for this 1st PC using a linear regression model from which the residuals were computed. These residuals obtained from the regression model represent the variability present in the data after removing the effect of this 1st PC. These adjusted expression levels were the ones considered in all analyses.
[0344] Rat Histomorphometric Cardiac Fibrosis
[0345] Histomorphometric measures of both interstitial and perivascular fibrosis were collected in the LV of the 30 RI strains (total n=180; n=5-7 per strain, males at 30 weeks of age). After fixation, short-axis heart slices were processed for paraffin embedding. Multiple 4 μm thick sections were de-paraffinized, rehydrated and picrosirius red-stained sections were prepared for evaluating fibrosis. The presence, type and extent of both interstitial and perivascular fibrosis was quantified via thresholding automated analysis using ImageJ 1.43 [9]. Blood pressure measurements were published in [10]. Indwelling aortic radiotelemetry transducers (Data Sciences International) at 8 weeks of age were implanted to measure arterial pressure in conscious, unrestrained rats. Radiotelemetry pressure was collected in 5-s bursts every 10 min and recorded over a period of 8 days, using 6-12 rats for each RI strain. The obtained blood pressure measurements were averaged within each RI strain and across eight sequential readings. Median blood pressure effects were removed from the interstitial and perivascular fibrosis measurements by performing standard multiple linear regression and then taking the residuals of the model.
[0346] Genetic Data in the Rat RI Strains
[0347] Within the rat RI panel, SNPs are a comprehensive source of genetic diversity available for genetic association studies. The genetic map of the rat BXH/HXB RI strains was generated by the STAR Consortium as described in [10]. This genetic map was generated from over 13,000 SNP that led to 1,384 unique blocks of adjacent SNPs with identical strain distribution patterns.
[0348] Human Dilated Cardiomyopathy (DCM) and Healthy Expression Data from Left Ventricle (LV) and Genetic Data
[0349] Human LV RNA-seq data was collected in 128 DCM patients and 106 controls. In addition, genotyping data was collected in 96 DCM and 91 controls samples. We performed a quality control step (QC) in the RNA-seq data and removed outlier samples (see section “b. RNA extraction, sequencing and RNA-seq data processing”). After QC, the final cohort size used for co-expression analyses was 126 DCM patients and 92 control samples. The number of samples used for genetic mapping was 96 DCM and 91 controls (i.e. samples for which we have both RNA-seq and genotyping data). This data were published by Heinig et al [11] and are available at the European Genome-phenome Archive under the accession number EGAS00001002454.
[0350] DCM/Controls LV Cohort Description
[0351] This data represents a retrospective cohort of heart patients diagnosed with DCM (obtained from the Royal Brompton and Harefield NHS Foundation Trust Tissue bank: EC Ref: 09/H0504/104+5) and healthy LV donors (healthy with respect to myocardial diseases such as DCM and HCM). Control left ventricular samples (healthy LV donors) were collected from healthy human hearts of non-related organ donors whose hearts were explanted to obtain pulmonary and aortic valves for transplant, valve replacement surgery or explanted for transplantation but not used due to logistical reasons. Control LV tissue was collected in four different centres: University of Szeged (Hungary), Vanderbilt University (Nashville, USA), University of Miami (USA), and the University of Sydney (Australia) [12]. In both DCM patients (“cases”) and controls only adult subjects 16 years old) were selected.
[0352] RNA Extraction, Sequencing and RNA-seq Data Processing
[0353] Reads were mapped to the human genome with TopHat 1.4.1 [4]. TopHat was run using annotation from Gencode release 19 (GRCh37.p13) annotation and allowing only 2 mismatches per 100 bp. In addition, TopHat was run with option −r 0, which specifies as zero the expected mate inner distance (as fragment size was 200 bp and read length was 100 bp). Option −M was also set, which removes multimapping reads before aligning to the transcriptome. Default options were chosen for the rest of parameters. TopHat mapping resulted in a mean of 173 M reads per sample being uniquely mappable (a mean of 92% of the total number of reads with all the samples having a mapping percentage higher than 60%) in the samples from DCM patients. Controls samples resulted in a mean of 185 M reads per sample being uniquely mappable (a mean of 90% of the total number of reads with all the samples had a mapping percentage higher than 60%).
[0354] RNA-seq read gene counts were computed with HTSeq software 0.5.3p3 [13]. In HTSeq the mode ‘intersection-nonempty’ was selected (mode suitable to quantify overlapping transcripts on different strands). To compute gene counts with HTSeq, we used the Gencode human annotation version 19 with a custom TTN annotation [11].
[0355] Steps followed for processing of the data: 1) Gene selection: From HTSeq output, only “protein coding” genes with “known” status in Ensembl genes (GRCh37.p13 data set) were consider in the analyses (n=19,456). 2) Filtering of low expressed genes: Transcript lengths were downloaded from Biomart Ensembl genes (GRCh37.p13 data set). FPKM was computed with DESeq2 1.6.3 R package [14] by considering for each gene, the average length of all the transcripts. A FPKM-based gene filtering was then applied (we only kept genes with FPKM>1 in at least 5% of the samples, considering DCM and controls samples together). This yielded a final number of 14,281 genes that were considered to be expressed in the samples cohort. 3) Normalisation and data transformation: After the FPKM filtering, gene raw counts were normalised and variance-stabilized transformed (VST) by using DESeq2 1.6.3 R package [14]. 4) Covariates adjustment: VST data was split into DCM patients and healthy controls. Then, the data was adjusted separately for relevant technical (“RIN score” and “library preparation day”) and clinical (“sex” and “age at tissue collection”) covariates by using a multivariate linear model. In the case of the healthy controls, the data was also adjusted for the centre from which the tissue had been collected by adding this information as an additional covariate. 5) Outlier samples removal (samples QC): Outliers samples were inspected by clustering the resulting samples gene expression levels. The DCM and control samples were independently clustered by hierarchical clustering with the R function flashClust from flashClust 1.01-2 R package [15], the agglomeration method used was set to “average”). The obtained dendrograms were cut at the 99th percentile of the dendrogram height distribution removing all the samples assigned to small clusters: 2 DCM patients and 14 healthy donors. By following these steps, we ended up with two LV adult cohorts of 126 DCM patients (cases) and 92 healthy donors (controls). Within the 126 DCM patients, 106 were males and 20 were females. In the Healthy Donors group there were 55 male and 37 female. The mean age in the DCM patients was 41.3±13.3 years (16 to 64), whereas the mean age in the controls was 42.6±13.6 years (17 to 72).
[0356] Genetic Data
[0357] A subset of the DCM and control LV samples from the CEU population (96 DCM samples and 91 control samples [11]), were typed by using genotyping arrays. After imputation and quality control, the final number of SNPs in the DCM patients and healthy controls LV samples was 1,309,892 SNPs.
[0358] In order to run genetic mapping, we carried out a pairwise linkage disequilibrium (LD) filtering (with an R.sup.2 threshold of 0.8) based on the 1,000 Genomes pilot 1 CEU population. Pairwise SNPs LD was computed using SNAP 2.2 [16] using as input the SNP identifiers. The SNP dataset chosen to run SNAP was the 1000 Genomes pilot 1 in the CEU reference population with an R2 threshold of 0.8 and a distance limit of 500bp. From SNAP output, SNPs returning a warning message as not present in the SNAP database were removed and not considered further analyses (the number of missing SNPs was 101,020 SNPs).
[0359] SNAP web tool outputs the input SNPs clustered in LD blocks (a total number of 126,922 LD blocks were returned by SNAP). To obtain a genomic map filtered by LD, we selected one single SNP for each of the LD blocks (R.sup.2=0.8). SNPs that were not included in any of the LD blocks (not clustered by SNAP) were added at the end. For each LD block we followed these steps: 1) Get all the SNPs in the LD block and compute their pairwise Spearman's ranked correlation. 2) As representative of the LD block, take the SNP that has the highest average Spearman's ranked correlation with the rest of SNPs included in the LD block.
[0360] With this procedure we obtained a set of 126,922 SNPs, each of them representing a SNPLD block. Finally, we added to these SNPs the ones that were not included by SNAP in any of the LD blocks (41,315 SNPs). Then, the final number of SNPs in our LD pruned (R.sup.2=0.8) genetic map was 168,237 SNPs.
[0361] Human Right Ventricle (RV) Cohort Data
[0362] This data is a cohort of RV RNA-seq samples with repaired Tetralogy of Fallot (rTOF) patients (fibrotic RV) and control samples. Tetralogy of Fallot (TOF) is a congenital heart disease (i.e. problem in the structure of the heart) characterized by by pulmonary artery stenosis, ventricular septal defect (a hole between the LV and RV and an overriding aorta which allows blood from both ventricles to enter the aorta (leading to cyanosis) and RV hypertrophy. TOF needs used to be treated surgically in the first year of life to increase the size of the pulmonary valve and arteries and repairing the septal defect. However, surgical repaired TOF is usually followed by cardiac fibrosis in both ventricles [17]. Additionally, TOF cases are operated several times during heart post-natal development because the prostatic material used at surgeries is unable to grow with the growing of the heart and valve. Therefore, surgery leftover samples can be collected at different ages of the patients.
[0363] rTOF Cohort Description
[0364] In this cohort there are 27 patients with rTOF (mean age 32.0±10.6 years, 78% male). All the patients aged ≥16 years old and were scheduled for elective pulmonary valve replacement. They were also under the care of the Adult Congenital Heart Disease service at the Royal Brompton Hospital, UK. For the rTOF patients, the clinical data from noninvasive investigations performed as part of the surgical work-up was available. This clinical information included electrocardiography, chest radiograph, echocardiography and CMR. RV myocardial tissue samples from the 27 patients were snap-frozen in liquid nitrogen at the time of tissue sampling intra-operatively.
[0365] RV tissue from 11 structurally normal hearts donated for cardiac transplantation was also collected (age donors: 34.0±13.0, Sex, Male/Female: 6/5). RV myocardial biopsies were made available via the Cardiovascular Biomedical Research Unit Biobank of the Royal Brompton & Harefield NHS Foundation Trust. Donor RV control tissue was collected and stored at the time of surgery following similar procedures as with the TOF tissue samples.
[0366] RNA Extraction, Sequencing and RNA-seq Data Processing
[0367] TRIzol (Life Technologies) was used for total RNA extraction from the frozen samples by following the manufacturer's protocol. RNA was quantified by ultraviolet spectrophotometry and RNA quality was assessed on the Agilent 2100 bioanalyser. RINs ranged from 6.3 to 9.1 (mean 8.2±0.6). 1 μg of total RNA was used to prepare the RNA-Seq libraries. RNA-Seq libraries were prepared with Illumina TruSeq RNA sample preparation kits by using the protocol for poly-A enriched mRNA. To avoid batch effects, samples were pooled (4-5 samples/pool, 2 lanes per pool). Finally, paired-end 2×100 bp sequencing was performed on the Illumina Hi-Seq platform (mean sequencing depth of 196M). TopHat 2.0.12 with Bowtie2 2.2.3 and Samtools 0.1.18 [4] was run by using human genome version GRch38 (hg38.78) reference genome. RNA-seq read counts were computed with HTSeq 0.6.1 [13]. The percentage of reads mapping to the human genome was higher that 80% (above 70% is considered an acceptable mapping percentage for paired-end sequenced reads).
[0368] Steps followed for the normalization, filtering and adjustment of the data: 1) Gene selection: From HTSeqoutput, only “protein coding” genes with status “known” in Ensembl genes (GRCh37.p13 data set) were selected (18,964 genes). 2) Filtering of low expressed genes: FPKMs were computed with DESeq2 1.6.3 R package [14] using the average transcript length of each gene, which were retrieved from Ensembl Biomart (GRCh37 version). A FKPM-based filtering criteria was applied by keeping only those genes with a value of FPKM>1 in at least 5% of the samples (in this case 2 samples). Following these criteria, the number of genes got reduced from 18,964 genes to 13,936 genes. 3) Data transformation: Size factors normalisation and VST was applied to the raw gene counts by using DESeq2 R package 1.6.3. Then, the gene counts that passed the filtering criteria described in the previous step were selected. 4) Covariates adjustment: After normalisation and filtering, the data was split into TOF patients and controls. Gene expression counts of TOF patients were adjusted for: 1) age at which the tissue was collected (age of operation) and 2) gender. This adjustment was performed by taking the residuals of a multivariate linear model in which both age and gender were added as predictors. Among the control samples, some had with missing values for age. More specifically, three samples had missing age. The gender of the sample with missing value was imputed by clustering the expression levels of selected sex specific genes (as described in [18]). Therefore, the gene expression counts of the control samples were adjusted for only sex by taking the residuals of a linear model in which gender was added as a predictor.
[0369] 1.2 Co-Expression Network Analysis in LV
[0370] Co-Expression Network Inference in Rat RI Strains and Human DCM LV Tissues
[0371] Co-expression networks were inferred by using WGCNA 1.42 [19]. Two independent WGCNA runs were carried out in the processed RNA-seq LV data (as described in previous sections): one in rat (RI panel, n=30) and one in the DCM cohort (n=126). In both rat and human runs, WGCNA was run using Turkey's biweight midcorrelation. Biweight midcorrelation estimates are more robust to outliers than the standard Pearson correlation as it assigns lower weights to points further from the centre of the distribution [20]. In the rat run, amongst all the soft threshold values (β) with R.sup.2>0.8, we chose the β that presented the highest mean connectivity (β=8). For the human run, the automatic value of β returned by the WGCNA function pickSoftThreshold was selected (β=6). In both cases, a network merge height of 0.25 was chosen as suggested in the original WGCNA guidelines. For the rest of WGCNA parameters, default settings were used. Once the WGCNA networks were obtained, they were alphabetically sorted by name and renamed as M1-M (number last network, Hs-M in the case of human networks) removing the grey cluster (the grey cluster contains all the non-clustered genes).
[0372] Gene-annotation enrichment analysis of the rat (n=41) and human (n=48) networks was performed by carrying out functional gene list enrichment analysis using the R DAVID Web Service 1.4.0 [21]. This tool provides an R interface to DAVID [22]. Rat networks were queried independently from human networks for Gene Ontology (Biological process, BP, Molecular Function, MF and Cellular Component, CC independently) [23] and KEGG [24] gene annotation categories. The gene reference background was set to the group of genes from which the gene co-expression networks were inferred. In the rat data this background was constituted by the genes robustly expressed in the rat LV data (i.e. after the filtering procedure described in previous sections, n=12,061), whereas in the human data the background used was the set of genes robustly expressed in the human LV RNA-seq cohort (i.e. after the filtering procedure previously described, n=14,281). In this analysis results were deemed significant if DAVID FDR/100<0.05 (i.e. % FDR).
[0373] Association of the Rat Co-Expression Networks with Fibrosis
[0374] Genome-wide gene expression levels in the RI strains LV were correlated with both interstitial and perivascular fibrosis using Spearman's ranked correlation (after correcting for average blood pressure effects as described previous sections). Here we used the measurements of interstitial and perivascular fibrosis in the rat after normal transformation (qqnorm R function used) and adjustment for mean blood pressure. Student p-values were obtained for each correlation estimate by using the function corAndPvalue from the R package WGCNA 1.42 [19].
[0375] All the rat co-expression networks were tested for enrichment of genes varying with fibrosis in the rat heart by carrying out GSEA. In this analysis, each of the rat co-expression networks was considered as a gene set. The corresponding Student p-values of the Spearman's ranked correlation between each of the robustly expressed rat genes and each of the fibrosis measurements were used for ranking all the rat genes (n=12,061) and run GSEA. GSEA 2.1.0 [25] was run in classic, pre-ranked mode with 10,000 iterations. To consider all the co-expression networks, maximum gene set size was set to 5,000 and minimum gene set size was set to 10. As the genes were ranked by p-value, in this GSEA test, significant negative normalised enrichment score (NES, i.e. FDR<0.05) denotes significant association.
[0376] Conservation of Rat and Human Co-Expression Networks
[0377] To assess whether the co-expression networks inferred in the rat were also inferred in the DCM patients, we computed the intersection between rat and human co-expression networks by carrying out Fisher's exact tests. The gene background used in these tests was composed by the genes with one-to-one human-rat ortholog relationship that were both robustly expressed in rat and human LV. For the generation of this gene background, rat human one-to-one orthologs relationships were downloaded from Ensembl archive (Ensembl 69). The common set of genes in both the rat LV expressed genes set (n=12,061) and human LV expressed genes (n=14,281), yielded a common set of 8,840 genes (this will be the rat-human orthologs background). Genes in the rat and human networks that were not included in this rat-human orthologs background were removed.
[0378] For each rat (n=41), human (n=48) network pair, a Fisher's exact test was computed with the R function fisher.test and setting the “alternative” parameter to “greater” (as we are interested in overrepresentation). The gene background used for computing the contingency table was the rat-human orthologs background (8,840 genes). Nominal Fisher's exact test pvalues were adjusted for the number of tests carried out (number of rat networks times number of human networks=1,968 tests) by using the R function p.adjust and the B&Y method.
[0379] Test for Differential Co-Expression of Human DCM LV Networks between DCM Patients and Controls
[0380] Differentially co-expressed networks (i.e. gene networks where the genes as a whole present divergent patterns of co-expression between cases and controls) can point to disease response for example. We carried out the following empirical differential coexpression test to assess the differential co-expression of the human DCM networks between DCM patients and controls samples. First, we computed Tukey's biweight pairwise gene-gene correlations [20] from the DCM and controls expression matrices. Then, for each of the DCM co-expression networks these steps were carried out: 1) Compute the network's dispersion value (DCM versus controls samples) as described in the section 3 of supplementary material of the reference [26]. The dispersion value quantifies the difference between the co-expression of the network in cases and controls; 2) Generate a null distribution of dispersion values for the network by randomly sampling networks with the same number of genes as the network being tested and then, compute the corresponding dispersion value of each randomly sampled network as described in step 1. 3) Compute the empirical p-value for the network under testing as [27]: P=(r+1)/(n+1), where n is the number of simulated dispersion values (number of permutations) and r is the number of simulated dispersion values that are higher than the actual dispersion value of the network of interest.
[0381] This differential co-expression test was run with 100,000 permutations, which yields a minimum nominal significance level of 1×10.sup.−5. Bonferroni-adjusted p-values were computed by correcting the nominal empirical p-values for the number of DCM co-expression networks tested (n=48). p.adjust R function was used to adjust the p-values for multiple testing.
[0382] This differential co-expression test was also carried out for the hECM-network in the rTOF/controls RV cohort. In this case the input was the sex-adjusted VST counts for the genes in the hECM-network. The test was also performed for 100,000 permutations.
[0383] 1.3 Analysis of the Human Networks
[0384] Enrichment of Human Networks for Cardiac Fibroblast and Myocyte Expression
[0385] Description of the Cardiac Data Used
[0386] This RNA-Seq data set consists of primary mouse cardiac fibroblasts (n=2) and cardiomyocytes (n=2). Neonatal cardiomyocytes and cardiac fibroblasts were isolated from 0-1 day old C57bl/6 mice using the procedure described in [28]. Cells were cultured at 37° C. for 48 h in the presence of 10% fetal calf serum in DMEM medium. Total RNA was purified using the RNeasy kit (Qiagen) and sequenced with the Illumina Genome Analyzer IIx, 2 replicates per condition at 75 bp, paired-end library, one lane per replicate (sequencing depth of roughly 40M reads).
[0387] TopHat2 (version 2.0.13) [29] was used to align the paired-end reads to the GRCm38 mouse genome using available transcript annotations from Ensembl release 76. Mean insert size and standard deviation were computed empirically from uniquely mapping, perfect matching mate pairs via a preliminary alignment with Bowtie2 (version 2.2.3) [30]. These were supplied as input parameters to TopHat2. Default parameterisations were used for the rest of options. Read counts for each gene were retrieved using HTSeq (version 0.6.1p1) with default parameters [13].
[0388] Enrichment Test
[0389] When inferring gene co-expression networks from complex tissue (i.e. LV), the obtained gene networks could be capturing transcriptional programs coming from specific cell types. We set out to explore this by analysing gene expression data from the two most abundant cell types in the heart: cardiac fibroblasts and myocytes [31]. We ran a test to assess whether any of the human DCM networks were overrepresented for genes with higher or lower expression levels in cardiac fibroblasts when comparing to cardiomyocytes. More specifically, we carried out a GSEA analysis of cell type gene expression data from the mouse cardiac fibroblasts and myocytes described hereinabove.
[0390] Mouse-human orthologs relationships was downloaded from Ensembl archive (Ensembl version GRCh37). Using this annotation, all mouse genes were mapped to their corresponding human ortholog by taking only those human genes with one-to-one (ortholog_one2one) human-mouse relationship. This resulted in 11,862 genes. The cardiac fibroblast-to-myocyte fold change was computed for each of these 11,862 genes by dividing the average expression counts in cardiac fibroblasts by the average expression counts in myocytes. This leads to a cardiac fibroblast-tomyocyte fold change value for each gene representing how many times is higher or lower the expression of that gene in fibroblasts in comparison to myocytes. Afterwards, all the human genes (11,862 genes) were ranked by their corresponding value of fibroblast-to-myocyte fold change. Then, this information was used as ranked list to run GSEA [25]. In this GSEA analysis we used the human DCM clusters (considering only the genes present in these 11,862 genes) as gene sets. We run GSEA (version 2.1.0) [25] in classic, pre-ranked mode with 10,000 iterations by setting the maximum gene set size to 5,000 and the minimum gene set size to 10.
[0391] Test for Conservation of the Human Networks in rTOF RV Tissue
[0392] To explore the conservation in rTOF RV patients of the prioritised human network inferred in human DCM LV, we implemented an empirical permutation test that assesses how likely is it to find by chance a set of genes of the same size as the network with a similar degree of average absolute correlation in the data set under testing. In this empirical we assess the conservation of each human network in rTOF expression data set as following: 1) From the background to test (i.e. expression data set in which we want to see whether our network is conserved), randomly sample without replacement a set of genes as the same size of the co-expression network under testing. Compute the mean absolute correlation of the sampled genes without considering self-correlations. We repeat this step for a number of times (e.g. 100,000) to generate a null distribution of simulated mean correlations. 2) Compute the mean absolute correlation (without considering self-correlations) of the genes in each network in the data set in which we are testing for conservation (rTOF). 3) Compute the corresponding empirical p-value by using the values obtained in steps 1 and 2 as in [27]: P=(r+1)/(n+1), where n is the number of simulations (number of permutations) and r is the number of simulated mean correlation values higher than the actual mean correlation value of the network of interest. We ran this test using Turkey's biweight midcorrelation and 100,000 permutations, which yields a minimum nominal significance level of 1×10−5. The pvalue obtained for the test of the hECM-network in rTOF RV patients was 1×10−5.
[0393] Test for Overrepresentation of Genes Differentially Expressed by TGFβ1 and TGFβ2 (24 hrs) in the Human Co-Expression Networks
[0394] Description of the Cells
[0395] Human ventricular cardiac fibroblasts (passage 5) were grown to ˜90% confluence in 1% fetal bovine serum medium for 24 h. Then, they were incubated in triplicates (9 samples in total) in 1% fetal bovine serum medium for 24 h with +/−: human TGFβ1 (R&D cat100B, 5 ng/ml), or human recombinant TGFβ2 (Millipore cat GF113, 5 ng/ml). Supplier of ventricular human cardiac fibroblasts: Lonza (Catalogue number: CC-2904), product name: NHCF-V human cardiac fibroblasts-ventricular. Media kit: Fibroblast Growth Medium-3 BulletKit™ Kit. Catalogue number: CC-4526.
[0396] RNA Extraction, Sequencing and RNA-seq Data Processing
[0397] Total RNA was extracted and then sequenced with the Solexa platform (2×100 bp pairedend). Reads were mapped to the human genome (homo sapiens GRCh37.73, human genome 19 version) using TopHat 2.0.6 [4] and gene counts were computed with HTSeq 0.5.4p3 [13].
[0398] Differential Expression Analysis
[0399] Raw human counts were used as input to run the R package DESeq2 1.6.3. [14]. DESeq R function was run, considering the following two treatments in the primary human cardiac fibroblasts data: 24 h induction with TGFβ1 (n=3) and 24 h induction with TGFβ2 (n=3). In both cases, as baseline, 24 h control human cardiac fibroblasts (n=3) were used. Due to the low number of samples, to obtain the adjusted DESeq2 DE p-values, the outlier correction parameter cooksCutoff from the DESeq2 package function results was set to false.
[0400] Enrichment Test for 24 hrs TGFβ DE Genes
[0401] In this analysis only genes expressed in human cardiac fibroblasts that were also robustly expressed in the DCM LV (i.e. WGCNA input set of genes) were considered for both the ranked list and human networks. Genes were ranked by the output statistic of the differential expression test (Wald statistic, as we were using default parameterizations). After this, GSEA was run twice using as gene set the co-expression networks built in human DCM LV, and as ranked list all the genes ranked by the statistic of the differential expression test after 24 h induction with either TGFβ1 or with TGFβ2. We run GSEA 2.1.0 in classic, pre-ranked mode with 10,000 iterations by setting the maximum gene set size to 5,000 and the minimum gene set size to 10.
[0402] hECM-Network Enrichment for Differentially Expressed Genes in DCM with Progression to Heart Failure
[0403] We carried out an enrichment analysis of the hECM-network by using the results of the study published by Burke et al [32]. In this study they use a mouse genetic model of DCM and perform RNA-seq and differentially expression analysis in both cardiomyocytes and non-myocytes from hearts of control mice, mice with DCM and heart failure. Two lists of nonmyocyte differentially expressed genes were downloaded from the Supplementary Table 3 from the study published by Burke et al (comparisons “non-myocyte DCM vs WT” and “nonmyocyte HF vs WT”) [32]. As this study was carried out in mouse, to test for overrepresentation of the hECM-network genes in these lists of differentially expressed genes, we first mapped the genes included in the hECM-network (683 human genes) to their one-to-one mouse orthologs that were also expressed in the mouse heart. For this we downloaded all the human-to-mouse one-to-one orthologs from Biomart and also only considered genes expressed in the mouse left ventricle with FPKM>1 in at least two mice (as described in section “Analysis of Wwp2.sup.Mut/Mut mouse RNA-seq data”). This resulted in a reduction of the 683 hECM-network genes to 415 genes as present in the network, expressed in the mouse heart left ventricle and with one-to-one human-mouse ortholog relationship. We then generated a background for this test composed by the set of mouse genes expressed in the mouse heart left ventricle also with one-to-one human-mouse ortholog relationship (10,271 genes). In the next step we selected only those genes in the differentially expressed gene lists provided by Burke et al that were present in these 10,271 mouse genes background. Finally, we carried out a Fisher's exact test testing for significant overlap between the hECM-network (now mapped to mouse, 415 genes) and the set of genes present in our 10,271 mouse gene background that were upregulated in noncardiomyocytes in DCM and HF differentially expressed gene lists separately (2,005 and 2,162 genes respectively). We used R function used fisher test and we set the alternative parameter to “greater” (as we tested for overrepresentation). This analysis resulted in 226 genes present in the hECM-network (now mapped to mouse) that were upregulated in DCM in non-myocytes and 221 genes upregulated in HF in non-myocytes (these represent 54% and 53% of the 415 network genes).
[0404] Enrichments of Experimentally Validated SMAD Targets in the hECM-Network
[0405] The ChEA TF curated database of experimental TF DNA binding mammalian data [33] was queried for experimentally validated SMAD1-8 targets. SMAD targets available for SMAD1-4 were also coming from the following references [34-37]. Overrepresentation of SMAD targets within the hECM-network was assessed by Fisher's exact test. Steps followed: 1) Remove SMAD targets that were not robustly expressed in DCM heart (here we considered the background from where the DCM network were inferred, n=14,281 genes). By applying this filtering, the initial number of SMAD targets (SMAD1=610, SMAD2=1936, SMAD3=3233 and SMAD4=5012) got reduced to SMAD1=386, SMAD2=1396, SMAD3=2076 and SMAD4=3164. 2) For each SMAD1-4 a Fisher's exact test was computed with the R function fisher.test and setting the alternative parameter to “greater” (as we are testing for overrepresentation). The gene background used for computing the contingency table was the set of genes from which the human DCM clusters were inferred (robustly expressed in the LV human cohort n=14,281).
[0406] 1.4 Network Quantitative Trait Locus (Network-QTL) Analysis
[0407] Co-expression networks suggest coordinated genetic regulation, which can be exploited to uncover genetic regulators of these transcriptional programs. Moreover, conserved genetic regulation can be driving fundamental biological mechanisms. In keeping with previous studies in the rat, where the BXH/HXB rat panel yielded increased power to carry out genetic mapping of gene networks [38]. We used multivariate Bayesian genetic mapping approaches to map the rat and human ECM-networks to the rat and human genomes. We first considered the expression of the rat ECM-network genes as a multivariate quantitative trait and jointly mapped this to the rat genome. Then, we inspected whether the regulatory locus identified in the rat was independently replicated in human DCM heart by joint mapping of the genes in the ortholog human network (e.g. hECM-network), to the human locus that is syntenic to the rat regulatory loci. This two-step strategy (mapping in rats first followed by mapping in humans) has been previously used to identify trans-acting genetic regulators of transcriptional networks underlying complex disease [1].
[0408] The mapping of the rat and human networks was carried out by using HESS [39]. HESS is a sparse Bayesian multiple linear regression method in which mRNA expression levels for multiple genes are regressed against all SNPs to identify the minimum (non-redundant) set of SNPs that predicts the mRNA expression variability. This method has the following features: 1) It takes into account the LD structure of the genotype data (the dependence of the genetic determinants or predictors). This allows to reduce the number of tests to be carried out and pinpoint the putative causal genetic variant. 2) It makes possible to map several responses in one single test. Therefore, with HESS is possible to map to the genome (i.e. all genome-wide genetic markers) expression levels of several genes jointly (for instance, map the expression levels of genes included in a co-expression network, without having to summarise their variability by PC analysis). 3) It exploits multidimensional dependencies within the responses (i.e. correlation of the gene expression levels). This can be used to boost moderate associations. The output of this method is a marginal posterior probability of inclusion (MPPI) for each gene-SNP pair tested, which represents the posterior probability of association of each SNP given the data. From this MPPI, the Bayes Factor (BF) can be computed. BF represents the evidence of genetic regulation versus no genetic control and it is defined as the ratio between the posterior odds and the prior odds or ratio between the strengths of these models [38]. In our case, the prior probability (π) for the jth SNP associated with the gth gene is defined as: π=E(p.sub.g)/p, where p is the number of SNPs we are testing and E(p.sub.g) is the a priori expected number of control points for the gth gene, in our case we fix E(p.sub.g)=2. For instance, in the case of the rat RI strains, as the number of genome wide SNPs we are testing is 1,394, p=1,384 and the prior probability becomes π=1.4×10.sup.−3. The BF is defined as the ratio between the posterior odds and the prior odds: BF=(MPPI.sub.gj/(1−MPPI.sub.gj))/(π/(1−π), where MPPI.sub.gj represents the marginal posterior probability of inclusion for the g th gene and the jth SNP. By using this BF formula, we can compute the BF for each response-predictor pair (i.e. gene and SNP under testing) from the output MPPIs of HESS.
[0409] Input Data for Genome-Wide Genetic Mapping of the Rat Networks in the RI Strains
[0410] The gene expression levels of the genes included in each of the rat co-expression networks built in the LV of interest were jointly mapped to the rat genome with HESS, i.e. rat networks conserved in human, overrepresented for genes correlating with fibrosis and with a pattern of co-expression not present in human control LV tissue: M1, M2 and M12 rat networks. The expression data used was the RNA-seq data in the RI strains (log2 transformed FPKM adjusted for the 1.sup.st PC). These runs were carried out with 1,384 genome-wide SNPs markers in 29 RI strains (instead of 30, as there is one RI strain with RNA-seq expression but no available genotype information).
[0411] Input Data to Carry Out Fine Mapping at the Syntenic Human Regulatory Loci
[0412] For each of the rat networks that had regulatory loci with median BF of the genes in the rat networks >100 (i.e. rat networks M1 and M2), the rat regulatory SNPs were selected. Two independent runs were carried out, one for the DCM patients and one for controls. The human expression data input to HESS was the expression level of the genes in the human networks that were significantly intersecting the rat modules of interest (RNA-seq data in the DCM/controls after the processing described above). This was done in the n=96 DCM and n=91 control samples for which we had both gene expression and genotype information. The genetic data input to HESS was the set of SNPs tagging the human locus syntenic to the identified rat locus in each case. To identify each the human syntenic locus, we follow these steps: 1) Obtain the start and end positions of the rat haplotype that contained the regulatory SNP (rat Ensembl version 69). 2) Compute the central genetic coordinate of the rat haplotype as: centre haplotype=start haplotype+(end haplotype-start haplotype)/2. 3) Get the closest rat gene to that coordinate, then get the human start/end coordinates of the human ortholog gene. 4) In human (human Ensembl version GRCh37), compute the centre of the selected gene: centre gene=start gene+(end gene-start gene)/2. Take a window of 10 Mb (±5 Mb) around the centre gene, which in the case of the hECM-network yielded the region Hs-chr16: 64415969 . . . 74415969 (human Ensembl version GRCh37). This was the region that we mapped in the human DCM data (region comprising 475 SNPs from our LDpruned genetic map).
[0413] Description of HESS Runs
[0414] HESS runs were performed with 20,000 sweeps, 5,000 burn ins and default parameterisations. From the output MPPIs, BFs were computed by using the formula described above. For each SNP, the interquantile range of the network genes BFs distribution was plotted by using the function errbar from the R package Hmisc 4.1-1.
[0415] 1.5 Correlation Analyses between WWP2 and hECM-Network Genes
[0416] WWP2 transcript levels in the LV cohorts (DCM/control) and RV cohorts (rTOF/control) were correlated separately in patients and controls to the rest of the genes included in the hECM network. We computed the Spearman's ranked correlation and p-value using the WGCNA R package function corAndPvalue WGCNA 1.42 [19]. The resulting correlation distributions were plotted as a density plot and tested for differences by using a two-sample nonparametric Mann-Whitney U test by using the R function wilcox.test. Nominal WWP2-gene correlation p-values were corrected for multiple testing by using the R function p.adjust (method “fdr”). In this multiple testing correction, we corrected for the number of genes in the hECM-network, 683. A core set of genes correlated to WWP2 was extracted by taking all the genes with correlation FDR<0.01 in both DCM and rTOF patients (326 genes). STRING protein-protein interaction database 10.0 [40] was queried with this set of genes on the May 9, 2018, here only experimental and co-expression connections were retrieved, confidence level 0.4. To obtain a network visualisation graph from STRING database, we only considered experimental connections and co-expression interactions with a minimum interaction score of 0.4. In the network graph, each node corresponds to a gene and colour was mapped to the correlation between WWP2 and each gene (the average correlation in DCM and rTOF patients). Genes annotated in String with the Gene Ontology CC term “extracellular matrix” were also retrieved and highlighted with thicker node border.
[0417] 1.6 eQTL Mapping of WWP2 Isoforms in the DCM Cohort
[0418] WWP2 isoform expression levels were quantified in the DCM patients from the TopHat output with Sailfish 0.6.3 [42]. Sailfish was run with default parameterisations. The data was adjusted for relevant technical (“RIN score” and “library preparation day”) and clinical (“sex” and “age at tissue collection”) covariates by using a multivariate linear model and then mapped to the regulatory locus identified in human chromosome 16 (1 Mb centred around WWP2). In this case we mapped the 27 SNPs tagging this region in our genetic map. From the output MPPIs, BFs were computed as described above.
[0419] Non-parametric Kruskal-Wallis Tests were computed by using the R function kruskal.test. The obtained p-values were corrected for the number tests carried out (i.e. the number of WWP2 isoforms inferred in the DCM heart with an average RPKM level higher than 0, 18 isoforms). P-values were corrected by using the R function p.adjust and the correction method “fdr”.
[0420] 1.7 Differential Expression Analysis between rTOF RV and Control RV Tissue
[0421] Differentially expressed genes between rTOF patients (n=27) and controls samples (n=11), were computed with the R package DESeq2 1.6.3 [14] adding gender as covariate in the model (age was not added as there were several RV control samples with missing age values). CooksCutoff DeSeq2 parameter was set to False. The rest of parameters were left to default.
[0422] 1.8 Animal Studies
[0423] Mice were bred and maintained in mice in a specific pathogen-free (SPF) environment and used according to guidelines issued by the National Advisory Committee on Laboratory Animal Research. Steps were taken to minimize animal suffering.
[0424] Wwp2.sup.Mut/Mut Mice
[0425] Wwp2.sup.Mut/wt mice were generated based on C57BL/6J background using CRISPR/Cas9 technology as previously reported [43, 44]. Briefly, mutant animals were generated by coinjection of Cas9 mRNA and individual gRNAs into one-cell mouse embryos. Founder animals carrying the indel mutations were identified first by PCR and T7 endonuclease I assay, and then by deep sequencing of the PCR products. Founders carrying the desired reading frame shift mutations were used to generate mutation-segregated heterozygous F1 animals by crossing with the wild type animals. Homozygous mutant animals were generated by heterozygote crossing and used for experiments in comparison with the wild type littermates. To look into specific function of individual Wwp2 isoforms, three gRNAs were designed to targeting coding Exon 2, aiming to introduce mutations in individual domains of the protein, which in humans have different functions and are encoded by the three different gene isoforms. Here, a reading frame shift mutation in Exon 2 would render Wwp2-FL and Wwp2-N to functionally null, but is unlikely to affect Wwp2-C function. The Wwp2.sup.Wt/Wt mice were crossbred to generate Wwp2.sup.Mut/Mut and Wwp2.sup.Wt/Wt (WT) mice in vivarium. The primers used for genotyping mice for Wwp2 are shown in SEQ ID NOs:63 and 64.
[0426] AngII (Angiotensin II) Infusion Model
[0427] Alzet mini osmotic pump (Model No. 1004, Durect Corporation) was subcutaneously implanted in eight-weeks old mice anaesthetized with 2% isoflurane. Miniosmotic pumps loaded with saline or Angiotensin II (Sigma Aldrich, #A9525) were implanted to deliver Ang II at 500 ng/kg/min for a period of 4 weeks [45].
[0428] Myocardial Infarction (MI) Model
[0429] MI was induced in 8-10 week-old mice after anaesthetizing with Ketamine and Xylazine and intubated with 22 G×1″ SURFLO Flash I.V. catheter (TERUMO) which was connected to an artificial rodent ventilator MINI VENT type 845 (Harvard Apparatus, USA). After exposing the heart via thoracotomy at the fourth left intercostal space, the left coronary artery (LAD) was permanently ligated with a 8-0 nylon monofilament suture [46]. The thorax was closed with 6-0 coated vicryl suture and mice were followed for 4 weeks after surgery.
[0430] For both models, mice were sacrificed after weighing them at indicated time points. Hearts were harvested for weight measurement, histological studies, collagen determination and molecular biology analyses. Samples sizes were determined by power analysis and n≥8 mice per group were used to account for the inherent variability in the fibrotic response of mice. Mice died undergoing surgery before the sample collection were excluded from statistical analysis. Data from the animal studies were collected in a blinded manner.
[0431] 1.9 Cell Culture and Treatment
[0432] Murine cardiac fibroblasts were taken from the left ventricle (LV) of mice. Minced LV pieces (1-3 mm.sup.3) were placed in 6 cm dishes with DMEM supplemented with 20% fetal bovine serum for less than 10 days to generate mice cardiac fibroblasts (P0) and passaged to P1 and P2 DMEM supplemented with 10% fetal bovine serum for experiments. Each experiment, all the cells from Wwp2.sup.Mut/Mut and Wwp2.sup.Wt/Wt (WT) hear were cultured at the same time with same generation. Human cardiac fibroblasts were isolated from right atrium (RA) appendage obtained from patients on cardiopulmonary-by-pass during cardiac surgery operations by digesting the tissue with Collagenase II. Cardiac fibroblasts are obtained by growing the homogenized tissue suspended in DMEM supplemented with 20% fetal bovine serum in a humidified atmosphere. C2C12 mouse myoblast cell line was grown in DMEM medium supplemented with 10% fetal bovine serum. The cells were passaged twice before being used for experiments.
[0433] In Vitro studies
[0434] To mimic the in vivo cardiac fibroblasts activity, cells were treated with transforming growth factor-b human (Sigma Aldrich, #T7039) at a concentration of 5 ng/μl for 16-72 hours. For siRNA and plasmid transfection, primary cardiac fibroblasts were seeded on a 6-well plate (˜70%) and were transiently transfected with siRNA duplexes (20 nM) designed for targeting 5′ or 3′ in WWP2 mRNA (Qiagen) using Lipofectamine RNAiMAX (Life technologies holdings, #13778075) in a serum free medium for 48-72 hours according to the manufacturer's instructions. In parallel, WWP2-FL and N plasmids were transiently transfected using Lipofectamine 2000 (Life Technologies Holdings, #11668019) for 48-72 hours according to the manufacturer's instructions. For siRNA transfection in human cardiac fibroblasts, primary human cardiac fibroblasts were seeded on a 12 well plate (70000 cells/well) and were transiently transfected with siRNA duplexes (20 nM) designed for targeting 5′ or 3′ in WWP2 mRNA (Qiagen) using Lipofectamine RNAiMAX (Life technologies holdings, #13778075) in a serum free medium for 24 hours according to the manufacturer's instructions. For ubiquitination analysis, cells were treated with proteosomal inhibitor MG132 (Sigma Aldrich, #M7449) following stimulation with TGFβ1 for 4 h before harvesting. SB430542 (Stem Cell Technologies, #72234) was used to inhibit TGFβ1 effect for respective time points. Clomipramine hydrochloride (Sigma Aldrich, #C7291) was used to block the activity of HECT domain 1 hour before TGFβ1 treatment.
[0435] 1.10 Echocardiography
[0436] Transthoracic echocardiography was performed on day 28 after Ang II infusion and MI model using Vevo 2100 (VisualSonics, VSI, Toronto, Canada) and a MS400 linear array transducer, 18- to 38-MHz under anesthetized condition. An average of 10 cardiac cycles of standard 2 dimensional (2D) were acquired and stored for subsequent analysis using Vevo Imaging Workstation version 1.7.2 (VisualSonics, VSI, Toronto, Canada). All images acquisition and analysis were performed by a blinded operator according to a previously described method [47].
[0437] For Ang II-infusion model, 2D guided M-mode of parasternal short-axis short (middle) were selected for visualization of the papillarymuscle during end systole and end diastole For MI model, the parasternal long-axis were analyzed at 3 levels (basal, mid and apical) and all measurements were averaged over three consecutive cardiac cycles. Left ventricular ejection fraction (EF) and fractional shortening (FS) were calculated using modified Quinone method, using the following formulas: LVEF=(LVIDed.sup.2−LVIDes.sup.2)/LVIDed2; FS=(LVIDed−LVIDes)/LVIDes; where LVIDed is Left ventricular internal diameter at end diastole and LVIDes is Left ventricular internal diameter at end systole.
[0438] 1.11 Analysis of Single-Cell RNA-Sequencing (scRNA-seq) Data from WT Mouse Heart with Angiotensin II Infusion
[0439] Mice Description and Experimental Protocol
[0440] Single cell suspension was prepared from the adult left ventricle of one mice with Angiotensin II infusion for 28 days, as previously described [48]. After removal of dead cells with MACS dead cell removal kit (Miltenyi Biotec, #130-090-101), cells were lysed and subsequently RNA was reverse-transcribed and converted into cDNA libraries for RNA-seq analysis using a Chromium Controller and a Chromium Single Cell 3′ v2 Reagent kit (Genomics 10×) following the manufacturer's protocol. The library was sequenced using the Illumina Hi-Seq3000 sequencing platform.
[0441] Single-Cell Data Analysis
[0442] The reads were mapped to the mouse genome (m38, Ensembl version 89) and quantified using Cell Ranger 2.1.1 (10× Genomics). We provided to Cell Ranger a custom built reference transcriptome generated by filtering the Ensembl transcriptome (Ensembl file: Mus_musculus.GRCm38.89.gtf) for the gene biotypes: protein coding, lincRNA and antisense. Cell Ranger was run with the expect number of cells parameter (expect-cells) set to 3000. Cell Ranger out filtered matrices (i.e. genes.tsv and barcodes.tsv) were then input into R and genes with zero counts in all cells were discarded.
[0443] Three cell quality control filtering steps were implemented. We removed: (a) cells with less than the 50th percentile of the distribution of the total cells library size, (b) cells with less than the 50th percentile of the distribution of total of number of detected genes, (c) cells with more than 50% of their total gene count coming from mitochondrial genes. This resulted in a final number of 508 cells.
[0444] After applying these cell filtering steps, we carried out five gene quality control steps on the remaining cells: (a) we only kept “detectable” genes, defined as genes detected with more than one transcript in at least two cells, (b) we removed genes with low average expression in the data (i.e. genes with an average expression below 0.01, this cutoff was set based on the total distribution of average gene expression across all cells and all genes), (c) We removed genes encoded on the mitochondrial genome and the gene “Malat1” as it was an outlier in the gene expression distribution. After all the gene quality control steps, the resulting number of genes was 6,728.
[0445] Gene counts were normalized with scran 1.8.4 R package [49]. Scran size factors were computed from cell pools by doing a pre-clustering of the data with the quickCluster function 45 (the output object of this function was provided to the computeSumFactors function and then run the normalize function was run, all of them with default parameterizations). t-SNE was computed by providing the Log2 scran-normalized data to the function plotTSNE from scater 1.8.4 R package [50]. Two tSNE components were computed setting a random seed of 123456 using automatic perplexity (after removing an imposed minimum of 50, i.e. floor(number cells/5)). In the t-SNE graph, each cell was coloured by Wwp2 expression level (Log2 scran-normalized gene counts). Heart cell subpopulations were identified by using known cell marker genes. Specifically, we used: Apinr and Pecam1 (endothelial cells); Lum (fibroblasts); Ttn (cardiomyocytes); Hbb-bs (eritrocytes); Ccr2, Cd163 and Ptprc (immune cells). 12.
[0446] 1.12 Analysis of Wwp2.sup.Mut/Mut Mouse RNA-seq Data
[0447] Generation of RNA-seq Data
[0448] Total RNA from tissue was isolated from left ventricles of 9 Wwp2.sup.Mut/Mut and 9 WT mice with 28 days of Angiotensin II infusion. mRNA libraries were constructed from poly(A)-selected RNA using the NEBNext Ultra Directional RNA library prep kit (Illumina, New England BioLabs) and sequenced on Illumina HiSeq 2500.
[0449] Data Processing, Differential Expression, Functional Enrichment Test and Test for Overrepresentation of Human Co-Expression Networks in the Differentially Expressed Genes
[0450] RNA-seq reads were assessed for quality, aligned to m38 (Ensembl Gene annotation build 89) using STAR 2.5.2b [51] and quantified with RSEM 1.2.31 [52]. The average mapping rate (unique and multimapping) was 94.5%. Gene annotation was retrieved from Ensembl version 89 (m38) using the R library biomaRt 2.30.0 [53]. Ribosomal genes (Ensembl gene biotype “rRNA”) and mitochondrial genes were removed (391 genes). Gene counts were rounded using the R function round and differential expression analysis was performed with DESeq2 1.14.1 [14] with a pre-filtering step in which we considered only genes with more than 1 count when summing up across all samples.
[0451] DESeq2 was run pairwise comparing Wwp2.sup.Mut/Mut with Angiotensin II against WT mice with Angiotensin II using Wald test, with the outlier correction parameter cooksCutoff set to false (default parameterizations for the rest of parameters). In the DESeq2 model, we added RNA concentration and sequencing lane as covariates. Functional enrichment analysis of the differential expression results was performed with Gene Set Enrichment Analysis (GSEA) software 2-2.2.2 [25]. From all genes included in DESeq2 output, we selected those with one-to-one mouse-human ortholog relationships (as downloaded from Biomart, 14820 genes) and then we mapped them to human gene symbols. Then ranked all the genes by the corresponding DESeq2 output Wald statistic (i.e. the estimate of the log2 fold change divided by its standard error). GSEA was run assessing overrepresentation of the following gene sets and pathways derived from the Molecular Signatures Database gene sets 5.1 [54], (gene sets were queried using gene symbols): Hallmark gene sets (i.e. coherently expressed gene signatures derived from the aggregation of many MSigDB gene sets to represent well-defined biological states or processes), Gene Ontology and Reactome databases. GSEA was run in classic pre-rank mode with 10,000 permutations to assess false discovery rate (FDR). In the GSEA runs, maximum gene set size was set to 5,000 and minimum gene set size was set to 10. In this test, upregulated processes and pathways in the Wwp2.sup.Mut/Mut will be positively enriched, whereas downregulated ones will be negatively enriched. Gene sets were deemed as enriched if FDR<0.05.
[0452] In addition, GSEA was run a second time using the same parameterization but this time testing for overrepresentation of all the human co-expression networks. In this last run, we further reduced the background to the common set of genes with one-one mouse-human ortholog relationship that were also present in both the DESeq2 mouse output and in the initial set of human genes considered for network inference in the human data. This background reduction was applied to both the ranked list and the human networks, resulting in a total number of 10,449 genes.
[0453] Differential Co-Expression Analysis of the hECM-Network in Wwp2.sup.Mut/Mut Mice
[0454] We tested whether the hECM-network genes displayed differential co-expression upon Angiotensin II infusion when comparing the Wwp2.sup.Mut/Mut mouse with control mice. To this aim, we followed the same procedure as we previously did to compute differential coexpression of the hECM-network genes in the two human cohorts (DCM/controls and rTOF/controls see previous sections). We applied a filtering and removed lowly expressed genes (i.e. we only took genes with an FPKM>1 in at least 2 out of the 18 samples), this led to 12,659 genes. Then, as the hECM-network was inferred in human, out of these 12,659 genes, we only considered the ones with one-to-one ortholog relationship to human (10,271 genes). From these genes we took the set of genes included in the hECM-network (415 genes). We added an offset of 1 and computed the Log.sub.2 FPKM. Then we computed genegene pair Turkey's biweight midcorrelation separately in the WT-Angiotensin II and Wwp2.sup.Mut/Mut-Angiotensin II mice and carried out the same test for differential co-expression as explain in the section “Test for differential co-expression of human DCM LV networks 47 between DCM patients and controls”.
[0455] 1.13 Histology and Immunofluorescence
[0456] Left ventricles harvested from the mice were fixed in 10% Neutral buffered formalin (NBF) for 24 hours at RT, processed with Leica automatic tissue processor, paraffin-embedded and sectioned with thickness of 5 μm. After dewaxing and rehydration, slides were stained with Sirius Red Collagen kit (Chondrex, Inc, #9046) and Masson's Trichrome staining kit (Sigma-Aldrich, #HT15) as per manufacturer's instructions. Sections were stained using anti-ACTA2 (1:100), anti-S100A4 (1:100) to identify cell and biochemical features. Bovine Anti Rabbit IgG-CFL 488 (Santacruz Biotechnology, #sc-362260) and Bovine Anti Mouse IgG-CFL 488 (Santacruz Biotechnology, #sc-362256) were used as secondary antibodies for immunofluorescence. Rhodamine Wheat Germ Agglutinin (WGA, Vector laboratories, #RL-1022) was used to stain the myocytes.
[0457] Cells were grown in 8-well chamber slide with a removable silicone chamber (ibidi) up to 70% confluence. After fixation with ice cold acetone and blocking with 1% BSA for 30 min at RT, the slides were incubated with primary antibodies anti-WWP2-FL/N(1:100), anti-ACTA2 (1:100), anti-S100A4(1:100), anti-Vimentin(1:100) and anti-FLAG(1:100) overnight at 4° C. Following washing steps, the slides were incubated with Bovine Anti Rabbit IgG-CFL 488 (Santacruz Biotechnology, #sc-362260) and Bovine Anti Mouse IgG-CFL 488 (Santacruz Biotechnology, #sc-362256) for 2 hours at RT. VectaShield Mounting Medium (Vector laboratories, #H-1200) with DAPI was used to stain the nuclei and the slides were covered by coverslip.
[0458] Slides were imaged on Leica fluorescence microscope and image was processed using ImageJ software with the Fiji package. With merge function, the positive Sirius red staining in the whole section of the left ventricle was quantified using custom semiautomated image analysis routine. To measure the fluorescence intensity of ACTA2 and COL1A1 in different cellular sections, images were taken using 20× Plan Fluor objective. Fluorescence intensity was measured by taking the integrated intensity of a region of interest and subtracting the background intensity, and normalized to cell number. For each group, at least six fields were analyzed per section. To measure fluorescence intensities of SMAD2, 1-μm Z-stacks through cells of fields interested were acquired [55]. A region was drawn around each cell and nucleus to be measured, and background without fluorescence was subtracted. The nuclear/cellular fluorescence intensity ratio was calculated. Each field represented around 8-10 cells and at least 4 fields were analyzed for each section.
[0459] 1.14 Hydroxyproline Assay
[0460] The amount of total collagen in the left ventricle was quantified using the Quickzyme Total Collagen assay kit (Quickzyme Biosciences). The assays were performed according to the manufacturer's protocol.
[0461] 1.15 Luciferase Assay
[0462] Cells were transfected with a luciferase reporter gene plasmid with SMAD binding sites (Yeasen, SMAD-Luc, #11543ES03), and co-transfected with pGMLR-TK (Yeasen, #11557ES03) as a normalization control. 30 hours after transfection, cells were treated with vehicle or TGFβ1 for 16 hours and harvested. Luciferase assays were performed using the Dual-Luciferase Reporter Assay System (Yeasen, #11402ES60).
[0463] 1.16 Cell Proliferation and Migration Assay
[0464] Cell proliferation was quantified by MTS assay (Promega) according to the manufacturers' protocol. For migration assay, cells were seeded at a density of 10,000 cells/well in a 96-well plate. A uniform, reproducible wound was created using Incucyte, Essen Bioscience (USA). The 96-well plate was placed in the Incucyte ZOOM apparatus and the images of cell migration was captured every 2 hours for up to a total of 48 hours [56].
[0465] 1.17 RT-gPCR
[0466] Total RNA was extracted from snap-frozen fibrotic cardiac tissue and primary cardiac fibroblasts using the RNeasy mini kit (Qiagen, #74106) and cDNA was prepared using iScript cDNA synthesis kit (primer specific, BIORAD, #170-8897) according to the manufacturer's instructions. Fast SYBR-Green master mix (BIORAD, #170-8880AP) was used for the analysis of gene expression using the BIORAD CFX RT-PCR system. 18S was used to normalize the relative gene expression and 2-.sup.ΔΔCt method was used to measure the fold change.
[0467] The primers used for RT-qPCR analysis are shown in SEQ ID NOs: 27 to 62.
[0468] 1.18 Western Blotting
[0469] Protein extracts were isolated from heart tissue and cells using RIPA buffer (Thermofischer, #89900) supplemented with protease (Sigma Aldrich, #11836170001) and phosphatase inhibitors cocktails (ROCHE, #PHOSS-RO). Nuclear and cytoplasmic extracts were obtained using NE-PER kit (Pierce, #78833) according to the manufacturer's instructions.
[0470] Co-Immunoprecipitation was performed with the cell lysates subjected to different treatment conditions with Pierce Direct Magnetic IP/CO-IP kit (Pierce, #88828) according to manufacturer's protocol. Immunoprecipitates were washed from conjugated beads and boiled in 4× SDS-PAGE buffer for further WB analysis.
[0471] After quantification with Bradford method, protein lysates were loaded onto a 4-12% acrylamide gel subjected to SDS-PAGE and then transferred onto a nitrocellulose membrane. After blocking in 5% nonfat dry milk, blotting was performed with anti-WWP2 targeting N-terminal region (Santa Cruz biotechnology, #sc30052, 1:500), anti-WWP2 targeting C-terminal region (Aviva Systems Biology, #ARP43089_P050, 1:500), anti-TGFb1 (Santa Cruz biotechnology, #sc52893, 1:500), anti-ACTA2 (Sigma-Aldrich, #A5228, 1:10,000), anti-S100A4 (Abcam, #ab41532, 1:500), anti-Vimentin (Abcam, #ab45939, 1:500), anti-Periostin (Novus bio, #NBP1-30042, 1:500), anti-Fibronectin (Sigma, #SAB4500974, 1:500), anti-pSMAD2 (CST, #18338, 1:500), anti-SMAD 2/3 (CST, #3102, 1:500), anti-SMAD-4 (Santa Cruz biotechnology, #sc-7966, 1:500), anti-Ubiquitin (CST, #3933, 1:500), anti-FLAG (Sigma-Aldrich, #F7425, 1:1000). Loading control was blotted with Anti-tubulin (Sigma-Aldrich, #T5168, 1:5000) and anti-GAPDH (Abcam, #ab8245, 1:5000) anti-Lamin (abcam, #ab8984, 1:5000) and anti-PARP (abcam, #ab6079, 1:5000) were used as nuclear controls. Blots were visualized by labeling with anti-Rabbit HRP (Bethyl laboratories, #A120-101P, 1:5000 or Thermo Fisher #101023, 1:1000) and anti-Mouse HRP (Bethyl laboratories, #A90-116P, 1:5000) and developed on a Kodak automated developer with the ECL and Femto Detection Systems (Pierce) and quantified using densitometry with Image J (version 2.0.0-rc-43).
[0472] 1.19 Statistical Analysis
[0473] All data were analyzed using the appropriate statistical analysis methods with SPSS software (version 21.0), and the data are expressed as the mean±s.d. mean (95%Cl or ±SD). The applied tests were dependent on the number of groups being compared and the study design. A two-tailed Mann-Whitney U test was used to compare two groups, with * denoting P<0.05, and ** denoting P<0.01. When comparing mice with different genotypes, male littermate mice were assigned to the WT and Mut/Mut groups according to the results of genotyping, and mice with the same genotype were randomly assigned to the control, AngII infusion or MI group using a simple random-sampling approach. All experiments requiring the use of animals, directly or as a source of cells, were subjected to randomization. The experimenters were blinded to the grouping information. All in vitro experiments were replicated at least three independent times.
1.20 References to Example 1
[0474] 1. Moreno-Moral, A. and E. Petretto, From integrative genomics to systems genetics in the rat to link genotypes to phenotypes. Dis Model Mech, 2016. 9(10): p. 1097-1110. [0475] 2. https://amstat.tandfonline.com/doi/abs/10.1080/01621459.1995.10476572-.XD7OSM8zb2K [0476] 3. Rintisch, C., et al., Natural variation of histone modification and its impact on gene expression in the rat genome. Genome Res, 2014. 24(6): p. 942-53. [0477] 4. Trapnell, C., L. Pachter, and S. L. Salzberg, TopHat: discovering splice junctions with RNA-Seq. Bioinformatics, 2009. 25(9): p. 1105-11. [0478] 5. Flicek, P., et al., Ensembl 2012. Nucleic Acids Res, 2012. 40(Database issue): p. D84-90. [0479] 6. Martin, J. A. and Z. Wang, Next-generation transcriptome assembly. Nat Rev Genet, 2011. 12(10): p. 671-82. [0480] 7. Stacklies, W., et al., pcaMethods—a bioconductor package providing PCA methods for incomplete data. Bioinformatics, 2007. 23(9): p. 1164-7. [0481] 8. Leek, J. T. and J. D. Storey, Capturing heterogeneity in gene expression studies by surrogate variable analysis. PLoS Genet, 2007. 3(9): p. 1724-35. [0482] 9. Mancini, M., et al., Mapping genetic determinants of coronary microvascular remodeling in the spontaneously hypertensive rat. Basic Res Cardiol, 2013. 108(1): p. 316. [0483] 10. Petretto, E., et al., Integrated genomic approaches implicate osteoglycin (Ogn) in the regulation of left ventricular mass. Nat Genet, 2008. 40(5): p. 546-52. [0484] 11. Heinig, M., et al., Natural genetic variation of the cardiac transcriptome in nondiseased donors and patients with dilated cardiomyopathy. Genome Biol, 2017. 18(1): p. 170. [0485] 12. Koopmann, T. T., et al., Genome-wide identification of expression quantitative trait loci (eQTLs) in human heart. PLoS One, 2014. 9(5): p. e97380. [0486] 13. Anders, S., P. T. Pyl, and W. Huber, HTSeq—a Python framework to work with highthroughput sequencing data. Bioinformatics, 2015. 31(2): p. 166-9. [0487] 14. Love, M. I., W. Huber, and S. Anders, Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol, 2014. 15(12): p. 550. [0488] 15. Langfelder, P. and S. Horvath, Fast R Functions for Robust Correlations and Hierarchical Clustering. J Stat Softw, 2012. 46(11). [0489] 16. Johnson, A. D., et al., SNAP: a web-based tool for identification and annotation of proxy SNPs using HapMap. Bioinformatics, 2008. 24(24): p. 2938-9. [0490] 17. Wald, R. M., A. M. Valente, and A. Marelli, Heart failure in adult congenital heart disease: Emerging concepts with a focus on tetralogy of Fallot. Trends Cardiovasc Med, 2015. 25(5): p. 422-32. [0491] 18. Heidecker, B., et al., The gene expression profile of patients with new-onset heart failure reveals important gender-specific differences. Eur Heart J, 2010. 31(10): p. 1188-96. [0492] 19. Zhang, B. and S. Horvath, A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol, 2005. 4: p. Article 17. [0493] 20. Hardin, J., et al., A robust measure of correlation between two genes on a microarray. BMC Bioinformatics, 2007. 8: p. 220. [0494] 21. Fresno, C. and E. A. Fernandez, RDAVIDWebService: a versatile R interface to DAVID. Bioinformatics, 2013. 29(21): p. 2810-1. [0495] 22. Huang da, W., B. T. Sherman, and R. A. Lempicki, Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc, 2009. 4(1): p. 44-57. [0496] 23. Ashburner, M., et al., Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet, 2000. 25(1): p. 25-9. [0497] 24. Kanehisa, M., et al., KEGG for integration and interpretation of large-scale molecular data sets. Nucleic Acids Res, 2012. 40(Database issue): p. D109-14. [0498] 25. Subramanian, A., et al., Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A, 2005. 102(43): p. 15545-50. [0499] 26. Choi, Y. and C. Kendziorski, Statistical methods for gene set co-expression analysis. Bioinformatics, 2009. 25(21): p. 2780-6. [0500] 27. North, B. V., D. Curtis, and P. C. Sham, A note on the calculation of empirical P values from Monte Carlo procedures. Am J Hum Genet, 2002. 71(2): p. 439-41. [0501] 28. Brand, N. J., et al., Analysis of cardiac myocyte biology in transgenic mice: a protocol for preparation of neonatal mouse cardiac myocyte cultures. Methods Mol Biol, 2010. 633: p. 113-24. [0502] 29. Kim, D., et al., TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome Biol, 2013. 14(4): p. R36. [0503] 30. Langmead, B. and S. L. Salzberg, Fast gapped-read alignment with Bowtie 2. Nat Methods, 2012. 9(4): p. 357-9. [0504] 31. Souders, C. A., S. L. Bowers, and T. A. Baudino, Cardiac fibroblast: the renaissance cell. Circ Res, 2009. 105(12): p. 1164-76. [0505] 32. Burke, M. A., et al., Molecular profiling of dilated cardiomyopathy that progresses to heart failure. JCI Insight, 2016. 1(6). [0506] 33. Lachmann, A., et al., ChEA: transcription factor regulation inferred from integrating genome-wide ChIP-X experiments. Bioinformatics, 2010. 26(19): p. 2438-44. [0507] 34. Chen, X., et al., Integration of external signaling pathways with the core transcriptional network in embryonic stem cells. Cell, 2008. 133(6): p. 1106-17. [0508] 35. Koinuma, D., et al., Chromatin immunoprecipitation on microarray analysis of Smad2/3 binding sites reveals roles of ETS1 and TFAP2A in transforming growth factor beta signaling. Mol Cell Biol, 2009. 29(1): p. 172-86. [0509] 36. Kim, S. W., et al., Chromatin and transcriptional signatures for Nodal signaling during endoderm formation in hESCs. Dev Biol, 2011. 357(2): p. 492-504. [0510] 37. Kennedy, B. A., et al., ChIP-seq defined genome-wide map of TGFbeta/SMAD4 targets: implications with clinical outcome of ovarian cancer. PLoS One, 2011. 6(7): p. e22606. [0511] 38. Heinig, M., et al., A trans-acting locus regulates an anti-viral expression network and type 1 diabetes risk. Nature, 2010. 467(7314): p. 460-4. [0512] 39. Bottolo, L., et al., Bayesian detection of expression quantitative trait loci hot spots. Genetics, 2011. 189(4): p. 1449-59. [0513] 40. Szklarczyk, D., et al., STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res, 2015. 43(Database issue): p. D447-52. [0514] 41. Shannon, P., et al., Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res, 2003. 13(11): p. 2498-504. [0515] 42. Patro, R., S. M. Mount, and C. Kingsford, Sailfish enables alignment-free isoform quantification from RNA-seq reads using lightweight algorithms. Nat Biotechnol, 2014. 32(5): p. 462-4. [0516] 43. Zine-EddineKherraf, et al., Creation of knock out and knock in mice by CRISPR/Cas9 to validate candidate genes for human male infertility, interest, difficulties and feasibility. Molecular and Cell Endocrinology, 2018. 468: p. 70-80. [0517] 44. Bao, Q., et al., Publisher Correction: Utf1 contributes to intergenerational epigenetic inheritance of pluripotency. Sci Rep, 2018. 8(1): p. 1664. [0518] 45. Lu, H., et al., Subcutaneous Angiotensin II Infusion using Osmotic Pumps Induces Aortic Aneurysms in Mice. J Vis Exp, 2015(103). [0519] 46. Ye, L., et al., Thymosin beta4 increases the potency of transplanted mesenchymal stem cells for myocardial repair. Circulation, 2013. 128(11 Suppl 1): p. S32-41. [0520] 47. Fard, A., et al., Noninvasive assessment and necropsy validation of changes in left ventricular mass in ascending aortic banded mice. J Am Soc Echocardiogr, 2000. 13(6): p. 582-7. [0521] 48. Yen-Rei A. Yu , E. G. O. K., Danielle F. Hotten, Matthew J. Kan, David Kopin, Erik R. Nelson, Loretta Que, Michael D. Gunn, A Protocol for the Comprehensive Flow Cytometric Analysis of Immune Cells in Normal and Inflamed Murine Non-Lymphoid Tissues. PLoS One, 2016. [0522] 49. Lun, A. T., D. J. McCarthy, and J. C. Marioni, A step-by-step workflow for low-level analysis of single-cell RNA-seq data with Bioconductor. F1000Res, 2016. 5: p. 2122. [0523] 50. McCarthy, D. J., et al., Scater: pre-processing, quality control, normalization and visualization of single-cell RNA-seq data in R. Bioinformatics, 2017. 33(8): p. 1179-1186. [0524] 51. Dobin, A., et al., STAR: ultrafast universal RNA-seq aligner. Bioinformatics, 2013. 29(1): p. 15-21. [0525] 52. Li, B. and C. N. Dewey, RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinformatics, 2011. 12: p. 323. [0526] 53. Durinck, S., et al., Mapping identifiers for the integration of genomic datasets with the R/Bioconductor package biomaRt. Nat Protoc, 2009. 4(8): p. 1184-91. [0527] 54. Liberzon, A., et al., The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst, 2015. 1(6): p. 417-425. [0528] 55. Potapova, T. A., et al., Mitotic progression becomes irreversible in prometaphase and collapses when Wee1 and Cdc25 are inhibited. Mol Biol Cell, 2011. 22(8): p. 1191-206. [0529] 56. Stuart T. Johnston, E. T. S., Lisa K. Chopin, D. L. Sean McElwaina nd Matthew J. Simpson, Estimating cell diffusivity and cell proliferation rate by interpreting IncuCyteZOOM assay data using the Fisher-Kolmogorov model. BMC Systems Biology, 2015.
Example 2: Identification of WWP2 as a Positive Regulator of a Network of ECM Genes Constituting a Pathological Fibrosis Program Conserved across Fibrotic Diseases Affecting Different Organs
[0530] 2.1 Identification of the hECM-Network
[0531] The inventors carried out a system genetics study that enabled the identification of pathological fibrosis programs (i.e. gene co-expression networks) conserved across species and common to several diseases. The analysis identified WWP2 gene as a regulator of the identified pathological fibrosis program.
[0532] The inventors set out to identify transcriptional programs conserved across species and associated to cardiac fibrosis, using a panel of 30 rat recombinant inbred (RI) strains (Hubner et al., 2005), allowing integrative analyses of cardiac gene expression with quantitative pathophysiological traits (e.g. cardiac fibrosis) and genome-wide genetic data (Petretto et al., 2008; Mancini et al., 2013; Langley et al., 2013). In addition, this panel of RI strains is an established model for cardiovascular traits and disease, including cardiac hypertrophy (Petretto et al., 2008), blood pressure (Pravenec et al., 2008) and heart remodeling (Mancini et al., 2013).
[0533] The inventors performed gene co-expression network inference in the rat RI strains left ventricle (LV) transcriptome using RNA-sequencing (RNA-seq) data. This identified 41 distinct gene co-expression networks, each with a different number of genes. The inventors then tested the association of these gene co-expression networks with quantitative histopathologic measurements of interstitial and perivascular fibrosis in the rat heart. Using Gene Set Enrichment Analysis (GSEA) (Subramanian et al., 2005), five gene co-expression networks were determined to be associated with both interstitial and perivascular cardiac fibrosis in rat LV tissue (adjusted P-value<0.05).
[0534] To identify transcriptional programs relevant to the fibrogenic processes in human heart disease, separate gene network analysis was also performed using LV RNA-seq data generated from a cohort of patients with DCM (n=126) and a cohort of control heart samples (n=92 organ donors whose hearts were explanted for transplantation, described in Heinig et al., 2017). Applying the same network methodology that was used for the rat analysis, the inventors inferred 48 genome-wide gene co-expression networks in the human DCM LV transcriptome.
[0535] The inventors next assessed which of the human DCM gene networks were conserved in the rat LV using a Fisher test. 14 human DCM networks were determined to have some degree of conservation with the rat networks (adjusted P-value<0.05).
[0536] In order to obtain additional evidence in support of a role for these networks in human heart disease, the inventors tested which gene co-expression patterns were present only in the LV from human DCM patients and not in LV from controls (i.e. they tested for differential co-expression between DCM and control hearts). The differential co-expression paradigm assumes that the disease state is linked to perturbations of the structure of the regulatory network itself, and might reflect the dysregulation of the underlying transcription factors in disease (Rotival and Petretto, 2014). Here, the inventors found that 8 human gene networks were both conserved across species and differentially co-expressed between human DCM and control LV tissues (adjusted P-value<0.05). Despite the fact that several networks were significantly conserved between rat and human DCM heart, only one human network (referred to hereafter as the hECM-network), containing 683 genes, was (i) significantly conserved in the rat (sharing 72 common genes with the rat ECM network (adjusted P-value=8.4×10.sup.−45), (ii) correlated with both interstitial and perivascular fibrosis levels in the rat heart, and (iii) differentially co-expressed in human DCM heart. The hECM-network was also functionally relevant for ECM regulation, and was enriched (False Discovery Rate (FDR)<0.05) for genes belonging to the specific biological pathways and processes: “ECM-receptor interaction”, “TGFβ signaling pathway” and “focal adhesion”, as determined by KEGG annotation analysis.
[0537] The hECM-network contains 237 co-expressed genes annotated to encode extracellular ECM region proteins; 21 encode collagens, several encode focal adhesion molecules such as ITGB5, COMP, MAPK10 and THBS4; several encode extracellular genes involved in TGFβ-signalling (e.g., DCN, CHRD, TGFβ3); 3 encode members of the BMP family (BMP4, BMP6 and BMP8B), and genes encoding other important matricellular proteins are also represented, such as CTGF and PDGFD, which contribute to the fibrogenic response (Wang et al., 2011; Zhao et al., 2013).
[0538] The full list of hECM-network 237 genes is as follows: ABI3BP, COL12A1, EPHB3, LRRC4C, PRCD, ACHE, COL14A1, EXT2, LSP1, PRCP, AEBP1, COL15A1, F2R, LTBP2, PSMB2, AGRN, COL16A1, F2RL2, LTBP3, PSMD11, AKR1C1, COL1A1, FAM26E, LTBP4, PTGS1, AKR1C3, COL1A2, FAP, LUM, PTHLH, ALDH1A1, COL22A1, FAT4, LYPD1, PTN, ANGPT1, COL24A1, FBLN1, MATN2, PTPN13, ANO1, COL3A1, FBLN2, MDK, PTPRA, ANTXRI, COL4A1, FBLN7, MFAP2, PTPRR, ANXA1, COL4A2, FBN1, MFAP4, PXDN, ANXA4, COL4A3, FGL2, MFAP5, QPCT, ART4, COL4A4, FLRT2, MGP, RECK, ASPN, COL4A5, FMOD, MMP14, S100A4, BEND7, COL5A1, FN1, MMP2, SCG2, BGN, COL5A2, FRMD4B, MMP28, SCRG1, BMP4, COL6A1, FRZB, MOXD1, SCUBE2, BMP6, COL6A2, FSCN1, MTMR11, SEPT11, BMP8B, COL8A1, FUT8, MXRA5, SEPT2, BST1, COL8A2, GAL3ST4, MXRA8, SERPINE2, C9ORF72, COL9A2, GDF10, MYH10, SERPINF1, CAPN5, COLEC11, GDF6, MYO1D, SFRP4, CBR3, COMP, GDNF, MYO1E, SH3BGRL, CCL28, CPE, GLIPR2, NAGLU, SLIT2, CD34, CPXM2, GNB1, NBL1, SLIT3, CD44, CRABP2, GNG2, NCS1, SOD2, CD9, CRISPLD1, GOLM1, NOTCH2, SOD3, CDH11, CTGF, GPM6A, NOTCH3, SOSTDC1, CDH6, CTHRC1, GPX3, NT5E, SPARC, CDK5RAP2, CTSK, GSN, NTM, SPON2, CHRD, CTSO, HAAO, OAF, SPTBN2, CHRDL1, CUL4B, HAPLN1, OGN, SSC5D, CHST14, CYBRD1, HMCN1, OLFML3, SULF1, CILP, CYTL1, HNMT, OMD, SUSD2, CLDN11, DCN, IGFBP6, PCDHGC3, TARS, CLEC11A, DDR2, IGFBP7, PCDHGC5, TGFB3, CNTN4, DNM1, IGSF10, PCOLCE, TGFBI, DPT, IL16, PCOLCE2, THBS3, DPYSL3, IL33, PCSK5, THBS4, DUOX2, ISLR, PCYOX1L, THY1, ECM1, ITGB5, PDGFD, TIMP2, ECM2, KIRREL, PDXK, TNXB, EFEMP2, LASP1, PENK, TOM1, EHD2, LEFTY2, PI16, TPPP3, ELN, LINGO1, PLA2R1, TRABD2B, EMILIN3, LOX, PLAT, TSPAN8, ENPP2, LOXL1, PLSCR4, UCHL1, ENTPD1, LOXL2, PLXDC2, VCAN, EPB41L2, LOXL3, PODN, WNT10B, EPHB2, LRRC17, POSTN and ZDHHC1.
[0539] Because the hECM-network was identified in DCM LV, the inventors further investigated whether the network was specific to the ECM remodelling processes taking place in DCM and/or in LV tissue. They considered a separate heart condition, specifically repaired tetralogy of Fallot (rTOF) (Villafañe et al., 2013), which has a very different etiology from DCM but that is characterized by the presence of cardiac dysfunction and diffuse and pathologic myocardial fibrosis of both the right ventricle (RV) and LV (Pradegan et al., 2016).
[0540] RNA-seq data were analysed from RV in a cohort of adult patients with rTOF (n=27) who underwent surgery for pulmonary valve replacement and age-matched control donor samples (n=11). The hECM gene co-expression network identified in DCM LV was found to be significantly conserved in rTOF RV (empirical P-value<1×10.sup.−5) but not in control RV (empirical P-value=1). The inventors next determined that the pairwise correlation pattern of the hECM-network genes in rTOF RV mirrored the pattern observed in DCM LV (i.e., strongest gene-gene correlation in disease), and this was significantly different between rTOF patients and controls.
[0541] The consistent differential co-expression in dilated cardiomyopathy (DCM) and rTOF heart suggests that the hECM-network is capturing common ECM remodeling processes taking place across a range of human cardiac fibrotic diseases, irrespective of the specific disease etiology and the heart tissue (i.e. LV/RV).
[0542] In addition, using published longitudinal cardiac transcriptome data from a mouse genetic model of DCM that progresses to heart failure (HF; Burke et al., 2016), the inventors determined that this hECM-network is enriched for genes upregulated in DCM that progresses to HF. Specifically, 54% of the hECM-network genes are upregulated in DCM compared with control heart (enrichment P-value=5.4×10.sup.−59) and 53% are upregulated in HF compared with control heart (enrichment P-value=2.4×10.sup.−49). These results in human and mouse DCM/HF suggest that the identified hECM-network recapitulates the maladaptive fibrotic remodeling that promotes heart failure and adverse cardiovascular outcomes.
[0543] Since the hECM-network was enriched for genes involved in the TGFβ-signalling (as determined by KEGG annotation analysis), the inventors investigated whether the hECM-network was downstream of TGFβ-signalling activation. RNA-seq data were analysed from three sets of primary cultures of human atrial cardiac fibroblasts exposed to TGFβ1 for 24 h, TGFβ2 for 24 h, or control media.
[0544] A significantly large proportion of the genes belonging to the hECM-network were transcriptionally regulated after 24 h of stimulation with TGFβ1 or TGFβ2, which induced the differential expression of 46% and 49% of the genes in the hECM-network (315 and 335 genes), respectively.
[0545] Given the role of SMAD transcription factors downstream of TGFβ receptor activation (Hu et al., 2018; Massague and Wotton, 2000), the inventors also investigated whether the hECM-network was enriched for SMAD target genes. To this aim, published ChIP-seq data (Lachmann et al., 2010) were analysed, and the hECM-network was found to be significantly enriched for SMAD-regulated genes (291 genes, 43% of the network), especially for SMAD2, SMAD3 and SMAD4.
[0546] Together, these results revealed a cross-species conserved gene network, which might recapitulate the pathological ECM remodeling undergoing downstream of TGFβ/SMAD signalling activation in the diseased heart.
[0547] Injury in any organ triggers multicellular molecular responses that lead to tissue fibrosis. There is a wide range of complex diseases that affect different organs and are associated with fibrosis. This suggests that there may be shared pathways involved in this pathological process. The hECM-network identified in the heart could be capturing transcriptional responses that also act in other fibrotic diseases.
[0548] In order to investigate this, the inventors evaluated whether the hECM-network is conserved in other fibrotic tissues, by analysing additional transcriptomic data sets with/without fibrosis of right ventricle (rTOF), lung (pulmonary hypertension-associated lung fibrosis vs. controls, GEO database entry GDS4549) and liver (hepatitis C virus-infected subjects with liver fibrosis vs. controls, GEO database entries GSE33650 and GSE61260 respectively).
[0549] An empirical permutation test that assesses whether it is likely to find by chance a network of genes in an independent data set (how likely is it to find by chance a set of genes of the same size as the network with the same value of average absolute correlation in the data set under testing) was employed. In this test: i) expression of a random set of genes of the same size of each of the human LV DCM networks is sampled, ii) the mean correlation of both the network being tested and the randomly sampled network is computed, iii) after repeating steps 1 and 2 at least 10,000 times, an empirical p-value is computed for each network and adjusted for the number of networks present in LV (n=48). Significant results in this test imply that the hECM-network is specifically conserved in that tissue and condition.
[0550] Expression of the hECM-network was found to be conserved in heart right ventricle, lung and liver, only in fibrotic disease and not in control tissue (FDR<0.05 indicates that the hECM-network is conserved in that condition):
TABLE-US-00002 Heart (Right Lung Ventricle) PH Liver rTOF Controls fibrosis Controls Fibrosis Controls (n = 27) (n = 11) (n = 62) (n = 22) (n = 72) (n = 38) hECM- 0.0005 1.0000 0.0005 1.0000 0.0005 1.0000 network
[0551] The hECM-network was found to be conserved in fibrotic diseases such as repaired tetralogy of fallot (fibrosis of the right ventricle), liver and pulmonary hypertension-associated lung fibrosis.
[0552] Co-expression networks suggest coordinated genetic regulation, and this was exploited by the inventors to identify key genetic regulators of the hECM transcriptional program in the rat and DCM heart.
[0553] Multivariate Bayesian genetic mapping (Bottolo et al., 2011; Lewin et al., 2016) of the ECM-network was performed in the rat and then in humans. In this analysis, in order to identify loci regulating the whole ECM-network the expression of the rat (or human) genes in the ECM-networks were treated as multivariate quantitative traits, and it was investigated whether the joint expression levels of the network genes are associated with genome-wide genetic variants (i.e. single nucleotide polymorphisms, SNPs). This Bayesian network-expression QTL (network-eQTL) mapping approach has previously been used to identify trans-acting genetic regulators of networks in several diseases, including type 1 diabetes (Heinig et al., 2010), epilepsy (Johnson et al., 2015) and inflammatory disease (Kang et al., 2014).
[0554] The rat ECM-network was mapped to the rat genome, and identified a single locus on rat chromosome 19 regulating 219 genes in the rat ECM-network (median Bayes Factor (BF)=181.7, where the Bayes Factor represents the strength of genetic regulation versus no genetic control).
[0555] 2.2 Identification of WWP2 as a Regulator of the hECM-Network
[0556] It was investigated whether this regulatory locus for the rat ECM-network was replicated and conserved in human DCM heart. The cohort of DCM patients in which the hECM-network was strongly co-expressed, was analysed to understand whether the genes in the hECM-network were jointly mapping to the human locus syntenic to the regulatory locus found in the rat (the human syntenic locus is located on chromosome 16). Network-eQTL mapping in human DCM heart detected a single regulatory SNP (rs9936589) located within an intron of the WWP2 gene, an E3 ubiquitin ligase. This SNP was strongly associated with the expression of the hECM-network in the DCM heart (median BF=2004 for the 683 genes in the hECM-network). Moreover, the network-eQTL was not detectable in the heart from control organ donors, suggesting that the genetic regulation of the hECM-network is present only in diseased heart. This regulatory SNP (rs9936589) and the WWP2 gene have not previously been associated with fibrotic disease or heart disease by GWAS or other genetic and eQTL mapping approaches.
[0557] Systems genetics analyses revealed a coordinated pro-fibrotic and disease-associated ECM transcriptional program in the diseased heart, whose expression was associated with a genetic variant within the WWP2 locus. The inventors therefore, investigated whether WWP2 was a potential transcriptional regulator of the hECM-network in human fibrotic heart disease.
[0558] Expression levels of WWP2 were correlated to the expression levels of the 683 genes forming the hECM-network in LV and RV fibrotic disease hearts (i.e., in DCM and rTOF) and controls separately. In both LV (DCM) and RV (rTOF) fibrotic heart, a positive and significant shift in the distribution of the correlations between WWP2 expression and the expression of the hECM-network genes was observed (
[0559] A core set of genes in the hECM-network was also identified that were positively and consistently correlated with WWP2 expression in the heart (FDR<1%) irrespective of heart tissue of origin (i.e., LV or RV) or heart condition (i.e., rTOF or DCM),
[0560] These findings suggest that increased expression of WWP2 is associated with elevated expression of pro-fibrotic genes in the diseased heart. Analysis of the expression levels of hECM-network genes stratified by the genotypes of the regulatory SNP rs9936589, showed increased expression of the hECM-network associated with the TT genotype (
[0561] To further analyse whether the regulatory effect of rs9936589 on the hECM-network was mediated by WWP2, the inventors analysed whether cardiac WWP2 expression was similarly regulated by this SNP (i.e., the inventors investigated whether expression of WWP2 was regulated in cis).
[0562] Three main WWP2 gene isoforms have been characterised, containing different protein domains: a full-length isoform (WWP2-FL, covering the entire gene), an N-terminal isoform (WWP2-N, corresponding to the 3′ end of the protein) and a C-terminal isoform (WWP2-C, containing the 5′ end of the protein) (Soond and Chantry, 2011)—see
[0563] All these three isoforms showed robust expression in the hearts of DCM patients. WWP2 isoform-specific cis-QTL mapping was performed, and showed that only the WWP2-N was regulated by the SNP rs9936589 (
Example 3: WWP2 Regulates Cardiac Fibrosis In Vivo
[0564] 3.1 Generation and Characterisation of Mice Lacking Expression of WWP2-FL and WWP2-N
[0565] In order to identify pathophysiological processes regulated by the N-terminal region of WWP2, the inventors generated Wwp2 mutant mice (Wwp2.sup.Mut/Mut) using CRISPR/Cas9 technology to introduce a 4-bp deletion (CTAC) in exon 2 of Wwp2, leading to disruption of Wwp2-FL and Wwp2-N isoforms. Consistent to the phenotype that was previously reported in global Wwp2-null mice (Zou et al., 2011), the Wwp2.sup.Mut/Mut mice showed reduced body weight relative to wildtype (P<0.001 at 10 weeks of age, repeated-measures t-test n=8 mice per group), abnormal craniofacial development and elongated teeth.
[0566] The 4-bp deletion in Wwp2 resulted in ablation of WWP2 isoforms containing the Wwp2 N-terminal region (that is, Wwp2-FL and Wwp2-N isoforms), as detected at the mRNA level by isoform-specific primer pair 1 (P1;
[0567] Western blot analysis confirmed lack of Wwp2-FL expression at the protein level of protein extracts prepared from cardiac fibroblasts obtained from cardiac tissue of Wwp2.sup.Mut/Mut mice (
[0568] 3.2 Analysis of the Effects of Loss of Wwp2-N/FL Isoform Expression in Mouse Models of Fibrotic Disease
[0569] Ang II Infusion Model
[0570] Systems genetics analysis indicated that the WWP2-N isoform positively regulates a pro-fibrotic transcriptional network in diseased heart (Example 2.2), and so the inventors hypothesised that loss of function (LOF) of WWP2-N/FL protein isoforms might inhibit the in vivo fibrogenic response.
[0571] Sections from the hearts of WT mice infused with Ang II had increased collagen content, as determined by increased Sirius Red staining relative to controls infused with saline (
[0572] Ang II infusion was also associated with ventricular remodelling and impaired cardiac function. Increased left ventricular mass index (LVMI), and decreased ejection fraction, and fractional shortening was observed in WT mice subjected to Ang II infusion relative to saline-infused controls (
[0573] RT-qPCR analysis also detected increased Wwp2 transcript levels (
[0574] In contrast to WT mice, Ang II-infused Wwp2.sup.Mut/Mut mice displayed significantly less fibrosis. LV sections from Ang II-infused Wwp2.sup.Mut/Mut mice showed reduced Sirius red and Masson's Trichrome staining relative to Ang II-infused WT mice (
[0575] Bulk RNA-seq analysis in Ang II-treated WT (n=8) and Wwp2.sup.Mut/Mut mice (n=8) revealed that the mouse orthologs of the hECM-network genes detected in DCM heart, showed a different co-expression pattern with increased gene-gene correlation in the LV of WT compared with Wwp2.sup.Mut/Mut mice (P=0.003). In Wwp2.sup.Mut/Mut mice, this pattern of differential co-expression was remarkably similar to the pattern observed between human fibrotic disease (DCM or rTOF) and controls; the Wwp2.sup.Mut/Mut mice showed a hECM-network co-expression pattern similar to control heart. Consistent with human data, the hECM-network was also significantly enriched for differentially expressed genes between Ang II-treated WT and Wwp2.sup.Mut/Mut mice. In addition, “TGFβ signaling” and “extracellular matrix” were two of the major downregulated pathways in Wwp2.sup.Mut/Mut mice following Ang II-infusion, (Normalized Enrichment Score (NES) of −2.71 and −2.8; see
[0576] Protein-based assays showed that Wwp2.sup.Mut/Mut mice had reduced levels of fibroblast activation and ECM protein markers as marked by α-smooth muscle actin (ACTA2;
[0577] Myocardial Infarction Model
[0578] The inventors next investigated whether WWP2-N/FL LOF had a protective effect on cardiac fibrosis and function post myocardial infarction (MI).
[0579] Histological analysis revealed less post-MI fibrotic remodelling in the LV of Wwp2.sup.Mut/Mut mice as compared to WT mice (
[0580] Taken together, the results in the Ang II infusion and MI models suggest that LOF of the WWP2 isoforms containing N-terminal region (i.e. WWP2-FL and WWP2-N) reduces cardiac fibrosis and improves cardiac function following Ang II-treatment or MI. This is in keeping with a role for WWP2 as a positive regulator of fibrosis in the diseased heart.
Example 4: WWP2 Regulates the TGFβ1-Induced Fibrotic Response in Primary Cardiac Fibroblasts
[0581] 4.1 Analysis of WWP2 Expression in Cardiac Cells
[0582] To investigate the regulation of WWP2 in the cardiac cells, the inventors imaged WWP2-expressing cells in heart sections by immunofluorescence. WWP2-positive cells did not show the morphology typical of a sarcomere-containing cardiomyocyte, and some WWP2-positive cells expressed fibroblast-specific protein 1 (FSP1) (
[0583] 4.2 Analysis of the Effects of TGFβ1 Stimulation
[0584] TGFβ1 stimulation of primary murine LV fibroblasts induced robust Wwp2 transcription at 72 hrs of treatment (
[0585] TGFβ1-stimulated WT fibroblasts presented a clear organization of ACTA2 into stress fibers, while Wwp2.sup.Mut/Mut-derived cells displayed diffuse expression of ACTA2 with rare incorporation into stress fibers (
[0586] 4.3 Analysis of the Functional Cellular Consequences of WWP2 Deficiency
[0587] Interestingly, WWP2 heterodimerizes with WWP1 (Chaudhary and Maddika, 2014), another HECT-type E3 ligase, which has been previously reported to induce ubiquitin-dependent degradation of TGFBR1 (Komuro et al., 2004). In addition, Wwp2.sup.Mut/Mut cardiac fibroblasts showed higher cell proliferation and migration compared to controls (
[0588] Isoform-eQTL analysis of WWP2 in the heart of DCM patients suggested a regulatory role for the WWP2 isoforms containing N-terminal region in fibrosis, and in agreement with this, murine cardiac fibroblasts responded to TGFβ1 treatment with increased protein levels of the WWP2-FL/N isoforms, but not of WWP2-C. Increased expression of WWP2-FL/N isoforms was consistent with the findings in the heart following Ang II-infusion in vivo. Wwp2.sup.Mut/Mut mice lack WWP2-FL and WWP2-N protein isoforms, and this is sufficient to alter the co-regulation of the hECM network genes and reduce cardiac fibrosis in vivo and inhibit conversion of fibroblasts to myofibroblasts in vitro. These data combined suggest a primary role for WWP2 gene isoforms containing N-terminal region of the protein (i.e., WWP2-FL and WWP2-N) in regulating the molecular program associated with cardiac fibrosis.
[0589] To confirm the differential effects of the WWP2 N-terminal region on cardiac fibroblast activity, siRNA sequences were designed, targeting either the 5′-terminal (siRNA-Wwp2-N′) or 3′-terminal (siRNA-Wwp2-C′) regions of the Wwp2 mRNA. siRNA transfection in human cardiac fibroblast successfully decreased the expression of WWP2-FL/N and WWP2-FL/C isoforms in cardiac fibroblasts, respectively (
[0590] The inventors then carried out a rescue experiment by re-introducing the two isoforms containing WWP2 N-terminal region (WWP2-FL and WWP2-N) separately in primary cardiac fibroblasts from Wwp2.sup.Mut/Mut mice (
Example 5: WWP2 Regulates the Nucleocytoplasmic Shuttling of the TGFβ-Signaling Transducer SMAD2
[0591] Subcellular Localisation of WWP2 Isoforms
[0592] The inventors further investigated the cellular mechanisms through which WWP2 regulates the fibrogenic response downstream of TGFβ signaling activation. WWP2 was found to weakly localize in both the cytoplasm and nuclei in quiescent cardiac fibroblasts; however, upon TGFβ1 simulation (16 hrs), increased WWP2 expression localized predominantly in the nucleus was observed (
[0593] The presence of WWP2 isoforms in the nucleus makes it plausible that WWP2 may be involved in regulating gene expression, possibly targeting transcription factors for ubiquitination as shown for other E3 ubiquitin ligases (Horwitz et al., 2007; Gao et al., 2016). Examples 3 and 4 herein provide evidence for WWP2 involvement in the downstream regulation of TGFβ-signalling in vivo and in vitro.
[0594] WWP2 Isoform Interaction with SMAD Proteins
[0595] It has been demonstrated that the pro-fibrotic transcriptional program (i.e., the hECM-network) positively regulated by WWP2 is enriched for transcriptional targets of the TGFβ-signalling transducer SMAD transcription factors. Genes downregulated by WWP2 in vivo in the heart following Ang II-infusion were significantly enriched for “SMAD binding” (FDR=0.003), “SMAD protein signal” (FDR=0.006) and “transcriptional activity of SMAD2/3/4 heterotrimer” (FDR=0.015) by GSEA.
[0596] No difference was observed at the mRNA level of SMAD2 transcription factor in Wwp2.sup.mut/Mut cardiac fibroblasts (
[0597] FLAG-tagged WWP2-FL and WWP2-N but not WWP2-C expressed in NIH-3T3 mouse embryonic fibroblast cells co-immunoprecipitated with SMAD2 protein (
[0598] These data suggest an endogenous physiological interaction between SMAD2 and WWP2-FL/N protein isoforms.
[0599] WWP2 was also found to interact directly with p-SMAD2 (
[0600] Analysis of WWP2-Mediated Ubiquitination of SMAD Proteins and Nucleocytoplasmic Shuttling
[0601] Ubiquitination assays followed by SMAD2 immunoprecipitation detected monoubiquitinated SMAD2 within 16 hours of TGFβ1 stimulation of the mouse NIH-3T3 fibroblast cell line (
[0602] Monoubiquitination of SMADs by ubiquitin E3 ligases reportedly affects their transcriptional activity in primary fibroblasts (Xie et al., 2014; Tang et al., 2011). SMAD2-dependent luciferase reporter activity assays were performed in primary cardiac fibroblasts from WT and Wwp2.sup.Mut/Mut mice to investigate whether WWP2 affects the transcriptional activity of SMAD2. TGFβ1-dependent SMAD2 reporter activity was significantly lower in Wwp2.sup.Mut/Mut cardiac fibroblasts compared to WT cells (
[0603] TGFβ-receptor activation promotes the nuclear accumulation of SMAD2/3/4 (Xu and Massagué, 2004) and this process is not necessarily accompanied by SMAD degradation in the nucleus, as SMADs are exported out of the nucleus upon dephosphorylation and dissociation of the SMAD complexes (Gareth J. Inman et al., 2002; Lin et al., 2006). Analysis of nuclear and cytoplasmic fractions obtained from cardiac fibroblasts showed nuclear accumulation of SMAD2 upon TGFβ1 stimulation (<16 hr). More SMAD2 was found to be localised to the nucleus in Wwp2.sup.Mut/Mut fibroblasts compared to WT cells (
[0604] Monoubiquitination is also important for the proper subcellular localization of SMADs, and in turn might regulate their transcriptional activity in the nucleus (Gareth J Inman et al., 2002; Schmierer and Hill, 2005). SB431542 (a selective inhibitor of TGFβ superfamily type I activin receptor-like kinase (ALK) receptors (Gareth J Inman et al., 2002)) was used to study the differential nuclear export and nucleocytoplasmic shuttling of SMAD2 (Schmierer and Hill, 2005). Using primary fibroblasts from WWP2.sup.Mut/Mut and WT mice, a delay in the nuclear export of SMAD2 in Wwp2.sup.Mut/Mut mice was observed (
[0605] Taken together, these data show that in primary cardiac fibroblasts WWP2 interacts with SMAD2, promoting its monoubiquitination, and modulates the nucleocytoplasmic shuttling and transcriptional activity of SMAD2 downstream of TGFβ-signalling activation (
Example 6: Analysis of Immunomodulatory Function of WWP2 in Fibrosis
[0606] Bulk RNA-seq analysis in the fibrotic heart (upon AngII-infusion) of Wwp2.sup.Mut/Mut and WT mice showed several pathways that are either up- or down-regulated by WWP2. The analysis revealed that both TGFβ-signalling and the coagulation/complement pathway were strongly downregulated by WWP2, the latter being the most significant pathway affected in vivo.
[0607] The complement system consists of more than 40 soluble and membrane bound proteins and is activated in several heart diseases. Different parts of the complement system play a role in both stable and unstable coronary heart disease (Oksjoki et al., 2007) and in idiopathic dilated and ischemic cardiomyopathies (Aukrust et al., 2001). One active component is the cleavage product of complement factor 5 (C5), complement factor 5a (C5a), which has chemotactic and inflammatory properties Oyer et al., 2011).
[0608] C5a Receptor Expression, Fibrosis and the Role of WWP2
[0609] Since C5a receptor (CSaR) is widely expressed in several different cell types in the liver, lung, heart, intestine, and kidney tissues (Haviland et al., 1995), the expression of CSaR was analysed in the fibrotic murine heart. The inventors found that CSaR expression was localized to non-myocytes cell in the heart tissues from both WT and Wwp2 LOF mice. AngII-infusion has been reported to activate CSaR signalling in circulating immune cells (Zhang et al., 2014).
[0610] AngII treatment was found to upregulate C5aR expression, and the number of C5aR+ cells induced by AngII-treatment was found to be significantly reduced in Wwp2 LOF mice (
[0611] Overall these data suggest that complement activation in heart tissue is involved in the fibrotic response, and that this process is regulated by WWP2 in vivo. Cardiac fibrosis has been proposed as an important therapeutic target in heart failure patients (Segura et al., 2014; Edgley et al., 2012), and so the identification of genes that regulate pathological fibrosis may provide new avenues to control the progression to heart failure.
[0612] WWP2 Expression by Immune Cells
[0613] To start exploring the role of WWP2 in regulating immune cell functions in cardiac fibrosis, the inventors first examined the specific immune cells expressing WWP2 in the heart. Preliminary SCS analysis in the heart suggested that WWP2 is expressed by several different types of cardiac cells, including fibroblasts, endothelial cells and various immune cells, such as M1 and M2 macrophages (
[0614] SCS data was also analysed to determine whether genes in the complement pathway are transcriptionally co-expressed with WWP2 in individual cardiac cells. Using different immune cell types (i.e. macrophages, T-cells, etc.) expressing WWP2, the expression levels of WWP2 were correlated with the expression levels of genes involved in the complement pathway in each immune cell-type. The strongest positive association between WWP2 and complement pathway genes was found in M2 and M1 macrophages, suggesting that macrophages are a main source contributing to the complement pathway in WT heart (which was most strongly downregulated in the heart of WWP2 LOF mice):
TABLE-US-00003 Immune cell types expressing WWP2 in the heart M2 M1 NK T macr. macr. cells cells Strength of WWP2 correlation with 1.32 1.28 0.70 0.46 expression of complement genes
[0615] The inventors also investigated WWP2 activation in the context of macrophage polarization. Macrophages are a heterogeneous population of tissue-resident professional phagocytes, characterized by phenotypic and functional diversities that play a crucial role in fibrogenesis (Wynn and Vannella, 2016). Since macrophage polarization states (pro-inflammatory M1 and alternatively activated anti-inflammatory M2) are important in mediating cardiac fibrosis (Zhou et al., 2017; Mylonas et al., 2015), bone marrow derived macrophages (BMDMs) were cultured and induced to differentiate to M1 and M2 macrophages by LPS/IFNγ and IL4/IL13 stimulation, respectively.
[0616] Both M1 and M2 polarization states were associated with increased expression of WWP2 (N-isoform) (
[0617] The inventors further tested the potential regulation of macrophage phenotype/polarization by Wwp2 using cultured BMDMs from both WT and Wwp2 LOF mice. Upon LPS/IFNγ stimulation, BMDMs showed an M1 phenotype, producing pro-inflammatory cytokines such as IL-6. This pro-inflammatory change in M1 macrophages was largely prevented in Wwp2 LOF BMDMs (
[0618] These findings suggest that Wwp2 LOF affects (at least in part) pro-inflammatory and pro-fibrotic phenotypes under M1/M2 polarization.
Example 7: WWP2 Regulates Kidney Fibrosis In Vivo
[0619] Renal fibrosis is a widespread pathological feature of progressive renal disease of virtually any etiology.
[0620] The inventors explored whether WWP2 regulates fibrosis in kidney fibrotic disease using a unilateral ureteral obstruction (UUO) model, which generates progressive renal fibrosis (Chevalier et al., 2009), in WT and Wwp2.sup.Mut/Mut mice.
[0621] After 14 days, increased levels of all WWP2 protein isoforms (i.e. WWP2-FL, WWP2-N and WWP2-C) was observed (
Example 8: Discussion
[0622] In the present studies, starting from the identification of a pro-fibrotic gene network conserved in rat and in human heart disease characterized by diffuse myocardial remodelling and fibrosis, the inventors identified WWP2 as a regulator of pathological fibrosis. The WWP2-reguated pro-fibrotic gene network was conserved across different fibrotic diseases and cardiac tissues, and was upregulated in dilated cardiomyopathy that progresses to heart failure in mice (Burke et al., 2016). The lack of regulation of the pro-fibrotic gene network by WWP2 in human control heart tissue suggests that WWP2 exerts its regulatory role on cardiac fibrosis upon disease.
[0623] WWP2 mRNA expression was only marginally increased in fibrotic heart disease (less than 2 folds in either human DCM, rTOF or mouse HF (Burke et al., 2016)). This might explain why WWP2 passed undetected by GWAS and other eQTL genetic mapping or cellular screening studies of fibrotic diseases. In the heart of DCM patients, isoform-specific eQTL analysis led the inventors to hypothesize that increased expression of the WWP2 N-terminal isoform was associated with the activation of a pro-fibrotic gene program downstream of TGFβ/SMAD signalling activation. This hypothesis was tested in primary cardiac fibroblasts and in vivo models of cardiac fibrosis.
[0624] The results provide the first indication of a role for WWP2 in regulating pathophysiological processes in the heart. In detail, the inventors have demonstrated that WWP2-N/FL LOF improves cardiac function and reduces myocardial fibrosis, in two established preclinical models of cardiovascular fibrosis (Example 3.2). Moreover, upon treatment with Ang II, a typical promoter of cardiac fibrosis, Wwp2.sup.Mut/Mut mice displayed increased oxidative phosphorylation, and exacerbated cell cycle and IFNα and IFNγ response in the heart. This suggests that depletion of the N-terminal part of WWP2 might have a wider role in myocardial fibrosis, preventing the glycolytic metabolic reprogramming required for myofibroblast differentiation (Bernard et al., 2015; Selvarajah et al., 2016), and potentially blocking the cell cycle arrest that has been shown to be a feature of fibrosis (Lovisa et al., 2015).
[0625] Depletion of the N-terminal part of WWP2 additionally increased the expression of IFNα and IFNγ response genes, which have been shown to induce apoptosis of myofibroblasts (Nedelec et al., 2001; Yokozeki et al., 1999).
[0626] Working in primary fibroblast cultures, the inventors have shown that WWP2 positively regulates the expression of established pro-fibrotic genes downstream of TGFβ/SMAD signalling activation. However, compared with in vivo results, the pro-fibrotic response to TGFβ1 in quiescent cardiac fibroblasts was more modest, possibly because some of the cultured fibroblasts have been already converted to myofibroblasts before TGFβ1 treatment (Santiago et al., 2010). Despite this, the phenotypic differences if the fibrotic response observed between WT and Wwp2.sup.Mut/Mut cardiac fibroblasts were consistent in all WWP2 loss-of-function, gain-of-function and rescue experiments.
[0627] Poly-ubiquitination and proteasome-mediated degradation of nuclear SMAD2 has already been described (Lo and Massagué, 1999). In addition, monoubiquitination can regulate nuclear accumulation and the nucleocytoplasmic shuttling of SMAD complexes, which are crucial for transduction of TGFβ-superfamily signals (Gareth J. Inman et al., 2002; Hill, 2009), and there is mounting evidence supporting the role of monoubiquitination of SMADs by E3 ubiquitin ligases in this context (Tang et al., 2011; Gareth J. Inman et al., 2002).
[0628] The inventors have shown for the first time that TGFβ1 stimulation promotes the translocation of WWP2 to the nucleus, where it interacts directly with SMAD2, possibly promoting its monoubiquitination, as shown for other ubiquitin E3 ligases (Komuro et al., 2004; Tang et al., 2011). The regulatory consequences of E3 ligase-mediated monoubiquitination appear to be complex and context specific. Monoubiquitination can regulate protein location, activity, and protein interactions with binding partners (Schnell and Hicke, 2003). Monoubiquitination has also been shown to be required for the activity and the intrinsic nuclear import of target transcription factor(s) (Zou et al., 2011; Trotman et al., 2007); while in other instances monoubiquitination has been reported to disrupt specific transcription factor interactions and their transcriptional activity (Inui et al., 2011).
[0629] In the present studies, the inventors did not observe SMAD2/3 degradation by endogenous WWP2 in primary fibroblasts, suggesting a mechanism of degradation-independent repression of SMAD2 activity downstream of TGFβ-signalling activation. The present data suggest that blocking WWP2 function can delay the TGFI31-induced nucleocytoplasmic shuttling of SMAD2 in primary fibroblasts.
Example 9: Conclusion
[0630] The present studies contribute to the understanding of the regulation of fibrosis and pathological inflammation, and identify a specific E3 ubiquitin ligase, WWP2, as a regulator of ECM accumulation downstream of TGFβ/SMAD signalling, and of immune cell phenotype.
[0631] In summary, using a systems genetics approach followed by functional validation in several cellular and pre-clinical animal model studies, the inventors have first identified and mechanistically explained that WWP2 is a novel and druggable therapeutic target with the potential to control fibrosis in pathological inflammation and tissue remodelling.
Example 10: References to Examples 2 to 8
[0632] Arpino, V. et al. (2015) The role of TIMPs in regulation of extracellular matrix proteolysis. Matrix Biology
[0633] Aschner, Y. et al. (2014) Protein tyrosine phosphatase a mediates profibrotic signaling in lung fibroblasts through TGF-β responsiveness. American Journal of Pathology.
[0634] Aukrust, P. et al. (2001) Complement activation in patients with congestive heart failure: Effect of high-dose intravenous immunoglobulin treatment. Circulation.
[0635] Baum, J. & Duffy, H. S. (2011) Fibroblasts and myofibroblasts: what are we talking about? Journal of cardiovascular pharmacology. 57 (4), 376-379.
[0636] Bernard, K. et al. (2015) Metabolic Reprogramming Is Required for Myofibroblast Contractility and Differentiation. The Journal of biological chemistry. 290 (42), 25427-25438.
[0637] Bottolo, L. et al. (2011) Bayesian detection of expression quantitative trait loci hot spots. Genetics. 189 (4), 1449-1459.
[0638] Brooks, C. L. et al. (2007) Mechanistic studies of MDM2-mediated ubiquitination in p53 regulation. Journal of Biological Chemistry.
[0639] Burke, M. A. et al. (2016) Molecular profiling of dilated cardiomyopathy that progresses to heart failure. JCI Insight.
[0640] Carvajal, L. A. et al. (2018) Dual inhibition of MDMX and MDM2 as a therapeutic strategy in leukemia. Science Translational Medicine.
[0641] Chaudhary, N. & Maddika, S. (2014) WWP2-WWP1 Ubiquitin Ligase Complex Coordinated by PPM1 G Maintains the Balance between Cellular p73 and Np73 Levels. Molecular and Cellular Biology.
[0642] Chevalier, R. L. et al. (2009) Ureteral obstruction as a model of renal interstitial fibrosis and obstructive nephropathy. Kidney International. 75 (11), 1145-1152.
[0643] Ciftciler, R. et al. (2017) The importance of serum biglycan levels as a fibrosis marker in patients with chronic hepatitis B. Journal of Clinical Laboratory Analysis. 31 (5), e22109.
[0644] DeBerge, M. et al. (2019) Macrophages in Heart Failure with Reduced versus Preserved Ejection Fraction. Trends in molecular medicine. 25 (4), 328-340.
[0645] Ding, L. et al. (2017) Inhibition of Skp2 suppresses the proliferation and invasion of osteosarcoma cells. Oncology Reports.
[0646] Edgley, A. J. et al. (2012) Targeting Fibrosis for the Treatment of Heart Failure: A Role for Transforming Growth Factor-β. Cardiovascular Therapeutics. 30 (1), e30-e40.
[0647] Fukumoto, C. et al. (2014) WWP2 is overexpressed in human oral cancer, determining tumor size and poor prognosis in patients: downregulation of WWP2 inhibits the AKT signaling and tumor growth in mice. Oncoscience. 1807.
[0648] Gao, R. et al. (2016) E3 ubiquitin ligase RLIM negatively regulates c-Myc transcriptional activity and restrains cell proliferation. PLoS ONE.
[0649] Haviland, D. L. et al. (1995) Cellular expression of the C5a anaphylatoxin receptor (C5aR): demonstration of C5aR on
[0650] nonmyeloid cells of the liver and lung. Journal of immunology (Baltimore, Md.: 1950). 154 (4), 1861-1869.
[0651] Heinig, M. et al. (2010) A trans-acting locus regulates an anti-viral expression network and type 1 diabetes risk. Nature. 467 (7314), 460-464.
[0652] Heinig, M. et al. (2017) Natural genetic variation of the cardiac transcriptome in non-diseased donors and patients with dilated cardiomyopathy. Genome Biology. 18 (1), 170.
[0653] Hill, C. S. (2009) Nucleocytoplasmic shuttling of Smad proteins. Cell Research
[0654] Horwitz, A. A. et al. (2007) A mechanism for transcriptional repression dependent on the BRCA1 E3 ubiquitin ligase. Proceedings of the National Academy of Sciences.
[0655] Hu, H.-H. et al. (2018) New insights into TGF-β/Smad signaling in tissue fibrosis. Chemico-Biological Interactions. 29276-83. Huang, X. & Dixit, V. M. (2016) Drugging the undruggables: exploring the ubiquitin system for drug development. Cell Research. 26 (4), 484-498.
[0656] Huang, X. L. et al. (2016) E3 ubiquitin ligase: A potential regulator in fibrosis and systemic sclerosis. Cellular Immunology
[0657] Hubner, N. et al. (2005) Integrated transcriptional profiling and linkage analysis for identification of genes underlying disease. Nature genetics. 37 (3), 243-253.
[0658] Inman, G. J. et al. (2002) Nucleocytoplasmic shuttling of Smads 2,3, and 4 permits sensing of TGF-β receptor activity. Molecular Cell.
[0659] Inman, G. J. et al. (2002) SB-431542 is a potent and specific inhibitor of transforming growth factor-beta superfamily type I activin receptor-like kinase (ALK) receptors ALK4, ALK5, and ALK7. Molecular pharmacology.
[0660] Inui, M. et al. (2011) USP15 is a deubiquitylating enzyme for receptor-activated SMADs. Nature Cell Biology.
[0661] Iyer, A. et al. (2011) Inhibition of inflammation and fibrosis by a complement C5a receptor antagonist in DOCA-salt hypertensive rats. Journal of Cardiovascular Pharmacology.
[0662] Jia, L. & Sun, Y. (2011) SCF E3 Ubiquitin Ligases as Anticancer Targets. Current Cancer Drug Targets.
[0663] Johnson, M. R. et al. (2015) Systems genetics identifies Sestrin 3 as a regulator of a proconvulsant gene network in human epileptic hippocampus. Nature communications. 6 (1), 6031.
[0664] Kang, H. et al. (2014) Kcnn4 is a regulator of macrophage multinucleation in bone homeostasis and inflammatory disease. Cell reports. 8 (4), 1210-1224.
[0665] Khalil, H. et al. (2017) Fibroblast-specific TGF-β-Smad2/3 signaling underlies cardiac fibrosis. Journal of Clinical Investigation. 127 (10), 3770-3783.
[0666] Kim, S. et al. (2018) Two Genetic Variants Associated with Plantar Fascial Disorders. International Journal of Sports Medicine. 39 (04), 314-321.
[0667] Komuro, A. et al. (2004) Negative regulation of transforming growth factor-β (TGF-β) signaling by WW domain-containing protein 1 (WWP1). Oncogene.
[0668] Lachmann, A. et al. (2010) ChEA: transcription factor regulation inferred from integrating genome-wide ChIP-X experiments. Bioinformatics (Oxford, England). 26 (19), 2438-2444.
[0669] Langley, S. R. et al. (2013) Systems-level approaches reveal conservation of trans-regulated genes in the rat and genetic determinants of blood pressure in humans. Cardiovascular research. 97 (4), 653-665.
[0670] Leach, H. G. et al. (2013) Endothelial Cells Recruit Macrophages and Contribute to a Fibrotic Milieu in Bleomycin Lung Injury. American Journal of Respiratory Cell and Molecular Biology. 49 (6), 1093-1101.
[0671] Leask, A. (2015) Getting to the Heart of the Matter: New Insights Into Cardiac Fibrosis. Circulation research. 116 (7), 1269-1276. Lewin, A. et al. (2016) MT-HESS: an efficient Bayesian approach for simultaneous association detection in OMICS datasets, with application to eQTL mapping in multiple tissues. Bioinformatics (Oxford, England). 32 (4), 523-532.
[0672] Lin, X. et al. (2006) PPM1A Functions as a Smad Phosphatase to Terminate TGFβ Signaling. Cell.
[0673] Lo, R. S. & Massagué, J. (1999) Ubiquitin-dependent degradation of TGF-β-activated Smad2. Nature Cell Biology.
[0674] Lovisa, S. et al. (2015) Epithelial-to-mesenchymal transition induces cell cycle arrest and parenchymal damage in renal fibrosis. Nature medicine. 21 (9), 998-1009.
[0675] Maddika, S. et al. (2011) WWP2 is an E3 ubiquitin ligase for PTEN. Nature cell biology. 13 (6), 728-733.
[0676] Mancini, M. et al. (2013) Mapping genetic determinants of coronary microvascular remodeling in the spontaneously hypertensive rat. Basic research in cardiology. 108 (1), 316.
[0677] Massague, J. & Wotton, D. (2000) NEW EMBO MEMBERS REVIEW: Transcriptional control by the TGF-beta/Smad signaling system. The EMBO Journal. 19 (8), 1745-1754.
[0678] Matsumura, S. I. et al. (2005) Targeted deletion or pharmacological inhibition of MMP-2 prevents cardiac rupture after myocardial infarction in mice. Journal of Clinical Investigation.
[0679] McVicker, B. L. & Bennett, R. G. (2017) Novel Anti-fibrotic Therapies. Frontiers in Pharmacology. 8318.
[0680] Moustakas, A. et al. (2001) Smad regulation in TGF-beta signal transduction. Journal of cell science. 114 (Pt 24), 4359-4369.
[0681] Mylonas, K. J. et al. (2015) The adult murine heart has a sparse, phagocytically active macrophage population that expands through monocyte recruitment and adopts an ‘M2’ phenotype in response to Th2 immunologic challenge. Immunobiology.
[0682] Nakamura, Y. et al. (2011) Wwp2 is essential for palatogenesis mediated by the interaction between Sox9 and mediator subunit 25. Nature communications. 2251.
[0683] Nedelec, B. et al. (2001) Myofibroblasts and apoptosis in human hypertrophic scars: The effect of interferon-a2b. Surgery. 130 (5), 798-808.
[0684] Oksjoki, R. et al. (2007) Receptors for the anaphylatoxins C3a and C5a are expressed in human atherosclerotic coronary plaques. Atherosclerosis.
[0685] Peterson, J. T. et al. (2000) Evolution of matrix metalloprotease and tissue inhibitor expression during heart failure progression in the infarcted rat. Cardiovascular Research.
[0686] Petretto, E. et al. (2008) Integrated genomic approaches implicate osteoglycin (Ogn) in the regulation of left ventricular mass. Nature genetics. 40 (5), 546-552.
[0687] Piersma, B. et al. (2015) Signaling in Fibrosis: TGF-β, WNT, and YAP/TAZ Converge. Frontiers in Medicine. 259.
[0688] Pradegan, N. et al. (2016) Myocardial histopathology in late-repaired and unrepaired adults with tetralogy of Fallot. Cardiovascular Pathology.
[0689] Pravenec, M. et al. (2008) Identification of renal Cd36 as a determinant of blood pressure and risk for hypertension. Nature genetics. 40 (8), 952-954.
[0690] Rockey, D. C. et al. (2015) Fibrosis—A Common Pathway to Organ Injury and Failure. The New England journal of medicine. 372 (12), 1138-1149.
[0691] Ross, S. & Hill, C. S. (2008) How the Smads regulate transcription. The International Journal of Biochemistry & Cell Biology. 40 (3), 383-408.
[0692] Rotival, M. & Petretto, E. (2014) Leveraging gene co-expression networks to pinpoint the regulation of complex traits and disease, with a focus on cardiovascular traits. Briefings in functional genomics. 13 (1), 66-78.
[0693] Santiago, J. J. et al. (2010) Cardiac fibroblast to myofibroblast differentiation in vivo and in vitro: Expression of focal adhesion components in neonatal and adult rat ventricular myofibroblasts. Developmental Dynamics.
[0694] Schafer, S. et al. (2017) IL11 is a crucial determinant of cardiovascular fibrosis. Nature. 552 (7683), 110.
[0695] Schmidt, M. et al. (2015) Controlling the Balance of Fibroblast Proliferation and Differentiation: Impact of Thy-1. Journal of Investigative Dermatology. 135 (7), 1893-1902.
[0696] Schmierer, B. & Hill, C. S. (2005) Kinetic analysis of Smad nucleocytoplasmic shuttling reveals a mechanism for transforming growth factor beta-dependent nuclear accumulation of Smads. Molecular and cellular biology.
[0697] Schnell, J. D. & Hicke, L. (2003) Non-traditional Functions of Ubiquitin and Ubiquitin-binding Proteins. Journal of Biological Chemistry.
[0698] Schwanekamp, J. A. et al. (2017) TGFBI functions similar to periostin but is uniquely dispensable during cardiac injury. PLoS ONE.
[0699] Segura, A. M. et al. (2014) Fibrosis and heart failure. Heart Failure Reviews.
[0700] Selvarajah, B. et al. (2016) Metabolic shift during TGF-β induced collagen synthesis. QJM: An International Journal of Medicine. 109 (suppl_1), S3-S3.
[0701] Soond, S. M. & Chantry, A. (2011) Selective targeting of activating and inhibitory Smads by distinct WWP2 ubiquitin ligase isoforms differentially modulates TGFβ signalling and EMT. Oncogene. 30 (21), 2451-2462.
[0702] Styrkarsdottir, U. et al. (2018) Meta-analysis of Icelandic and UK data sets identifies missense variants in SMO, IL11, COL11A1 and 13 more new loci associated with osteoarthritis. Nature Genetics. 50 (12), 1681-1687.
[0703] Subramanian, A. et al. (2005) Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences of the United States of America. 102 (43), 15545-15550.
[0704] Tang, L. Y. et al. (2011) Ablation of Smurf2 reveals an inhibition in TGF-β signalling through multiple mono-ubiquitination of Smad3. EMBO Journal.
[0705] Trotman, L. C. et al. (2007) Ubiquitination Regulates PTEN Nuclear Import and Tumor Suppression. Cell.
[0706] Uderhardt, S. et al. (2019) Resident Macrophages Cloak Tissue Microlesions to Prevent Neutrophil-Driven Inflammatory Damage. Cell. 0 (0).
[0707] Villafañe, J. et al. (2013) Hot topics in tetralogy of fallot. Journal of the American College of Cardiology
[0708] Wang, Q. et al. (2011) Cooperative interaction of CTGF and TGF-β in animal models of fibrotic disease. Fibrogenesis & tissue repair. 4 (1), 4.
[0709] Wynn, T. A. & Vannella, K. M. (2016) Macrophages in Tissue Repair, Regeneration, and Fibrosis. Immunity
[0710] Xie, F. et al. (2014) Regulation of TGF-β Superfamily Signaling by SMAD Mono-Ubiquitination. Cells.
[0711] Xu, L. & Massagué, J. (2004) Nucleocytoplasmic shuttling of signal transducers. Nature Reviews Molecular Cell Biology
[0712] Yang, Y. et al. (2013) E3 ligase WWP2 negatively regulates TLR3-mediated innate immune response by targeting TRIF for ubiquitination and degradation. Proceedings of the National Academy of Sciences. 110 (13), 5115-5120.
[0713] Yokozeki, M. et al. (1999) Interferon-gamma inhibits the myofibroblastic phenotype of rat palatal fibroblasts induced by transforming growth factor-beta1 in vitro. FEBS letters. 442 (1), 61-64.
[0714] Zhang, C. et al. (2014) Complement 5a receptor mediates angiotensin II-induced cardiac inflammation and remodeling. Arteriosclerosis, Thrombosis, and Vascular Biology.
[0715] Zhao, T. et al. (2013) Platelet-derived growth factor-D promotes fibrogenesis of cardiac fibroblasts. American journal of physiology. Heart and circulatory physiology. 304 (12), H1719-26.
[0716] Zhou, D. et al. (2017) Macrophage polarization and function: new prospects for fibrotic disease. Immunology and Cell Biology. 95 (10), 864-869.
[0717] Zou, W., Chen, X., Shim, J.-H., Huang, Z., Brady, N., Hu, D., Drapp, R., Sigrist, K., Glimcher, L. H., et al. (2011) The E3 ubiquitin ligase Wwp2 regulates craniofacial development through mono-ubiquitylation of Goosecoid. Nature cell biology. 13 (1), 59-65.
Example 11: Identification of WWP2 Inhibitors
[0718] Inhibitors of WWP2 isoforms WWP2-FL and WWP2-N were successfully identified.
[0719] 11.1 Identification of Test Compounds that Bind to WWP2
[0720] WWP2-FL, WWP2-N and WWP2-C were expressed as GST fusion proteins in bacteria, and purified using glutathione sepharose 4B resin (GE Healthcare) according to the manufacturer's instructions. Protein levels were determined by Coomasie staining and Western blot prior to use in Octet Binding assays.
[0721] Octet binding assays were performed using the Octet RED96e System (FortéBio Inc., Menlo Park, Calif.). Proteins were biotinylated using NHS-PEO4-biotin (Pierce). Super-streptavidin (SSA) biosensors (FortéBio Inc., Menlo Park, Calif.) were coated in a solution containing 1 μM of biotinylated protein for 4 hours at 25° C. A duplicate set of sensors was incubated in an assay buffer (1× kinetics buffer of ForteBio Inc.) with 2.5% DMSO without protein for use as a background binding control. Both sets of sensors were blocked with a solution of 10 μg/ml Biocytin for 5 minutes at 25° C. A negative control of 2.5% DMSO was also used.
[0722] Binding of test compound samples (250 μM) to coated and uncoated reference sensors was measured over 120 seconds.
[0723] Data analysis on the FortéBio Octet RED instrument was performed using a double reference subtraction (sample and sensor references) in the FortéBio data analysis software. The analysis accounts for non-specific binding, background, and signal drift and minimizes well-based and sensor variability.
[0724] Test compounds determined to bind to WWP2-N, and determined not to bind to WWP2-C were selected for further characterisation.
[0725] 11.2 Identification of Test Compounds that Inhibit WWP2-Mediated Ubiquitination
[0726] In vitro ubiquitination assays are performed. The assays contain 150 ng E1 (BIOMOL, Plymouth Meeting), 150 ng E2 (UbcH7 or UbcH5c, BIOMOL; 500 ng UbcH7 with Smurf2), 10 μg ubiquitin (BIOMOL), and 20 ng E3 (WWP2-FL or WWP2-C).
[0727] Fora quantitative readout, untagged ubiquitin (10 μg) is mixed with N-terminally biotinylated ubiquitin (1 μg), and 1 μg of GST-tagged SMAD2 substrate is adsorbed to glutathione-coated (or for some experiments, anti-GST coated) ELISA plates (Pierce; maximum capacity, 10 ng protein) for 45 min, followed by incubation with HRP-coupled Streptavidin for 20 min. Ubiquitinated SMAD2 is quantified after addition of 100 μL tetramethylbenzidine solution (Sigma-Aldrich) on a TECAN Infinite F200 microplate reader.
[0728] Candidate inhibitor test compounds are pre-incubated with E3 for 10-30 min at a range of concentrations prior to addition to the other components of the ubiquitination assay. Inhibition of WWP2FL/C ubiquitination activity by the test compounds is compared relative to vehicle control conditions.
[0729] 11.3 Evaluation of Test Compounds for Inhibition of Cellular Fibrosis
[0730] Primary cultured cardiac fibroblasts are plated at a density of 350 cells per well on poly-D-lysine (PDL, 10 μg/mL)-coated 96-well plates. After 24-h serum starvation, medium is incubated with test compounds (5 μM) for 30 minutes, and which are then added to the cardiac fibroblasts, and the cells are stimulated by culture in the presence of TGF-β1 (5 ng/mL; Sigma).
[0731] After 48 h, cells are fixed with a 4% paraformaldehyde solution for 10 min, subsequently blocked for 30 min (5% FBS, 0.3% Triton X-100 in PBS) and then stained overnight at 4° C. with primary mouse anti-αSMA antibody (in 1% FBS, 0.1% Triton X-100 in PBS). After washing with PBS, cells are stained with secondary antibody (donkey anti-mouse Alexa Fluor 488, Invitrogen; 1:500) for 2 h at room temperature in the dark. After three washes with PBS, the plate is sealed for imaging.
[0732] High-content imaging is performed with an Operetta High Content Imaging system (PerkinElmer). Four imaging fields are captured per well with a 5× imaging objective, to enable visualization of the entire well.
[0733] Relative cell area and relative staining intensity of α-SMA are determined using internal algorithms in the Cellomics Scan software package, and thresholds fitted using multiple wells of positive (SB-431542, 10 μM) and negative (DMSO, 0.1%) controls present in each screening plate.
Example 12: Treatment of Fibrosis in Various Different Tissues Using WWP2 Inhibitors
[0734] The utility of WWP2 inhibitors to treat/prevent fibrosis is demonstrated in in vivo mouse models of fibrosis for various different tissues.
[0735] 12.1 Heart Fibrosis
[0736] An AngII infusion model is established as described in Example 1.8. A myocardial infarction model is established as described in Example 1.8.
[0737] Mice are treated with WWP2 inhibitor or vehicle (negative control). At the end of the experiment, hearts are harvested and analysed for correlates of fibrosis.
[0738] Mice treated with WWP2 inhibitor are found to display a reduced level of heart fibrosis as compared to vehicle-treated controls.
[0739] 12.2 Kidney Fibrosis
[0740] A mouse model of kidney fibrosis is established by unilateral ureteral obstruction (UUO), as described in Chevalier et al., Kidney International (2009) 75 (11), 1145-1152.
[0741] Mice are treated with WWP2 inhibitor or vehicle (negative control). At the end of the experiment, kidneys are harvested and analysed for correlates of fibrosis.
[0742] Mice treated with WWP2 inhibitor are found to display a reduced level of kidney fibrosis as compared to vehicle-treated controls.
[0743] 12.3 Liver Fibrosis
[0744] A mouse model of liver fibrosis is established.
[0745] Mice are treated with WWP2 inhibitor or vehicle (negative control). At the end of the experiment, livers are harvested and analysed for correlates of fibrosis.
[0746] Mice treated with WWP2 inhibitor are found to display a reduced level of liver fibrosis as compared to vehicle-treated controls.
[0747] 12.4 Lung Fibrosis
[0748] A mouse model of lung fibrosis is established.
[0749] Mice are treated with WWP2 inhibitor or vehicle (negative control). At the end of the experiment, lungs are harvested and analysed for correlates of fibrosis.
[0750] Mice treated with WWP2 inhibitor are found to display a reduced level of lung fibrosis as compared to vehicle-treated controls.
[0751] 12.5 Skin Fibrosis
[0752] A mouse model of skin fibrosis is established.
[0753] Mice are treated with WWP2 inhibitor or vehicle (negative control). At the end of the experiment, the skin is analysed for correlates of fibrosis.
[0754] Mice treated with WWP2 inhibitor are found to display a reduced level of skin fibrosis as compared to vehicle-treated controls.
[0755] 12.6 Eye Fibrosis
[0756] A mouse model of eye fibrosis is established.
[0757] Mice are treated with WWP2 inhibitor or vehicle (negative control). At the end of the experiment, the eyes are analysed for correlates of fibrosis.
[0758] Mice treated with WWP2 inhibitor are found to display a reduced level of eye fibrosis as compared to vehicle-treated controls.
[0759] 12.7 Bowel Fibrosis
[0760] A mouse model of bowel fibrosis is established.
[0761] Mice are treated with WWP2 inhibitor or vehicle (negative control). At the end of the experiment, the bowel is analysed for correlates of fibrosis.
[0762] Mice treated with WWP2 inhibitor are found to display a reduced level of bowel fibrosis as compared to vehicle-treated controls.
Example 13: WWP2 Regulates Immune Function of Cardiac Macrophages in Fibrosis
[0763] The inventors first assessed the macrophages in the heart following Ang II infusion. Ang II infusion was performed as described in Example 1.8 but the duration of infusion was 7 days. Hearts from sham or AngII-infused animals were analysed by flow cytometry. Ang II induced significant accumulation of macrophages (CD45.sup.+/CD11b.sup.+/Ly6G.sup.−/F4/80.sup.+/CD64.sup.+) at 7 days. In contrast to WT mice, Wwp2.sup.Mut/Mut mice showed a significant attenuation of Ang II-induced macrophages (
[0764] Ly6C.sup.high cells are often referred to as “inflammatory monocytes”. The inventors observed an increase of Ly6C.sup.high macrophages in fibrotic hearts, as well as a suppressing effect of Wwp2 LOF in this macrophage population (
[0765] Using CellphoneDB (a publicly available repository of curated receptors, ligands and their interactions) on the single-cell RNA-sequencing data, the inventors found significant Ligand-Receptor interactions in inflammatory macrophages for both WT and Wwp2 LOF mice. Wwp2 LOF inhibited activation of the CCL5 pathway (
[0766] Differential expression analysis between WT and Wwp2 LOF inflammatory macrophages was performed in order to investigate the regulation of cardiac macrophages by Wwp2 (
[0767] Analysis of cardiac macrophages sorted from fibrotic heart validated that Wwp2 LOF suppressed the inflammatory response (e.g. CCL12, S100a9) (
Example 14: WWP2 LOF Protects from Lung Fibrosis in Mice
[0768] Bleomycin-induced lung fibrosis is a widely used mouse model for human idiopathic pulmonary fibrosis (IPF)—see Liu et al., Methods Mol Biol (2017) 1627:27-42. Briefly, lung fibrosis was induced by oropharyngeal treatment of bleomycin: 8-10 weeks old female mice weighing approximately 20 to 22 g were anesthetized by isoflurane inhalation and then bleomycin (Sigma-Aldrich) was administered oropharyngeal at a dose of 1 mg/kg body weight. Saline administration was treated as sham control. Mice were observed daily for body weight and activity levels and harvested on day 21 post-bleomycin challenge. Administration of bleomycin to the lung was found to cause moderate long-term mortality. The inventors observed that fibrosis is fully developed with extensive and diffuse tissue involvement of the lung at 21 days of bleomycin treatment.
[0769] When bleomycin was administered to Wwp2.sup.Mut/Mut mice, the animals showed decreased mortality (P=0.044) when compared to WT mice (
[0770] Further, in Wwp2.sup.Mut/Mut mice, histological analysis revealed decreased pulmonary fibrosis (
[0771] Masson's Trichrome staining of lung tissue showed that more prominent staining corresponding to mature collagen was distributed in the alveolar septa or interstitial and peribronchial connective tissue in the bleomycin-injured WT lung than that in the Wwp2.sup.Mut/Mut lung (
Example 15: WWP2 is Upregulated in Human Chronic Kidney Disease, and WWP2 LOF Protects Mice from Renal Fibrosis
[0772] Considering that fibrosis is the common pathway in chronic kidney disease (CKD), the inventors assessed the expression of WWP2 kidney tissue from patients with CKD. Immuno-histological analysis showed that expression of WWP2 was obviously higher in the kidney tissue of patients with CKD than in the healthy control at the tubulointerstitial region (
[0773] The inventors further investigated the effects of Wwp2 knockout in the unilateral ureteral obstruction (UUO) model, a well-established rodent model of progressive renal fibrosis. The UUO model was established as described in Chevalier et al., Kidney International (2009) 75 (11), 1145-1152, and resulted in diffuse fibrosis in the kidneys of WT mice after 14 days. By contrast, Wwp2.sup.Mut/Mut mice had significantly less fibrosis throughout the kidney (
[0774] The inventors also investigated whether inhibiting WWP2 reduced pro-fibrotic phenotypes in renal fibroblasts in vitro. TGFβ1-stimulated renal fibroblasts from WT mice displayed clear organization of ACTA2 into stress fibers, while Wwp2.sup.Mut/Mut-derived cells displayed diffuse expression of ACTA2, with rare incorporation into stress fibers (
Example 16: Inhibition of WWP2 Using a Small Molecule Inhibitor Reduces Fibrosis
[0775] 16.1 Methods
[0776] 16.1.1 Evaluation of Binding to WWP2 by Octet RED Analysis
[0777] The WWP2 isoforms WWP2-FL, WWP2-N and WWP2-C were expressed as a GST fusion protein in bacteria and purified using glutathione sepharose 4B resin (GE Healthcare) according to the manufacturer's instructions. Expression and purity of the expressed WWP2 proteins was analysed by Coomassie Blue staining of the gel after separation by SDS-PAGE and western blotting (WB).
[0778] Binding assays were performed using the Octet RED96e System (FortéBio Inc., Menlo Park, Calif.). WWP2-FL, WWP2-N and WWP2-C proteins were biotinylated using NHS-PEO4-biotin (Pierce). Super-streptavidin (SSA) biosensors (FortéBio Inc., Menlo Park, Calif.) were coated in a solution containing 1 μM of biotinylated protein for 4 hours at 25° C. A duplicate set of sensors were incubated in assay buffer (1× kinetics buffer of ForteBio Inc.) with 5% DMSO without protein, for use as a background binding control. Both sets of sensors were blocked with a solution of 10 ug/ml Biocytin for 5 minutes at 25° C. A negative control of 5% DMSO was also used.
[0779] The compounds analysed in the assay were as follows: [0780] (i) 2-[11-(2,4-dimethoxyphenyl)-10,12-dioxo-7-thia-9,11-diazatricyclo[6.4.0.02.6]dodeca-1(8),2(6)-dien-9-yl]-N-(furan-2-ylmethyl)acetamide (PubChem CID: 3240890; Molecular Formula: C24H23N3O6S), referred to hereafter as EP1. The chemical structure for EP1 is shown in
[0782] Clomipramine hydrochloride (Sigma Aldrich, #C7291) was employed as a positive control, having previously been demonstrated to be an inhibitor of HECT domain-containing E3 ligase activity (see e.g. Rossi et al., Cell Death Dis. 2014 May; 5(5): e1203, hereby incorporated by reference in its entirety).
[0783] Binding of samples containing the different compounds (at a concentration of 250 μM) to coated and uncoated reference sensors was measured over 120 seconds. Analysis of the data was performed using a double reference subtraction (sample and sensor references) in the FortéBio data analysis software. The analysis accounts for non-specific binding, background, and signal drift and minimizes well-based and sensor variability.
[0784] 16.1.2 Cellular Fibrosis Assays for the Small Molecules.
[0785] Primary renal fibroblasts were obtained from the renal cortex of mice. Minced renal cortex pieces (1-3 mm.sup.3) were placed in 6 cm dishes with DMEM supplemented with 20% fetal bovine serum for less than 10 days to generate mice renal fibroblasts (P0), and passaged to P1 and P2 in DMEM supplemented with 10% fetal bovine serum for experiments. To mimic the in vivo activation of renal fibroblasts in fibrosis, cells were treated with human transforming growth factor-β1 (TGFβ1) (Sigma Aldrich, #T7039) at a concentration of 5 ng/μl for 72 hours.
[0786] To test the anti-fibrotic effect of candidate small-molecules, cells were treated with each compound (0.3 mM) for 1 hour, followed by stimulation with TGFβ1 before harvesting. After fixation with ice-cold acetone and blocking with 1% BSA for 30 min at RT, the slides were incubated with primary anti-smooth muscle a actin (ACTA2) antibody (ACTA2 is a widely recognised myofibroblast marker, indicating fibroblast activation) (Sigma-Aldrich, #A5228, 1:100) overnight at 4° C. Following washing steps, slides were incubated with Bovine Anti-Mouse IgG-CFL 488 secondary antibody (Santacruz Biotechnology, #sc-362256, 1:100) for 2 hours at RT. VectaShield Mounting Medium (Vector laboratories, #H-1200) with DAPI was used to stain the nuclei and the slides were covered by coverslip. Slides were imaged on Leica fluorescence microscope (Leica).
[0787] 16.2 Results
[0788] Binding of Clomipramine hydrochloride at 250 μM to WWP2-N and WWP2-C was 0.08 nm and 0.0839 nm, respectively. Thus, a binding level of ≥0.08 nm can be considered to be indicative of binding WWP2.
[0789] EP1 displayed stronger binding to WWP2-N than Clomipramine hydrochloride (0.1253 nm), and weaker binding to WWP2-C than Clomipramine hydrochloride (0.0061 nm) as indicated in the table below:
TABLE-US-00004 Binding level (nm) to WWP2 N to WWP2 C Anti-fibrotic response Compound name isoform isoform in renal fibroblasts EP1 0.1253 0.0061 Reduced expression of ACTA2 in TGFβ1- stimulated fibroblasts Clomipramine HCl 0.0800 0.0839 Positive control drug
[0790] TGFβ1 stimulation of primary renal fibroblasts increased fibrosis as evidenced by increased ACTA2 expression, as determined by fluorescence microscopy (
[0791] Treatment of the fibroblasts with EP1 prior to stimulation with TGFβ1 was associated with reduced expression of ACTA2 (
[0792] Thus EP1 is identified as an inhibitor of WWP2, and demonstrates that small molecule inhibitors of WWP2 (in particular, small molecule WWP2 inhibitors which bind to the WWP2 N-terminal isoform comprising the C2 domain) are useful for the treatment/prevention of fibrosis, as evidenced by its ability to antagonise TGFβ1-mediated activation of fibroblasts to myofibroblasts (which are effectors of fibrosis).