Modulation of lncRNA links endothelial glycocalyx to vascular dysfunction of tyrosine kinase inhibitor

Abstract

Aims

Novel cancer therapies leading to increased survivorship of cancer patients have been negated by a concomitant rise in cancer therapies-related cardiovascular toxicities. Sunitinib, a first line multi-receptor tyrosine kinase inhibitor, has been reported to cause vascular dysfunction although the initiating mechanisms contributing to this side effect remain unknown. Long non-coding RNAs (lncRNAs) are emerging regulators of biological processes in endothelial cells (ECs); however, their roles in cancer therapies-related vascular toxicities remain underexplored.

Methods and results

We performed lncRNA expression profiling to identify potential lncRNAs that are dysregulated in human-induced pluripotent stem cell-derived ECs (iPSC-ECs) treated with sunitinib. We show that the lncRNA hyaluronan synthase 2 antisense 1 (HAS2-AS1) is significantly diminished in sunitinib-treated iPSC-ECs. Sunitinib was found to down-regulate HAS2-AS1 by an epigenetic mechanism involving hypermethylation. Depletion of HAS2-AS1 recapitulated sunitinib-induced detrimental effects on iPSC-ECs, whereas CRISPR-mediated activation of HAS2-AS1 reversed sunitinib-induced dysfunction. We confirmed that HAS2-AS1 stabilizes the expression of its sense gene HAS2 via an RNA/mRNA heteroduplex formation. Knockdown of HAS2-AS1 led to reduced synthesis of hyaluronic acid (HA) and up-regulation of ADAMTS5, an enzyme involved in extracellular matrix degradation, resulting in disruption of the endothelial glycocalyx which is critical for ECs. In vivo, sunitinib-treated mice showed reduced coronary flow reserve, accompanied by a reduction in Has2os and degradation of the endothelial glycocalyx. Finally, we identified that treatment with high molecular-weight HA can prevent the deleterious effects of sunitinib both in vitro and in vivo by preserving the endothelial glycocalyx.

Conclusions

Our findings highlight the importance of lncRNA-mediated regulation of the endothelial glycocalyx as an important determinant of sunitinib-induced vascular toxicity and reveal potential novel therapeutic avenues to attenuate sunitinib-induced vascular dysfunction.

Graphical Abstract

iPSCs, Cardio-oncology, Vascular toxicity, lncRNAs, Glycocalyx

Time of primary review: 25 days

Translational perspective

Sunitinib is a multi-targeted receptor tyrosine kinase inhibitor used in the treatment of various human malignancies. Despite its effectiveness, most sunitinib-treated patients develop cardiac and vascular dysfunction. Emerging studies have reported the vital role of long non-coding RNAs (lncRNAs) in the vascular system. Therefore, the identification of lncRNAs that regulate the function of the vascular endothelium in the presence of sunitinib is crucial to understanding and evolving potential new strategies that might protect the vasculature. Here, we describe and dissect the functional importance of the lncRNA HAS2-AS1 both in vitro and in vivo. Modulation of the HAS2-AS1 axis may be a novel approach to attenuate sunitinib-induced vascular dysfunction.

1. Introduction

Cancer-related mortality has been on the decline, partly due to the emergence of molecular targeted therapies. Unfortunately, the success of these drugs, including tyrosine kinase inhibitors (TKIs), has been tempered by a concomitant rise in the prevalence of cancer therapy-related cardiovascular toxicities¹. TKIs target kinases and their downstream pathways hijacked by cancer cells, but these pathways also often play critical roles in the homeostasis of the cardiovascular system.² One such example is sunitinib malate, a multi-receptor TKI widely used to treat solid tumours, including metastatic renal cell carcinoma (RCC),³ gastrointestinal stromal tumours,⁴ and pancreatic neuroendocrine tumours.⁵ Systemic hypertension,⁶ capillary rarefaction,⁷ decreased left ventricular ejection fraction,⁸ and congestive heart failure⁹ are among the most often reported side effects of its use. While the mechanisms of cardiomyocyte (CM) toxicity have been identified,^10,11 much less is known about the impact of sunitinib on vascular dysfunction, which also contributes to heart failure and directly causes vascular diseases such as hypertension and stroke.^12,13 Thus, a better understanding of sunitinib-induced vascular dysfunction and its underpinning mechanisms is critical to mitigate its risk.

Previously considered an inert and static scaffold, the extracellular matrix (ECM) is now known to harbour a variety of signalling regulators involved in both physiologic and pathophysiologic processes.¹⁴ The endothelial glycocalyx (GCX) is a layer of ECM lining the apical surface of endothelial cells (ECs) of which hyaluronic acid (HA), a ubiquitous glycosaminoglycan, is a major constituent. HA is synthesized at the plasma membrane via HA synthases (HAS1-3), with HAS2 being the main enzyme involved.¹⁵ The GCX maintains the vasculature by regulating vascular permeability and microvascular tone, preventing thrombosis, and regulating leukocyte adhesion.^16,17 Treatment of the endothelium with hyaluronidase leads to the loss of the capillary filtration barrier to plasma proteins.^18,19 Degradation of the GCX by hyaluronidase also decreases nitric oxide production, a hallmark of endothelial dysfunction.²⁰ Furthermore, GCX has been demonstrated to have a critical role in regulating cardiovascular diseases.²¹ However, despite its importance in endothelial homeostasis, whether anticancer drugs affect the GCX and thereby cause vascular dysfunction remains undefined.

Breakthroughs in high-throughput sequencing have revealed that only 5% of the human genome is transcribed into protein-coding RNA, the remaining of which consists of non-coding RNAs (ncRNAs). NcRNAs are subdivided into small ncRNAs (<200 nt) such as microRNAs and long non-coding RNAs (lncRNAs) (>200 nt), the latter being less well characterized. On a functional level, lncRNAs have emerged as potent regulators of numerous cellular processes through their ability to interact with RNA, DNA, or protein depending, in part, on their cellular localization.²² Although most of these studies have focused on the contribution of lncRNAs in cellular development and differentiation, emerging evidence suggests an additional contribution of lncRNAs in cardiovascular homeostasis and diseases, including roles in endothelial biology.^23,24 Whether lncRNAs are involved in cancer therapy-induced toxicities remains a major knowledge gap, with the existing few studies focused primarily on CMs^25–29 and not ECs.

To date, functional studies of endothelial lncRNAs are often focused on those conserved in mice for experimental tractability, which is especially challenging as there is poor conservation between mouse and human lncRNAs, necessitating the need for human-based models for studying the biology of lncRNAs. Additionally, there is a strong need for a model that can predict drug-induced vascular toxicity, illuminate the causative mechanisms, and assess potential therapeutics. Although current standard preclinical models such as mice and dogs are indispensable for toxicological evaluations, it is increasingly recognized that existing animal models alone are inadequate for the accurate prediction of drug toxicity in humans due to differences in physiology, metabolism, and molecular functions between species. The advent of human-induced pluripotent stem cell (iPSC) technology presents a valuable opportunity to solve these conundrums by offering a human cell model. In this approach, it is possible to generate a virtually unlimited supply of all the major cell lineages of the human cardiovascular system, including CMs, ECs, smooth muscle cells, and cardiac fibroblasts. These cells functionally and structurally resemble their counterpart primary cells. Harnessing this technology, we have previously highlighted the utility of this approach by using human iPSC-CMs to investigate the molecular mechanisms contributing to cancer drugs-induced cardiac toxicities.^30,31

In the present study, we expand our use of human iPSCs to investigate whether lncRNAs are involved in regulating sunitinib-induced endothelial dysfunction and the underlying mechanisms. We demonstrate the critical role of lncRNA HAS2-AS1 in regulating sunitinib-induced vascular dysfunction. We found that sunitinib down-regulates HAS2-AS1 by hypermethylation, leading to the reduced production of HA and increased expression of ADAM metallopeptidase with thrombospondin type 1 motif 5 (ADAMTS5), an ECM-degrading enzyme, resulting in the degradation of GCX. Notably, we reveal that CRISPR-mediated activation (CRISPRa) of HAS2-AS1, or in vitro high molecular weight hyaluronic acid (HMW-HA) supplementation, protects ECs against sunitinib-induced vascular injury, which is associated with restoration of the GCX. Moreover, our in vivo sunitinib-induced vascular dysfunction mouse model demonstrates the down-regulation of lncRNA Has2os (mouse analogue of HAS2-AS1), and supplementation of HMW-HA in sunitinib-treated mice prevents sunitinib-induced vascular toxicity.

2. Methods

A detailed description of materials and methods can be found in the Supplemental Material.

All animal procedures conformed to the guidelines from either the Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes or the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Mice were anaesthetized with isoflurane at doses of 1%, 2.5%, or 3% as stated within the aforementioned detailed description and were euthanized by cervical dislocation.

The investigation also conformed to the principles outlined in the Declaration of Helsinki. All procedures conformed to the University of Illinois Chicago institutional review board-approved protocol.

3. Results

3.1 Human iPSC-ECs recapitulate sunitinib-induced endothelial dysfunction

Using our chemically defined protocol^32,33 (see Supplementary material online, Figure S1A), we differentiated four healthy human iPSC lines into ECs that showed typical cobblestone features, expressed endothelial markers, including CD31 and VE-cadherin, and could take up acetylated low-density lipoprotein (see Supplementary material online, Figure S1B). To investigate whether these iPSC-ECs can recapitulate sunitinib-induced endothelial dysfunction, we treated these cells with sunitinib followed by downstream characterization and functional assessment (Figure 1A). We found that sunitinib treatment decreased cell viability in a concentration-dependent manner, with 1 µM causing a reduction of ∼20% after 72 h of treatment (Figure 1B). Consistent with our findings, previous studies have shown that 1 µM sunitinib is a clinically relevant dose.³⁴ To further characterize the endothelial injury induced by sunitinib treatment, we assessed the tube formation potential of iPSC-ECs using 0.5 and 1 µM sunitinib. We found that treatment with 1 µM sunitinib significantly impaired iPSC-ECs in their ability to form highly organized networks compared to DMSO-treated cells (Figure 1C), whereas the effects of 0.5 µM sunitinib were negligible, further confirming the deleterious effects of sunitinib. Similarly, cell migration of iPSC-ECs following 1 µM but not 0.5 µM sunitinib treatment was also found to be significantly impaired when compared to DMSO-treated cells (Figure 1D). These effects were also evident in HAECs and MCECs treated with 1 µM sunitinib (see Supplementary material online, Figure S1C–H). We also determined the transendothelial electrical resistance (TEER) across the endothelial monolayer in DMSO- and sunitinib-treated iPSC-ECs and found that only 1 µM sunitinib caused a significant decrease in TEER (Figure 1E), suggesting a potential loss of endothelial barrier integrity, which is consistent with a previous report showing that sunitinib causes increased microvascular permeability in mice.³⁵ Additionally, VE-cadherin localization at the cell membrane was sporadic and internalized, with visible gaps along with increased stress fibres around the cell periphery at both concentrations of sunitinib-treated iPSC-ECs (Figure 1F). Overall, these data demonstrate that human iPSC-ECs represent a valid model for studying the molecular determinants of sunitinib-induced endothelial dysfunction. As we consistently detected significant perturbations of endothelial functions using 1 µM sunitinib, subsequent experiments were performed at this concentration.

Figure 1

iPSCs as a model to study sunitinib-induced endothelial dysfunction. (A) Schematic representation of the study workflow. (B) Determination of cell viability after 72 h of sunitinib (SUN) treatment at different concentrations (0.25–1.5 μM) compared to the DMSO control group using the CellTiterGlo assay (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (C) Determination of the tube formation efficiency of iPSC-ECs following treatment with DMSO, 0.5 or 1 μM sunitinib for 24 h (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (D) The wound healing ability of iPSC-ECs following treatment with DMSO, 0.5 or 1 μM sunitinib for 24 h (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (E) Determination of TEER in an iPSC-EC monolayer following treatment with DMSO or 0.5 and 1 µM sunitinib (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (F) Phalloidin, VE-cadherin, and nuclei immunostaining of iPSC-ECs treated with 0.5 and 1 µM sunitinib for 6 and 24 h. The arrows indicate the accumulation of stress fibers, and weakened adherens junctions in phalloidin staining, and VE-cadherin staining, respectively. All data are represented as the mean ± SD. *P < 0.05; **P < 0.01, ***P < 0.001; ****P < 0.0001.

3.2 Identification of dynamically regulated lncRNAs in sunitinib-treated iPSC-ECs

Next, we sought to identify lncRNAs whose expression levels were affected by sunitinib treatment as some of these differentially regulated lncRNAs may contribute to sunitinib-induced endothelial injury. RNA was extracted from four independent human iPSC-EC lines treated with either DMSO or 1 µM sunitinib for 72 h and subjected to microarray-based lncRNA profiling. We identified 180 differentially expressed lncRNAs in sunitinib-treated iPSC-ECs with a fold change of ±1.5 and a false discovery rate (FDR) < 0.05, comprising 59 up-regulated and 121 down-regulated lncRNAs, respectively (Figure 2A, Supplementary material online, Figure S2A and Table S1). Box-whisker plotting indicated a similar distribution of the data from all eight samples (see Supplementary material online, Figure S2B). Among all differentially regulated lncRNAs, 40% were annotated as intergenic, 28% as natural antisense transcript lncRNAs, and 32% as other lncRNAs (see Supplementary material online, Figure S2C).

Figure 2

Identification of dynamically regulated lncRNAs in sunitinib-treated iPSC-ECs. (A) Workflow of genome-wide lncRNA profiling. (B) Heatmap of the top 10 down-regulated lncRNAs in sunitinib-treated iPSC-ECs. (C) Expression of HAS2-AS1 was verified using RT–qPCR in sunitinib-treated iPSC-ECs (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line, two-way ANOVA). (D) Expression of HAS2-AS1 in iPSC-CMs treated with 1 µM sunitinib for 72 h (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (E) ENCODE HUVEC ChIP-seq signals in regions surrounding HAS2-AS1. (F) Gene expression of HAS2-AS1, MALAT1, DANCR, and GAPDH in subcellular fractions from iPSC-ECs in the absence and presence of sunitinib as measured by qPCR, plotted as percentages in association with nucleoplasm (Nuc) and cytoplasm (Cyt) (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line). (G) Coding probability of HAS2-AS1, XIST, GAPDH, and ACTB as measured by the CPAT webtool. (H) Expression of HAS2-AS1-ORF-Flag in iPSC-ECs transduced with lentivirus expressing HAS2-AS1-ORF-Flag for 48 h. (I) Comparison of human HAS2-AS1 and mouse Has2os loci. (J) Expression of Has2os in MCECs treated with 1 µM sunitinib for 72 h (N = 4 independent experiments, unpaired t-test). All data are represented as the mean ± SD. On the graphs, **P < 0.01; ****P < 0.0001.

We chose to evaluate down-regulated lncRNAs, as we hypothesized that failure to induce these lncRNAs might contribute to the development of sunitinib-induced EC dysfunction. A previous study reported that LINC00460 is up-regulated in clear cell renal cell carcinoma (ccRCC), a cancer type in which sunitinib acts as a first-line anti-cancer drug.³⁶ Therefore, it is plausible that sunitinib reduces the levels of LINC00460. As such, overexpression of LINC00460 as a therapeutic target to mitigate the vascular toxicity of sunitinib may not be an ideal approach since it may induce resistance against sunitinib in ccRCC patients. The lncRNA RP11-463J10.3 is not yet well characterized, and its full-length sequence is not available in the National Center for Biotechnology Information database, suggesting that it may be a pseudogene. Therefore, we decided to study the lncRNA HAS2-AS1, which ranked third among the down-regulated lncRNAs upon sunitinib exposure (Figure 2B). The microarray results of HAS2-AS1 were further verified by reverse transcription–quantitative polymerase chain reaction (RT–qPCR), which confirmed its reduced expression following sunitinib treatment (Figure 2C). Since CMs and fibroblasts are also abundantly present in the heart along with ECs, we also treated iPSC-CMs and human cardiac fibroblasts (HCF) with sunitinib along with HAECs to check whether sunitinib-induced down-regulation of HAS2-AS1 is specific to ECs. We observed significantly reduced expression of HAS2-AS1 in HAECs but not in HCFs upon sunitinib treatment (see Supplementary material online, Figure S2D). Despite exhibiting cardiac damage as shown by reduced levels of ATP (see Supplementary material online, Figure S2E and F), we did not observe reduced expression of HAS2-AS1 in sunitinib-treated iPSC-CMs (Figure 2D), suggesting that reduced HAS2-AS1 upon sunitinib exposure is endothelial specific. Sunitinib is a multi-targeted TKI. To check whether sunitinib down-regulates HAS2-AS1 by inhibiting VEGFR2, we treated iPSC-ECs with ZM 323881, a selective VEGFR2 inhibitor with no activity on VEGFR1, PDGFRβ, FGFR1, EGFR, and ErbB2,³⁷ which did not alter the expression of HAS2-AS1 (see Supplementary material online, Figure S2G), indicating that sunitinib may down-regulate HAS2-AS1 through other receptors.

HAS2-AS1 is located on the plus strand of human chromosome 8 and consists of four exons, with exon 2 sharing sequences complementary to the first exon of the HAS2 gene, located on the opposite strand. ChIP-seq data from HUVECs shows that the HAS2-AS1 locus is also enriched with marks of enhancers (i.e. H3K27ac and H3K4me1 peaks) and active promoters near transcription start sites (H3K4me3) (Figure 2E). To gain insights into the biological role of HAS2-AS1, we first determined its subcellular localization using a cell fractionation assay. The purity of the subcellular fractions was validated by nuclear (Histone H3) and cytoplasmic (GAPDH) markers (see Supplementary material online, Figure S2H). HAS2-AS1 was predominantly detected in the cytoplasm of iPSC-ECs, with a minor fraction in the nucleus. Sunitinib did not affect its localization, indicating that its biological role is mainly in the cytoplasm (Figure 2F). MALAT1 (metastasis associated lung adenocarcinoma transcript 1), a well-known nuclear-enriched lncRNA, and DANCR (differentiation antagonizing non-protein coding RNA), a cytoplasm-enriched lncRNA, were used as controls. Genotype-tissue expression data indicated that HAS2-AS1 was present in cardiovascular and endothelial-enriched tissues (see Supplementary material online, Figure S2I). According to the widely used coding-potential assessment tool CPAT, the coding potential of HAS2-AS1 is low (CPAT coding probability: 0.01), similar to the well-characterized lncRNA XIST (X inactive specific transcript), whereas genes including GAPDH and ACTB were predicted to be protein-coding (Figure 2G). Nonetheless, CPAT analysis indicated the possible presence of a 243 nt ORF within exon 2 of HAS2-AS1. To test whether this ORF is translated into a micropeptide, we generated a lentiviral construct in which a FLAG tag was inserted at the C-terminus of the predicted 243 nt ORF (HAS2-AS1-ORF-FLAG) (see Supplementary material online, Figure S2J). When this construct was transduced into iPSC-ECs, immunoblotting with a FLAG antibody detected the corresponding peptide (Figure 2H), suggesting that this small ORF within HAS2-AS1 is translated, although whether this peptide is functional remains to be tested. Interestingly, there appears to be a mouse ortholog, Has2os, which contains four exons but is only 485 bp long and located on chromosome 15 (Figure 2I). Similar to our observations in sunitinib-treated human iPSC-ECs, we confirmed that sunitinib treatment of MCECs also repressed Has2os (Figure 2J), further supporting the notion that this lncRNA may be involved in sunitinib-induced endothelial dysfunction.

3.3 Sunitinib represses HAS2-AS1 by hypermethylation

To gain insight into the mechanism by which HAS2-AS1 is down-regulated by sunitinib, we analysed the half-life of HAS2-AS1 by inhibiting RNA polymerase II transcription with actinomycin D. The levels of HAS2-AS1 mRNA were unaltered in sunitinib-treated iPSC-ECs compared to DMSO-treated cells, arguing against posttranscriptional regulation of HAS2-AS1 (Figure. 3A). Based on the UCSC Genome Browser, we recognized CpG islands and H3K4Me1, H3K4Me3, and H3K27Ac marks within the promoter and transcript regions of HAS2-AS1 (Figure 2E), indicating a potential regulation of HAS2-AS1 by DNA methylation or histone acetylation. To explore these possibilities, HAS2-AS1 was measured in iPSC-ECs treated with sunitinib alone or in conjunction with either 5-azacytidine (5-AzaC), a DNA methyltransferase (DNMT) inhibitor, or SAHA, a pan-histone deacetylase inhibitor. Interestingly, 5-AzaC but not SAHA eliminated the inhibitory effects of sunitinib on the expression of HAS2-AS1 (Figure 3B). This suggests that increased DNA methylation rather than decreased histone acetylation may be responsible for the blunted expression of HAS2-AS1. Hypermethylation can be caused by increased DNMT-mediated methylation or reduced ten–eleven translocation (TET) dioxygenase-mediated demethylation.³⁸ To address the involvement of DNMTs and TETs in methylating HAS2-AS1, we measured the enzymatic activity of DNMTs and TETs in nuclear extracts of DMSO- and sunitinib-treated iPSC-ECs. DNMT activity was significantly increased upon sunitinib treatment (Figure 3C), whereas TET activity was markedly reduced by sunitinib (Figure 3D), suggesting the involvement of both DNMTs and TETs in regulating HAS2-AS1 methylation. To conclusively verify that sunitinib affects the methylation status of HAS2-AS1, we designed several primer sets targeting the predicted CpG islands located within the promoter and exon 1 region of HAS2-AS1 for methylated DNA immunoprecipitation followed by polymerase chain reaction analysis (MeDIP-PCR) (Figure 3E). Accordingly, we confirmed that the methylation levels of HAS2-AS1 were higher in sunitinib-treated cells than in control cells (Figure 3F).

Figure 3

DNA methylation is involved in the down-regulation of HAS2-AS1. (A) The RNA stability of HAS2-AS1 in DMSO or sunitinib-treated iPSC-ECs was determined in the presence of 5 μg/mL actinomycin D (ActD) (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line). (B) The expression of HAS2-AS1 was measured in sunitinib-treated iPSC-ECs in the presence and absence of 5-AzaC and SAHA (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (C) The activity of DNMTs was determined in iPSC-ECs treated with sunitinib in the presence and absence of 5-AzaC for 72 h (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (D) The measurement of TET activity in iPSC-ECs treated with DMSO and 1 μM sunitinib for 72 h (N = 4 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (E) Representative figure showing the presence of CpG islands. Scheme showing the primer design in the promoter (P2, P1) and transcript regions (R1, R2) of HAS2-AS1. (F) Representative image showing the electrophoresis of the PCR products of selected regions in the promoter and transcript of HAS2-AS1 using MeDIP and input DNA from iPSC-ECs after sunitinib exposure for 72 h. Immunoprecipitation of methylated DNA was verified by PCR using primer pairs targeting the endogenous hypermethylated promoter region of testis-specific histone 2B (TSH2B), which serves as a positive control for the presence of hypermethylation. P1 and P2 represent promoters, and R1 and R2 represent transcript regions of HAS2-AS1. All data are represented as the mean ± SD. On the graphs, *P < 0.05; **P < 0.01; ***P < 0.001.

3.4 HAS2-AS1 is essential for endothelial health

Since HAS2-AS1 was significantly decreased by sunitinib, we sought to explore whether its down-regulation was associated with endothelial dysfunction. To this end, we first used siRNAs to inhibit HAS2-AS1 in iPSC-ECs (Figure 4A). Crucially, inhibition of HAS2-AS1 phenocopied the effects of sunitinib in causing endothelial dysfunction, as revealed by impaired angiogenesis (Figure 4B) and delayed wound healing (Figure 4C). Notably, the proliferation of iPSC-ECs appeared to be unaffected by HAS2-AS1 knockdown, as no significant difference in Ki-67 staining was observed compared to control cells (see Supplementary material online, Figure S3A). Analysis of barrier leakage by TEER also revealed a significant increase in vascular permeability (Figure 4D), consistent with immunofluorescence staining showing increased VE-cadherin internalization and the formation of peripheral stress fibres in iPSC-ECs with HAS2-AS1 inhibition (Figure 4E). Knockdown of HAS2-AS1 in HAECs also led to endothelial dysfunction as indicated by impaired tube formation and delayed wound healing (see Supplementary material online, Figure S3B–D). Collectively, these findings indicate that HAS2-AS1 is necessary for maintaining normal endothelial function and support the hypothesis that sunitinib-induced endothelial dysfunction is linked to the repression of HAS2-AS1.

Figure 4

HAS2-AS1 is an important regulator of endothelial function and may prevent sunitinib-induced endothelial dysfunction. (A) Successful siRNA-mediated knockdown of HAS2-AS1 was confirmed in iPSC-ECs (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test) (B) The tube formation efficiency of iPSC-ECs following knockdown of HAS2-AS1 was measured (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (C) The wound healing ability of iPSC-ECs following knockdown of HAS2-AS1 was determined (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (D) Measurement of TEER in a monolayer of iPSC-ECs following HAS2-AS1 inhibition (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line). (E) Phalloidin, VE-cadherin, and nuclei immunostaining of iPSC-ECs following knockdown of HAS2-AS1. The arrows indicate the accumulation of stress fibers and weakened adherens junctions in phalloidin staining, and VE-cadherin staining, respectively. (F) Schematic representation of the CRISPRa system used for HAS2-AS1 overexpression. (G) Confirmation of ectopic HAS2-AS1 overexpression using two different gRNAs targeting the promoter region of HAS2-AS1 compared to control gRNA (NTC) (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (H) TEER was measured in HAS2-AS1-ECs and CTRL-ECs in the presence and absence of sunitinib (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (I) The tube formation efficiency of HAS2-AS1-ECs and CTRL-ECs treated with DMSO or sunitinib for 24 h (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). All data are represented as the mean ± SD. On the graphs, *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Given that inhibition of HAS2-AS1 leads to endothelial dysfunction, we reasoned that overexpression of HAS2-AS1 in iPSC-ECs may protect against sunitinib-induced endothelial dysfunction. For unclear reasons, our attempts at performing 5′ and 3′ rapid amplification of cDNA ends (RACE) to determine the full-length sequence of HAS2-AS1 for overexpression studies were unsuccessful (data not shown), which is in accordance with a previous study that also failed to perform RACE of HAS2-AS1.³⁹ To circumvent this issue, we opted to use synergistic activation mediator (SAM)-mediated CRISPR activation (CRISPRa) to induce HAS2-AS1 overexpression. The SAM complex comprises three components that drive transcription: a guide RNA (gRNA) incorporating two MS2 RNA aptamers, a catalytically inactive dCas9-VP64 fusion protein, and an MS2-P65-HSF1 activator fusion protein (Figure 4F). Two different gRNAs targeting the promoter region of HAS2-AS1 were designed along with non-targeting control (NTC) gRNAs to generate stable HAS2-AS1-overexpressing iPSC lines. As we found gRNA #2 to have better overexpression efficacy than gRNA #1 (Figure 4G), subsequent experiments were performed using gRNA #2, and iPSC-ECs generated from NTC gRNA and HAS2-AS1 gRNA #2 were categorized as CTRL-ECs and HAS2-AS1-ECs, respectively. In the presence of sunitinib, HAS2-AS1-ECs showed significantly better wound healing capacity than CTRL-ECs, with comparable cell viability between both groups (see Supplementary material online, Figure S3E and F). Likewise, TEER measurement of barrier integrity revealed that HAS2-AS1-ECs were able to maintain barrier integrity upon sunitinib treatment compared to CTRL-ECs, which exhibited increased permeability (Figure 4H). Congruent with these findings, HAS2-AS1-ECs retained their angiogenic capacity compared to CTRL-ECs in the presence of sunitinib (Figure 4I). Thus, these data show that HAS2-AS1 is a critical regulator of vascular homeostasis and CRISPRa-mediated up-regulation of HAS2-AS1 effectively counteracts the detrimental effects of sunitinib on ECs.

3.5 HAS2-AS1 interacts with HAS2 via RNA duplex formation and regulates hyaluronic acid production

Antisense lncRNAs regulate the expression and function of their corresponding protein-coding sense genes, as is the case for HAS2-AS1, which has been shown to regulate its sense gene HAS2 in osteosarcoma cells although the precise mechanisms remain unclear.³⁹ Hyaluronan synthases (HAS) are enzymes that exist as three isoforms (HAS1, HAS2, and HAS3) that are required for synthesizing HA.⁴⁰ To assess whether HAS2-AS1 primarily regulates HAS2 in iPSC-ECs, the expression levels of all three HAS isoforms were determined following HAS2-AS1 inhibition. Only HAS2 was down-regulated upon HAS2-AS1 inhibition (Figure 5A). We also tested the expression levels of multiple neighbouring genes of HAS2-AS1, which might be coregulated, but none of the tested genes showed differential changes upon HAS2-AS1 inhibition (Figure 5A, Supplementary material online, Figure S4A), demonstrating the specificity of HAS2-AS1 in regulating HAS2. HAS2-AS1 inhibition-induced down-regulation of HAS2 was further confirmed at the protein level by immunoblotting (Figure 5B). To evaluate whether HAS2-AS1-mediated down-regulation of HAS2 leads to the reduced synthesis of HA, we performed immunofluorescence staining of HA with biotinylated HA binding protein (HABP), which comprises the G1 domain of aggrecan and linkage proteins that have an affinity to interact with HA domains.⁴¹ As expected, knockdown of HAS2-AS1 led to a significant decrease in the staining intensity of HABP (Figure 5C). Conversely, there was an increase of HAS2 in CRISPRa-mediated HAS2-AS1 overexpression iPSC-ECs (see Supplementary material online, Figure S4B), resulting in a corresponding increased synthesis of HA (see Supplementary material online, Figure S4C), supporting HAS2 as a direct downstream target of HAS2-AS1. Transduction of iPSC-ECs with lentivirus overexpressing HAS2-AS1-ORF-FLAG did not reverse the reduced levels of HABP caused by HAS2-AS1 knockdown (see Supplementary material online, Figure S4D), suggesting that the HAS2-AS1-encoded micropeptide does not play a role in HA synthesis. Indeed, overexpression of the HAS2-AS1 ORF in iPSC-ECs did not affect the levels of HAS2-AS1 and HAS2 under basal conditions (see Supplementary material online, Figure S4E) or restore HAS2 when exposed to sunitinib (see Supplementary material online, Figure S4F). Accordingly, the HAS2-AS1 ORF did not confer protection when overexpressed in sunitinib-treated iPSC-ECs (see Supplementary material online, Figure S4G and H), suggesting that this micropeptide is not involved in the HAS2-AS1/HAS2 axis. Consistent with the effects of HAS2-AS1 knockdown, sunitinib-treated iPSC-ECs exhibited significantly reduced levels of HAS2 at both the mRNA and protein levels (see Supplementary material online, Figure S4I and J). Sunitinib treatment of iPSC-ECs led to decreased HA synthesis, as reflected by reduced HABP staining intensity and ELISA measurement of secreted HA in the supernatant (see Supplementary material online, Figure S4K and L). HA size separation analysis was also performed to evaluate whether sunitinib-associated reduced HA is linked to reduced synthesis or increased degradation of HMW-HA or breaking down HMW-HA into low molecular weight (LMW)-HA. Whereas sunitinib-treated iPSC-ECs showed a reduced presence of HMW-HA compared to control cells, we did not observe a significant increase in LMW-HA (see Supplementary material online, Figure S4M), suggesting an overall reduction in HA synthesis. Sunitinib-induced down-regulation of HAS2 (see Supplementary material online, Figure S4N) and its detrimental effects following knockdown of HAS2 were also evident in HAECs (see Supplementary material online, Figure S4O). Similarly, sunitinib-induced down-regulation of Has2 leading to impaired HA synthesis was also confirmed in MCECs treated with sunitinib for 72 h (see Supplementary material online, Figure S4P and Q).

Figure 5

HAS2-AS1 stabilizes HAS2 via RNA duplex formation and regulates HA production. (A) Relative expression of neighbouring genes in the locus of HAS2-AS1 following knockdown of HAS2-AS1 in iPSC-ECs (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, two-way ANOVA). (B) Expression of HAS2 at the protein level following knockdown of HAS2-AS1 in iPSC-ECs (bar graph, N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (C) The levels of HA were determined in siHAS2-AS1-treated iPSC-ECs using HABP immunofluorescence staining (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (D) Primer positions for the ribonuclease protection assay (top). Electrophoresis image of PCR products showing that the overlapping region of HAS2-AS1/HAS2 was protected from degradation by RNase treatment (bottom). (E) Minimum free energy secondary structure of HAS2-AS1 as predicted by bioinformatics resource (the RNAstructure webserver). The red line indicates the targeted location of antisense DNA oligonucleotide probe 11. (F) qPCR validation of HAS2-AS1 and HAS2 enrichment after HAS2-AS1 RNA pull-down compared to scrambled RNA pull-down (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line). (G) Successful siRNA-mediated knockdown of HAS2 was confirmed in iPSC-ECs (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (H) The tube formation efficiency of iPSC-ECs following knockdown of HAS2 was determined. (I) Measurement of TEER in monolayer of iPSC-ECs following knockdown of HAS2 (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (J) The levels of HA as determined using HABP immunostaining in iPSC-ECs treated with 1 mM 4-MU for 24 h (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (K) The tube formation efficiency of iPSC-ECs treated with 1 mM 4-MU for 24 h was measured (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (L) Measurement of TEER in a monolayer of iPSC-ECs treated with DMSO or 1 mM 4-MU (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). All data are represented as the mean ± SD. On the graphs, *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Exon 2 of HAS2-AS1 overlaps with the first exon of HAS2. This prompted us to explore the possibility that HAS2-AS1 forms an RNA duplex with HAS2, leading to increased stability of the transcribed HAS2 mRNA by protection against ribonuclease degradation. To test this hypothesis, primers were designed to target overlapping and non-overlapping regions of HAS2-AS1 with HAS2 (Figure 5D, Supplementary material online, Table S2). Ribonuclease protection assay confirmed that in the presence of RNase A/T1 mix, which degrades single-stranded RNA, only the overlapping part of both transcripts was protected from degradation (Figure 5D), indicating that HAS2-AS1 and HAS2 mRNA form an RNA duplex. The interaction between HAS2-AS1 and HAS2 mRNA was further validated by RNA pull-down experiments using biotinylated triethylene glycol antisense DNA oligonucleotides targeting multiple regions of HAS2-AS1. Of the 11 unique probes that we designed (see Supplementary material online, Table S2), probe #11 was found to be highly enriched for HAS2-AS1. Using the RNAstructure Web Server,⁴² this probe was found to have a low probability of internal base pairing while displaying a strong affinity for the targeted region. We then performed RNA pull-down using this probe and found significant enrichment of HAS2 compared to the negative control probe based on a scrambled sequence that displays no affinity in the genome (Figure 5E and F), further confirming the RNA–RNA interaction between HAS2-AS1 and HAS2.

To verify the biological relevance of reduced HA as a direct contributor to the endothelial dysfunction seen in both sunitinib-treated and HAS2-AS1 knockdown iPSC-ECs, we artificially reduced HA levels in iPSC-ECs by both genetic (siRNA targeting HAS2) and pharmacological (4-methylumbelliferone, 4-MU, an inhibitor of HA synthesis) methods. As predicted, knockdown of HAS2 (Figure 5G) led to impaired tube formation (Figure 5H) and loss of barrier integrity, as assessed by TEER (Figure 5I). Similarly, iPSC-ECs treated with 4-MU exhibited a dose-dependent decrease in cell viability (see Supplementary material online, Figure S4R), reduced HABP staining intensity reflecting impaired HA synthesis (Figure 5J), impaired tube formation (Figure 5K), and increased vascular permeability (Figure 5L). To explore the potential of restoring HA as a means to counteract sunitinib-induced endothelial dysfunction, iPSC-ECs were treated with sunitinib with or without HMW-HA supplementation. Importantly, treatment with HMW-HA successfully alleviated the detrimental effects of sunitinib on iPSC-ECs, as indicated by restored tube formation, improved wound healing, and maintenance of barrier integrity (see Supplementary material online, Figure S5A–C). These results suggest that sunitinib induces endothelial dysfunction in a HAS2-AS1/HAS2/HA synthesis-dependent manner and confirm the vascular protective role of HA against sunitinib.

3.6 The HAS2-AS1/HAS2 axis is required for maintaining endothelial glycocalyx integrity

Disruption of the GCX has been observed in multiple diseases, including diabetes,⁴³ inflammatory conditions,⁴⁴ and chronic kidney disease.⁴⁵ Since our results have shown that perturbation of the HAS2-AS1/HAS2 axis leads to decreased HA, an essential constituent of the GCX, we reasoned that HA loss might compromise the integrity of the GCX, leading to endothelial dysfunction, as seen in sunitinib-treated iPSC-ECs. To test this hypothesis, we investigated the intactness of GCX in sunitinib-treated iPSC-ECs using wheat germ agglutinin (WGA) staining as previously described.⁴⁶ Since some studies have demonstrated that ECs cultured in vitro lack GCX,^47–49 we took advantage of a recent development that utilizes cyclosporin A to improve the functional maturation of iPSC-ECs, allowing for the formation of a luminal GCX in vitro.⁵⁰ To this end, cyclosporin A was added from Day 3 of iPSC-EC differentiation. We confirmed that sunitinib treatment also led to repression of HAS2-AS1 and its downstream target HAS2 in these cyclosporin A-treated iPSC-ECs (see Supplementary material online, Figure S6A), similar to our earlier results. As expected, luminal GCX thickness was significantly reduced by sunitinib treatment (Figure 6A). Consistent with previous reports that loss of the GCX is associated with endothelial activation and inflammation,⁵¹ we also observed that several inflammatory and endothelial activation-associated genes, including E-selectin, VCAM-1, and ICAM-1, were significantly up-regulated in sunitinib-treated cells (Figure 6B). Notably, supplementation with HMW-HA, which we have shown to protect iPSC-ECs against sunitinib, led to the preservation of luminal GCX thickness in the presence of sunitinib (Figure 6A), confirming an association between the GCX integrity and endothelial homeostasis. Likewise, CRISPRa of HAS2-AS1 also significantly blunted the detrimental effects of sunitinib on the intactness of GCX (Figure 6C). Taken together, these data suggest that sunitinib-induced endothelial dysfunction may be associated with impaired GCX integrity.

Figure 6

HAS2-AS1 maintains the integrity of endothelial GCX through HA-mediated regulation of ADAMTS5. (A) Representative side-view confocal images of iPSC-ECs stained for WGA and DAPI after 72 h of treatment with 1 µM sunitinib in the presence and absence of 300 µg/mL of HMW-HA. Endothelial cells GCX integrity was evaluated by measuring GCX thickness via WGA staining using ImageJ (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (B) Relative RNA expression of vascular-associated inflammatory markers in iPSC-ECs treated with 1 µM sunitinib for 24 h, as measured by RT–qPCR (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, two-way ANOVA). (C) Representative side-view confocal images of CTRL-ECs or HAS2-AS1-ECs stained for WGA and DAPI after 72 h of treatment with DMSO or 1 µM sunitinib (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). (D) Workflow of RNA-seq performed on both sunitinib-treated and siHAS2-AS1 iPSC-ECs along with their respective controls. (E) Expression of the top 20 statistically significant differentially regulated genes among 40 genes that were identified as common between both sunitinib-treated and siHAS2-AS1 iPSC-ECs. (F) STRING analysis was performed using the 40 differentially regulated genes. Representative image shows only genes with strong interactions, e.g. HAS2 and ADAMTS5. P < 0.05. (G) Expression of ADAMTS5 was validated in iPSC-ECs treated with sunitinib for 72 h (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, unpaired t-test). (H) The expression of ADAMTS5 was validated in iPSC-ECs by immunofluorescence in both sunitinib-treated and siHAS2-AS1 iPSC-ECs. (I) GCX thickness of siHAS2-AS1 iPSC-ECs in the presence or absence of 1 µM ADAMTS5 inhibitor (ADAMTS5i) was determined (N = 3 iPSC lines, each datapoint represents the average of three independent experiments per line, one-way ANOVA). All data are represented as the mean ± SD. On the graphs, *P < 0.05; **P < 0.01; ***P < 0.001, ****P < 0.0001.

To discover the potential mechanisms by which sunitinib leads to impaired GCX in a HAS2-AS1-dependent manner, we performed two independent sets of RNA-sequencing (RNA-seq) experiments on sunitinib-treated iPSC-ECs and HAS2-AS1 knockdown iPSC-ECs along with their respective controls (Figure 6D). Upon overlapping the results of both datasets, 40 common differentially expressed genes (DEGs, 23 up- and 17 down-regulated genes) were identified (see Supplementary material online, Table S3), with HAS2 among one of the down-regulated genes as expected (Figure 6E). These DEGs were enriched in cellular compartments related to hyaluronan cable, extracellular region, platelet-derived growth factor complexes (see Supplementary material online, Figure S6B), and molecular functions related to receptor–ligand and cytokine activity (see Supplementary material online, Figure S6C). Gene enrichment analysis also revealed the association of network pathways related to hyaluronan metabolism, hyaluronan biosynthesis and export, and hyaluronan uptake and degradation (see Supplementary material online, Figure S6D). We then performed interaction analysis on these 40 DEGs using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database and found a significant interaction between HAS2 and one of the up-regulated genes ADAMTS5 (Figure 6F). ADAMTS5 belongs to the ADAMTS family of proteinases that are secreted and responsible for degrading ECM proteins such as proteoglycans and has been shown to contain HA-binding motifs within its disintegrin domain.^52,53 Previous studies have reported that the loss of HA or hyaluronidase treatment is accompanied by the activation of several metalloproteases, including ADAMTS5.^54–56 To evaluate whether increased ADAMTS5 is associated with sunitinib-induced disruption of the GCX, we first validated our RNA-seq results by RT–qPCR and confirmed that ADAMTS5 is increased in sunitinib-treated iPSC-ECs (Figure 6G, cyclosporin A iPSC-ECs, Supplementary material online, Figure S6E). Immunofluorescence staining revealed that ADAMTS5 was localized in the nucleus of iPSC-ECs treated with either DMSO or siControl (Figure 6H). However, after treatment with either sunitinib or siHAS2-AS1, we observed increased expression of ADAMTS5 along with diffusion into the cytoplasm (Figure 6H), suggesting that ADMATS5 may degrade the components of GCX by being secreted into the cytoplasm. Consistent with our findings, others have reported the localization of ADAMTS5 in the cytoplasm of cells in a diffuse pattern.⁵⁷ Interestingly, the disruption of the GCX seen in sunitinib-treated or HAS2-AS1 knockdown iPSC-ECs was partially alleviated in the presence of ADAMTS5 inhibitor (ADAMTS5i), supporting the potential role of ADAMTS5 in destabilizing the GCX (Figure 6I, Supplementary material online, Figure S6F and G). These results suggest that sunitinib-induced endothelial dysfunction arises from the combined down-regulation of HA/GCX maintenance and synthesis pathways and the activation of HA/GCX proteases such as ADAMTS5.

3.7 HMW-HA supplementation alleviates vascular dysfunction in sunitinib-treated mice

Our in vitro data showed that sunitinib induces endothelial dysfunction by suppressing the HAS2-AS1/HAS2 axis, leading to disrupted GCX integrity because of reduced HA synthesis and increased ADAMTS5 activity. To examine whether this proposed mechanism holds true in vivo, we took advantage of a previously reported mouse model of sunitinib-induced vascular dysfunction in which C57BL/6 mice were treated with sunitinib at 40 mg/kg/day for 3 weeks (Figure 7A), which resembles the typical clinical regimen.³⁵ To confirm the presence of vascular dysfunction in sunitinib-treated mice, we first measured coronary flow reserve (CFR), which is an indicator of coronary microvascular function. Previous studies have shown that impaired CFR serves as a predictor of cardiovascular diseases⁵⁸ and is strongly correlated with increased mortality in patients with cardiovascular comorbidities.⁵⁹ As expected, sunitinib-treated mice showed a significant decrease in CFR compared to vehicle-treated mice (Figure 7B), associated with the inability to accommodate haemodynamic changes that occur during the hyperemic state (see Supplementary material online, Figure S7A), supporting the notion that sunitinib treatment leads to vascular dysfunction. Moreover, using isolectin-IB4 to stain vasculature in the hearts of mice, we observed significantly reduced vascular density in sunitinib-treated mice compared to control mice (Figure 7C), further confirming the presence of vascular dysfunction in the heart. The mRNA levels of Klf2 were also found to be down-regulated in isolated cardiac ECs of sunitinib-treated mice compared to controls, further confirming endothelial-specific dysfunction^60,61 (Figure 7D). Of note, we observed no significant changes in hypertension, which is in agreement with results from this previously reported mouse model of sunitinib-induced vascular dysfunction (see Supplementary material online, Figure S7B).³⁵ Since we observed reduced HA and GCX destabilization following sunitinib treatment in vitro, we next assessed the HA levels and intactness of GCX in both aortas and hearts isolated from vehicle- and sunitinib-treated mice. Immunofluorescence staining of aortic cross-sections from sunitinib-treated mice revealed reduced HABP staining that overlapped with CD31 staining, indicating that sunitinib primarily decreases the intactness of GCX in ECs (see Supplementary material online, Figure S7C and D). This was further supported by staining for WGA, which binds to all components of the GCX, including HA, heparan sulfates (HS), and sialic acids, which showed a significantly reduced presence upon sunitinib treatment (Figure 7E, Supplementary material online, Figure S7E). In line with these results, we observed similar findings in the blood vessels of cardiac cross-sections from sunitinib-treated mice in which the endothelial GCX appeared to be degraded compared to controls (see Supplementary material online, Figure S8A). Conversely, the GCX of PDGFRα⁺ fibroblast-like cells appeared to be unaffected (see Supplementary material online, Figure S8B). To determine whether sunitinib-induced GCX degradation leads to increased vascular permeability, we quantified the leakage of Evans Blue dye (EBD) into the aortic tissue of mice. Sunitinib-treated mice showed a significantly increased amount of EBD in aortic tissue compared to vehicle-treated mice (Figure. 7F). Moreover, we examined the levels of has2os in the mouse aorta and heart to explore its functional and disease relevance. As shown in Figure 7G and H, the levels of endothelial Has2os were significantly reduced in sunitinib-treated mice compared to control mice, which agrees with our in vitro findings that sunitinib reduces HAS2-AS1 in iPSC-ECs (Figure 2B) and Has2os in MCECs (Figure 2J).

Figure 7

Protective effects of HMW-HA supplementation in a mouse model of sunitinib-induced vascular dysfunction. (A) Schematic representation of HMW-HA supplementation in sunitinib-treated mice. (B) Representative ultrasound tracings of dilated (induced with 2.5% isoflurane) and basal (with 1% isoflurane) coronary flow after 21 days of treatment. Quantification of coronary flow reserve (CFR) (dilated/basal flow) in sunitinib-treated mice in the presence and absence of HMW-HA and corresponding vehicle-treated mice (N = 8, one-way ANOVA). (C) Representative staining images of mouse heart cross-sections with isolectin-IB4 and DAPI. Capillary density was evaluated by quantifying IB4⁺ tubular structures per square millimetre (N = 5 mice/group, one-way ANOVA) (D) Expression of Klf2 in isolated cardiac ECs of sunitinib treated mice (N = 6, unpaired t-test). (E) Representative staining images (i) of mouse aorta cross-sections with WGA, DIO (3,3′-dioctadecyloxacarbocyanine perchlorate), and DAPI. (ii) An enlargement of the white rectangular area from (i). The arrows above the nucleus denotes the GCX; the arrow below the nucleus side indicate the internal elastic lamina (IEL). (iii) and (iv) show the representative images stained with WGA and cell membrane staining with DIO, respectively. (F) Quantitation of EBD extravasation in mouse aortic tissues isolated from sunitinib-treated mice in the presence and absence of HMW-HA and corresponding vehicle-treated mice (N = 5, one-way ANOVA). (G) Expression of Has2os in RNA isolated from mouse aortas treated with sunitinib (N = 5, unpaired t-test). (H) Expression of Has2os in cardiac ECs of sunitinib-treated mice (N = 5, unpaired t-test). All data are represented as the mean ± SD. On the graphs, *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

To further extrapolate our in vitro results demonstrating the protective effects of HMW-HA against sunitinib-induced endothelial dysfunction, sunitinib-treated mice were concurrently given intraperitoneal injections of HMW-HA (30 mg/kg/day) throughout the entire treatment duration (Figure 7A). Indeed, the impairment of CFR seen in sunitinib-treated mice was successfully alleviated by HMW-HA compared to mice receiving only vehicle (Figure 7B). Similarly, the vascular density was significantly improved by HMW-HA compared to mice receiving sunitinib alone (Figure 7C). Of note, HMW-HA by itself does not appear to increase vascular density. Interestingly, this improvement was coupled to intact endothelium GCX in both the aorta (Figure 7E) and blood vessels of the heart (see Supplementary material online, Figure S8A). Additionally, we observed reduced EBD leakage (Figure 7F) in aortic tissues of HMW-HA but not vehicle-treated mice, further affirming the protective role of HMW-HA against sunitinib.

4. Discussion

Vascular toxicities are the second most common cause of mortality in cancer survivors.⁶² Among cancer treatment modalities, TKIs, including sunitinib, exhibit substantial detrimental effects on the vasculature.^63,64 In this study, we present evidence using both in vitro and in vivo models for the dynamic regulation of lncRNA HAS2-AS1 in mediating sunitinib-induced vascular toxicity. Using human iPSC-ECs as a model of sunitinib-induced vascular dysfunction, we revealed that sunitinib led to the reduced expression of the lncRNA HAS2-AS1, which was dependent on increased DNA methylation. We show that the inhibition of HAS2-AS1 in iPSC-ECs caused endothelial dysfunction, whereas CRISPRa-mediated HAS2-AS1 overexpression successfully reversed sunitinib-induced vascular toxicity. Our results confirm that the loss of HAS2-AS1 led to the decreased expression of its sense gene HAS2, which is required for the synthesis of HA, an essential component of endothelial GCX. Accordingly, degradation of endothelial GCX was a consequence of sunitinib-treatment, which was partially restored by either HAS2-AS1 overexpression or supplementation with HMW-HA. Importantly, using a mouse model of sunitinib-induced vascular dysfunction, we confirmed that sunitinib treatment resulted in impaired microvascular dysfunction along with reduced has2os and degraded GCX within the aorta, effects that were successfully alleviated by HMW-HA supplementation.

We and others have previously demonstrated the use of iPSCs to investigate the cardiotoxic effects of cancer drugs.^30,31,65,66 In this study, we decided to use 1 µM sunitinib for lncRNA profiling, although the C_max of sunitinib is ∼0.2 µM because of the following considerations: (i) sunitinib is taken daily for 3–4 weeks, which may lead to accumulation, causing higher exposure at a steady state as opposed to C_max; (ii) sunitinib is poorly soluble in water, hence requiring a higher concentration when used in cell culture (aqueous) media; and (iii) several studies have shown the most accurate prediction of toxicity when using in vitro concentrations up to 30× the in vivo efficacious dose.^67,68 This concentration of sunitinib is also in line with previous in vitro studies.^34,69 Based on this concentration of sunitinib that causes endothelial dysfunction but without excessive cytotoxicity, we focused on HAS2-AS1, which was down-regulated. HAS2-AS1 has been studied in some cancer cell lines, including glioma^70,71 and epithelial ovarian cancer cells.⁷² However, its role in the endothelium, especially in the context of sunitinib-induced vascular injury, remains unknown. Our study demonstrates HAS2-AS1 as a novel regulator of endothelial health and may help to prevent sunitinib-induced vascular dysfunction. As multiple lncRNAs were identified in our profiling to be differentially expressed upon sunitinib exposure, future work will be required to investigate additional lncRNAs that may be involved in sunitinib-induced vascular dysfunction. Additionally, it should be noted that one limitation of this study is that only healthy but not patient-specific iPSC lines were used to explore the role of lncRNAs in sunitinib-induced vascular dysfunction.

Epigenetic regulation is a fundamental mechanism that can influence the expression of lncRNAs.⁷³ In this study, we observed that sunitinib treatment led to increased methylation of HAS2-AS1, which is coordinated by increased DNMT and reduced TET activities, demonstrating a previously undescribed mechanism of HAS2-AS1 regulation. Importantly, knockdown of HAS2-AS1 phenocopied the detrimental effects of sunitinib on iPSC-ECs, confirming our initial hypothesis that sunitinib-mediated repression of HAS2-AS1 contributes to endothelial dysfunction. Conversely, CRISPRa-mediated HAS2-AS1 overexpression was found to alleviate the detrimental effects of sunitinib. With the increasing maturity of CRISPR technology, this approach of activating endogenous lncRNAs may represent an attractive approach in cardio-oncology in the foreseeable future.

Cis-regulation mediated by lncRNA genes can be achieved through various molecular mechanisms.⁷⁴ We demonstrated that HAS2-AS1 cis-regulates its sense gene HAS2 by forming an RNA duplex increasing the latter’s stability, a mechanism that has previously been described in cancer cells but never before in ECs.⁷⁵HAS2 is responsible for synthesizing HA, a ubiquitous glycosaminoglycan that, along with HS, constitutes most of the GCX. The GCX has emerged as a crucial target in cardiovascular diseases.⁷⁶ Degradation of GCX, which is required in vascular integrity and homeostasis, where it regulates mechanotransduction, vascular permeability, coagulation, and inflammation, has been linked to various chronic diseases, including diabetes,^43,77 atherosclerosis,⁷⁸ hypertension,⁷⁹ and sepsis.⁸⁰ To date, there are no studies linking GCX perturbation to cardiotoxicities associated with cancer drugs. Our results suggest that mechanistically, sunitinib disrupts the HAS2-AS1/HAS2 axis, leading to inadequate HA synthesis and ultimately compromising the GCX integrity, leading to endothelial dysfunction. Transcriptomic profiling also revealed the increased expression of genes such as ADAMTS5 that may be involved in ECM degradation. Importantly, both CRISPRa-mediated HAS2-AS1 overexpression and HMW-HA supplementation were found to preserve the GCX in the presence of sunitinib, further confirming the importance of GCX as an important mediator of sunitinib-induced endothelial dysfunction. These results posit the maintenance of GCX as an actionable therapeutic strategy that can potentially be targeted in the future for preventing cancer drug-associated vascular toxicities.

Compared to protein-coding genes, lncRNAs are often not well conserved between humans and mice. In the case of HAS2-AS1, we detected the down-regulation of a mouse homolog, has2os, which also consists of four exons and is located on the opposite strand of has2 in sunitinib-treated MCECs. Using a previously reported mouse model of sunitinib-induced vascular dysfunction,³⁵ we successfully confirmed that sunitinib treatment for 3 weeks led to microvascular dysfunction, as indicated by reduced CFR. Notably, we also observed reduced expression of has2os within the aorta and heart along with the presence of degraded GCX, corroborating our in vitro results. These results are in accordance with studies performed in both mice and humans linking impaired endothelial GCX to disrupted CFR,^81,82 consistent with the role of HA as a primary component of the endothelial GCX. For example, a recent study reported that impaired endothelial GCX led to increased arterial stiffness resulting in disrupted CFR.⁸³ Delivery of HMW-HA in vivo successfully restored CFR and led to partial recovery of the GCX, supporting the notion that adequate HA is essential for maintaining the integrity of the GCX to support endothelial homeostasis. Of note, the mouse model used in this study did not exhibit hypertension, which is among one of the most significant side effects of sunitinib clinically, suggesting fundamental differences between mice and humans in response to sunitinib exposure. Overall, our results indicate that the HAS2-AS1/HAS2/HA synthesis axis is an important regulator of sunitinib-induced vascular toxicity and that targeting this axis may represent an actionable therapeutic strategy for alleviating the deleterious effects of sunitinib on the cardiovascular system. However, given that HAS2 has been reported to be enriched in selected cancers, although it appears not to be the case for RCC based on our analysis of The Cancer Genome Atlas (see Supplementary material online, Figure S9), one could specifically restore HAS2 or HA levels in the endothelium and thus reduce the risks of collateral effects in tumour cells. For example, a recent study revealed the use of nanoparticles to deliver CRISPR–Cas9 plasmid DNA to the adult vascular endothelium for genome editing,⁸⁴ which could potentially be adapted to mediate HAS2-AS1/HAS2 activation in a controlled manner.

In conclusion, sunitinib exposure leads to decreased HAS2-AS1, which is required for HAS2 stabilization. Perturbation of the HAS2-AS1/HAS2 axis negatively affects the synthesis of HA, which is needed for maintaining GCX integrity, ultimately resulting in endothelial dysfunction. Thus, restoration of the GCX and, in a broader context, targeting the ECM may represent an attractive option for preventing sunitinib-induced vascular dysfunction.

Supplementary material

Supplementary material is available at Cardiovascular Research online.

Authors’ contributions

Conceptualization: S.B.N., W.H.L., and S.G.O. Methodology and investigation: S.B.N., J.J., G.Y., Z.H., Y.C., Y.K., K.G., and N.Y.S. Bioinformatics: C.L. and N.Y.S. Discussion: S.B.N., S.B.O., S.P., J.R., W.H.L., and S.G.O. Supervision: S.B.O., W.H.L., and S.G.O. Writing: S.B.N., S.B.O., W.H.L., and S.G.O.

Acknowledgements

We would like to thank Dr Irena Levitan for helpful discussion.

Funding

This work was supported by grants from the National Institutes of Health (R00 HL130416 and R01 HL148756 to S.G.O., R01 HL162584 to S.P., T32 HL007829 to J.J); American Heart Association (917176 to Z.H, 23POST1029855 to G.Y.); the Research Grants Council of Hong Kong, the Chinese University of Hong Kong (CUHK) and the Lui Che Woo Institute of Innovative Medicine of Hong Kong (CUHK 24110822, 2020.035, PIEF/Ph2/COVID/08, Improvement on Competitiveness in Hiring New Faculties Funding Scheme to S.B.O.).

Data availability

All data are available in the main text or the supplementary materials. Sequencing data files have been deposited in the Gene Expression Omnibus database under the accession number GSE216454 and will be made publicly available upon acceptance of the manuscript.

Author notes

Sarath Babu Nukala, Jordan Jousma, Gege Yan and Zhenbo Han contributed equally to this work.

Conflict of interest: None declared.