HuR-mediated posttranscriptional modication of Cx40 and coronary microvascular dysfunction in type 2 diabetes

Diabetic patients with coronary microvascular disease (CMD) exhibit higher cardiac mortality than patients without CMD. However, the molecular mechanism by which diabetes promotes CMD is poorly understood. RNA-binding protein HuR is a key regulator of mRNA stability and translation of many genes, and there is growing evidence showing the potential role of HuR in cardiovascular disease. In this study, we investigated the role of HuR and its target genes in the development of CMD in type 2 diabetic mice. Type 2 diabetes was induced in male mice by a high-fat diet combined with a single injection of low-dose streptozotocin. We assessed coronary ow velocity reserve (CFVR, a determinant of coronary microvascular function) in vivo and isolated cardiac endothelial cells (CECs) from those mice for in vitro experiment. Coronary endothelial function was evaluated in the 3rd order of coronary arteries using a wire myograph. Human CECs from 4 control subjects and 4 diabetic patients were purchased from the company.

The age was matched between diabetic or transgenic mice and their control mice. Male mice were used in this study due to the difference in the onset of hyperglycemia and diabetic complications between male and female mice.

Isolation of mouse cardiac endothelial cells (CECs)
Mouse CECs were isolated using a method previously described [11,12,18,35]. Brie y, after ushing blood from the heart, the heart was dissected, minced, and incubated with M199 containing 1 mg/ml collagenase II and 0.6 U/ml dispase II for 1 h at 37°C. The digested material was collected and incubated with magnetic beads that were prepared as follows: Dynabeads® Sheep Anti-Rat IgG were incubated with rat anti-mouse CD31 monoclonal antibody (1 μg/ml) at 4°C overnight. The cell suspension was incubated with beads for 1 h at 4°C, and then CECs were captured and isolated by the Dynal magnet (Thermo Fisher Scienti c, MA, USA). The purity of the CEC population in cells isolated from hearts was tested by DiI-acLDL (Thermo-Fisher Scienti c) uptake and Bandeiraea Simplicifolia lectin-FITC (BS-l, Sigma Aldrich, Inc. MO, USA) or CD144 staining (Additional File 3: Fig. S2). E cient isolation yields approximately 10 4 cells from one heart, with over 80% purity. Western blot and real-time PCR were conducted with freshly isolated CECs from mice. For the immunohistochemistry experiment, we cultured CECs after isolation, and experiments were performed within 5 days without passing the cells.

Human CECs
Human CECs from 4 control and 4 T2D patients were purchased from commercial suppliers (Additional File 1: Materials) and cultured in EC media composed of M199 supplemented with 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, 20 μg/ml ECGS, and 16 U/ml heparin. All experiments were conducted before passage 10.
Isolation of mouse cardiac myocytes and aortic smooth muscle cells Mouse cardiac myocytes (CMs) were collected after removing ECs from the digested materials of the hearts. After removing ECs, the majority of cells in the digested material are cardiac myocytes; however, other types of cells (i.e., smooth muscle cells [SMCs] and broblasts) might be present in the samples at a very small percentage. The samples of aortic SMCs were obtained from an aorta after removing ECs by gently scrubbing the inner layer of the aortic lumen using a cotton tip.
Coronary ow velocity reserve measurement Coronary ow velocity reserve (CFVR) was used to assess coronary microvascular function [18], instead of coronary ow reserve, because of the di culty in precisely measuring coronary arterial diameter in mice [39]. Coronary blood ow velocity (CFV) was measured using a Vevo 2100 system (FUJIFILM Visual Sonics, Inc. Toronto, Canada. Additional File 3: Fig. S3). Mice were anesthetized with iso urane and kept on the heating pad at 37°C. The resting level of CFV was obtained at 1% iso urane. CFVR was de ned as maximal hyperemic CFV (induced by 2.5% iso urane) divided by resting CFV (1% iso urane) [18,40].
Each experiment was completed within 40 min, and the heart rate was kept above 400 bpm. If the procedure took a longer time, or the heart rate was dropped lower than the criteria, the data was eliminated without analysis.

Immuno uorescence experiment
The evaluation of capillary density and EC apoptosis in the left ventricle (LV) were conducted as described previously [11,12,18]. Brie y, the heart was dissected, embedded in OCT compound, frozen in 2-methylbutane precooled with liquid nitrogen, and then kept at -80°C until being sectioned. Sections (6 µm in thickness) were xed in 4% formaldehyde for 5 min, blocked with 5% BSA for 30 min, and incubated with Bandeiraea Simplicifolia lectin-FITC (BS-l, Sigma Aldrich, Inc.) for 30 min. BS-l was used to probe the terminal β-galactosyl saccharides associated with endothelial cells on the surface of arterioles and venules as well as capillaries. Apoptotic cells were detected using a TUNEL assay (an in situ cell death detection kit, Roche).
After isolation of mouse CECs, cells were stained with HuR antibody and followed by anti-mouse Alexa488. The images were captured with a Nikon Eclipse Ti-E 3D Deconvolution microscope (Nikon Corp. Tokyo, Japan) with a 20x objective lens (for EC apoptosis and capillary density) or 60x objective lens (for HuR staining) in a blinded fashion. The uorescence intensities were calculated using ImagePro-PLUS 7.0 software (Media Cybernetics Inc. MD, USA).

Western blot analysis
Protein levels were analyzed using SDS-PAGE separation and electrophoretic transfer to nitrocellulose membranes. Primary antibodies used in this study are listed in Additional File 1: Materials.
Real-time PCR mRNA from mouse CECs was isolated using a miRNeasy Mini Kit (QIAGEN, CA, USA), and cDNA was made by RT 2 First Strand Kit (QIAGEN). We chose 92 genes (including Actb and Gapdh) that are highly expressed in ECs and play crucial roles in endothelial functions for analysis by real-time PCR, including a)  Table S3 for the gene list). The custom PCR plates were made by QIAGEN based on the selected genes (SABIO Number CAPA38128-6:CLAM25240). One 384-well plate includes quadruplicate wells for one gene (for the gene of interest and internal control) and replicates genomic DNA controls, reverse-transcription controls, and positive PCR controls. Primer sets used for the PCR plate are authenticated by the company. Real-time PCR was conducted using the CFX384 Touch Real-Time PCR Detection System (Bio-Rad Laboratories, CA, USA). GAPDH was used as an internal control. The transcript levels of the Gene of Interest were quanti ed according to the cycle threshold (ΔCt) method. Ct values > 35 were not included in the analysis and were considered as negative. Note that the primer set for Elavl1 (HuR) on the plate (Product # PPM30921A, QIAGEN) detects exon 5, not exon 2; therefore, real-time PCR was repeated using an exon 2-speci c primer (Additional File 2: Table S2).

Ribonucleoprotein immunoprecipitation (RIP)
To assess the association of endogenous HuR protein with endogenous Cx40 mRNA, immunoprecipitation (IP) of ribonucleoprotein complexes was performed as previously described [25] (Additional File 3: Fig. S4). Mouse CECs were isolated and lysed with lysate buffer (100 mM KCl, 5 mM MgCl 2 , 10 mM HEPES [pH7.0], 0.5% Igepal, 1 mM DTT, 1% protease inhibitor cocktail, 1% phosphatase inhibitor cocktail, 100 U/ml in RNase free water). Prior to RIP, the IP matrix was conjugated with HuR antibody or IgG, and cell lysate was incubated with IP matrix overnight. RNA in IP materials was used for reverse transcription, followed by real-time PCR analysis (Fig. 4). The data of Cx40 mRNA bound to HuR protein was normalized by Cx40 mRNA bound to IgG.

Isometric tension measurement in coronary arterial ring
Isometric tension measurement in coronary arteries (CAs) was performed as described previously [11,35,36]. Brie y, third-order small CAs were dissected from the hearts and then cut into 1-mm segments. The CA rings were mounted on a myograph (DMT-USA, Inc. MI, USA) using thin stainless wires (20 μm in diameter), and the resting tension was set at 100 mg. CAs were allowed to equilibrate for 45 min with intermittent washes every 15 min. After equilibration, each CA ring was contracted by treatment with PGF 2α to generate a similar contraction level in all groups. The concentration of PGF 2α used for precontraction was 5.34 ± 0.19 in Wt and 5.63 ± 0.16 in Cx40 -/mice (shown in -log M). The diameter of the vessels was 134.6 ± 7.1 µm in Wt and 136.1 ± 5.4 µm in Cx40 -/mice. There was no signi cant difference in either PGF 2α concentration or vessel diameter between Wt and Cx40 -/mice. Acetylcholine (ACh) or sodium nitroprusside (SNP, an NO donor) was administrated in a dose-dependent manner (1 nmol/l to 100 μmol/l). The degree of relaxation was shown as a percent of PGF2α-induced contraction.

Tube formation assay in human CECs
We used Cx40 adenovirus (Cx40-Adv) to overexpress the Cx40 gene [11]. HuR downregulation was achieved by HuR siRNA transfection (Santa Cruz Biotechnology Inc. Dallas, TX). Human control CECs (10 5 cells) were seeded on a 3 cm plate, and Control-or Cx40-Adv was added to the cells at the titer of 100 pfu/cell on the following day. Twenty-four hours later, the viruses were washed, and cells were transfected with control siRNA or HuR siRNA at 100 nM using lipofectamine 3000 reagent (Thermo Fisher Scienti c). Speci c protein knockdown was veri ed with Western blotting 48 hours after transfection (Additional File 3: Fig. S5). For tube formation assay [18], cells were detached, and 4×10 4 cells were seeded on the Matrigel-coated 4-well chamber. Twenty-four hours after plating cells, 4 microscopic elds, selected at random, were photographed using an EVOS FL Auto Cell Imaging System with 4x objective lens (Thermo Fisher Scienti c) in a blinded fashion. Meshes number, total meshes area, junction number, segments number, and total segments length were analyzed using Angiogenesis Analyzer in NIH ImageJ 1.51k software.

Cytosolic ROS measurement in human CECs
HuR and Cx40 were downregulated using HuR siRNA or Cx40 siRNA, respectively (Santa Cruz Biotechnology Inc.). Human control CECs (2 x 10 4 cells) were seeded on a 4-well glass chamber and then transfected with control-, HuR-or Cx40-siRNA at 100 nM using lipofectamine 3000 reagent. Cytosolic ROS was detected using the uorescent probe dihydroethidium (DHE). Cells were preloaded with 50 μmol/l DHE for 30 min before capturing images. Cytosolic DHE exhibits blue uorescence; once it is oxidized by ROS, it illuminates red (ethidium bromide [EB]). The images were captured with a Nikon Eclipse Ti-E 3D Deconvolution microscope with a 60x objective lens in a blinded fashion. The uorescence intensity was calculated using ImagePro-PLUS 7.0 software. The background uorescence intensity was subtracted from the cell intensity. The index of cytosolic ROS concentration is described as a ratio of EB and DHE.

Statistics
We conducted data analysis in a blinded fashion wherever possible and set proper controls for every experimental plan. The mouse numbers and independent experiment numbers are described in the gure legends. Statistical analysis was performed using GraphPad Prism 9 (La Jolla, CA, USA). Data are presented as mean ± SEM. After the data passed a normality test (Kolmogorov-Smirnov), the two-tailed Student's t-test was used for comparisons of two groups, and one-way ANOVA was used for multiple comparisons. If the data did not pass the normality test, a non-parametric test (Mann-Whitney for two groups, Kruskal-Wallis for multiple comparisons) was used. Bonferroni's multiple comparisons test was used as a post hoc test for one-way ANOVA and Dunn's test for the Kruskal-Wallis test. Statistical comparison between dose-response curves was made by two-way ANOVA with Bonferroni post hoc test. Differences were considered to be statistically signi cant when P<0.05.

Coronary microvascular dysfunction in diabetic mice
We used type 2 diabetic (T2D) model mouse generated by high-fat diet feeding and a single low-dose injection of STZ. Our diabetic mice exhibited increased body weight, abnormal glucose tolerance, and dyslipidemia ( Fig. 1, Table 1, and Table 2). Coronary ow velocity reserve (CFVR) was measured as a determinant of coronary microvascular function (Additional File 3: Fig. S3) [18,39,40]. Decreased CFVR indicates that mice are suffering from CMD and prone to ischemic heart disease [18]. We found that CFVR was signi cantly reduced in diabetic mice compared to control (Fig. 1C). In the LV, diabetic mice showed signi cantly lower capillary density (Fig. 1D) and higher EC apoptosis (Additional File 3: Fig. S6) than control mice. These results indicate that coronary microvascular function, due potentially to reduced capillary density, is signi cantly attenuated in mice with experimental diabetes.
Decreased HuR protein level in CECs isolated from diabetic mice and patients Freshly isolated mouse CECs were used to detect HuR levels in control and diabetic mice. HuR levels were signi cantly decreased in CECs from diabetic mice compared to those from control mice, as determined by Western blot (Fig. 2A) and immuno uorescence study (Fig. 2B). In line with inducible T2D mouse data, CECs from spontaneous T2D mouse model (Tallyho [TH] mice) and CECs from diabetic patients exhibit a signi cant decrease in HuR protein level compared to their controls (Fig. 2C-D). On the other hand, HuR levels in cardiac cells (remain heart cells after depletion of ECs) and aortic SMCs were not different between control and diabetic mice (Additional File 3: Fig. S7). Taken together, HuR protein is selectively downregulated in CECs in diabetic mice, and endothelial downregulation of HuR is a potential contributor to reduced capillary density in the heart and decreased coronary microvascular function in diabetes.
However, Tie2-HuR -/mice displayed decreased CFVR (Fig. 3F) and capillary density (Fig. 3G), and increased apoptotic ECs compared to Wt mice (Additional File 3: Fig. S6). We found that there was no signi cant difference in apoptotic cell number in other cell types (total apoptotic cells # -apoptotic endothelial cell #) between Wt mice and Tie2-HuR -/mice (Wt, 2.4 ± 0.4 NA/mm 2 ; Tie2-HuR -/-, 2.9 ± 0.5 NA/mm 2 ; n mice =5 per group), indicating that cell apoptosis induced by HuR deletion was con ned to CECs. These data provide strong evidence that EC-speci c deletion of HuR induces the reduction of capillary density and ultimately attenuates coronary microvascular function.
Target genes with altered expression in CECs from diabetic and Tie2-HuR -/mice To de ne the target genes altered by HuR deletion and by diabetes, we conducted the real-time PCR on 92 genes (the gene list is shown in Additional File 2: Table S3) and compared the mRNA levels of these genes between control and diabetic mice (Additional File 2: Table S4) and between Wt and Tie2-HuR -/mice (Additional File 2: Table S5). The genes altered by diabetes and/or HuR deletion are summarized in a Venn Diagram (Fig. 4A). We chose the 92 genes that are expressed in ECs and play crucial roles in endothelial functions, such as a) endothelium-derived relaxing factors and their regulators; b) modi ers of cytosolic [Ca 2+ ], mitochondrial [Ca 2+ ], and endoplasmic reticulum [Ca 2+ ]; and c) regulators of EC proliferation, migration, and apoptosis. The expression levels of 3 genes: Elav1 (HuR), Gja5 (connexin40, Cx40), and nicotinamide adenine dinucleotide phosphate reduced oxidases 4 (Nox4) were signi cantly decreased in CECs from diabetic mice and in CECs from Tie2-HuR -/mice compared to their controls. These results demonstrate that a selective group of genes (i.e., HuR, Cx40, and Nox4) is concomitantly downregulated in mice with diabetes and mice genetically deleted endothelial HuR. The next set of experiments was designed to examine whether downregulated HuR in diabetic mice may directly regulate Cx40, an important gap junction protein required for normal EC function, to decrease coronary microvascular function. Downregulated Cx40 level and decreased Cx40 mRNA binding to HuR protein in CECs of diabetic mice We rst con rmed that Cx40 protein level was downregulated in CECs from diabetic mice (Fig. 4B) and Tie2-HuR -/mice (Fig. 4C) compared to their controls. Next, we performed ribonucleoprotein immunoprecipitation to examine the binding of Cx40 mRNA to HuR protein and found that Cx40 mRNA binding to HuR protein was signi cantly lower in CECs from diabetic mice than in CECs from control mice (Fig. 4D). Taken together, downregulated Cx40 in diabetic EC is potentially due to reduced HuR level and decreased HuR binding to Cx40 in diabetic mice.

Deletion of Cx40 decreases CFVR by reducing capillary density
Since HuR downregulation results in decreased Cx40 protein level, we examined the role of Cx40 in the development of CMD. Cx40 is a component of gap junction that acts as a tunnel for small molecules (<1 kDa) and electrical propagation during endothelium-derived hyperpolarization (EDH)-dependent vascular relaxation. There was no difference in body weight between Wt and Cx40 -/mice; however, Cx40 -/mice exhibited a slight, but statistically signi cant increase in plasma glucose level and MAP (Table 2). Figure   4E demonstrated that Cx40 protein level was diminished in CECs of Cx40 -/mice. Endothelium-dependent relaxation was assessed by acetylcholine (ACh)-induced relaxation, and EDH-dependent relaxation was evaluated by ACh-induced relaxation in the presence of L-NAME (an eNOS inhibitor) and indomethacin (a cyclooxygenase inhibitor). We con rmed that endothelium-dependent relaxation (Fig. 4F) and EDHdependent relaxation (Fig. 4G), but not EC-independent relaxation (Fig. 4H), were signi cantly attenuated in CAs isolated from Cx40 -/mice compared to Wt mice. Those data indicate that deletion of Cx40 su ciently inhibited gap junction function. Importantly, Cx40 -/mice exhibited a signi cant decrease in CFVR (Fig. 4I) and capillary density (Fig. 4J), and an increase in EC apoptosis (Additional File 3: Fig. S6) compared to Wt. These results indicate that Cx40, which is directly regulated by HuR, is required for maintaining normal coronary endothelial function.

Inhibition of HuR attenuates tube formation in control CECs, and overexpression of Cx40 increases tube formation in HuR-reduced CECs
To examine how HuR deletion and Cx40 overexpression alter angiogenic capability of ECs, we conducted an ex vivo tube formation assay in human CECs. Cx40 was overexpressed using Cx40-adenovirus (Adv), and HuR was inhibited by HuR siRNA transfection in human CECs (Additional File 3: Fig. S5). Figure 5 demonstrates that HuR inhibition in control CECs signi cantly reduced tube formation, suggesting that HuR regulates endothelial angiogenic capability. Cx40 overexpression in control CEC did not affect angiogenic capability; however, Cx40 overexpression in HuR-inhibited CECs slightly, but signi cantly, increased tube formation. Taken together, the decreased Cx40 due to HuR downregulation in CECs contributes to inhibiting EC-driven angiogenesis, reducing capillary density, and ultimately attenuating coronary microvascular function.
Overexpression of Cx40 in diabetic mice improves coronary microvascular function Type 2 diabetes was induced in Wt mice and EC-speci c Cx40-overexpressing mice [38]. The mice were used for the experiments 16 weeks after diabetic induction. Under diabetic conditions, Cx40 transgenic mice demonstrated increased Cx40 protein levels in CECs compared to Wt (Fig. 6A). Cx40 overexpression did not alter body weight (Fig. 6B) or glucose tolerance (Fig. 6C) in diabetic mice but decreased plasma glucose level at non-fasting conditions ( Table 2). Endothelium-dependent relaxation (Fig. 6D), but not smooth muscle-dependent relaxation (Fig. 6E), was signi cantly augmented by Cx40 overexpression in CAs of diabetic mice. Furthermore, Cx40 overexpression signi cantly increased capillary density (Fig. 6F) and CFVR (Fig. 6G) in diabetic mice. We here provide strong evidence that overexpression of Cx40 is su cient to restore EC-dependent vasodilation, capillary density, and CFVR in diabetic mice. These observations suggest that increasing Cx40 expression and/or function is a potential strategy for treating CMD in diabetes in which HuR level is downregulated.

Discussion
Clinical data indicate that diabetic patients with CMD exhibit high cardiac mortality [19,20]; however, the molecular mechanisms by which diabetes leads to CMD is poorly understood. To investigate microvascular function in diabetes, we used an inducible T2D mouse model generated by administrating a single injection of low-dose STZ and feeding a high-fat diet. This mouse model has given us reproducible data with hyperglycemia and hyperinsulinemia [35,36]. T2D mice not only exhibited increased body weight and abnormal glucose tolerance, but also suffered from dyslipidemia (Figs. 1A, and 1B, and Tables 1 and 2). However, lipid plaque formation has never been detected in this model. Therefore, the reduction of CFVR (Fig. 1C) is due solely to coronary microvascular dysfunction, and decreased CFVR indicates that mice suffer from CMD. In line with the result from inducible T2D mice, TH mice (spontaneous T2D mice) exhibited reduced CFVR [18] without detectable plaque formation, suggesting that chronic hyperglycemia is the risk factor of CMD. Since capillary density positively correlates with coronary ow reserve [8,41,42], decreased capillary density in the heart could be a cause of reduced CFVR. Figure 1D demonstrates that T2D mice displayed decreased capillary density in the LV.
Capillary density can be reduced by 1) augmented endothelial cell apoptosis, 2) attenuated cell migration and/or proliferation of neighboring mature ECs, and/or 3) reduced mobilization of circulating endothelial progenitor cells (EPCs) in sites of the vascular wall where ECs are damaged and/or lost. We and other investigators demonstrated that CECs are apoptotic in diabetes [12,18,43,44] (Additional File 3: S6) and characterized by a diminished ability to migrate and proliferate in diabetes [18,45]. The number of EPCs is also decreased, and the function of EPCs is attenuated in diabetic patients and diabetic animal models [46,47]. This study was designed to identify the key genes that in uence capillary density and ultimately change microvascular function in the diabetic heart.
Like other RNA-binding proteins, HuR binds to many RNAs and changes their fates. The data from systemic and tissue-speci c deletion of HuR gene indicate that HuR plays a critical role in embryonic development and the regulation of physiological function [23][24][25][26]. Therefore, it was natural to hypothesize that an abnormal level of HuR may be involved in the development or progression of cardiovascular disease. The conclusive data came from cancer research at rst; HuR expression is increased with cancer and aids in the progression of angiogenesis in tumor tissue [31]. However, the contribution of HuR to other diseases is still controversial. Zhou et al. demonstrated that HuR protein level was signi cantly reduced in the heart from the patients with heart failure compared to control patients [29]. They also showed that myocardial infarction (MI) decreased HuR protein level in mouse hearts, and overexpression of HuR by AAV-HuR injection reduced infarct size and improved cardiac function [29]. On the other hand, Krishnamurthy et al. found that HuR level was increased after MI in mice, and downregulation of HuR by HuR-shRNA lentivirus injection restored cardiac function [48]. Unfortunately, we could not nd the reason for the discrepancy between these studies. Recent reports demonstrate that HuR upregulation is implicated in the development of atherosclerosis [49,50] and diabetic nephropathy [51].
We found that HuR level was signi cantly decreased in CECs from diabetic mice compared to control mice (Figs. 2A, 2B). This phenomenon was also observed in CECs from a spontaneous T2D mouse model, TH mice (Fig. 2C), and CECs from diabetic patients (Fig. 2D). Tie2-HuR −/− mice exhibited similar microvascular functions to T2D mice (reduced CFVR, decreased capillary density, and increased EC apoptosis in the LV compared to Wt, see Fig. 3 and Additional File 3: Fig. S6), suggesting that decreased HuR expression in CECs is one of the leading causes of CMD in diabetes.
Some might think that it would be essential to examine the effect of HuR overexpression on EC function in diabetic mice, and we agree. However, we encountered a problem when overexpressing HuR in CECs. In ex vivo studies, we found that the working concentration of HuR overexpression is very narrow. Overexpression of HuR easily killed CECs, and we had a very di cult time controlling HuR levels during the experiment. This implies that HuR may not only interact with mRNAs that are important for EC angiogenesis, but also bind to mRNAs that regulate cell death. Therefore, we decided to examine the target genes of HuR, which are also involved in CMD in diabetes. We examined 92 (Additional File 2: Table S3) genes by real-time PCR using a PCR plate custom-made by QIAGEN. We are aware that there are other genes which are not included in the plate but are also important for EC function. We did not use a microarray or RNAseq in this study because: 1) these experiments would require more animals to obtain a su cient amount of mRNA, and 2) they would provide an overwhelming amount of information for the studies at the time. We believe that real-time PCR with 92 genes still gave us su cient information to move to the next step. Interestingly, we found that only 3 genes out of 92 were altered in CECs by HuR deletion and diabetes: HuR, Cx40, and Nox4 ( Fig. 4 and Additional File 2: Table S4 and S5). Cx40 protein levels were signi cantly decreased in diabetes and Tie2-HuR −/− mice compared to those controls and HuR-bound Cx40 mRNA was signi cantly lower in diabetes than in control (Figs. 4B-4D). These data indicate that Cx40 could be a potential target of HuR. It is important to note that we examined the protein level of Nox4 and found that Nox4 protein level was not altered in CECs of diabetic mice compared to the control (Control, 1.01 ± 0.06; Diabetic, 1.16 ± 0.32. N mice =7 per group. P = 0.65). Therefore, we did not conduct a further experiment to examine the role of Nox4 in this study. We were indeed surprised to see that only 3 genes were shared in diabetic and Tie2-HuR −/− mice after screening genes in an unbiased way. These results suggested that HuR overexpression in diabetic mice may not be ideal since it would potentially lead to unnecessary alterations in many other genes besides Cx40. Thus, we believe overexpression of Cx40, a downstream HuR-sensitive gene, in diabetes could be a safer and better option to treat diabetic cardiovascular complications than overexpression of modi cation of HuR.
Cx40 is a major gap junction protein in ECs, and decreased gap junction activity due to reduced Cx40 expression attenuates EDH [52] and endothelial migration [53]. We and other investigators demonstrated that ECs in type 1 diabetic mice exhibited a signi cant decrease in Cx40 protein level [11,54]. However, there is no report examining the role of Cx40 in CECs of T2D mice to the best of our knowledge. The results from Fig. 4 suggest that HuR regulates Cx40 gene expression in CECs and decreased HuR protein level and HuR-binding to Cx40 mRNA are, at least in part, the causes for downregulated Cx40 expression in CECs in diabetes. Therefore, we obtained Cx40 −/− mice [37] (Fig. 4E) to examine whether mice without Cx40 exhibit similar coronary microvascular function shown in diabetes. First, we examined EDHdependent relaxation in 3rd order of CAs to show the functional change of gap junction by Cx40 deletion. EDH was abolished by Cx40 deletion in CAs (Fig. 4G) without any change in smooth muscle-dependent relaxation (Fig. 4H), suggesting that Cx40 deletion is functionally working. Next, we examined coronary microvascular function and found that CFVR was signi cantly reduced in Cx40 −/− mice (Fig. 4I) accompanied by decreased capillary density and increased EC apoptosis (Figs. 4J and Additional File 3: Fig. S6). This is the rst report to demonstrate that the loss of Cx40 leads to CMD. We, therefore, hypothesized that the CMD seen in T2D mice might result from decreased Cx40 expression in CECs due to downregulated HuR expression.
It has been known for decades that coronary endothelial dysfunction is implicated in the development of obstructive CAD. However, the Women's Ischemia Syndrome Evaluation (WISE) study shed light on endothelial dysfunction in patients with non-obstructive CAD (CMD) in 2004. Their data suggest that the decrease in coronary microvascular function predicts adverse cardiovascular outcomes independent of CAD severity [55]. In addition, the treatment with ranolazine (a late sodium current inhibitor that is commonly used for obstructive CAD) does not show any bene cial effect on ischemia in patients with CMD [56]. These reports emphasize the necessity to develop new drugs speci c for patients with CMD.
We believe that CX40 is an excellent therapeutic target for CMD based on the following reasons: 1) there are more myoendothelial gap junctions that are composed of connexins (Cxs) in smaller resistant vessels than in large vessels [57][58][59], suggesting that Cx40 upregulation could be more effective in small vessels; 2) Cx40 is predominantly expressed in ECs [57,60]; therefore, increased Cx40 expression and activity lead to a speci c effect on EC function; and 3) overexpression of Cx40 does not lead to abnormal angiogenesis as shown in tumor tissues [61]. We demonstrate here that overexpression of Cx40 in ECs augments endothelium-dependent relaxation, increases capillary density, and results in improved CFVR in diabetic mice (Fig. 6). These results provide strong evidence that overexpression of Cx40 is a useful therapeutic strategy for CMD in diabetes.
Overexpression of Cx40 augmented angiogenesis in HuR-de cient CECs ex vivo (Fig. 5) and in vivo (Fig. 6E). Other investigators also show that Cx40 positively regulates cell migration and angiogenesis [53]; however, the detailed mechanisms of how exactly Cx40 enhances angiogenesis is still unknown. It has been reported that Cx43 overexpression increases [62,63] or decreases [64,65] angiogenesis or cell migration independently of gap junction activity. The deletion of Cx37 promotes angiogenesis [66]. These data suggest that the effect of Cxs on cell migration and angiogenesis seems to be different among the subtypes of Cxs and that enhanced EC angiogenesis by Cx40 overexpression might be not only due to increased gap junction activity, but also due to unknown mechanisms through Cx40 overexpression. It has been reported that Cx43 regulates the expression of other genes [67]; therefore, it is possible that Cx40 can also regulate the expression of other genes.
HuR or Cx40 deletion in ECs led to endothelial apoptosis (Additional File 3: Fig. S6). In this study, we focused on angiogenic capability of ECs (Fig. 5) rather than endothelial apoptosis. However, increased cell apoptosis by HuR or Cx40 deletion would contribute to decreased capillary density in the heart. Excess production of ROS is one of the leading causes of cell apoptosis. We previously reported that ROS formation in CECs was considerably increased in diabetes [12,35,68,69]. Additional File 3: Fig. S8 demonstrates that the deletion of HuR or Cx40 gene increased cytosolic ROS formation in human CECs. Excess ROS production by HuR inhibition was sort of expected since HuR deletion signi cantly downregulates Opa1 expression ( Fig. 4 and Additional File 2: Table S5). Opa1 is a mitochondrial fusion protein, and reduced mitochondrial fusion (or increased mitochondrial ssion) increases mitochondrial ROS formation, followed by the rise of cytosolic ROS concentration [70]. Increased ROS generation by Cx40 inhibition surprised us and would require further experiments to identify the molecular mechanisms. Other investigators have also investigated the potential mechanisms in which the inhibition of HuR leads to cell apoptosis. HuR binds to Mdm2, a primary negative regulator of p53; therefore, deletion of HuR increases p53 and leads to cell apoptosis [71]. Inhibition of HuR also increases caspase 3 expression [26] and other proapoptotic factors [71] that ultimately induces cell apoptosis. We have reported that p53 is one of major causes of coronary endothelial apoptosis in diabetes, and inhibition of p53 improves coronary microvascular function [18]. It has to be noted that the mechanisms to induced endothelial cell apoptosis in Tie2-HuR −/− and diabetic mice might be the same or different; therefore, it requires additional experiments to identify detailed molecular mechanisms in which HuR deletion-and hyperglycemiainduced endothelial-apoptosis in the heart.
We found that Cx40 −/− mice exhibited a signi cant increase in plasma glucose level, and Cx40 overexpression in diabetes displayed a slight but signi cant decrease in plasma glucose level (Table 2). In the endocrine system, Cx40 is well known to regulate the function of renin-producing cells in the kidneys. Therefore, the increase in blood pressure by Cx40 deletion (Table 2) might be partly led by increased renin secretion [72]. However, there is no report showing that Cx40 contributes to insulin secretion and/or glucose tolerance. Cx36 is expressed in β-cells, and Cx36-KO mice develop glucose intolerance via attenuation of glucose-stimulated insulin secretion from the β-cells [73]. Although Cx40 is not expressed in β-cells, there may be mechanisms by which Cx40 regulates plasma glucose level (e.g., altered in ltration of in ammatory cells in adipose tissues). We will further investigate this unique phenomenon in future studies.

Conclusions
This study demonstrates for the rst time that diabetes leads to downregulation of HuR, an RNA-binding protein, which subsequently decreases expression of Cx40, a gap junction channel protein, in cardiac ECs, and attenuates coronary microvascular function in diabetes. Overexpression of Cx40 increases capillary density and restores coronary microvascular function determined by CFVR in diabetes. These data indicate that Cx40 is a promising therapeutic target for developing novel and unique treatment for CMD in diabetic patients.