Pregestational Diabetes Induces Fetal Coronary Artery Malformation via Reactive Oxygen Species Signaling

Pregestational diabetes is a risk factor for congenital heart defects in infants (1,2). Clinical studies have shown that pregestational diabetes increases the risk of congenital heart defects in the offspring by three- to fivefold compared with nondiabetic pregnancies (1–3). To date, analysis of congenital heart malformation in the newborn is mainly restricted to major cardiac structures, which include the aorta, pulmonary artery, atrioventricular septum, cardiac valves, and myocardium, but coronary arteries are not routinely examined (4). Congenital malformation can occur in the coronary arteries, leading to null coronary artery or hypoplastic coronary artery disease (5). Although null coronary artery is embryonically lethal, hypoplastic coronary artery disease is a rare congenital coronary abnormality defined by malformation of one or more major branches of the coronary arteries with a marked decrease in luminal diameter and length. Hypoplastic coronary artery disease can be asymptomatic but often is associated with myocardial infarction and sudden cardiac death during intense physical activity (6). However, the etiology and pathogenesis of hypoplastic coronary artery disease remain undefined. Furthermore, it is not known whether pregestational diabetes results in coronary artery malformation in offspring. Isolated cases of congenital coronary artery abnormalities have been identified in infants by autopsy. However, whether the mothers of these infants had pregestational diabetes was not disclosed in these reports (7,8). A large multicenter case-control study showed a significant association between pregestational diabetes and a wide range of congenital heart defects (1). Because coronary arteries were not routinely examined, defects of coronary arteries were not reported in this study. Of note, a recent population-based study reported cases of coronary artery anomaly in infants born to mothers with obesity; however, mothers with established diabetes were excluded from this study (9).

Coronary arteries are developed from several sources of progenitors, which include endothelial cells of the sinus venosus, proepicardial organ, and endocardium (10–12). Current understanding is that cells from the proepicardial organ migrate toward the myocardium and form the epicardium. Concurrent with formation of the epicardium, some epicardial cells undergo epithelial-to-mesenchymal transition (EMT), migrate to the subepicardial space and myocardium, and become epicardial-derived cells (EPDCs). Simultaneously, a subpopulation of endothelial cells of the sinus venosus also migrates to the subepicardial space, forming subepicardial endothelial progenitors (11). When they reside in the myocardium, the EPDCs give rise to vascular smooth muscle cells and fibroblasts (13), whereas subepicardial endothelial progenitors give rise to coronary endothelial cells (14). Together, EPDCs and subepicardial endothelial cells contribute to the development of coronary arteries with a minor contribution from the endocardium.

Epicardial EMT is regulated by several epicardial transcription factors and myocardial-induced signaling molecules (15). An important factor in vasculogenesis and heart development is hypoxia-inducible factor 1α (HIF-1α) (16), which regulates many downstream target genes, including Wilms tumor 1 (Wt1), a critical transcription factor expressed in epicardial progenitors and EPDCs (17,18). Studies have shown that Wt1 directs epicardial progenitors to become smooth muscle cells and fibroblasts and forms coronary arteries (19). Pregestational diabetes increases reactive oxygen species (ROS) production (20,21) and downregulates HIF-1α expression in embryos (22). We have shown that treatment with N-acetylcysteine (NAC) increases reduced glutathione (GSH) levels and decreases ROS production in the hearts of fetuses of mice with pregestational diabetes (20). The current study tested the hypothesis that pregestational diabetes impairs epicardial EMT and induces coronary artery malformation in offspring. We further hypothesized that these abnormalities are mediated by ROS signaling and can be rescued by NAC treatment in diabetic females during gestation.

C57BL/6J wild-type Wt1^CreERT2/+ and Rosa26^mTmG mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Animals were handled in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH publication no. 85-23, revised 1996). All procedures involving mouse handling and manipulation were in accordance with the guidelines of the Canadian Council of Animal Care and approved by the Animal Use Subcommittee at the University of Western Ontario.

Eight-week-old C57BL/6 female mice were induced to become diabetic by streptozotocin (STZ) 80 mg/kg body weight/day i.p. injections for 3 consecutive days as described previously (20,23,24). STZ was dissolved in sodium citrate (pH 4.0), and mice treated with sodium citrate served as controls. Nonfasting blood glucose levels were determined 1 week after STZ injection using a glucose meter (OneTouch Ultra2; LifeScan Canada, Burnaby, BC). Mice with blood glucose levels >11 mmol/L were considered diabetic and bred with normal adult males. Vaginal plugging was monitored as a sign of successful mating, and the day of vaginal plugging was recorded as day 0.5 of gestation (E0.5). A subgroup of control and diabetic mice received 4 mg/mL NAC in drinking water from E0.5 to the end of gestation or harvesting of the embryos (20). Water and food intake of pregnant mice was measured and normalized to body weight.

Fate mapping of the Wt1⁺ lineage was performed using Wt1^CreERT2/+ (Stock #10912, The Jackson Laboratory) and the double-fluorescent Cre reporter line Rosa26^mTmG (Stock #7576). Wt1^CreERT2/+ is a heterozygous Wt1^−/+ mouse, where exon 1 of the Wt1 gene is deleted and replaced by a CreERT2 sequence under the control of the Wt1 promoter (19). CreERT2 is created by fusion of Cre recombinase to a modified human estrogen receptor ligand–binding domain (ERT2). The Rosa26^mTmG reporter mice possess loxP sites on either side of a membrane-targeted tomato (mT) cassette followed by a membrane-targeted green fluorescent protein (GFP) (mG) cassette. In the absence of Cre, mT (red fluorescence) is expressed in all tissues. In the presence of Cre, the mT cassette is deleted and mG (green fluorescence) expressed only in the targeted cells. Female Rosa26^mTmG mice were induced to become diabetic by STZ 80 mg/kg body weight/day i.p. for 3 days as already described and were bred with Wt1^CreERT2/+;Rosa26^mTmG males. Tamoxifen 50 mg/kg body weight i.p. was injected into diabetic or control females at E10.5. Upon tamoxifen administration, Cre is activated in Wt1-expressing cells and induces recombination to express GFP, which is one of the two fluorescent proteins in the Wt1^CreERT2/+;Rosa26^mTmG offspring. Embryos were harvested at E14.5. Double-transgenic Wt1^CreERT2/+;Rosa26^mTmG embryos were sectioned and immunostained for GFP. The Wt1⁺ derivatives, which are GFP⁺ cells and structures, were quantified and compared between the offspring of diabetic and control mice.

To assess oxidative stress, embryonic heart tissues were harvested at E14.5. Frozen samples were cut into 10-μm sections using a cryostat (CM1950; Leica). Heart sections were stained with 4-hydroxynonenal (4-HNE) antibody (1:300; abm Inc.) followed by CY3-conjugated anti-goat IgG (1:1,000; Jackson ImmunoResearch). 4-HNE florescence signals were detected using a florescence microscope (Observer D1; Zeiss, Oberkochen, Germany). At least five images were captured using a fixed exposure time for each heart, and intensities of florescence signals per myocardial area were quantified using AxioVision software (Zeiss).

Immunohistochemical analysis was performed to study the formation of the coronary vasculature and its precursors. In brief, the thorax of fetuses and neonates were harvested and processed in 4% paraformaldehyde overnight, embedded in paraffin, and sectioned transversely in 5-μm sections. Before immunostaining, antigen retrieval was performed in citric acid buffer (0.01 mol/L, pH 6.0) for 12 min at 94°C using a microwave oven (BP-111; Microwave Research & Applications, Inc., Carol Stream, IL). Samples were incubated with primary antibody overnight, as follows: anti-α-smooth muscle actin (1:3,000, Sigma-Aldrich), biotinylated lectin-1 (1:250, Vector Laboratories), anti-Wt1 (1:300, Calbiochem), anti-E-cadherin (1:200, Santa Cruz), anti-GFP (1:1,000, Abcam), anti-phosphohistone H3 (pHH3) (1:500, Abcam), and anticleaved caspase-3 (1:800, Cell Signaling) followed by one of the following secondary antibodies (Vector Laboratories) for 1 h: biotinylated goat anti-rabbit IgG (1:500), biotinylated goat anti-mouse IgG (1:500), or biotinylated donkey anti-goat IgG (1:500). Signals were amplified by incubation with avidin/biotin complex reagent (Vector Laboratories) and visualized using 3-3′diaminobenzidine tetrahydrochloride (Sigma-Aldrich). Heart sections were counterstained with hematoxylin (Thermo Scientific), and images were captured using a light microscope (Observer D1; Zeiss). To visualize and analyze coronary artery volume, images of every 25-μm heart section immunostained with α-smooth muscle actin were captured and imported into the Amira software program for three-dimensional reconstructions. Using the analytical tools of Amira, pixels were converted to volume (μL), and a ratio of coronary artery volume to myocardial volume was obtained.

EMT was assessed ex vivo. Ventricles of E12.5 embryos from control dams were harvested and cultured on collagen gel (25). Collagen (1 mg/mL, type I collagen of rat’s tail; BD Bioscience) was prepared in M199 media (M5017; Sigma) containing 5 mmol/L d-glucose or in M199 media with an additional 25 mmol/L d-glucose, making the final d-glucose concentration 30 mmol/L. Casted collagen was hydrated by Opti-MEM media plus 1% FBS and insulin-transferrin-selenium for 30 min at 37°C. E12.5 ventricles were explanted and incubated at 37°C overnight. The following day, 5 or 30 mmol/L d-glucose in M199 media with 10% FBS were added to the heart explants. The glucose concentration was ∼5 mmol/L higher than the blood glucose of diabetic dams at E10.5. Heart explant cultures were treated with or without 0.5 mmol/L NAC. The number of spindle-shaped cell outgrowths from the explanted ventricles was quantified 4 days postculturing. Images were captured using a phase contrast microscope (Zeiss).

Total RNA was extracted from individual E11.5 fetal hearts using RNeasy Mini Kit (QIAGEN, Burlington, ON) per the manufacturer’s instructions. cDNA was synthesized from 0.1 μg total RNA using Maloney murine leukemia virus reverse transcriptase in 10 μL reactions, which were diluted by three times for genes of interest and 500 times for housekeeping gene 28S. Two microliters diluted cDNA were used for real-time PCR amplification using EvaGreen qPCR MasterMix (Applied Biological Materials, Vancouver, BC). Specific primers were designed for Hif-1α, Snail1, Slug, basic fibroblast growth factor(bFgf),β-catenin, and Aldh1a2 (Supplementary Table 1). Samples were amplified for 35 cycles using Eppendorf realplex (Eppendorf, Hamburg, Germany). The mRNA levels in relation to 28S rRNA were determined (25).

Data are presented as mean ± SEM. Statistical analysis was performed using one-way ANOVA followed by Newman-Keuls test (GraphPad Prism version 3.0 software). The incidence of malformation was assessed by χ² test. P < 0.05 was considered statistically significant.

To examine the effect of pregestational diabetes on coronary artery formation, adult female mice were induced to become diabetic by STZ, and their nonfasting blood glucose levels were determined during gestation. The data show that blood glucose levels in diabetic females were significantly higher than that in normal controls (P < 0.01) (Table 1). NAC treatment did not significantly alter blood glucose levels in the diabetic and normal dams. Incidence of coronary artery malformation among fetal hearts of mice with pregestational diabetes was 46%, which was decreased to 6% by NAC treatment (P < 0.01) (Table 1). Food and water intake in the diabetic females was about eight times higher than that in control mice. Following NAC treatment, food consumption was similar between control and diabetic mice. However, water intake in the NAC-treated diabetic mice was still twice the amount in control mice (P < 0.05) (Table 1). The estimated daily dose of NAC was 0.6 and 1.2 g/kg in control and diabetic mice, respectively.

We recently showed that superoxide generation is increased in E12.5 fetal hearts of mice with pregestational diabetes, which is normalized by treatment with NAC in drinking water in the diabetic dams during gestation (20). To further assess oxidative stress, immunostaining of 4-HNE, a product of lipid peroxidation (26), was performed in E14.5 hearts (Fig. 1A). Myocardial 4-HNE levels were significantly increased in the offspring of pregestational diabetic mice, which was abrogated by NAC treatment (Fig. 1B). These data suggest that pregestational diabetes increases ROS and oxidative stress in the fetal heart, which are effectively inhibited by NAC treatment.

Formation of the coronary artery and capillaries was evaluated following immunostaining of α-smooth muscle actin and lectin-1, respectively. Offspring of diabetic mothers showed reduced left- and right-side coronary artery diameter and abundance at postnatal day 0 (P0) (Fig. 1C–E, G–I) and at E16.5 (data not shown). Additionally, coronary arteries in the offspring of diabetic mice had fewer smooth muscle cells surrounding the vessels (Fig. 1E) and decreased capillary abundance at P0 (Fig. 1F and J). Of note, abnormalities in coronary artery and capillary development in the offspring of diabetic mice were rescued by NAC treatment (Fig. 1C–J). Impaired coronary artery development is illustrated in three-dimensional reconstructions of coronary arteries, which demonstrate smaller arteries with reduced arborization in the offspring of diabetic mice at E16.5 and P0 (Fig. 2A). The impaired coronary artery development was not a result of changes in myocardial volume because the ratio of total coronary artery volume to myocardial volume was decreased in the offspring of diabetic mice at P0 and E16.5 (Fig. 2B and C).

To gain a better understanding of cellular and molecular events that caused coronary artery malformation in the offspring of diabetic mice, the epicardium and proepicardial organ were studied during embryogenesis using immunohistochemistry to detect Wt1 as a marker. The data show that the epicardium in fetal hearts of pregestational diabetic mice formed a sparse and loosely attached cell layer to the myocardium with significant reductions in the number of Wt1⁺ epicardial and subepicardial cells at E12.5 (Fig. 3A, D, and E). Impaired Wt1 expression in fetal hearts of diabetic mothers was persistent at E14.5, as evidenced by a reduced number of Wt1⁺ cells in the epicardial cell layer and in the compact myocardium (Fig. 3B, F, and G). Concurrent with reductions in the number of epicardial Wt1⁺ cells, the number of E-cadherin–positive epicardial cells was increased in fetal hearts of diabetic mothers at E12.5 (Fig. 3C and H). Of note, NAC treatment during gestation reduced epicardial E-cadherin expression and restored the number of Wt1⁺ cells in the epicardium, subepicardium, and myocardium in fetal hearts (Fig. 3A–H). However, pregestational diabetes did not alter the number of Wt1⁺ cells in the proepicardial organ at E9.5 (Fig. 3I).

To directly evaluate the derivatives of epicardial Wt1⁺ cells, fate mapping analysis was performed using Rosa^mTmG female mice in which diabetes was induced by STZ and then bred to tamoxifen-inducible Wt1^CreERT2/+;Rosa^mTmG males. Following tamoxifen treatment in the female mice at E10.5, Wt1⁺ derivatives, including coronary vessels, were observed in the hearts of Wt1^CreERT2/+;Rosa^mTmG embryos at E14.5 (Fig. 3J). In agreement with our hypothesis, pregestational diabetes diminished Wt1⁺ lineage and coronary vessels in the hearts of Wt1^CreERT2/+;Rosa^mTmG embryos (Fig. 3J–L).

To further understand ROS signaling in epicardial EMT, the expression of Hif-1α, a master regulator of vasculogenesis and epicardial EMT (27), and its downstream signaling molecules essential for coronary development were evaluated by RT-PCR. The expression of Hif-1α was significantly diminished in E11.5 fetal hearts of diabetic dams (Fig. 4A). Additionally, mRNA levels of Snail1 and Slug, which are key regulators of EMT, were significantly decreased in fetal hearts of diabetic dams (Fig. 4B and C). Furthermore, retinoic acid signaling may be impaired in the fetal hearts of diabetic dams as evidenced by reduced mRNA levels of Aldh1a2, which encodes the retinoic acid synthesis enzyme retinaldehyde dehydrogenase 2 (RALDH2) (28) and its downstream target bFGF (Fig. 4D and E). Of note, these changes in the expression of Hif-1α, Snail1, Slug, Aldh1a2, and bFGF were restored in the hearts of offspring of diabetic females by NAC treatment (Fig. 4A–E). However, β-catenin mRNA levels were not significantly altered by pregestational diabetes or NAC treatment (Fig. 4F).

To investigate whether hyperglycemia impairs epicardial EMT in fetal hearts of diabetic mothers, E12.5 hearts were cultured on collagen gel in both high glucose (30 mmol/L) and normal glucose (5 mmol/L) conditions for 4 days (Fig. 5A). The number of spindle-shaped cells, which are epicardial cells that had undergone EMT, was quantified. The data show that the number of spindle-shaped cells was significantly decreased in high-glucose compared with normal-glucose cultures. The response was restored by NAC 0.5 mmol/L treatment (Fig. 5A and B).

To assess the effect of pregestational diabetes on epicardial cell proliferation, immunostaining for pHH3, a marker of mitotic cells, was performed in fetal hearts. The results show that the number of pHH3⁺ epicardial cells was significantly decreased in fetal hearts of diabetic dams at E12.5 and E14.5 (Fig. 6A–D). Of note, treatment with NAC in diabetic dams completely restored epicardial cell proliferation at both time points (Fig. 6C and D). Assessment of the apoptotic epicardial cells in E12.5 hearts using immunostaining for cleaved caspase-3 show no significant difference between the control offspring (0.37 ± 0.16%) and the offspring of diabetic mice (0.33 ± 0.12%, n = 8–10 hearts/group).

The current study demonstrates for the first time to our knowledge that pregestational diabetes impairs development of the coronary artery vasculature in fetal hearts in an animal model. We further demonstrate that pregestational diabetes increases oxidative stress, diminishes HIF-1α and Wt1 expression, and decreases Snail/Slug and RALDH2/bFGF signaling pathways, leading to the disruption of epicardial EMT and malformation of fetal coronary arteries (for a schematic diagram of the proposed pathway, see Fig. 7). Of note, inhibition of ROS production by NAC restores epicardial EMT and prevents malformation of coronary arteries in the fetal hearts of mice with pregestational diabetes. The study suggests a critical role of ROS signaling in coronary artery malformation during pregestational diabetes.

The embryonic epicardium is a major contributor to coronary artery development (29,30). In this regard, cells from the epicardium undergo EMT and become EPDCs, which then differentiate into vascular smooth muscle cells and cardiac fibroblasts, leading to the formation of coronary vessels. To study whether pregestational diabetes affects epicardial formation, immunohistochemical analysis was performed. Although there were no significant changes in proepicardial organ progenitor cell numbers, the number of epicardial cells was reduced, and the epicardium was detached from the myocardium with decreases in epicardial cell proliferation at E12.5 and E14.5 in the hearts of fetuses of diabetic mothers. Epicardial attachment to the myocardium is critical to epicardial cell proliferation and formation of the epicardium (31). The current study shows that pregestational diabetes inhibits epicardial development likely through an interruption of epicardial attachment to the myocardium and a reduction of epicardial cell proliferation. The number of epicardial cells undergoing apoptosis was low (0.3%), and no significant difference was observed at E12.5 between control and the offspring of diabetic mice. However, whether pregestational diabetes increases epicardial cell apoptosis in the offspring at other stages of development remains to be determined.

Epicardial EMT is a critical process in coronary artery development (32). In the current study, we demonstrate that pregestational diabetes decreases epicardial EMT. This is supported by the following experimental data. First, the number of Wt1⁺ cells, which are EPDCs from epicardial EMT, was reduced in the subepicardial space and in the compact myocardium in the offspring of diabetic mice. Second, high glucose levels inhibited epicardial cell outgrowth in cultured heart explants, and third, fate mapping analysis revealed a significant reduction of Wt1⁺ cell lineage in the fetal hearts of pregestational diabetic dams. The Wt1^CreERT2 line has been shown to label the epicardium and its derivatives (33). The decreased Wt1⁺ lineage in fetal hearts is consistent with our hypothesis that pregestational diabetes impairs epicardial EMT. To our knowledge, this is the first experimental evidence to show that pregestational diabetes inhibits epicardial EMT in the fetal heart.

HIF-1α is a transcription factor that promotes vasculogenesis during embryonic development (34). To this end, HIF-1α signaling has been shown to regulate epicardial EMT and EPDC migration into the myocardium, both of which are critical in patterning the coronary vasculature during early cardiac vasculogenesis (27). The effects of HIF-1α are mediated by the expression of factors essential for coronary artery development, including Wt1 (17,35). Importantly, HIF-1α is also ROS sensitive. Although low levels of ROS increase HIF-1α expression and promote cardiovascular differentiation, high levels of ROS may inhibit HIF-1α activity by inhibiting the binding of coactivator p300 to HIF-1α (36,37). Furthermore, diabetes increases ROS levels through hyperglycemia and decreases the expression and activity of HIF-1α (38,39). In the current study, ROS levels were significantly increased, whereas HIF-1α expression and the number of Wt1⁺ epicardial cells and EPDCs were decreased in the hearts of embryos of pregestational diabetic mice. Of note, these changes were abrogated by NAC treatment, which reduces ROS levels in the embryonic heart. The data suggest that pregestational diabetes impairs the HIF-1α/Wt1 signaling pathway through elevated ROS levels in the fetal heart (Fig. 7).

Wt1 has been shown to regulate epicardial EMT through the expression of its downstream targets of Snail1, Slug, and RALDH2 (40–42). In line with our observation of reduced epicardial EMT, we further demonstrate that the expression of Snail1 and Slug is decreased in the embryonic hearts of pregestational diabetic mice. Retinoic acid feeds into an essential signaling pathway crucial for epicardial formation, epicardial attachment to the myocardium, myocardial growth and proliferation, and coronary artery development (43,44). bFGF is an important mediator in retinoic acid signaling that promotes epicardial EMT and vasculogenesis in the embryonic heart. The current results show that the expression levels of Aldh1a2, which encodes RALDH2, and its downstream target bFGF were diminished, suggesting impaired retinoic acid signaling in the fetal hearts of diabetic mothers. Thus, pregestational diabetes impairs both Wt1/Snail/Slug and Wt1/RALDH2/bFGF signaling pathways, which may contribute to decreased epicardial EMT and malformation of coronary arteries in the fetal heart (Fig. 7).

Under physiological conditions, ROS signaling regulates vasculogenesis (45). Although basal endogenous ROS levels are critical to normal vascular development, excessive ROS production during embryogenesis may impair vasculogenesis. In this regard, oxidative stress induced by maternal diabetes impairs vascularization in the yolk sac of rat embryos (46). NAC is a precursor of cysteine and decreases ROS levels through increases in GSH synthesis and antioxidant capacity (47). The current study shows that ROS and oxidative stress are increased in the fetal hearts of mice with pregestational diabetes. Of note, NAC treatment in diabetic females during gestation diminished ROS levels and restored expression of critical factors essential for epicardial growth and EMT. Importantly, NAC treatment rescued coronary artery malformation induced by pregestational diabetes. This study suggests a critical role of elevated ROS and its signaling in mediating coronary artery malformation in pregestational diabetes. However, whether NAC treatment normalizes GSH levels in fetal hearts is not known because GSH levels were not determined and other effects of NAC beyond ROS reductions cannot be excluded. Of note, antioxidant treatments prevent congenital heart defects induced by hyperglycemia or pregestational diabetes (20,48–50). The results are consistent with our recent findings that elevated ROS levels contribute to the abnormal development of congenital heart defects in the offspring of mice with pregestational diabetes, which is rescued by NAC treatment (20). Thus, NAC may have therapeutic potential in the prevention of congenital heart defects, including coronary artery malformation, in the offspring of mothers with pregestational diabetes.

In conclusion, the current study demonstrates that pregestational diabetes impairs epicardial EMT and coronary artery development in mice. These abnormalities are associated with increased ROS production and decreased HIF-1α/Wt1 signaling (Fig. 7). Of note, these abnormalities were all prevented by NAC treatment, suggesting a key role of ROS signaling in the malformation of coronary arteries resulting from pregestational diabetes. The coronary artery phenotype observed in the current study simulates human hypoplastic coronary artery disease, which carries a high risk of spontaneous myocardial infarction and sudden cardiac death (6). The current study suggests that pregestational diabetes could cause hypoplastic coronary arteries in mice. However, further clinical studies are required to determine whether pregestational diabetes increases the incidence of hypoplastic coronary artery disease in humans.

Funding. This study was supported by grant 000505 to Q.F. from the Heart and Stroke Foundation of Canadahttp://dx.doi.org/10.13039/100004411. Q.F. is a Heart and Stroke Foundation of Canada Career Investigator.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.M. contributed to the research design, experiments, data analysis, and drafting and revision of the manuscript. X.L. contributed to the research design, experiments, data analysis, and revision of the manuscript. M.L. contributed to the experiments. Q.F. contributed to the study concept and research design, interpretation of the results, and revision of the manuscript. H.M. and Q.F. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.