Hyperglycaemia-induced impairment of the autorhythmicity and gap junction activity of mouse embryonic stem cell-derived cardiomyocyte-like cells

Diabetes mellitus with hyperglycaemia is a major risk factor for malignant cardiac dysrhythmias. However, the underlying mechanisms remain unclear, especially during the embryonic developmental phase of the heart. This study investigated the effect of hyperglycaemia on the pulsatile activity of stem cell-derived cardiomyocytes. Mouse embryonic stem cells (mESCs) were differentiated into cardiac-like cells through embryoid body (EB) formation, in either baseline glucose or high glucose conditions. Action potentials (APs) were recorded using a voltage-sensitive fluorescent dye and gap junction activity was evaluated using scrape-loading lucifer yellow dye transfer assay. Molecular components were detected using immunocytochemistry and immunoblot analyses. High glucose decreased the spontaneous beating rate of EBs and shortened the duration of onset of quinidine-induced asystole. Furthermore, it altered AP amplitude, but not AP duration, and had no impact on neither the expression of the hyperpolarisation-activated cyclic nucleotide-gated isoform 4 (HCN4) channel nor on the EB beating rate response to ivabradine nor isoprenaline. High glucose also decreased both the intercellular spread of lucifer yellow within an EB and the expression of the cardiac gap junction protein connexin 43 as well as upregulated the expression of transforming growth factor beta 1 (TGF-β1) and phosphorylated Smad3. High glucose suppressed the autorhythmicity and gap junction conduction of mESC-derived cardiomyocytes, via mechanisms probably involving TGF-β1/Smad3 signalling. The results allude to glucotoxicity related proarrhythmic effects, with potential clinical implications in foetal diabetic cardiac disease.


Introduction
Cardiovascular complications of diabetes mellitus with inadequately controlled hyperglycaemia are a major cause of death in both adults and children (Dal Canto et al. 2019;Schafer et al. 2020). Children, in particular offspring of diabetic mothers, are at a great risk of developing life-long complications such as cardiomyopathy and malignant dysrhythmias (Hoodbhoy et al. 2019). Furthermore, foetal and neonatal dysrhythmias as a result of maternal gestational diabetes are more prevalent than generally reported (Pike et al. 2013), partly because, unlike cardiac structural defects, dysrhythmias are not readily diagnosed and not often investigated (Killen et al. 2014).
Dysrhythmias in foetuses exposed to diabetic conditions occur through a maladaptive pathological electrical remodelling of the foetal heart in response to hyperglycaemia and other diabetic metabolic abnormalities (Basu and Garg 2018;Nakano et al. 2021;Pedra et al. 2002), but the mechanisms remain unclear. Although such foetal cardiac complications persist and progress in later life, they are often only detected when other super-added features of diabetes such as myocardial infarction and heart failure become apparent (Mehta et al. 2020). In adult hearts, a key process in diabetic cardiac remodelling is the re-activation of the foetal gene programme to try to induce cardiomyocyte proliferation and replace damaged cells (Lipsett et al. 2019;Rosa et al. 2013). However, adult hearts differ from those of foetuses in that cardiomyocytes exit the cell cycle at birth (Bergmann et al. 1 3 2015), and therefore adult cells lose the capacity to proliferate and repair optimally during diabetic remodelling (Foglia and Poss 2016;Lazar et al. 2017). As such, there is limited knowledge on therapeutic interventions for diabetic foetal cardiac diseases, since the cardiac developmental pathophysiology is not fully understood.
We previously established and characterised a stem cellderived cardiac cellular model that mimics cardiac cellular development, and showed that hyperglycaemia suppressed the cardiac differentiation of mouse embryonic stem cells (mESCs) and induced cardiomyocyte contractile dysfunction and disruption of the myofibrillar architecture, features that were attributed to oxidative stress and mitochondrialdependent apoptosis (Aboalgasm et al. 2021a, b). However, the effect of hyperglycaemia on the spontaneous beating characteristics of those cardiac-like cells remains unclear. In the present study, we used the mESC-derived cardiac cellular hyperglycaemic model to study effects on autorhythmicity and gap junction function, and explored the possible underlying mechanisms.
The mESCs were differentiated into cardiomyocyte-like cells using the hanging drop method and embryoid body (EB) formation as previously described (Aboalgasm et al. 2021b). Briefly, mESCs were seeded in LIF-free differentiation medium (Lee et al. 2011;Wobus et al. 2002) containing DMEM/25 mM glucose, supplemented with 10% FBS, 1% glutamax, 1% penicillin/streptomycin, and 0.1% β-mercaptoethanol. In order to minimise the heterogeneity of the EBs in terms of the proportion of cardiomyocytes per EB, a fixed number of mESCs (1000 cells) were seeded into 20 µl differentiation medium (Lee et al. 2011;Wobus et al. 2002). The EBs were expanded further in suspension culture and thereafter plated onto 0.1% gelatin-coated coverslips or glass-bottomed imaging dishes for further culture. The first day of the hanging drop protocol was taken as day one of the differentiation protocol.

Experimental treatments
The control mESCs were differentiated in DMEM medium containing baseline (25 mM) glucose, which although higher than physiological blood glucose (e.g., 5.5 mM), is the default concentration suitable for mESC culture (Ali et al. 2004;Crespo et al. 2010;Guan et al. 1999). Compared to 5.5 mM glucose (equivalent to physiological blood glucose), for in vitro mESC cardiac differentiation studies, a glucose level of 25 mM has been shown to provide beneficial reactive oxygen species that are needed for cardiac differentiation (Ali et al. 2004) and to produce a more stable cardiomyocyte phenotype with non-disrupted sarcomeric organisation (Guan et al. 1999). The high glucose mESCs were differentiated in 50 mM-glucose (Han et al. 2015;Xu et al. 2020), which, when considering the starting level of 25 mM glucose, is only double the baseline glucose, similar to the diabetic threshold (i.e., from 5.5 mM to 11 mM blood glucose). Osmolality changes due to glucose addition were not adjusted for in this and in other in vitro studies (Aboalgasm et al. 2021a, b;Han et al. 2015;Xu et al. 2020), thereby mimicking in vivo diabetic hyperglycaemia with uncorrected osmolality changes. However, to verify the impact of osmolality changes, a subset of EBs was cultured in 25-mM glucose combined with 25-mM d-mannitol. There was no difference in the beating profiles of EBs between that group and 25-mM glucose (% beating EBs [mean, SD]: 76 ± 15) for 25-mM glucose alone versus 71 ± 8 for 25-mM glucose combined with 25-mM d-mannitol; P > 0.05, N = 6 EBs per group, 3 independent cell culture batches). As such, we chose not to control for osmolality changes in subsequent experiments. Measurements on EBs were performed after approximately 3 weeks (days 17-19) of the differentiation protocol, at a stage where the differentiated cardiomyocytes have been shown to have foetal-and neonatal-like cardiac myofibrillar structural-and electrophysiological properties . Each test drug (dissolved in water) ivabradine (10 μM), isoprenaline (12 μM), or quinidine (1 μM) was applied in the culture medium under standard conditions. Still images and time-lapse images of beating EBs were captured on an EVOS™ M5000 imaging system (ThermoFisher Scientific). Images were analysed with ImageJ (NIH, USA) and a motion-detecting macro Myo-cyter™ (Grune et al. 2019), which produces a graphical output of the amplitude of cellular contraction of over time.

Action potential measurements
Action potentials (APs) were recorded using a voltage-sensitive fluorescent dye di-4-ANEPPS (#D1199, ThermoFisher Scientific) that is used to study AP profile (Maeda et al. 2007). EBs (plated on gelatin-coated glass-bottomed imaging dishes) were loaded with 10-µM di-4-ANEPPS (dissolved in DMSO) in the culture medium and incubated for 15-20 min and mounted in an incubation chamber (5% CO 2 at 37 ℃) on the stage of a Carl Zeiss LSM880 Airyscan™ confocal microscope for imaging using the ZEN software (Zeiss.com). Fluorescent signals were analysed with ImageJ (NIH, USA) using the LC_Pro Plugin (Francis et al. 2014), from which AP amplitude and AP duration at 50% repolarisation (APD 50 ) and at 90% repolarisation (APD 90 ) were quantified as the averaged data of APs recorded over 6 s. The relative fluorescence intensity was expressed as fluorescence (F) normalised to the baseline fluorescence (F 0 ).

Scrape-loading dye transfer assay
The scrape-loading dye transfer assay adapted from (Babica et al. 2016) was used to assess gap-junction intercellular communication within an EB. A clean linear cut was made across an individual EB plated on a glass coverslip using a curved surgical scalpel blade. The EB was then loaded with lucifer yellow fluorescent dye added into the culture medium (1 mg/ml, dissolved in water; Sigma, SA) and incubated for 5 min under standard culture conditions. The EB on the coverslip was rinsed, fixed with 10% formalin solution, and imaged on the EVOS™ M5000 fluorescence microscope. The distance of lucifer yellow dye spread was analysed using ImageJ (NIH, USA).

Immunocytochemistry
Immunocytochemistry was performed on adherent EBs as previously described (Aboalgasm et al. 2021a, b). Samples were fixed with 4% paraformaldehyde, permeabilized with ice-cold methanol, and blocked using 3% bovine serum albumin (BSA) with 0.01% Triton X-100 in phosphatebuffered saline (PBS). Samples were incubated overnight (at 4 ℃ in PBS with 1% BSA) with primary antibodies directed against α-actinin 2 (dilution 1:250; #701914, ThermoFisher Scientific) or connexin 43 (dilution 1:200; #13-8300, ThermoFisher Scientific). According to the supplier's technical datasheet, the α-actinin 2 monoclonal antibody was verified using relative expression, whereas the connexin 43 monoclonal antibody was verified via the knock-out of the target protein. To exclude non-specific binding of secondary antibodies, primary antibodies were omitted in negative controls. The samples were then incubated with a fluorophore-conjugated secondary antibody (Alexa Fluor 488, dilution 1:5000; #715-546-150, Amersham or Cy3, dilution 1:1000; #711-166-152, Amersham) for 2 h at room temperature. The samples were counterstained with Hoechst 33258 (0.5 μg/ml; Sigma, SA) and imaged on a Carl Zeiss LSM880 Airyscan™ confocal microscope using an oil immersion lens (63× magnification; 1.4 numerical aperture). Double-coloured images (Fig. 1b) were generated using lasers with excitation wavelengths of 405 nm and 561 nm, whereas triple-coloured images (Fig. 4e) included an additional laser with an excitation wavelength of 488 nm. Confocal microscopy images were acquired using ZEN software (Zeiss.com), and the image-acquisition settings were kept constant across different slides via the "Reuse" configuration feature. Images were analysed using ImageJ (NIH, USA). High glucose impairs embryoid body (EB) beating characteristics. a Representative transmitted light-microscopy image of an EB cultured under baseline (25 mM) glucose conditions. b Confocal microscopy merged image of a cell dissociated from an EB and stained with α-actinin 2 and Hoechst. c Representative tracing of an EB pulsatile activity (as detected by the Myocyter™ programme), under baseline (25 mM) glucose conditions. Double arrowheads indicate the duration of beat-to-beat intervals of two consecutive beats. d-f Summary data (from three independent cell culture batches) of EB beating rate (N = 20 EBs per group), the standard deviation of the EB beat-to-beat interval (N ≥ 20 EBs per group), and onset of 1-µM quinidine-induced asystole (N = 4 EBs per group) under different glucose conditions. Data are shown as box plot. Scale bar 600 µm (a) and 10 µm (b). Glu, glucose; a.u., arbitrary unit

Data analysis
Data are expressed as box plot and N indicates the number of EBs studied. Graphical artwork was made using Microcal Origin 6.1 programme. Statistical analysis was conducted using Statistica (version 13). A Shapiro-Wilk test for normality was used to test the distribution of variables. For parametric data, an unpaired t test was used to compare differences between glucose groups, whereas a Mann-Whitney U test was used for non-parametric data. A value of P ≤ 0.05 was considered statistically significant.

Stem cell cardiac differentiation and effect of hyperglycaemia on beating characteristics
Differentiated mESCs formed distinct, spherically-shaped spontaneously beating EBs (Fig. 1a). The EBs were made up of cells that stained positively for the cardiac sarcomeric protein α-actinin 2, which showed a typical striated pattern (Fig. 1b). The EB pulsatile activity, shown as a graphical output of the Myocyter™ (Grune et al. 2019) programme, had distinct peaks and troughs, with identifiable beat-to-beat intervals (Fig. 1c). A high glucose concentration in the culture medium decreased the EB beating rate (P = 0.005 versus baseline glucose; Fig. 1d). This result is consistent with the autorhythmic defects in diabetic adult hearts (Howarth et al. 2007;Malone et al. 2007;Zhang et al. 2019) and in other mESC-derived cardiac-like cells (Yang et al. 2016). However, high glucose did not alter the standard deviation of the beat-to-beat interval (P = 0.23 versus baseline glucose; Fig. 1e), which is an indicator of the intrinsic variability of cellular pacemaker activity attributed to fluctuations in the baseline transmembrane potential (Ponard et al. 2007) and has been shown to be present in mESC-derived cardiomyocytes (Niehoff et al. 2016). However, in animal studies, pacemaker cells isolated from diabetic rat hearts showed large fluctuations in the beat-to-beat intervals that manifested clinically as arrhythmias (Albarado-Ibanez et al. 2013;Soltysinska et al. 2014). Furthermore, the application of quinidine (1 µM), which among its actions, prominently blocks voltage-gated Na + channels (Roden 2014), induced asystole significantly quicker in EBs cultured under high glucose compared to those in baseline glucose (P = 0.040; Fig. 1f). Although the role of Na + channel was not further evaluated in this study, taken together with the altered beating rate, the results suggest pro-arrhythmic effects of hyperglycaemia.
We further evaluated the sensitivity of EBs to either an inhibitor of the cardiac pacemaker HCN channel ivabradine (Thollon et al. 2007) or the adrenergic receptor/cAMP-mediated pacemaker channel stimulant isoprenaline (Bucchi et al. 2003). High glucose had no significant effect on the extent of neither the 10-µM ivabradine-induced decrease in the EB beating rate nor the 12-µM isoprenaline-induced increase in beating rate (P > 0.05 versus baseline glucose for each drug; Fig. 2a, b). Furthermore, immunoblot analysis showed no significant difference in the expression of the main HCN channel isoform found in the sinoatrial node HCN4 (Hennis et al. 2022) between the two glucose groups (P = 0.61; Fig. 2c, d). These results suggest that the decreased beating rate in high glucose seemed to be unrelated to changes in pacemaker channel sensitivity or expression. These findings are in contrast with those in an animal study in which pacemaker tissue isolated from diabetic rats showed a diminished response to ivabradine (Huang et al. 2017). However, since only the expression of the major sinoatrial node pacemaker channel HCN4 was evaluated, the involvement of the other pacemaker channels such as T-type Ca 2+ cannot be ruled out.

Alteration of action potentials (APs)
Given that hyperglycaemia moderately reduced (rather than abolish) the EB pulsatile activity, we hypothesised that hyperglycaemia induced defects in the generation or conduction of electrical activity in otherwise viable cells. For APs, the fluorescence signals emitted by a voltage-sensitive dye (di-4-ANEPPS) in EBs of both glucose groups typically showed AP waveforms with rapid depolarisation and repolarisation phases, without a noticeable plateau phase (Fig. 3a, b). The AP amplitude was significantly larger in high glucose conditions compared to baseline glucose (P = 0.007; Fig. 3c). However, there was no significant difference in the APD 90 /APD 50 ratio, which is the ratio of the AP duration (APD) at 50% repolarisation (APD 50 ) to the APD at 90% repolarisation (APD 90 ) (P = 0.41; Fig. 3d).
In contrast to the present findings, diabetic adult rat pacemaker tissue generally show decreased AP amplitude and prolonged AP duration (Gallego et al. 2021;Howarth et al. 2007;Liu et al. 2012;Ozturk et al. 2021), features attributed to the downregulation of several ion channels contributing to the AP (Ferdous et al. 2016;Howarth et al. 2018). As such, the differences between the AP parameters in the present study and those in the other studies could reflect the unique differences in the glucose sensitivity in developing cells versus adult cardiac cells. However, the reason for the increased AP amplitude observed in the present study is still unclear. A possible explanation could be a change in the resting membrane potential; however, in the present study, a voltage-sensitive dye was used to record APs (which has a limitation in that it only detects the relative changes in depolarisation or depolarisation), so the absolute resting potential could not be determined. Alternatively, we speculate that there could be a change in the depolarising currents such as the voltage-gated Na + current, given the abovementioned altered quinidine sensitivity. Finally, because of the downregulation of repolarising K + currents reported in diabetes (Howarth et al. 2018), the unopposed depolarising AP upstroke currents could produce a greater AP overshoot.

Disruption of gap junction conduction
Although there was no significant difference in the size of the beating area in EBs cultured in high glucose compared to those in baseline glucose (P = 0.59; Fig. 4a, b), high glucose significantly decreased the distance of the spread of lucifer yellow within an EB (P = 0.014 versus baseline glucose; Fig. 4c, d). Furthermore, qualitatively, there was a decrease in the expression of the major cardiac gap junction protein connexin 43, since connexin 43 was detected by immunocytochemistry only in two out of five EBs in high glucose conditions, whereas the protein could be detected in all seven EBs in baseline glucose conditions (Fig. 4e). No connexin 43 signal was detected in the negative control in which . c, d Summary data of AP amplitude and APD 50 /APD 90 (N = 5 EBs per group). AP data were obtained using the LC_Pro programme, and although the baseline potential had some instabilities due to background noise, the signal-to-noise ratio was optimal for the detection of AP parameters. Data are shown as box plot. APD 50 and APD 90 , AP duration at 50% and 90% of repolarisation, respectively. Glu, glucose primary antibody was omitted (not illustrated). Nonetheless, similar to the EBs in baseline glucose conditions, the EBs in high glucose that lacked connexin 43 still stained positively for the cardiac sarcomeric protein α-actinin 2, which identifies cardiomyocytes within an EB (Fig. 4e). Similarly, immunoblot analysis showed a significant reduction in the expression of connexin 43 protein by high glucose (P = 0.03 versus baseline glucose; Fig. 4f, g).
These results indicate intercellular conduction defects that are consistent with the hyperglycaemic cell structural disruption previously observed in the same model (Aboalgasm et al. 2021a). The downregulation of connexin 43 has also been reported in cultured H9C2 cells exposed to high glucose (Han et al. 2015), and in the rat diabetic heart (Zhang et al. 2019). Notably, the disruption of gap junctions could, in part, also account for the beating rate disturbances seen in this study, since connexins are proposed to mediate impulses within the pacemaker tissue itself, as part of the impulse generation process (Boyett et al. 2006;Ferdous et al. 2016).

Modulation of the transforming growth factor beta 1 (TGF-β1) pathway
We further investigated the effect of hyperglycaemia on TGF-β1, a cytokine-like growth factor that is implicated in the pathological cardiac remodelling that involves the disruption of the cardiac conduction system (Derangeon et al. 2017;Dzialo et al. 2018). Western blot analysis showed that high glucose upregulated the expression of TGF-β1, relative to that of the housekeeping protein α-tubulin P = 0.04 versus baseline glucose; Fig. 5a, c). In addition, high glucose significantly increased the expression (relative to β-actin) of phosphorylated Smad3, a TGF-β1 downstream pathway signalling molecule (P = 0.02 versus baseline glucose; Fig. 5b, d). The finding is consistent with that in diabetic rat hearts Effect of high glucose on the transforming growth factor beta 1 (TGF-β1) pathway. a, b Representative Western blot images of TGF-β1 and phosphorylated Smad3 (p-Smad3) and the housekeeping proteins α-tubulin and β-actin. c, d Summary data of TGF-β1 expression normalised to α-tubulin (N = 3 replicates per group) and of p-Smad3 normalised to β-actin (N = 3 replicates per group). Data are shown as box plot. Glu, glucose 1 3 in which the TGF-β1/Smad3 pathway was also activated in inducing fibrosis (Jia et al. 2020), and suggests that the pathway could play a key role in the high glucose effects observed in the present study. In diabetic hearts, TGF-β1 is known to be involved in the remodelling of cardiac conduction tissue, via the induction of interstitial fibrosis (Liu et al. 2012) or cardiomyocyte hypertrophy, with altered transcellular axial electrical resistance (Dobaczewski et al. 2011). However, the mechanisms underlying TGF-β1mediated effects seen in the present study require further investigations.

Limitations of the study
Limitations of this study include the clinical relevance of the unphysiologically high glucose (50 mM). However, such a high glucose has been used as an appropriate concentration in other in vitro model studies (Han et al. 2015;Xu et al. 2020) where the standard baseline glucose is supraphysiological (e.g., 25 mM). Nonetheless, the present results still need further validation in an in vivo model. Furthermore, although the heterogeneity of EBs was minimised by fixing the number of mESCs seeded into each hanging drop (Lee et al. 2011;Wobus et al. 2002), the EBs may contain other active non-cardiomyocyte cells, of which the nature was not determined. In order to offset the contamination by non-cardiomyocytes in in vitro models, other high-cardiomyocyte yield stem cell differentiation protocols have been developed (Lian et al. 2013) that would be most applicable where purely cardiomyocyte-to-cardiomyocyte interactions are to be studied. However, given that in a real heart the proportion of cardiomyocytes is only about 30% of the total cells present (Banerjee et al. 2007), the EB model (even with a relatively low cardiomyocyte yield) provides a multicellular, cardiac tissue-like environment for cardiomyocytes to interact with both cardiomyocytes and non-cardiomyocyte cells, as long as the proportion of cardiomyocytes is kept relatively constant.

Conclusions
In conclusion, hyperglycaemia suppressed the autorhythmicity and gap junction function of mESC-derived cardiac-like cells, probably via the TGF-β1/Smad3 pathway. The results suggest that hyperglycaemia induces unique proarrhythmic effects in the developing cardiac cells, with potential clinical implications in foetal diabetic cardiac disease.