Hypoglycemia Induced Mitochondrial Connexin-43 Accumulation Aggravates Diabetic Cardiomyopathy

Background: Diabetic cardiomyopathy (DCM) is a complex multifaceted disease 25 responsible for elevated hospitalization and mortality in patients with diabetes mellitus 26 (DM). DCM patients exhibit subclinical diastolic dysfunction, progression towards 27 systolic impairment, and abnormal electrophysiology. Hypoglycemia events that occur 28 spontaneously or due to excess insulin administration threaten the lives of DM patients 29 – with the increased risk of sudden death. However, the molecular underpinnings of 30 hypoglycemia-aggravated DCM remain to be elucidated. 31 Methods and Results: Here we used the established streptozotocin-induced type 1 33 diabetic cardiomyopathy (T1 DCM) murine model to investigate how hypoglycemia 34 aggravates DCM progression. We showed that chronic hyper- or hypoglycemic 35 challenges dampened cardiac diastolic function in vivo as well as myocardial 36 contractility and calcium handling in isolated cardiomyocytes. Similar contractile 37 defects were recapitulated using neonatal mouse ventricular myocytes (NMVMs) 38 under glucose fluctuation challenges. Using immunoprecipitation mass spectrometry, 39 we identified and validated that hypoglycemia challenge activates the MEK/ERK and 40 PI3K/Akt pathways which results in Cx43 phosphorylation by Src protein in 41 cardiomyocytes. Cx43 dissociation and accumulation at mitochondrial inner 42 membrane was confirmed both in human and murine cardiomyocytes. To determine 43 causality, we overexpressed a mitochondrial targeting Cx43 (mtCx43) using AAV2. At 44 normal blood glucose levels, mtCx43 overexpression recapitulated cardiomyocytes contractile deficiencies, cardiac diastolic dysfunction as well as aberrant 46 electrophysiology both in vitro as well as in vivo . 47 48 Conclusions: Hypoglycemia challenges results in the accumulation of mtCx43 49 through the MEK/ERK/Src and PI3K/Akt/Src pathways. We provide evidence that Cx43 50 mislocalization is present in diabetes mellitus patient hearts, STZ-induced DCM murine 51 model, and glucose fluctuation challenged NMVMs. Mechanistically, we demonstrated 52 that mtCx43 is responsible for inducing aberrant contraction and disrupts 53 electrophysiology in cardiomyocytes and our results support targeting of mtCx43 in 54 treating DCM. 55 56 Translational perspective: Severe hypoglycemia drives cardiac dysfunction and 57 aggressive ventricular arrhythmias in patients with DCM that leads to sudden cardiac 58 death. Here we demonstrate that Cx43 mislocalization to mitochondria occurs upon 59 hypoglycemic challenge and mtCx43 accumulation is responsible for cardiac diastolic 60 dysfunction, cardiomyocyte contractile dysfunction, and aberrant electrophysiology in 61 vivo . Our findings give support for therapeutic targeting of MEK/ERK/Src and 62 PI3K/Akt/Src pathways to prevent mtCx43-driven DCM.


Abstract 24
Background: Diabetic cardiomyopathy (DCM) is a complex multifaceted disease 25 responsible for elevated hospitalization and mortality in patients with diabetes mellitus 26 (DM). DCM patients exhibit subclinical diastolic dysfunction, progression towards 27 systolic impairment, and abnormal electrophysiology. Hypoglycemia events that occur 28 spontaneously or due to excess insulin administration threaten the lives of DM patients 29 -with the increased risk of sudden death. However, the molecular underpinnings of 30 hypoglycemia-aggravated DCM remain to be elucidated. are present in patients with DM and is clinically defined as diabetic cardiomyopathy 72 (DCM) [2]. DCM hearts are characterized by cardiac remodeling, early onset of 73 diastolic dysfunction followed by systolic impairment, and eventually progresses to 74 heart failure with reduced ejection fraction (HFrEF) [3]. Compared to coronary artery 75 disease induced dilated cardiomyopathy, DCM has worse prognosis [4,5]. Compared 76 to other cardiomyopathies, DCM exhibit metabolic dysfunction, electrophysiology, and 77 insulin resistance [6, 7] -making DCM management much more challenging. Moreover,78 our understanding of the molecular underpinnings of DCM remains limited. 79 80 Hypoglycemia is a major challenge in DM management [8] that significantly increases 81 the mortality rate in both type 1 and type 2 DM patients [9,10]. Severe hypoglycemia,82 defined as blood glucose ≤ 3.0 mM, can result in cognitive confusion, loss of 83 consciousness, seizures, and even death in both young and elderly DM patients [ 11 -84 14]. Prolonged hypoglycemia induced neuroglycopenia are rare; most fatal 85 hypoglycemic episodes result in cardiac dysfunction, especially ventricular arrhythmias 86 [14,15]. Experimentally induced hypoglycemic events in patients with type 1 or type 2 87 DM resulted in pro-arrhythmogenic cardiac repolarization with prolonged QT intervals 88 [16,17]. In rodent models, severe hypoglycemia (blood glucose < 1.0 mM) leads to 89 prolonged QT interval, ventricular ectopy, and high-degree atrioventricular blockage 90 [18]. How hypoglycemia can induce cardiac dysfunction and abnormal 91 electrophysiology remains to be elucidated. 92 93 Gap junctions (GJs) are channels that directly connect two adjacent cells with two 94 hemichannels, allowing passage of ions (Na + , K + , Ca 2+ ), and proteins and molecules 95 less than 1.5 kDa [19] and are tightly regulated in response to intraceullular and 96 extracellular signals [20,21] To study the effect of hypoglycemia on cardiac function, we first induced DCM (Fig. 1a) 143 in mice using the established STZ model (150mg/kg). Two weeks post-injection, the 144 mice exhibited an average blood glucose level of 6.06 ± 0.30 and 29.38 ± 1.15 mM in 145 Control and DCM groups, respectively (Additional file: Figure S1a). Cardiac function 146 was evaluated using echocardiography at 2-and 8-weeks post STZ injection (Fig. 1b). 147 At 2-weeks post STZ injection, no significant differences in left ventricle (LV) ejection 148 fraction (EF), fractional shortening (FS), early to late ventricular filling velocity ratio (E/A 149 ratio), or early mitral inflow velocity to early diastolic mitral annulus velocity ratio (E/E' 150 ratio) between DCM and Control mice (Fig. 1b, c) was observed. 8-weeks post STZ 151 injection, DCM animals exhibited diastolic dysfunction marked by a significant 152 decrease in E/A ratio and an increase in E/E' ratio, but no significant difference in EF 153 and FS compared to Control animals (Fig. 1c). DCM animals were significantly 154 emaciated and depilated after 10 weeks of continued STZ injection (Additional file: 155 Figure S1b, c). 156

157
We injected DCM animals with insulin to mimic hypoglycemic challenge (HDCM 158 group; Fig. 1a). 10 DCM mice were randomly selected and injected with insulin to 159 induce hypoglycemia for 120 min prior to sacrifice. Unlike DCM and Control animals, 160 HDCM mice exhibited a significant decrease in blood glucose levels (Control: 6.35 ± 161 0.34 mM, DCM: 26.48 ± 1.30 mM, HDCM: 2.28 ± 0.20 mM; Additional file: Figure S1d).  Figure S2a). Since DCM cardiomyocytes 177 experienced a hyperglycemic environment prior to isolation, we expected that a HG 178 culturing condition would be better tolerated by DCM cardiomyocytes while a low 179 glucose culturing condition would mimic a hypoglycemia challenge. First, 180 cardiomyocytes isolated from Control animals, HG or LG culturing conditions did not 181 alter sarcomere shortening (Fig. 2c) or contraction peak height (Fig. 2d). HG medium 182 significantly decreased contraction time to peak 90 (  Further, these data indicates that severe hyperglycemic or hypoglycemia challenge is 202 well tolerated by healthy cardiomyocytes but is loss in DCM cardiomyocytes. 203 204

Recapitulation of aberrant contraction and electrophysiology in HLG challenged 205 neonatal mouse ventricular myocytes 206
To further study the underlying mechanism that drives cardiac dysfunction upon 207 hypoglycemic challenge, we established isolated NMVMs and treated the 208 cardiomyocytes with either NG, HG, or high glucose following low glucose group (HLG) 209 medium that mirrored in vivo blood glucose conditions (Fig. 2g). Our NMVM cultures 210 contained ≥ 80% cTnT + cardiomyocytes evaluated by immunofluorescent staining 211 (Additional file: Figure S3a). Using live cell imaging we measured the mean contraction 212 velocity and beating frequency of NMVMs using a previously established algorithm [30,213 31]. In a 4-hour time course study, we subjected NMVMs to either NG, HG, or HLG 214 challenge and measured changes in contractility (Additional file: Figure S3b, c). Upon 215 HG challenge, there was a slight decrease in mean velocity and beat rate compared 216 to HG baseline (0 h) and NG (Additional file: Figure S3b, c). HLG treatment resulted in 217 a significant reduction in both contraction velocity and beat rate compared to NG or 218 HG treatment (Additional file: Figure S3b, c). 219 Next, we evaluated excitation-contraction coupling in NMVMs by measuring 220 impedance (IMP; surrogate for contractile frequency) and extracellular field potential 221 (EFP, surrogate for cell surface voltage) under exogenous electrical stimulation (14 222 volts) at 1Hz frequency ( Fig. 2h, i). Under continuous pacing, there was no difference 223 in the IMP baseline among the three groups (Additional file: Figure S3d). Compared to 224 NG, HG treatment did not induce any changes in IMP amplitude (Fig. 2j) or beat rate 225 (Fig. 2k), but a significant decrease in EFP (Fig. 2l). In the HG phase of the HLG group, 226 we observed no differences in IMP amplitude and beat rate but a significant decrease 227 in EFP (Fig. 2j, l). Upon hypoglycemic challenge (media exchange at 2-hour mark), we 228 observed a significant decrease in IMP amplitude, beating frequency and EFP 229 compared to NG and Controls (Fig. 2j with NG and HG condition [28]. To further illustrate it, we used immunoelectron 242 microscopy to examined Cx43 protein localization in Control, DCM and HDCM murine 243 cardiomyocytes (Fig. 3a). In Control cardiomyocytes, Cx43 localized to cell-cell 244 junctions (Fig. 3a, white arrows). Interestingly, we observed mtCx43 aggregation in 245 DCM and HDCM murine cardiomyocytes but not in Controls (Fig. 3a, yellow arrows). 246 Compared to Controls, DCM and HDCM murine cardiomyocytes exhibited a significant 247 decrease in number of Cx43 located at the cell-cell junction (Fig. 3b). There was a 248 significant increase in Cx43 migratory distance away from cell membrane interface in 249 HDCM cardiomyocytes compared to DCM and Control cardiomyocytes (Fig. 3c). Next,250 we quantified number of mtCx43 per mitochondrial cross-sectional area as a readout 251 of mtCx43 density. We found although abnormal aggregation of mtCx43 could be 252 observed in cardiomyocytes of both DCM and HDCM mice, mtCx43 density is 253 significantly greater in HDCM cardiomyocytes compared to DCM cardiomyocytes (Fig.  254 3d). These data indicates that a certain amount of mtCx43 in DCM mice can maintain 255 a compensatory balance, but too much mtCx43 in HDCM mice breaks this balance, 256 resulting in a rapid decline in myocardial cell function. 257

258
To verify and quantify our immunoelectron microscopy findings, we performed 259 immunofluorescence staining for N-cadherin (marking cell-cell junctions), translocase 260 of outer mitochondrial membrane 20 (Tomm20; marking mitochondria), and Cx43 in 261 NMVMs (Fig. 3e). Similar to our in vivo observations, Cx43 and N-cadherin co-262 localization in NMVMs decreased significantly post 4-hour HG treatment and was 263 further exacerbated in the HLG group (Fig. 3f). A similar trend was observed for the 264 mean Cx43 fluorescence intensity (Additional file: Figure S4a, b), while no change in 265 N-cadherin fluorescence intensity was observed (Additional file: Figure S4c). Notably,266 Cx43 and Tomm 20 co-localization significantly increased upon HLG treatment (Fig. 267 3g). Furthermore, we observed a loss of membrane Cx43 by immunoblotting of the 268 hydrophobic fraction at 2-hour post-hypoglycemic challenge (Fig. 3h). To verify 269 whether Cx43 dissociation and mitochondrial aggregation upon hypoglycemic 270 challenge is conserved in humans, we performed a similar analysis on DM and Non-271 DM patient cardiac sections (Fig. 3i). Consistent with our results, we observed a 272 significant decrease in Cx43 and N-cadherin co-localization (Fig. 3j), a significant 273 increase in Cx43 and Tomm 20 co-localization (Fig. 3k), and a significant decrease in 274 Cx43 mean fluorescence intensity (Fig. 3l)  To identify proteins responsible for Cx43 internalization and translocation to 282 mitochondria upon hypoglycemia challenge, we performed co-immunoprecipitation 283 coupled with mass spectrometry (Co-IP/MS) using antibody against Cx43 protein using 284 NMVMs cultured under NG, HG and HLG conditions. Proteins mapped were used for 285 downstream gene ontology enrichment analysis. Compared to those under NG and 286 HG conditions, kinase activity, protein phosphorylation, cell-cell adhesion, and 287 cytoskeleton-associated proteins were significantly increased under HLG conditions 288 (Fig. 4a). The percentage and propensity score matching coverage of mitochondria-289 associated proteins bound to Cx43 increased significantly upon HLG induction (Fig.  290 4b, c and Additional file: Figure S5a) These results are consistent with the increase in 291 mtCx43 accumulation upon hypoglycemia induction (Fig. 4a-g) Furthermore, Src and 292 Src-interacting proteins were enriched in our Cx43 Co-IP/MS analysis (Fig. 4d, e, 293 Additional file: Figure S5b, c). In addition, immunofluorescence staining data showed 294 that HLG culture could significantly increase the correlation coefficient between Src 295 and Cx43 (Additional file: Figure S5d, e). To test if Cx43 translocation was mediated 296 by Src, we treated HLG NMVMs with Saracatinib, a Src inhibitor (Fig. 4f). Saracatinib 297 did not change the overall expression of Cx43 (Fig. 4g), which suggest that Saracatinib 298 does not interfere with Cx43 turnover. Moreover, Saracatinib significantly increased 299 Cx43 and N-cadherin co-localization (Fig. 4h) and decreased Cx43 and Tomm20 co-300 localization ( Fig. 4i) in NMVMs cultured under HLG conditions. Lastly, Saracatinib 301 restored Cx43 protein levels located at the membrane fraction (Fig. 4j). 302

Overexpression of mitochondrial Cx43 results in worse cardiac dysfunction and 341 risk of arrhythmia susceptibility 342
To determine whether mtCx43 is sufficient to increase the risk of cardiac dysfunction 343 and susceptibility of arrhythmias, we constructed an AAV overexpression vector 344 containing a mitochondrial localization sequence fused to either Cx43 or EGFP (Fig.  345 7a; Additional file: Figure S6a). First, we confirmed that overexpression of mtCx43 in 346 NMVMs showed an increase in the Cx43 signal at mitochondria by 347 immunofluorescence (Additional file: Figure S6b, yellow arrows). Functionally, 348 overexpression of mtCx43 in NMVMs resulted in a significant increase in base IMP 349 (Fig. 7b) and decrease in IMP (Fig. 7c) as well as beat rate (Fig. 7d), but no change in 350 EFP (Fig. 7e) compared to untreated or mtEGFP Controls. We also observed a 351 significant decrease in the contraction velocity and beat rate of mtCx43 overexpressing 352 NMVMs under no pacing stimulation (Additional file: Figure S6c To further validate whether overexpression of mtCx43 can lead to cardiac 355 dysfunction, wild-type animals were injected with AAV2 viruses overexpressing 356 mtCx43, expression was confirmed by immunoelectron microscopy (Additional file: 357 Figure S6f). Cardiac functions were evaluated by echocardiography (Fig. 7f). In 358 mtCx43 overexpression animals, we did not observe any differences in EF, FS, or the 359 E/A ratio (Fig. 7g-i) compared to mtEGFP Controls; however, we observed a significant 360 decrease in the E/E' ratio in the mtCx43 overexpression group compared to that in the 361 mtGFP group in week 2 (Fig. 7j), suggestive of mild diastolic function. As expected, 362 mtCx43 overexpression induced QTc, QT interval, and JT interval prolongation (Fig.  363 7k, l, m, n), suggestive of ventricular arrhythmia susceptibility. 364

365
In summary, we demonstrated that hypoglycemic challenge causes diastolic 366 cardiac dysfunction and increases the risk of ventricular arrhythmia in an STZ-induced 367