In patients with diabetes, impaired activity of ADAMTS13 (a disintegrin and metalloprotease with thrombospondin type 1 repeats, member 13), the plasma metalloprotease that cleaves highly thrombogenic von Willebrand factor multimers, is a major risk factor of cardiovascular events. Here, using Adamts13−/− mice made diabetic by streptozotocin, we investigated the impact of the lack of ADAMTS13 on the development of diabetes-associated end-organ complications. Adamts13−/− mice experienced a shorter life span than their diabetic wild-type littermates. It was surprising that animal death was not related to the occurrence of detectable thrombotic events. The lack of ADAMTS13 drastically increased the propensity for ventricular arrhythmias during dobutamine-induced stress in diabetic mice. Cardiomyocytes of diabetic Adamts13−/− mice exhibited an aberrant distribution of the ventricular gap junction connexin 43 and increased phosphorylation of Ca2+/calmodulin-dependent kinase II (CaMKII), and with the consequent CaMKII-induced disturbance in Ca2+ handling, which underlie propensity for arrhythmia. In vitro, thrombospondin 1 (TSP1) promoted, in a paracrine manner, CaMKII phosphorylation in murine HL-1 cardiomyocytes, and ADAMTS13 acted to inhibit TSP1-induced CaMKII activation. In conclusion, the deficiency of ADAMTS13 may underlie the onset of lethal arrhythmias in diabetes through increased CaMKII phosphorylation in cardiomyocytes. Our findings disclose a novel function for ADAMTS13 beyond its antithrombotic activity.
Introduction
The risk of cardiovascular (CV) disease for subjects with type 2 diabetes is two- to threefold greater than that among the general population (1,2). Excess CV risk is even higher in patients with diabetes with renal involvement, to the extent that patients with diabetes may die of CV complications before they develop renal and other end-organ dysfunction (3,4). ADAMTS13 (a disintegrin and metalloprotease with thrombospondin type 1 repeats, member 13) is a protease produced by liver stellate cells and endothelial cells (5) that cleaves highly thrombogenic von Willebrand factor (VWF) multimers into smaller, less active forms (6). ADAMTS13 activity tightly modulates the thrombotic process, as shown by evidence that acquired or genetically determined ADAMTS13 deficiency associates with uncontrolled intravascular thrombosis of thrombotic thrombocytopenic purpura (7). Low ADAMTS13 levels and a high VWF-to-ADAMTS13 ratio were associated with vascular complications in diabetes (8). The ADAMTS13 gene is highly polymorphic, and several ADAMTS13 single nucleotide polymorphisms were associated with altered protein secretion and activity in vitro (9,10). We found that in patients with type 2 diabetes the Pro618Ala polymorphism of ADAMTS13 was associated with less ADAMTS13 proteolytic activity than wild-type homozygous Pro618 allele and was a major independent predictor of CV events (11). How impaired ADAMTS13 activity translates into enhanced risk of CV complications in patients with diabetes is still unclear.
Using a model of diabetic Adamts13-deficient (Adamts13−/−) mice, we investigated the impact of the lack of ADAMTS13 on the development of diabetes-associated end-organ complications. Here we show that Adamts13−/− mice had a shorter life span than did their diabetic wild-type (WT) littermates, which was attributable to increased propensity for arrhythmias. In search of the mechanisms linking ADAMTS13 deficiency to arrhythmia, we found in cardiomyocytes of diabetic Adamts13−/− mice an aberrant distribution of the ventricular gap junction connexin 43 (Cx43), associated with increased phosphorylation of Ca2+/calmodulin-dependent kinase II (CaMKII) and consequent altered expression of its downstream targets, as pThr17-phospholamban (PLN), sarcoplasmic reticulum Ca2+-ATPase (Serca-2a), Ca2 + channel β-subunit (CaVβ2), and Na+-Ca2+ exchanger (NCX), which regulate Ca2+ handling (12–14).
Although it is well established that increased CaMKII phosphorylation confers a proarrhythmic substrate to the heart in diabetes (15–17), the intracellular mechanisms underlying CaMKII activation are still to be elucidated. There is evidence that in cardiomyocytes the thrombospondin 1 (TSP1)–CD47 axis promotes CaMKII phosphorylation, which in turn is responsible for left ventricular heart failure (18,19). TSP1 is highly expressed in large vessels and in myocardium of diabetic animals (20,21), and it has been shown to bind ADAMTS13 to form protein complexes (22). Using cultured cardiomyocytes, we show that ADAMTS13 acts to inhibit TSP1-induced CaMKII phosphorylation. Our data reveal a novel role of ADAMTS13 in diabetes that goes beyond antithrombotic activity and is not dispensable for cardiac function.
Research Design and Methods
Animals
We used Adamts13−/− mice (gifted by Dr. D.G. Motto, Iowa University) (23) and WT littermates (C57BL/6 genetic background) with initial body weights of 25–30 g. Mice were maintained in a specific pathogen-free facility with a 12-h dark/12-h light cycle and a constant temperature, and were given free access to standard diet and water. Male WT and Adamts13−/− mice received a single intraperitoneal injection of streptozotocin (STZ; 200 mg/kg body wt; Sigma-Aldrich, St. Louis, MO) to induce diabetes or received saline. The presence of diabetes was confirmed 2 days later on the basis of glucose levels in tail blood. WT and Adamts13−/− mice with diabetes (n = 19 or 20 mice/group) and corresponding nondiabetic controls (n = 9 or 10 mice/group) were followed to determine survival. Kidney and heart morphology and immunohistochemistry were studied 4 months after injection of STZ or saline in additional WT and Adamts13−/− mice (n = 6–8 mice/group). Electroechocardiographic analyses were performed in 3–5 mice per group. Additional nondiabetic and diabetic WT and Adamts13−/− mice (n = 3 mice/group) were sacrificed 4 months after STZ or saline injection to evaluate ex vivo blood prothrombotic activity.
Laboratory Parameters
Blood glucose levels were assessed through the use of a reflectance meter (OneTouch UltraEasy; LifeScan). Cholesterol and triglyceride plasma levels were measured by the Reflotron test (Roche Diagnostic Corp., Basel, Switzerland). Urinary albumin excretion was measured in 24-h urine samples with ELISA (Albuwell M test kit; Exocell, Philadelphia, PA). Urinary creatinine was measured with a Cobas Mira autoanalyzer (Roche Diagnostic Systems).
Cardiac Function and Dobutamine Stress Test
Cardiac function in mice anesthetized with 2 vol% isoflurane in O2 (0.5 L/min) was assessed by transthoracic echocardiography through the use of a commercially available echocardiograph equipped with a 12-MHz probe (MyLab30; Esaote, Genova, Italy) (24,25). Warmed echo gel was placed on the shaved chest, and each parameter was measured through B-mode–guided M-mode imaging in the parasternal short-axis view at the level of the papillary muscles. Images were analyzed offline by readers who were blinded to the animal groups. Global cardiac function was calculated as the percentage of left ventricular fractional shortening (LVFS). Left ventricular end-systolic wall thickening (LVESWT), defined as the percentage increase in wall thickness from end diastole to end systole, was evaluated as a marker of regional left ventricular systolic function. The end-diastolic thickness was measured as the index of left ventricular remodeling. Heart rate (HR) and rhythm were monitored simultaneously by three-lead electrocardiography (ECG). Cardiac function was evaluated at rest and at peak HR after intraperitoneal injection of a single bolus of dobutamine (1.5 μg/g body wt); this is an established method to assess the myocardial contractile reserve (26). Echocardiographic scans and electrocardiographic recordings were repeated immediately after the single-bolus injection and then periodically over a 10-min period, until the peak HR response was reached and HR began to decline again. ECG was recorded for 3 min at rest and at peak HR under dobutamine-induced stress. A trained specialist (V.L.) used R-wave peaks to detect QRS complexes and the R-R interval to calculate HR. The incidence and severity of arrhythmias was quantified by applying a previously published arrhythmic score to each mouse: 0, no arrhythmic events; 1, one premature ventricular contraction; 2, bigeminy and/or salvos; 3, ventricular tachycardia; 4, ventricular fibrillation; 5, spontaneous ventricular fibrillation (16).
Ex Vivo Evaluation of Prothrombotic Activity
When mice were sacrificed, whole blood was collected into heparin tubes (10 units/mL final concentration) and perfused for 2 min over slides coated with collagen (derived from equine tendon, 200 mg/mL; Mascia Brunelli, Milan, Italy) at shear stress of 60 dyne/cm2 (27). Platelet thrombi that were adherent to collagen were fixed and stained with May-Grünwald Giemsa stain. The percentage of the surface occupied by thrombi was quantified as the percentage of positive area through the use of ImageJ software.
Histologic Analyses
Renal sections (3 µm thick) fixed in Duboscq-Brasil fluid and embedded in paraffin were stained with periodic acid Schiff reagent (Bio Optica Milano Spa, Milan, Italy). For ultrastructural analysis of the kidney, ultrathin sections were stained with uranyl acetate and lead citrate and examined with a Morgagni 268D transmission electron microscope (Philips, Amsterdam, the Netherlands). Cardiac sections (3 µm thick) fixed in formalin and embedded in paraffin were stained with hematoxylin-eosin to evaluate cardiomyocyte number and cross-sectional area, as described previously (28). Periodic acid Schiff staining was used to detect glycogen storage in cardiomyocytes. Interstitial collagen was determined in picrosirius red–stained sections. All biopsies were reviewed by a pathologist blinded to the animal groups. Details are provided in the Supplementary Data.
Immunofluorescence Analyses
Frozen renal and cardiac sections (3 μm thick) were incubated with fluorescein isothiocyanate–conjugated polyclonal rabbit antifibrin(ogen) antibody (1:50; F0111; Dako, Glostrup, Denmark), or with polyclonal rabbit anti-VWF antibody (1:3,000; A0082; Dako) followed by goat antirabbit CY3-conjugated antibody (1:200; Jackson ImmunoResearch Laboratories Inc., West Grove, PA), or with biotinylated isolectin B4 (1:50; Vector Laboratories, Burlingame, CA) followed by mouse anti-biotin Cy3-conjugated antibody (1:100, Jackson ImmunoResearch Laboratories Inc.). Sections stained for VWF and isolectin B4 were counterstained with fluorescein isothiocyanate–conjugated wheat germ agglutinin (Vector Laboratories); see the Supplementary Data for further details.
Immunoperoxidase Analyses
Cardiac sections (3 μm thick) fixed in formalin and embedded in paraffin were stained for nitrotyrosine (rabbit polyclonal antinitrotyrosine IgG [1:500]; 06–284; Upstate Biotechnology, Lake Placid, NY), Cx43 (rabbit anti-Cx43 [1:50]; no. 3512; Cell Signaling Technology, Danvers, MA), desmin (rabbit polyclonal antidesmin IgG [1:100]; H-76; Santa Cruz Biotechnology, Santa Cruz, CA), and TSP1 (mouse monoclonal anti-TSP1 [1:20], clone A6.1; Thermo Fisher Scientific, Rockford, IL). The stainings were visualized with the use of diaminobenzidine (Biocare Medical), and sections were counterstained with Mayer hematoxylin. Details are provided in the Supplementary Data.
Western Blotting Analyses
Heart tissue homogenates were resolved with SDS-PAGE under reducing conditions. Blots were probed with the following antibodies: phosphorylated Cx43 (Ser368; 3511; Cell Signaling Technology), CaMKII (D11A10; Cell Signaling Technology), phosphorylated CaMKII (Thr286; D21E4; Cell Signaling Technology), pThr-17-PLN (A010–13AP; Badrilla, Leeds, U.K.), pSer-16-PLN (A010–12AP; Badrilla), PLN (A010–14; Badrilla), NCX (C2C12; Thermo Fisher Scientific), Serca-2a (A010–20; Badrilla), and β1-adrenoreceptor (ADR) (V-19; Santa Cruz Biotechnology). Then the blots were analyzed with the Odyssey Fc infrared image system (see the Supplementary Data for further details).
Quantitative RT-PCR Analysis
CaVβ2 transcript levels were evaluated in total RNA isolated from frozen heart tissue. Quantitative RT-PCR was performed with the use of SYBR Green (Thermo Fisher Scientific) and the specific primers detailed in the Supplementary Data.
In Vitro Studies With the HL-1 Cardiomyocyte Line
Mouse atrial HL-1 myocytes (a gift from Professor W. Claycomb, Department of Biochemistry and Molecular Biology, LSU Health Sciences Center, New Orleans, LA) were cultured in Claycomb medium (Sigma-Aldrich Chemical Co.) supplemented with 10% FBS (Sigma-Aldrich Chemical Co.), 1% norepinephrine, 1% penicillin-streptomycin, and 1% l-glutamine. HL-1 myocytes were seeded in p60 plates at a density of 3 × 105 cells per well. After 24 h, cells were serum-starved overnight and then treated for 6 h with vehicle (PBS), TSP1 (0.2 nmol/L) purified from human platelets (29), recombinant human (rh) ADAMTS13 (0.3 µg/mL; 6156-AD; R&D Systems, Minneapolis, MN), or both TSP1 and rhADAMTS13. Samples were evaluated through the use of Western blotting analysis. In additional experiments, HL-1 cells were exposed to low (5 mmol/L) or high (30 mmol/L) glucose concentration in the presence or absence of rhADAMTS13 (0.3 µg/mL). Serum-free supernatants were collected to measure TSP1 production by ELISA (Cusabio, Hubei, China); see the Supplementary Data for further details.
Measurement of ADAMTS13 Activity
Serum ADAMTS13 activity was measured by residual collagen-binding assay (11). Details are provided in the Supplementary Data.
Statistics
Data are presented as the mean ± SEM. Data were analyzed with Prism software (GraphPad Software Inc., La Jolla, CA). Comparisons were made with the use of ANOVA with the Tukey post hoc test, or with the paired Student t test. Statistical significance was defined as P < 0.05.
Study Approval
All procedures involving animals were performed in accordance with institutional guidelines in compliance with national (Decreto Legislativo n.26, March 4, 2014) and international (Directive 2010/63/EU on the protection of animals used for scientific purposes) laws and policies and were approved by the Institutional Animal Care and Use Committees of IRCCS – Istituto di Ricerche Farmacologiche Mario Negri.
Results
ADAMTS13 Deficiency Does Not Affect Major Metabolic Parameters, Even Though It Causes Premature Death in Diabetic Mice
Adamts13−/− mice with STZ-induced diabetes died before their diabetic WT littermates. Log-rank analysis showed significant (P = 0.0022) differences between the survival curves of the two groups of diabetic mice (Fig. 1A). All diabetic Adamts13−/− mice had died by 6 months after diabetes induction, whereas 53% of the diabetic WT mice had died. No mortality was recorded in two groups of age-matched nondiabetic Adamts13−/− and nondiabetic WT mice studied in parallel. Monitoring of relevant metabolic parameters did not show signs of worsening diabetes that could predict premature death in Adamts13−/− mice. As expected, induction of diabetes with STZ was associated with a reduction in body weight (30), which was comparable in Adamts13−/− and WT mice (Table 1). The two groups of diabetic mice had higher levels of blood glucose and plasma triglycerides than those in the corresponding nondiabetic control mice. No differences in plasma cholesterol levels were observed among the experimental groups (Table 1). Measurement of serum ADAMTS13 activity showed similar levels in nondiabetic (10 ± 2.3%) and diabetic (8.7 ± 1.3%) WT mice at 4 months.
Adamts13 deficiency impairs survival of diabetic mice. A: Kaplan-Meier curves comparing survival of STZ-induced diabetic WT (n = 19) and Adamts13−/− (n = 20) mice and their corresponding nondiabetic controls (n = 9 or 10 mice/group) throughout the study period. The P value represents the result of the log-rank test. Original magnification ×200. B: Urinary albumin-to-creatinine ratio (Ualb/Ucreat) measured at 4 months in WT and Adamts13−/− mice with diabetes and in age-matched nondiabetic mice. C: Representative images (left panel) and quantification (score; right panel) of mesangial matrix accumulation at 4 months in kidney tissue stained with periodic acid Schiff. Original magnification ×630. Scale bars = 50 μm. Data for B and C are the mean ± SEM (n = 7 or 8 mice/group) and were analyzed by ANOVA with the Tukey post hoc test.
Adamts13 deficiency impairs survival of diabetic mice. A: Kaplan-Meier curves comparing survival of STZ-induced diabetic WT (n = 19) and Adamts13−/− (n = 20) mice and their corresponding nondiabetic controls (n = 9 or 10 mice/group) throughout the study period. The P value represents the result of the log-rank test. Original magnification ×200. B: Urinary albumin-to-creatinine ratio (Ualb/Ucreat) measured at 4 months in WT and Adamts13−/− mice with diabetes and in age-matched nondiabetic mice. C: Representative images (left panel) and quantification (score; right panel) of mesangial matrix accumulation at 4 months in kidney tissue stained with periodic acid Schiff. Original magnification ×630. Scale bars = 50 μm. Data for B and C are the mean ± SEM (n = 7 or 8 mice/group) and were analyzed by ANOVA with the Tukey post hoc test.
Systemic and laboratory parameters in mice 4 months after diabetes induction
. | WT mice . | Adamts−/− mice . | ||
---|---|---|---|---|
Nondiabetic (n = 10) . | Diabetic (n = 13) . | Nondiabetic (n = 9) . | Diabetic (n = 8) . | |
Body weight (g) | 33.00 ± 0.93 | 25.31 ± 0.96*** | 32.33 ± 1.62 | 26.38 ± 1.15** |
Blood glucose (mg/dL) | 133.40 ± 2.74 | 502.23 ± 19.92*** | 132.33 ± 3.27 | 549.50 ± 16.54*** |
Cholesterol (mg/dL) | 108.44 ± 2.88 | 113.85 ± 3.41 | 109.78 ± 3.91 | 123.88 ± 5.56 |
Triglycerides (mg/dL) | 87.10 ± 5.42 | 142.92 ± 17.68* | 86.22 ± 3.2 | 141.63 ± 9.84* |
. | WT mice . | Adamts−/− mice . | ||
---|---|---|---|---|
Nondiabetic (n = 10) . | Diabetic (n = 13) . | Nondiabetic (n = 9) . | Diabetic (n = 8) . | |
Body weight (g) | 33.00 ± 0.93 | 25.31 ± 0.96*** | 32.33 ± 1.62 | 26.38 ± 1.15** |
Blood glucose (mg/dL) | 133.40 ± 2.74 | 502.23 ± 19.92*** | 132.33 ± 3.27 | 549.50 ± 16.54*** |
Cholesterol (mg/dL) | 108.44 ± 2.88 | 113.85 ± 3.41 | 109.78 ± 3.91 | 123.88 ± 5.56 |
Triglycerides (mg/dL) | 87.10 ± 5.42 | 142.92 ± 17.68* | 86.22 ± 3.2 | 141.63 ± 9.84* |
Data are expressed as the mean ± SEM and were analyzed by ANOVA with the Tukey post hoc test.
*P < 0.05; **P < 0.01; ***P < 0.0001 vs. the corresponding nondiabetic control mice.
To evaluate whether diabetes could drive the prothrombotic phenotype of Adamts13−/− mice to the point of causing their death, we compared the ability of whole blood taken from ad hoc groups of Adamts13−/− and WT mice (with and without diabetes) to form platelet thrombi. Mice were studied 4 months after diabetes induction, when 60% mortality had already been reached among Adamts13−/− mice with diabetes. As shown in Fig. 2A, no difference was observed in the area covered by thrombi between diabetic and nondiabetic WT mice. As expected, Adamts13 deficiency was associated with increased capability to form thrombi (P < 0.01 vs. WT mice), which was not, however, heightened by diabetes.
Diabetes does not exacerbate the prothrombotic status of Adamts13−/− mice. A: Representative images (left panels) and experimental design (right top panel) of the assay for ex vivo evaluation of blood thrombotic activity. Blood was taken from nondiabetic and diabetic WT and Adamts13−/− mice at 4 months, and its ability to form platelet thrombi was tested on immobilized collagen under shear stress. The graph (right bottom panel) shows the percentage of the surface of immobilized collagen occupied by thrombi. Data are the mean ± SEM (n = 3 mice/group) and were analyzed by ANOVA with the Tukey post hoc test. B: Representative immunofluorescence images (left panel) and quantification (score; right panel) of VWF accumulation in the heart at 4 months. Original magnification ×400. C: Representative immunofluorescence images (left panel) and quantification (right panel) of capillary density (number of capillaries per 100 cardiomyocytes) at 4 months. Original magnification ×400. Data are the mean ± SEM (n = 5–8 mice/group) and were analyzed by ANOVA with the Tukey post hoc test. Scale bars = 50 μm.
Diabetes does not exacerbate the prothrombotic status of Adamts13−/− mice. A: Representative images (left panels) and experimental design (right top panel) of the assay for ex vivo evaluation of blood thrombotic activity. Blood was taken from nondiabetic and diabetic WT and Adamts13−/− mice at 4 months, and its ability to form platelet thrombi was tested on immobilized collagen under shear stress. The graph (right bottom panel) shows the percentage of the surface of immobilized collagen occupied by thrombi. Data are the mean ± SEM (n = 3 mice/group) and were analyzed by ANOVA with the Tukey post hoc test. B: Representative immunofluorescence images (left panel) and quantification (score; right panel) of VWF accumulation in the heart at 4 months. Original magnification ×400. C: Representative immunofluorescence images (left panel) and quantification (right panel) of capillary density (number of capillaries per 100 cardiomyocytes) at 4 months. Original magnification ×400. Data are the mean ± SEM (n = 5–8 mice/group) and were analyzed by ANOVA with the Tukey post hoc test. Scale bars = 50 μm.
Diabetic Adamts13−/− and diabetic WT mice had developed albuminuria to a similar extent 4 months after disease induction (Fig. 1B). In this early phase of the disease, morphological analysis of the kidney showed the presence of mild glomerular mesangial expansion in both diabetic Adamts13−/− and diabetic WT mice (Fig. 1C). VWF deposits had similarly increased in diabetic mice of both genotypes (Supplementary Fig. 1A). Fibrinogen deposits were rare or absent within glomeruli in both groups of diabetic mice (Supplementary Fig. 1B). Furthermore, ultrastructural analysis of the kidney did not reveal the presence of platelet thrombi or clumps in the glomeruli of diabetic mice (data not shown).
ADAMTS13 Deficiency Increases Propensity for Arrhythmias in Diabetic Mice
We next evaluated whether cardiac abnormalities could be the cause of the premature death of diabetic Adamts13−/− mice. Analysis of cardiac tissue from all experimental groups, through the use of hematoxylin-eosin staining, did not reveal signs of ischemia or thrombosis. Moreover, glycogen deposition, a marker of ischemic injury (31), was similar in all the experimental groups (Supplementary Fig. 2). Consistent with previous observations (32), VWF levels were higher in the hearts of diabetic WT mice than in the hearts of nondiabetic WT mice, and the lack of ADAMTS13 did not result in a further increase in VWF staining (Fig. 2B). Cardiac fibrinogen staining was similar in all the experimental groups (Supplementary Fig. 3). Cardiac microvascular density was significantly lower in WT mice with diabetes than in WT mice without diabetes (Fig. 2C). The number of capillaries in the myocardia of diabetic Adamts13−/− mice was similarly lower than that in nondiabetic Adamts13−/− mice (Fig. 2C).
We then evaluated the impact of ADAMTS13 deficiency on myocyte hypertrophy, a common feature of cardiac remodeling and dysfunction in diabetes. We observed a marked enlargement of the cardiomyocyte area in diabetic compared with nondiabetic WT mice (Fig. 3A). The lack of Adamts13 per se was associated with increased cardiomyocyte hypertrophy compared with that in WT mice, which was not further exacerbated by diabetes. The total number of cardiomyocytes in the left ventricle of diabetic WT mice was significantly lower (by 20%) than the number in nondiabetic WT mice. Myocyte numbers in nondiabetic and diabetic Adamts13−/− mice were similar to those observed in diabetic WT mice (Fig. 3B). The expression of nitrotyrosine, a marker of oxidative damage, was significantly increased in cardiomyocytes and vessels of diabetic WT mice (Fig. 3C and D). High nitrotyrosine staining was present in cardiomyocytes of diabetic Adamts13−/− mice (P = 0.05 vs. nondiabetic Adamts13−/− mice) (Fig. 3C). In vessels, oxidative damage was already detected in nondiabetic Adamts13−/− mice and remained at similar levels in diabetic mice (Fig. 3D).
Cardiac tissue analyses. Cardiac hypertrophy is expressed as the area of cardiomyocytes (A) and the number of cardiomyocytes (B) in diabetic WT and diabetic Adamts13−/− mice and in age-matched nondiabetic WT and nondiabetic Adamts13−/− mice at 4 months. Oxidative stress at 4 months was quantified (scored) on the basis of nitrotyrosine staining in cardiomyocytes (C) and vessels of cardiac tissue (D). E: Representative images of cardiac interstitial collagen by Sirius red staining (left panel) and quantification of deposition area (square micrometers; right panel). Original magnification ×400. Scale bars = 50 μm. F: Representative images of cardiac TSP1 staining by immunoperoxidase (left panel) and its quantification (score; right panel). Original magnification ×630. Scale bars = 20 μm. Data are the mean ± SEM (n = 5–8 mice/group) and were analyzed by ANOVA with the Tukey post hoc test.
Cardiac tissue analyses. Cardiac hypertrophy is expressed as the area of cardiomyocytes (A) and the number of cardiomyocytes (B) in diabetic WT and diabetic Adamts13−/− mice and in age-matched nondiabetic WT and nondiabetic Adamts13−/− mice at 4 months. Oxidative stress at 4 months was quantified (scored) on the basis of nitrotyrosine staining in cardiomyocytes (C) and vessels of cardiac tissue (D). E: Representative images of cardiac interstitial collagen by Sirius red staining (left panel) and quantification of deposition area (square micrometers; right panel). Original magnification ×400. Scale bars = 50 μm. F: Representative images of cardiac TSP1 staining by immunoperoxidase (left panel) and its quantification (score; right panel). Original magnification ×630. Scale bars = 20 μm. Data are the mean ± SEM (n = 5–8 mice/group) and were analyzed by ANOVA with the Tukey post hoc test.
The hearts of diabetic mice of both genotypes showed a similar increase in interstitial collagen deposition when compared with deposition in nondiabetic mice (Fig. 3E). Similar to fibrosis, significant TSP1 upregulation was observed in diabetic mice of both genotypes (Fig. 2F). Staining for TSP1 localized in perivascular and interstitial areas, whereas it was not detectable in cardiomyocytes (Fig. 2F).
We created acute stress in mice by administering a low dose of dobutamine during echocardiography and ECG monitoring to evaluate contractile reserve, an effective predictor of survival. As shown in Fig. 4, global cardiac function at rest was similar in each experimental group. Although the HR increased upon administration of the dobutamine dose in both groups of WT mice (Fig. 4A), dobutamine increased LVFS and LVESWT exclusively in nondiabetic WT mice (Fig. 4B and C). It is interesting that acute dobutamine-induced stress triggered increases in HR, LVFS, and LVESWT in nondiabetic and diabetic Adamts13−/− mice, although differences in HR and LVFS in nondiabetic Adamts13−/− mice did not reach statistical significance (Fig. 4A–C). Even though arrhythmic events were undetectable before the dobutamine injection, we observed premature ventricular contractions and ventricular salvos in diabetic Adamts13−/− mice during dobutamine-induced stress, which were undetectable in the other groups (Fig. 4D and E). The arrythmia score did indeed show that the absence of ADAMTS13 drastically increased susceptibility to ventricular arrhythmias after dobutamine challenge in diabetic mice (P < 0.05 vs. diabetic WT mice). Of note, a similar dose of dobutamine did not elicit premature ventricular contractions and ventricular salvos in nondiabetic and diabetic WT mice, nor in nondiabetic Adamts13−/− mice (Fig. 4D).
Adamts13 deficiency preserves the myocardial contractile reserve and increases the propensity for arrhythmic events during dobutamine challenge in diabetic mice. HR (A), percentage LVFS (B), and percentage LVESWT (C) were evaluated at rest (pre) and at peak after intraperitoneal injection of a single bolus of dobutamine (1.5 μg/g body wt) (post) at 4 months. D: Arrhythmias (arrhythmia score) were evaluated at peak after dobutamine injection in all experimental groups. E: Representative ECG traces coupled to corresponding M-mode images at the peak of dobutamine action during stress echocardiography. Data are the mean ± SEM (n = 3–5 mice/group). Statistical analyses were performed with the paired Student t test (A–C) and ANOVA with the Tukey post hoc test (D).
Adamts13 deficiency preserves the myocardial contractile reserve and increases the propensity for arrhythmic events during dobutamine challenge in diabetic mice. HR (A), percentage LVFS (B), and percentage LVESWT (C) were evaluated at rest (pre) and at peak after intraperitoneal injection of a single bolus of dobutamine (1.5 μg/g body wt) (post) at 4 months. D: Arrhythmias (arrhythmia score) were evaluated at peak after dobutamine injection in all experimental groups. E: Representative ECG traces coupled to corresponding M-mode images at the peak of dobutamine action during stress echocardiography. Data are the mean ± SEM (n = 3–5 mice/group). Statistical analyses were performed with the paired Student t test (A–C) and ANOVA with the Tukey post hoc test (D).
ADAMTS13 Deficiency Is Associated With Phosphorylation and Dislocation of Cx43 in Cardiomyocytes of Diabetic Mice
Cx43 is the major structural protein of the ventricular gap junctions localized in the intercalated disc that enables appropriate electrical coupling in healthy hearts (33). Phosphorylation of Cx43 causes pathological remodeling, which may lead to sudden arrhythmia and death (34). We evaluated, using Western blotting, the extent of Ser368 phosphorylation of Cx43 in cardiac tissue at 4 months. Levels of phosphorylated Cx43 showed a trend toward an increase in diabetic compared with nondiabetic WT mice (Fig. 5A). Lack of ADAMTS13 in diabetic mice significantly enhanced Cx43 phosphorylation (P < 0.01 vs. nondiabetic Adamts13−/− mice). We then used immunohistochemistry to assess the expression and localization of Cx43. In nondiabetic WT and nondiabetic Adamts13−/− mice, Cx43 was organized and aligned with myocardial architecture at intercellular junctions (Fig. 5B). Higher levels of Cx43 were detected in hearts of diabetic WT mice (P < 0.05 vs. nondiabetic WT mice) (Fig. 5C); in these hearts, however, Cx43 localization was still restricted within myocyte gap junctions (Fig. 5B and D). Cx43 expression further increased in cardiomyocytes of diabetic Adamts13−/− mice (P < 0.01 vs. nondiabetic Adamts13−/− mice), and its distribution was highly disorganized: Cx43 staining became wider and less linear, with more prevalence in the mid–myocytic regions (Fig. 5B and D). Cx43 was highly disorganized in cardiac tissue of diabetic WT mice that were sacrificed at the end of the study (i.e., at 6 months, when 60% mortality was reached in this group; Cx43 dislocation: 49.3 ± 6.8%) to an extent similar to that in diabetic Adamts13−/− mice at 4 months (Supplementary Fig. 4). This finding reveals that the cardiac abnormalities observed in Adamts13−/− diabetic mice, which may be responsible for animal death, develop later in diabetic WT mice. To demonstrate that cardiac arrythmia changes in diabetic Adamts13−/− mice specifically depend on Cx43 remodeling, we analyzed the expression and distribution of desmin, an intermediate filament protein expressed by cardiac muscle cells, whose alteration is also associated with arrhythmogenic cardiomyopathy (35). Immunoperoxidase staining did not show differences in desmin levels among the experimental groups (Supplementary Fig. 5).
Adamts13 deficiency associates with increased and displaced Cx43 expression in cardiomyocytes of diabetic mice. A: Representative Western blotting (left panel) and densitometric analysis (right panel) of Ser368-phosphorylated Cx43 (p368-Cx43) protein levels in heart tissue from nondiabetic and diabetic WT and Adamts13−/− mice at 4 months. GAPDH was used as the loading control. B: Representative images of immunoperoxidase staining of Cx43 on heart specimens from nondiabetic and diabetic WT and Adamts13−/− mice. Scale bars = 50 μm. Original magnification ×400. C: Quantification of Cx43 staining expressed as the percentage of mean positive area per high-power field. D: Quantification of dislocated Cx43 staining expressed as the percentage of positive fields on total fields. Data are the mean ± SEM (n = 4–8 mice/group) and were analyzed by ANOVA with the Tukey post hoc test.
Adamts13 deficiency associates with increased and displaced Cx43 expression in cardiomyocytes of diabetic mice. A: Representative Western blotting (left panel) and densitometric analysis (right panel) of Ser368-phosphorylated Cx43 (p368-Cx43) protein levels in heart tissue from nondiabetic and diabetic WT and Adamts13−/− mice at 4 months. GAPDH was used as the loading control. B: Representative images of immunoperoxidase staining of Cx43 on heart specimens from nondiabetic and diabetic WT and Adamts13−/− mice. Scale bars = 50 μm. Original magnification ×400. C: Quantification of Cx43 staining expressed as the percentage of mean positive area per high-power field. D: Quantification of dislocated Cx43 staining expressed as the percentage of positive fields on total fields. Data are the mean ± SEM (n = 4–8 mice/group) and were analyzed by ANOVA with the Tukey post hoc test.
Cx43 Remodeling in Diabetic Adamts13−/− Mice Is Associated With CaMKII Phosphorylation
In search of intracellular pathways implicated in Cx43 phosphorylation, we focused on CaMKII, a multifunctional serine-threonine protein kinase that is abundant in the heart and has been implicated in diabetes-associated arrhythmias (15,36). CaMKII phosphorylation sites have been identified in the carboxyl terminus of Cx43 (37), and CaMKII has been recognized as an important signal for controlling Cx43 localization and conduction velocity in the myocardium (38). CaMKIIδ is the predominant isoform within the myocardium, but expression of the CaMKIIγ isoform also has been detected (39,40). We explored the activation of CaMKII by assessing expression levels and the extent of CaMKII phosphorylation in the left ventricles of mice in the nondiabetic and diabetic groups. Western blotting demonstrated the presence of the CaMKII isoforms CaMKIIδC (56 kDa), CaMKIIδB (plus CaMKIIδ9, which differs by only three amino acids; 58 KDa), and CaMKIIγ (60 kDa) (Fig. 6A) in all groups of mice. No differences were observed in the levels of total CaMKII expression (Supplementary Fig. 6A). It is notable that increases in the phosphorylation of all the CaMKII isoforms were observed in diabetic Adamts13−/− mice (P < 0.001 vs. diabetic WT mice), as indicated by the ratio of phosphorylated to total CaMKII expression (Fig. 6C).
Adamts13 deficiency associates with CaMKII activation and Ca2+ handling protein disturbance in cardiomyocytes of diabetic mice. A: Representative Western blotting of phosphorylated and total CaMKII. The three major bands seen for each sample represent the isoforms CaMKIIγ (62 kDa), CaMKIIδb plus δ9 (58 kDa), and CaMKIIδc (56 kDa) (left panel). The graph (right panel) quantifies this with the use of densitometric analysis of CaMKII activation, expressed as the ratio of phosphorylated to total CaMKII. Representative Western blotting show phosphorylated Thr17-PLN (CaMKII site) (B, left panel) and phosphorylated Ser16-PLN (PKA site) (C, left panel), and the graphs quantify this through the use of densitometric analysis of phosphorylated Thr17-PLN (B, right panel) and Ser16-PLN (C, right panel), expressed as the ratio of phosphorylated to total PLN. D and E: Representative Western blotting (left panels) and quantification (right panels) of Serca-2a (D) and NCX protein (E). GAPDH was used as the loading control. Protein expression was evaluated in heart tissue from nondiabetic and diabetic WT and Adamts13−/− mice at 4 months. Data are expressed as the mean ± SEM (n = 4 or 5 mice/group) and were analyzed by ANOVA with the Tukey post hoc test.
Adamts13 deficiency associates with CaMKII activation and Ca2+ handling protein disturbance in cardiomyocytes of diabetic mice. A: Representative Western blotting of phosphorylated and total CaMKII. The three major bands seen for each sample represent the isoforms CaMKIIγ (62 kDa), CaMKIIδb plus δ9 (58 kDa), and CaMKIIδc (56 kDa) (left panel). The graph (right panel) quantifies this with the use of densitometric analysis of CaMKII activation, expressed as the ratio of phosphorylated to total CaMKII. Representative Western blotting show phosphorylated Thr17-PLN (CaMKII site) (B, left panel) and phosphorylated Ser16-PLN (PKA site) (C, left panel), and the graphs quantify this through the use of densitometric analysis of phosphorylated Thr17-PLN (B, right panel) and Ser16-PLN (C, right panel), expressed as the ratio of phosphorylated to total PLN. D and E: Representative Western blotting (left panels) and quantification (right panels) of Serca-2a (D) and NCX protein (E). GAPDH was used as the loading control. Protein expression was evaluated in heart tissue from nondiabetic and diabetic WT and Adamts13−/− mice at 4 months. Data are expressed as the mean ± SEM (n = 4 or 5 mice/group) and were analyzed by ANOVA with the Tukey post hoc test.
CaMKII Phosphorylation in Adamts13−/− Diabetic Mice Disturbs Ca2+ Handling Proteins
The finding that CaMKII activity was enhanced in cardiac tissue of diabetic Adamts13−/− mice prompted us to assess Ca2+ handling proteins that are regulated directly by CaMKII and contribute to the triggered arrhythmia (14). Western blotting showed more phosphorylation of a specific CaMKII target site on PLN (pThr17-PLN) in diabetic Adamts13−/− than in diabetic WT mice (Fig. 6B). The hearts of diabetic Adamts13−/− mice had reduced expression of Serca-2a (Fig. 6D), which transports calcium to the sarcoplasmic reticulum (14). We then evaluated the expression of NCX, which is considered to be the dominant Ca2+ efflux mechanism in cardiomyocytes and, together with Serca-2a, is responsible for reducing the amount of intracellular Ca2+ (41). Results showed a significant reduction of NCX levels in diabetic compared with nondiabetic Adamts13−/− mice (Fig. 6E). The aforementioned alterations in Ca2+ handling proteins we observed in diabetic Adamts13−/− mice were associated with lower gene expression of the Ca2 + channel β subunit (CaVβ2) (Supplementary Fig. 7), a strong modulator of l-type calcium currents in adult cardiomyocytes (42).
Next, given that the activation of the β-ADR signaling pathway has been implicated in Cx43 synthesis (43,44) and CaMKII activation (45) in cardiomyocytes, we investigated β-ADR expression and activity in the hearts. Western blotting did not show differences in β1-ADR protein levels among the experimental groups (Supplementary Fig. 8). As shown in Fig. 6C, phosphorylation of PLN at Ser16, a target of cAMP-dependent protein kinase (PKA) (14), the predominant regulator of β-ADR–mediated enhancement of cardiac contractility, was not statistically different among the experimental groups, which rules out the possibility that β-ADR signaling has a role in dictating CaMKII activation in our setting.
ADAMTS13 Affects CaMKII Phosphorylation in Cultured Cardiomyocytes
Evidence shows that, in cardiomyocytes, the TSP1-CD47 axis promotes CaMKII activation, which in turn is responsible for left ventricular heart failure (18,19). In vitro analyses showed that TSP1 and ADAMTS13 form protein complexes (22). We hypothesized that in cardiomyocytes of diabetic mice, which are exposed to higher TSP1 levels (21), circulating ADAMTS13 could bind TSP1 and hamper TSP1-CD47 signaling, thereby limiting CaMKII activation. To verify this hypothesis in vitro, we used murine HL-1 cardiomyocytes, an established model of contractile adult cardiomyocytes (46). Cardiomyocytes were exposed to TSP1 (0.2 nmol/L), rhADAMTS13 (0.3 μg/mL), or both combined for 6 h. Western blotting showed more phosphorylated CaMKII after stimulation with either TSP1 or rhADAMTS13 than in unstimulated (control) HL-1 cells. This effect disappeared when TSP1 and rhADAMTS13 were added together, as indicated by densitometric analyses (Fig. 7). To investigate whether hyperglycemia directly affects TSP1 expression and CaMKII activity in cardiomyocytes, HL-1 cells were exposed to a low or high glucose concentration. High glucose did not affect the release of TSP1 in the cell supernatant (Supplementary Fig. 9), and as a consequence, no changes in CaMKII phosphorylation were observed with respect to low glucose (Supplementary Fig. 10).
ADAMTS13 and TSP1 combined limit CaMKII phosphorylation in cultured cardiomyocytes. A: Representative Western blotting of phosphorylated and total CaMKII protein expression in HL-1 cells exposed for 6 h to rhADAMTS13 (0.3 µg/mL), TSP1 (0.2 nmol/L), or both. GAPDH was used as the loading control. B: Quantification by densitometric analysis of total CaMKII (T-CaMKII) expression. C: CaMKII activation expressed as the ratio of phosphorylated CaMKII (p-CaMKII) to T-CaMKII. Data are the mean ± SEM (n = 4 samples/group) and were analyzed by ANOVA with the Tukey post hoc test.
ADAMTS13 and TSP1 combined limit CaMKII phosphorylation in cultured cardiomyocytes. A: Representative Western blotting of phosphorylated and total CaMKII protein expression in HL-1 cells exposed for 6 h to rhADAMTS13 (0.3 µg/mL), TSP1 (0.2 nmol/L), or both. GAPDH was used as the loading control. B: Quantification by densitometric analysis of total CaMKII (T-CaMKII) expression. C: CaMKII activation expressed as the ratio of phosphorylated CaMKII (p-CaMKII) to T-CaMKII. Data are the mean ± SEM (n = 4 samples/group) and were analyzed by ANOVA with the Tukey post hoc test.
Discussion
Impaired ADAMTS13 activity has been associated with higher risk of CV complications in patients with diabetes (11) and provided the starting point for our study. Hence, we investigated Adamts13 knockout mice made diabetic by STZ with the aim of gaining a better understanding of the link between defective ADAMTS13 activity and excess CV events in diabetes. Adamts13 deficiency is tolerated well in mice. They exhibit normal development and survival, without signs of histopathological lesions in the organ tissues (47). However, diabetes induced by STZ injection caused premature death in Adamts13−/− mice. To unravel the most plausible determinant of the survival disadvantage observed in diabetic Adamts13−/− mice, we first considered thrombosis. The results unexpectedly indicated that shortened life span was not related to the occurrence of thrombotic events. Diabetic Adamts13−/− mice died before they developed overt nephropathy, thereby mimicking the clinical situation of patients with diabetes who can die of cardiovascular complications before they develop renal insufficiency.
We found that lack of Adamts13 increased propensity for ventricular arrythmic events in diabetic mice at the peak of the dobutamine challenge, which may expose them to arrhythmogenic risk related to increased myocardial oxygen consumption (48). It is interesting to note that diabetic Adamts13−/− mice did not show a deficient contractile reserve in response to the dobutamine bolus. Conversely, the magnitude of the chronotropic response to dobutamine stress was similar in all experimental groups. Converging with previous clinical findings (49,50), our data suggest that mortality related to diabetes occurs in the presence of a preserved inotropic state and opens new avenues for understanding mechanisms underlying sudden cardiac arrest caused by ventricular arrhythmias in patients with preserved left ventricular systolic function (51). Evidence indicates that early cardiac dysfunction in diabetes manifests as arrhythmogenic mechanoelectrical instability, which leads to premature death (52). Moreover, the altered propagation of the electrical signal in diabetic hearts is dependent on changes in intercellular gap junctions. A major component of ventricular gap junction is the protein Cx43, whose alteration can lead to slowed conduction velocity (53). In the diabetic rat heart, phosphorylation and lateralization of Cx43 are associated with nonfunctional gap junctions, which translate to a reduced conduction reserve and thereby an increase in the incidence of arrhythmias (53,54). We showed that Adamts13 deficiency strongly enhances the phosphorylation and the consequent lateralization of Cx43 in diabetic mice. Even a modest disarrangement of Cx43 can result in profound conduction impairment when it occurs in conjunction with other proarrhythmogenic factors, such as fibrosis and alterations in intracellular calcium homeostasis (55). Indeed, in the hearts of diabetic Adamts13−/− mice, Cx43 lateralization was associated with increased interstitial fibrosis, suggesting that in diabetic Adamts13−/− mice an altered ventricular conduction velocity could contribute to the occurrence of ventricular arrhythmias.
To the best of our knowledge, no other reports have described the role of ADAMTS13 in modulating the molecular mechanisms that underlie the propagation of cardiac impulses in diabetes. The next step was to find a link between Adamts13 deficiency and Cx43 phosphorylation/dislocation in diabetic cardiac tissue. We focused on CaMKII, a key enzyme in many cardiac pathologies that lead to contractile dysfunction and arrhythmias. CaMKII phosphorylation sites have been identified in the carboxyl terminus of Cx43 (37), and evidence indicates that CaMKII is an important signal for controlling Cx43 localization and conduction velocity in the myocardium (38). Canonically, CaMKII activation requires Ca2+/calmodulin binding as an initiating step, and the duration of CaMKII activation is dependent on the frequency of calcium release events. It is well established that prolonged association of Ca2+/calmodulin with CaMKII allows for CaMKII activation through posttranslational modifications, including oxidation, O-linked glycosylation, and phosphorylation (56), and it confers a proarrhythmic substrate to the diabetic heart (15,16,36).
Relevant studies using a comprehensive multilevel approach, from single cells to whole hearts, have consistently demonstrated that CaMKII activation regulates Ca2+ handling proteins, thereby directly contributing to arrhythmias (12–14,17). Our data showing that hearts of diabetic Adamts13−/− mice had enhanced phosphorylation of CaMKII downstream target proteins, such as serine 368–Cx43 and threonine 17–PLN, along with low expression of Serca-2a, NCX, and CaVβ2, strongly indicate that changes in contractility in diabetic Adamts13−/− mice are mainly attributable to CaMKII-induced disturbance in pathways that mediate Ca2+ handling. Further studies to characterize in more depth the molecular determinants underlying CaMKII-induced arrhythmias will include RyR2 phosphorylation, which is involved in arrhythmogenic mechanisms in early-stage diabetes (17).
Next, we evaluated β-ADR signaling in order to elucidate the intracellular mechanisms underlying CaMKII activation in cardiomyocytes of diabetic Adamts13−/− mice. The finding that PKA-dependent phosphorylation of PLN was not altered in diabetic Adamts13−/− mice suggests that β-ADR signaling does not have a role in the propensity for arrhythmia in diabetic Adamts13−/− mice. However, we cannot rule out the possibility that other as yet unrecognized β-ADR-mediated mechanisms may contribute to the arrhythmogenic phenotype elicited in diabetic Adamts13−/− mice. Our observation is in line with that in a recent study of experimental and human heart failure that demonstrated that the rise of CaMKII activity in an oxidative microenvironment did not depend on β-ADR stimulation (57), suggesting that other factors may be involved. Evidence exists that TSP1 expression is upregulated in the myocardium of diabetic animals (20,21), and through immunohistochemistry experiments we found that TSP1 staining localizes in perivascular and interstitial areas, but not in cardiomyocytes, in diabetic mice. TSP1 changes may play a critical pathophysiological role in heart failure, as suggested by a study showing that activation of the cardiomyocyte TSP1-CD47 axis promotes CaMKII phosphorylation (19). It is important to note that our in vitro data show that TSP1 promotes, in a paracrine manner, CaMKII phosphorylation in the contractile murine HL-1 cardiac cell line, and they demonstrate that ADAMTS13 helps to inhibits TSP1-induced CaMKII activation, possibly through TSP1 binding and sequestration (22). It is conceivable that in the absence of ADAMTS13 in diabetes, TSP1-CD47 signaling may be potentiated with the consequent aberrant activation of CaMKII and Cx43, eventually resulting in lethal arrhythmias. Although we conducted cutting-edge in vivo, ex vivo, and in vitro analyses to investigate the relations between Adamts13 deficiency and arrhythmogenic substrates of CaMKII activity in diabetic mice, we acknowledge that further studies are necessary to investigate in depth cardiac electrical abnormalities.
To summarize, we found the following results: 1) The lack of Adamts13 shortens the life span of diabetic mice. 2) Animal death is not associated with the occurrence of thrombotic events. 3) Diabetic Adamts13−/− mice show a propensity for cardiac arrhythmias, which is associated with increased phosphorylation of cardiac CaMKII and consequent lateralization of Cx43 and disturbance of Ca2+ handling proteins. Our data provide evidence that impaired ADAMTS13 expression may increase the risk for arrhythmias in diabetes and suggest a novel role of ADAMTS13 in diabetes that goes beyond antithrombotic activity.
Article Information
Acknowledgments. The authors thank Daniela Rottoli and Patrizia Borsotti for technical assistance and Miriam Galbusera and Elena Gagliardini (IRCCS – Istituto di Ricerche Farmacologiche Mario Negri) for helpful advice. The authors are grateful to Vincenzo Duino (Azienda Socio-Sanitaria Territoriale) for his critical review of the manuscript. Manuela Passera and Kerstin Mierke (IRCCS – Istituto di Ricerche Farmacologiche Mario Negri) helped to prepare and edit the manuscript.
Funding. M.L. has received a fellowship from Fondazione Aiuti per la Ricerca sulle Malattie Rare, Bergamo, Italy.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. P.C., G.R., A.B., and C.Zo. conceived and designed the study, discussed data, and wrote the manuscript. D.Ce. and V.L. performed experiments, discussed and analyzed data, and wrote the manuscript. C.Za., D.Co., F.G., R.N., S.C., M.L., and G.T. performed experiments and analyzed data. V.C., M.M., S.V., and S.G. performed experiments. P.C., D.Ce., C.Za., D.Co., V.L., F.G., R.N., S.C., V.C., M.M., M.L., G.T., S.V., S.G., G.R., A.B., and C.Zo. approved the final version of the manuscript. P.C. and C.Zo. 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.