Nerve Growth Factor Gene Therapy Using Adeno-Associated Viral Vectors Prevents Cardiomyopathy in Type 1 Diabetic Mice

Infectious recombinant AAV particles were generated in human embryonic kidney 293 cells by a cross-packaging approach (16,17) whereby the vector genome was packaged into AAV capsid serotype 2 (18) or capsid serotype 9 (19,20). Viral stocks were obtained by CsCl2 gradient centrifugation. For AAV–human NGF (hNGF) preparation, the complete coding sequence of hNGF was excised from a pCMV-hNGF plasmid (LGC Promochem, Teddington, U.K.) and subcloned in the shuttle vector pAAV-MCS (Agilent Technologies, Edinburgh, U.K.). The plasmids containing the AAV backbones were obtained from Agilent. AAV2-hNGF and AAV9-hNGF contain a cytomegalovirus promoter, a β-globin intron, and a human growth hormone polyA signal. AAV–β-Gal drives the expression of β-galactosidase from a cytomegalovirus promoter and contains an intron from human growth hormone and a polyA signal derived from SV40.

In vivo experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources) and covered by U.K. Home Office licenses. Type 1 diabetes was induced in 8-week-old male CD1 mice (30–33 g body wt; Harlan, Blackthorn, U.K.) by five consecutive daily intraperitoneal injections of streptozotocin (40 mg/kg/day in 0.1 mol/l sodium citrate buffer, pH 4.5; Sigma, Dorset, U.K.) (21). Age- and sex-matched nondiabetic control mice were injected with the streptozotocin buffer. Diabetes was confirmed 14 days after the first streptozotocin injection by measurement of glycosuria using Test strips (Clinistix; Bayer, Basel, Switzerland) and reassessed at the moment the mice were killed. Fasting blood glucose levels were additionally measured at 4, 8, and 12 weeks in AAV9-injected mice by using a blood glucose meter (Accu-Check; Roche Diagnostics, Mannheim, Germany). Survival of mice was monitored for the entire duration of the experiments.

Gene transfer of hNGF or β-Gal control was performed in diabetic mice 14 days after the first streptozotocin injection. AAV2-hNGF or AAV2–β-Gal was delivered into the left ventricle wall by four injections (total dose of 1 × 10¹¹ viral particles in 20 μL) (5). AAV9-hNGF and AAV9–β-Gal were injected in the tail vein (total dose of 1.5 × 10¹² viral particles in 100 μL). Age-matched nondiabetic mice were injected with either AAV2–β-Gal or AAV9–β-Gal and used for reference.

Efficiency of cardiac transduction at 2 and 12 weeks after AAV2–β-Gal intramyocardial injection was evaluated by X-Gal staining. Moreover, the presence of AAV2-hNGF genome was determined by PCR on total DNA isolated from hearts at 2 weeks post–gene transfer. AAV9-hNGF–mediated efficiency and duration of transduction in the left ventricle were evaluated by real-time RT-PCR at 2, 4, 8, and 12 weeks post–gene transfer. hNGF biodistribution was additionally determined at 12 weeks post–gene transfer by real-time RT-PCR in the right ventricle, liver, kidney, spleen, and adductor muscles. hNGF cDNA level was normalized to ribosomal protein L32 cDNA level. PCR primers are shown in Supplementary Table 1. In addition, transgenic NGF protein levels were measured by ELISA (Promega, Southampton, U.K.) in mouse plasma at 2 and 12 weeks after gene transfer with either AAV2 or AAV9 and in protein extracts of left ventricle, right ventricle, liver, kidney, spleen, and adductor muscles taken at 12 weeks from AAV9 injections.

Murine NGF mRNA levels were evaluated by real-time RT-PCR in the left ventricle of streptozotocin-injected diabetic mice (12 weeks) and nondiabetic controls. PCR primers are shown in Supplementary Table 1, available in the Supplementary Data. Western blot analyses of left ventricle for total and phospho(Ser⁴⁷³)–Akt-B, total and phospho(Thr³²)-Foxo3a, and tubulin (all from Cell Signaling, Danvers, MA) were performed at 12 weeks post–gene transfer.

A high-resolution echocardiography system (Vevo 770; Visual Sonics, Toronto, Canada) was used in anesthetized (880 mmol/kg i.p. tribromoethanol; Sigma) mice to measure the following parameters before gene transfer (baseline) and at 4, 8, and 12 weeks thereafter: heart rate (bpm), left ventricle ejection fraction (LVEF) (measured as percent), left ventricle fractional shortening (LVFS) (percent), left ventricle chamber volume (microliters), and left ventricle internal diameter (LVID) (millimeters) during both systole and diastole (5). Additionally, peak systolic left ventricle pressure (LVP) (millimeters of Hg), maximal rate of LVP rise (dP/dt_max) (millimeters of Hg per second), and minimal rate of LVP fall (dP/dt_min) (millimeters of Hg per second) were measured with a miniaturized 1.4F Millar tip catheter (Millar Instruments, Houston, TX) at 12 weeks post–gene transfer (5). Moreover, in mice treated with AAV9 vectors, the early-to-atrial velocity ratio (E-to-A ratio) (index of diastolic function) was measured by echocardiography and pulsed Doppler at the level of the mitral valve (22).

Absolute myocardial blood flow (millimeters per minute per gram of tissue) was determined at 12 weeks post–gene transfer using fluorescent microspheres (Invitrogen, Paisley, U.K.) as previously described (5,22). Briefly, anesthetized mice (Tribromoethanol) were intubated, and a polyethylene catheter (PE10) was inserted into the right carotid artery and connected to a syringe pump for collection of reference blood. Then, the chest was opened and microspheres (0.02 μm in diameter) were injected into the left ventricle cavity (200 μl of total volume). Reference blood was collected at a rate of 0.15 mL/min. Mice were killed 2 min later, the heart was removed, and the left ventricle was separated, weighed, cut in small pieces, and digested in 10 mL of 2M ethanolic KOH (Sigma) at 60°C for 48 h. Finally, microspheres were collected and fluorescence intensity was determined by a fluorometer.

At 12 weeks post–gene transfer, hearts were arrested in diastole and perfusion fixed for paraffin embedding. Capillary and arteriole densities were evaluated in left ventricle transverse sections (20× magnification) after staining with isolectin-B4 (endothelial cell marker, 1:100; Invitrogen) and for α-smooth muscle actin (for marking vascular smooth muscle cells, 1:400; Sigma) (5). Apoptosis was determined by transferase-mediated dUTP nick-end labeling (TUNEL) assay (40× magnification) (in situ cell death detection kit Fluorescein; Roche, Indianapolis, IN) (5). TUNEL-positive apoptotic nuclei of endothelial cells were identified on the basis of cell morphology and vascular localization, while cardiomyocytes were considered apoptotic when TUNEL-positive stained nuclei were inside the cardiomyocyte contour. Fibrosis was evaluated using Picrosirius Red staining and revealed by using a polarized light (20× magnification). For each staining, fifteen left ventricle fields were randomly examined and averaged. Infiltrating hematopoietic cells were evaluated (20× magnification) after CD45 (BD Bioscience) staining developed using 3,3′-diaminobenzidine (Dako) followed by Mayer’s hematoxylin to counterstain the nuclei.

All data are expressed as means ± SEM. ANOVA was used, followed, when appropriate, by unpaired t test. Survival curves were analyzed by log-rank test. Analyses were performed using the SigmaStat 3.1 software. A P value <0.05 was interpreted to denote statistical significance.

Glycosuria (data not shown) and fasting glycemia (measured only in the AAV9 protocol [Supplementary Table 2]) increased in diabetic mice and were not affected by hNGF gene transfer. Relative mRNA expression of murine NGF decreased by 7.1-fold in the left ventricle of mice with 12 weeks of diabetes (P < 0.01 vs. age-matched nondiabetic controls [Supplementary Fig. 1]).

X-Gal staining of whole hearts injected 2 and 12 weeks in advance with AAV2–β-Gal confirmed successful transduction (Supplementary Fig. 2A–D). Additionally, hNGF DNA (Supplementary Fig. 2E) and hNGF mRNA (Supplementary Fig. 2G) were found in hearts that had received AAV2-hNGF. hNGF protein was present in the mouse plasma at 2 and 12 weeks after either AAV2-hNGF (Supplementary Fig. 2F) or AAV9-hNGF (Supplementary Fig. 3D). hNGF plasma levels were similar in the two groups at 2 weeks (P = not significant) but higher in AAV9-injected mice at 12 weeks (P < 0.05). Supplementary Fig. 3A shows the presence of hNGF mRNA in the left ventricle of AAV9-hNGF–injected mice at different time points. Supplementary Fig. 3B shows the hNGF mRNA tissue biodistribution at 12 weeks from intravenous AAV9-mediated gene transfer. After AAV9-hNGF, hNGF mRNA expression was higher in the heart (left and right ventricles), adductor muscles, and liver but present in all examined tissues. hNGF protein expression was found in the left and right ventricles, liver, and adductor muscles (Supplementary Fig. 3C).

A progressive deterioration of cardiac function was observed in β-Gal–injected diabetic mice (Tables 1 and 2). In AAV2–β-Gal diabetic mice, LVEF and LVFS gradually decreased over time (Table 1), with a significant reduction at 12 weeks post–gene transfer (P < 0.05 vs. nondiabetic mice for both comparisons). In AAV9–β-Gal diabetic mice (Table 2), LVEF was already diminished at 8 weeks post–gene transfer. By contrast, AAV-mediated hNGF gene transfer preserved both LVEF and LVFS (Tables 1 and 2). Ventricular dilatation was evident in diabetic mice compared with nondiabetic controls at 12 weeks post–gene transfer, whereas diabetes-associated left ventricle chamber dilatation and increase in LVID were preserved by either AAV2-hNGF or AAV9-hNGF (Tables 1 and 2). Heart rate gradually decreased over the time in diabetic mice but was preserved by NGF overexpression (Tables 1 and 2). Figure 1 shows a decrease in E-to-A ratio, indicative of diastolic dysfunction, already at 4 weeks post–gene transfer in AAV9–β-Gal–injected diabetic mice. Importantly, AAV9-hNGF improved diastolic function (Fig. 1 and Supplementary Table 3). In addition, NGF overexpression improved LVP, dP/dt_max, and dP/dt_min in diabetic mice (Fig. 2A–C for AAV2- and Fig. 2D–F for AAV9-injected mice).

Diabetes causes microvascular rarefaction in the myocardium (23). Indeed, at 12 weeks of diabetes, hearts injected with either AAV2–β-Gal or AAV9–β-Gal showed reduced densities of capillaries and small (diameter <50 μm) arterioles (Fig. 3). Conversely, cardiac microvasculature was preserved in diabetic hearts overexpressing NGF (Fig. 3A–C for AAV2- and Fig. 3E and F for AAV9-injected mice). The preserved microvessel density was associated with improved cardiac perfusion after hNGF gene transfer. In fact, at 12 weeks of diabetes the β-Gal–expressing hearts showed a significant reduction of blood flow in comparison with that shown without diabetes. By contrast, both AAV2-hNGF and AAV9-hNGF improved cardiac perfusion to values similar to those observed in nondiabetic controls (Fig. 3D and G).

Diabetes is associated with increased apoptosis in the heart (24). Accordingly, left ventricle sections at 12 weeks after diabetes induction and either AAV2–β-Gal or AAV9–β-Gal showed an increased number of TUNEL-positive apoptotic endothelial cells and cardiomyocytes compared with that shown in nondiabetic controls (Fig. 4A–F). By contrast, AAV2-hNGF and AAV9-hNGF reduced apoptosis of both cardiovascular cell types (Fig. 4A–C and E for AAV2-treated mice and Fig. 4D and F for AAV9-treated mice), providing evidence that sustained cardiac NGF expression improves cardiac cell survival under diabetic conditions.

The transcription factor Foxo3a triggers apoptosis and negatively regulates angiogenesis (4,25). It acts downstream of Akt signaling, and Akt-dependent phosphorylation of Foxo3a leads to its inactivation and degradation, thus promoting cell survival (4,26) and angiogenesis (25). We have previously shown that NGF activates the Akt/Foxo3a pathway in cultured cardiomyocytes and in the mouse infarcted heart. We additionally reported that Akt/Foxo3a signaling plays a functional role in NGF-induced cardiomyocyte survival and myocardial angiogenesis (4,5). Moreover, the Akt/Foxo signaling is known to be altered in the diabetic heart (27). As shown in Fig. 6, phospho-Akt (active form) and phospho-Foxo3a (inactive form) were significantly reduced at 12 weeks of diabetes in mice injected with β-Gal. Importantly, NGF gene transfer normalized the phosphorylated-to-total ratio of both Akt and Foxo3a in the diabetic hearts (Fig. 5A–F). These data suggest that the Akt/Foxo3a pathway may be involved in the regulation of the antiapoptotic and proangiogenic effects of NGF in the diabetic heart.

Cardiac interstitial fibrosis was significantly increased in the heart of diabetic control mice—an effect that was prevented by NGF overexpression (Fig. 6A–C). These data together with the improved perfusion in the hNGF-treated heart, decreased apoptosis of endothelial cells, and preserved cardiac microvasculature support the idea that NGF gene transfer is able to preserve the functionality of small vessels in the diabetic heart. Of note, no gross evidence of interstitial inflammation was observed in any of the investigated groups (Supplementary Fig. 4).

In the current study, survival of mice monitored for 12 weeks showed a substantial increased mortality of both AAV2– and AAV9–β-Gal-injected diabetic mice compared with that in AAV2\x{2013} and AAV9\x{2013}β-Gal-injected nondiabetic mice (Supplementary Fig. 5A and B). AAV2-hNGF did not reduce diabetes-associated mortality. By contrast, diabetic mice given AAV9-hNGF did not show higher mortality than nondiabetic controls. Although hallmarks of diabetic cardiomyopathy have been observed during the progression of diabetes, we cannot exclude that noncardiac events, including kidney failure and stroke, may have contributed to mortality in the diabetes groups. This would be in line with the improved survival after AAV9-hNGF, which determines a wider hNGF biodistribution (Supplementary Figs. 2G and 3B) and higher circulating hNGF level (Supplementary Figs. 2F and 3D) in comparison with AAV2-hNGF.

Diabetes and its several complications are a major public health problem (28). Accumulating evidence from epidemiological data and animal and clinical studies shows that the risk of heart failure is considerably increased by diabetes, and diabetes has been recognized as one of the major risk factors for the development of heart failure (28,29). Nonetheless, the pathophysiology of diabetic cardiomyopathy remains uncertain and novel therapeutic strategies for its prevention and rescue are needed. Several experimental mouse models of type 1 and type 2 diabetes are available, and some, including mice with streptozotocin-induced type 1 diabetes and type 2 diabetic db/db mice, have already been shown to develop cardiomyopathy (22,30,31).

Here, we report that AAV-mediated cardiac NGF overexpression inhibits the development of cardiomyopathy in mice with streptozotocin-induced type 1 diabetes. NGF, which was discovered in the nervous system and initially considered with respect to its neural functions (32), was later shown to exert prosurvival actions in the cardiovascular system and to promote angiogenesis, including in the infarcted heart (4,5). Cardiac NGF expression is altered during various pathological conditions. For example, NGF protein levels increase in the peri-infarct area of human and mouse hearts early after coronary occlusion (5,33). By contrast, myocardial levels of NGF were reported to be decreased in the setting of diabetes (8,9)—a finding that we have confirmed in the current study. Moreover, the reduction in cardiac NGF was shown to be a major contribution to cardiac diabetic neuropathy, which could be reverted by NGF overexpression (9).

The aim of this study was to evaluate the effect of NGF gene transfer on diabetes-induced cardiac dysfunction and microvascular rarefaction. Since diabetes is a chronic disease and the levels of cardiac NGF decrease progressively in the diabetic heart (10), to sustain NGF expression we chose an AAV-based gene transfer approach. In fact, AAVs have the ability to transduce nondividing cells (including cardiomyocytes) in vivo and to achieve a sustained transgene expression. AAV serotypes have minor differences in their capsids, which determine their tropism (12,13). The discrepancies in tissue tropism among the different serotypes are partly due to different mechanisms of uptake into target cells but also to postentry mechanisms involving endosome trafficking, nuclear uptake, and efficiency of viral capsid uncoating (34,35). For example, AAV2 has a broad tropism and infects a wide range of tissues with good efficiency, while AAV9 displays the impressive feature of being able to pass through the endothelial cell barrier of blood vessels when injected systemically, showing substantial levels of transduction in the heart muscle (15). In the current study, we first provided evidence of the beneficial effects of cardiac-specific overexpression of NGF after local intramyocardial injection by using an AAV2. Then, since this method requires an invasive surgery and is therefore not optimal as a potential therapy in cardiac patients or for the welfare of research animals, we additionally tested the cardiac responses to intravenous delivery of AAV9-hNGF.

Left ventricle dysfunction is a common feature of the diabetic heart. Throughout the development of diabetes, the contractile capacity of the heart progressively reduces and the heart shows signs of diastolic and systolic alterations. Left ventricle diastolic dysfunction is considered one of the earliest hallmarks occurring in the setting of diabetic cardiomyopathy, and it starts before the onset of systolic dysfunction (36,37). In the current study, cardiac function was measured by echocardiography and pulsed Doppler analyses before diabetes induction and then every 4 weeks. Moreover, at the end of the experiment left ventricle function was additionally measured by left ventricle catheterization. In our diabetic mice, echocardiography and pulsed Doppler analyses showed the presence of diastolic dysfunction at early stages of diabetes, which confirms our previous findings (22,30). In fact, β-Gal–injected mice exhibited a pronounced decrease in the E-to-A ratio (starting at 4 weeks of diabetes) and of left ventricle dP/dt_min (which was assessed as a terminal procedure 12 weeks after diabetes induction). Indexes of left ventricle systolic dysfunction were evident at 8 weeks (reduced LVEF) and 12 weeks (reduced dP/dt_max, LVEF, and LVFS) after diabetes induction. Taken together, in line with previous reports these results support the presence of diabetes-induced early diastolic dysfunction followed by systolic dysfunction (30,36). Moreover, the diabetic heart is also characterized by structural changes and, in particular, by increased ventricular dilatation. Notably, echocardiography revealed significant increase in LVID as well as left ventricle chamber volume dilatation in β-Gal–injected diabetic hearts. Importantly, NGF overexpression in the diabetic heart prevented left ventricle diastolic and systolic dysfunction and reduced left ventricle chamber dimensions. These results suggest that lack of NGF is detrimental for cardiac performances in the diabetic heart at both earlier and later stages of cardiomyopahty. On the other hand, NGF gene transfer appears utilitarian for improving cardiac function in diabetes. Diabetes is often characterized by cardiac autonomic neuropathy that results in sympathetic nerve damage and consequent alteration of circulating catecholamine levels and heart rate (38). Given that NGF regulates cardiac sympathetic innervation (9), it is possible that NGF overexpression could prevent nerve damage in the diabetic heart. In line with this, heart rate was higher in diabetic mice treated with hNGF. We did not evaluate the effects of NGF overexpression on cardiac innervations or plasma catecholamines, which represents a limitation of our study. Microvascular rarefaction is another common hallmark of the diabetic heart. Diabetic myocardial microangiopathy has been shown in both humans and rodents with streptozotocin-induced diabetes (39,40). Changes in the cardiac microvascular architecture are causally associated with increased apoptosis of endothelial cells, fibrosis deposition, and reduction of cardiac perfusion. Immunohistological analyses of heart sections at 12 weeks of diabetes confirmed a reduced number of capillaries and small-size arterioles and increased endothelial cell and cardiomyocyte apoptosis, confirming previous reports in diabetic patients and animals (41,42). The antiapoptotic and proangiogenic effects of NGF are well documented in several tissues and in different types of pathologies (4,5,7,43,44). The serine-threonine kinase Akt plays a crucial role in the prosurvival and proangiogenic effects of NGF (7,45,46). Phosphorylation/activation of Akt downstream of the NGF high-affinity receptor TrkA leads to the phosphorylation/inactivation of the Forkhead transcription factor Foxo3a, resulting in cardiomyocyte survival and angiogenesis (4,25). Moreover, Akt-Foxo signaling is altered in the diabetic heart (27). In agreement with our previous reports demonstrating the involvement of Akt/Foxo3a signaling in the prosurvival and proangiogenic effects of NGF (4,5), here we show that NGF overexpression increases Akt and Foxo3a phosphorylation in the diabetic heart, which may explain the cardioprotective effects. In the current study, we have found that AAV-mediated NGF cardiac overexpression preserves the vascular architecture of the diabetic myocardium. These significant effects of NGF gene transfer were also emphasized by increased myocardial perfusion in the presence of diabetes. Finally, another frequent feature of the diabetic heart is increased interstitial fibrosis, which results in increased left ventricle stiffness, thus compromising the ability of the heart to contract and relax efficiently (47,48). In line with this, histological analyses performed at 12 weeks of diabetes demonstrated accumulation of fibrotic content in the interstitial myocardium of β-Gal–injected diabetic mice, with a significant reduction after NGF gene transfer, indicating that NGF overexpression confers protection against cardiac fibrosis induced by diabetes.

At 12 weeks from gene transfer, similar cardiac hNGF mRNA expression was found in mice treated with AAV2 and AAV9. However, hNGF plasma levels were higher after AAV9. This is in line with the different tissue biodistribution observed in the two gene transfer approaches and suggests that the heart is not the only source of circulating hNGF after AAV9-hNGF injection. As the two gene transfer approaches resulted in comparable cardiac effects, it is possible that the amount of hNGF produced by AAV2-hNGF was sufficient to ensure optimal cardiac protection.

In conclusion, our results suggest that NGF gene transfer might have a tremendous therapeutic potential for the treatment of diabetic cardiomyopathy and encourage further translational efforts for the final benefit of diabetic patients. However, before this gene transfer approach can be tested in patients, additional preclinical studies need to be performed (including in larger animals) not only to verify the efficiency and the safety of AAV-mediated NGF overexpression in the setting of type 1 diabetes but also to find the most efficient AAV serotype, as well as the optimal dose and delivery route to be used. Moreover, since NGF is secreted and TrkA is present in several cell types, a better understanding of NGF gene transfer impact on noncardiac tissues and organs needs to be developed. Moreover, potential side effects common to all other proangiogenic and prosurvival factors (like tumor and ocular angiogenesis and inflammatory angiogenesis of joints) should also be considered. Finally, it seems important to preclinically evaluate the result of NGF gene transfer in association with insulin and other therapeutic treatments commonly used in diabetic patients.

This study was funded by a European Programme in Type 1 Diabetes research grant from the European Foundation for the Study of Diabetes together with the Juvenile Diabetes Research Foundation and the Novo Nordisk (to C.E.). It was also funded by the British Heart Foundation senior research fellowship grant (to C.E.). I.F. was recipient of a bursary from the Sardinian regional government (“Master and Back” scheme). No other potential conflicts of interest relevant to this study were reported.

M.M. researched data and wrote the manuscript. B.D., A.C., and I.F. researched data. L.Z. prepared AAV vectors and reviewed and edited the manuscript. M.G. reviewed and edited the manuscript. C.E. ideated the study, obtained the funds for the research, and wrote the manuscript.

Parts of this study were presented in abstract form at the American Heart Association (AHA) Scientific Sessions, Orlando, Florida, 14–18 November 2009; at the British Microvascular Society Meeting, Exeter, U.K., 19\x{2013}20 April 2010; and at the AHA Scientific Sessions 2010, Chicago, Illinois, 13–17 November 2010.

Drs. Rajesh Katare and Graciela-Sala Newby (both from University of Bristol) helped with measuring E-to-A ratio and with NGF vector preparation, respectively.