The effect of controlled expression of VEGF by transduced myoblasts in a cardiac patch on vascularization in a mouse model of myocardial infarction

Cardiomyocytes

Neonatal rat cardiomyocytes (RCM) were isolated from 2 day-old Sprague Dawley rats (Harlan, South Easton, MA) as previously described [18] according to an institutionally approved IACUC protocol. Briefly, heart ventricles were quartered, incubated at 4°C in a 0.06% (w/v) trypsin in Hank’s Balanced Salt Solution (HBSS, Gibco, Carlsbad, CA), and subjected to a series of digestions (3–4 min, 37°C, 150 rpm) in a 0.1% (w/v) solution of collagenase type II (Worthington Biochemical Corp., Lakewood, NJ) in HBSS. Cell suspensions from the first 4–5 digestions were collected, and preplated in polystyrene culture flasks for 75 min to remove the readily adhering cardiac fibroblasts. The non-adherent cells, that are enriched for cardiac myocytes, were counted using a hemocytometer and suspended in cardiac medium (CM) consisting of high glucose Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco) supplemented with 10% heat inactivated fetal bovine serum, as previously described [18].

Transduced skeletal myoblasts

Primary skeletal myoblasts (SM) were isolated from neonatal (2–5 day old) C57BL/6 mice (Jackson Laboratory, Bar Harbor, ME) as previously described [29]. Briefly, both forelimb and hind limb muscles were collected, minced to a slurry and digested with 1.5 U/ml of collagenase D (Roche Molecular Biochemicals, Indianapolis, IN), 2.4 U/ml of dispase II (Roche Molecular Biochemicals) and 2.5 mM CaCl₂ at 37°C. After filtration through an 80-µm nylon mesh (Becton Dickinson, Franklin Lakes, NJ) cells were plated on collagen-coated dishes, fibroblasts were allowed to attach for 45 minutes and non-adhering myoblasts were removed to a fresh dish.

Myoblast stocks were transduced with a retrovirus expressing the β-gal marker gene [29], expanded and frozen until use. Aliquots were then infected at high efficiency as previously described [27,30] with retroviral vectors expressing either VEGF linked to truncated CD8a as a FACS-quantifiable cell surface marker (pAMFG-VICD8), or CD8a alone as a control (pAMFG-ICD8). Briefly, high-titer retroviral supernatants were obtained from the helper-free, ecotropic ϕNX packaging cell line after transient transfection with the retroviral construct plasmids [31]. Skeletal myoblasts (4×10⁵ per one 60-mm dish) were plated, incubated in 4 ml of fresh viral supernatant supplemented with 8 µg/ml polybrene (Sigma-Aldrich, St. Louis, MO) for 15 minutes at 37°C, and centrifuged in a microplate carrier at 1,100g for 30 minutes at room temperature.

Myoblast populations were expanded at 37°C in 5% CO₂ on collagen-coated dishes, in growth medium consisting of 40% F10, 40% Dulbecco’s modified Eagle’s medium, and 20% fetal bovine serum, supplemented with 2.5 ng/ml FGF-2, as previously described [32]. FACS-purification of transduced myoblast populations expressing a controlled VEGF level was performed as previously described [27]. Briefly, transduced myoblasts were stained with an antibody specific to mouse CD8a and directly conjugated to allophycocyanin (clone 53–6.7 from BD Biosciences, Basel, Switzerland). Optimized conditions included the use of 0.8 µg of antibody/ 10⁶ cells in 200 µl of PBS with 5% BSA for 20 minutes on ice. Cell sorting was performed with a FACS Vantage SE cell sorter (BD Biosciences). Both VEGF-expressing and control transduced myoblasts were expanded for 2 passages after FACS-purification, frozen in 10% DMSO in calf serum and stored in liquid nitrogen.

Channeled elastomer scaffolds

Porous poly(glycerol sebacate) (PGS) scaffolds were prepared as previously described [18], with open interconnected pores ranging from 75 – 150 µm, a void volume of 89%, an elastic modulus in tension of 282 ± 25 kPa and a compressive modulus of 2.35 ± 0.03 kPa [17]. An array of cubically packed channels (250 µm in diameter, 500 µm wall-to-wall spacing) was created using a computerized 40W carbon dioxide laser (Epilog, Pololu Robotics & Electronics, Las Vegas, NV). Scaffolds were cored into disks (8 mm diameter, 1 mm thick), sterilized by autoclave and ethanol gradient (75, 50, 25, and 0% EtOH), and soaked overnight in DMEM containing 10% FBS.

Cell preparation and scaffold seeding

Engineered constructs were prepared using either (i) RCM alone or (ii) a mixture of RCM with control SMs (CD8-SM), or (iii) VEGF-expressing SM (VEGF-SM). One day prior to cell seeding, the scaffolds were coated with laminin (10µg/ml, BD Biosciences) [19, 33] and seeded by perfusing a suspension of cells as in our previous studies [17]. The inner 5-mm cores of laminin-coated scaffolds were directly perfused with the suspensions of RCM, CD8-SM or VEGF-SM as specified above, at a velocity of 1 mm/s for 2 hours. In all groups, the seeding density of RCM has been maintained at 1.48×10⁸ RCM/cm³ (corresponding to 3.3×10⁶ cells per scaffold). The ratio of SM/RCM in the mixture was 0, 1:2 or 1:6, corresponding to 0, 1.65×10⁶ or 0.55×10⁶ SM/scaffold, respectively.

The cell-seeded constructs were then cultured at unidirectional perfusion flow rate of 0.1 µL/s for an additional 7 days. Depending on the experiment, cells were cultured on the scaffolds in either of the following media: (a) cardiac medium (Cm) consisting of DMEM containing 4.5 g/L glucose supplemented with 10% FBS, (b) skeletal muscle differentiation medium (Diff) DMEM containing 1 g/L glucose supplemented with 2% horse serum (HyClone, South Logan, UT), and (c) a 1:1 mix of the Cm and Diff (Mix). All media were supplemented with 10mM HEPES, 2mM L-glutamine, and 100 units/mL penicillin (Gibco).

Animal model of acute cardiac ischemia

After 8 days of cultivation, the cardiac patches were implanted into a nude mouse (Harlan) model of cardiac infarction according to an institutionally approved IACUC protocol [6]. Briefly, the left anterior descending coronary artery (LAD) was exposed through a left thoracotomy and tied with a prolene suture. One week after ischemia, the patch was sutured onto the infarct bed (n=6/group). The perioperative mortality was around 50%. Four weeks later, mice were sacrificed and the hearts were harvested.

Cardiac function

Cardiac function was assessed by high- frame-rate, high-resolution echocardiography before myocardial infarction (normal baseline), 1 week after inducing infarcts by LAD ligation (infarct baseline), and 4 weeks after patch implantation. The high-resolution myocardial elastography technique that was previously developed [34, 35] was used. Briefly, a 30-MHz ultrasound transducer (RMV-707B, VisualSonics Inc., Toronto, Canada) was placed on the mouse chest in the parasternal position to obtain a long-axis view and a short-axis view of the left ventricle (LV) of the heart. The field of view was 12mm×12mm, the axial resolution was 55µm, and the lateral resolution was 115µm. The transducer operated on a line-by-line basis. A two-channel, 14-bit waveform digitizer (CompuScope 14200, Gage Applied Technologies Inc., Lachine, Canada) was used to simultaneously acquire the ultrasound radiofrequency (RF) signals and associated ECG at 200 MHz and 8 kHz, respectively. The acquisition involved 180 to 192 RF signals over the entire field of view and lasted approximately 7 min. Using retrospective ECG gating the acquired RF signals were registered off-line to reconstruct the RF image sequence for a complete cardiac cycle at the extremely high frame rate of 8 kHz.

The incremental axial displacements of the myocardium were estimated between consecutive RF frames using a time-efficient, 1-D normalized cross-correlation technique previously described [36]. A window size of 480µm and a window overlap of 95% were used. By automatically tracking the same regions of the LV and accumulating the incremental displacements over time [35], the cumulative displacements were obtained representing the total motion from the first frame to any other frame of the cine-loop. A linear Savitzky-Golay digital differentiator with a length of seven samples (140µm) [37] was applied to the cumulative displacements to estimate the cumulative strains. The cumulative strains over one cardiac cycle were extracted from a region-of-interest (ROI) that was manually segmented to correspond to the infarcted area that received the cardiac patch. The peak cumulative strains representing the total deformation of the myocardium during systole were determined. In the short-axis view, only the strains in the regions where the radial direction of the myocardial deformation aligned with the axial direction of the ultrasound beam were shown, since only 1-D, i.e. axial, displacements and strains were estimated [34, 35].

Histological evaluation

Whole hearts were harvested, bisected along the short axis, embedded in OCT compound (Sakura Finetek, Torrance, CA), frozen in freezing isopentane, and sectioned for histological evaluation. For general histomorphological evaluation, sections were stained with hematoxylin and eosin (H&E) and Masson trichrome, according to standard protocols. Color images were acquired on a Q-imaging camera mounted onto a light microscope (Olympus, Center Valley, PA).

Immunofluorescence

Tissue sections were fixed in acetone, incubated for 40 minutes in 10% goat serum (at room temperature), and then for 1 hr (at room temperature) in the following primary antibodies: (1) rat monoclonal anti-mouse CD31 (clone MEC 13.3; BD Biosciences) at 1:100, (2) rabbit polyclonal anti-NG2 (Chemicon, Temecula, CA, US) at 1:200, (3) mouse anti-cardiac troponin I (TnI; Millipore, Billerica, MA) at 1:150. Subsequently, tissue sections were incubated for 1 hr (at room temperature) in fluorescently labeled Alexa546 anti-rat, Alexa647 anti-mouse and Alexa488 anti-rabbit secondary antibodies (Invitrogen, Carlsbad, CA) at 1:200. Antibodies were diluted to 1.5% in horse serum and 0.5% in Tween 20 PBS.

Tissue morphometry

The thickness of the different regions in the myocardium (infarct bed, implanted tissue constructs and adjacent native myocardium) was evaluated using tissue sections stained with Masson trichrome. The infarct region could be clearly identified by the blue stained collagenous scar tissue. The implanted constructs were measured as the region right above the infarction. The underlying myocardium was identified as the region around the infarction (beneath or immediately adjacent) and under the patch. Vascular density was measured in tissue sections immunostained with fluorescent antibodies for endothelium (CD31, red) and pericytes (NG2, green). The vessel density was quantified with Image J software (NIH, Bethesda, MD) as the percent area that positively stained for CD31 separately in the three regions: fibrous scar tissue, tissue constructs, and the underlying myocardium for all experimental groups.

Statistics

Data were analyzed using the Prism 4.0a software (GraphPad, La Jolla, CA) and results were presented as means ± standard error. The significance of differences was evaluated by analysis of variance (ANOVA) followed by a Bonferroni test. P < 0.05 was considered statistically significant.

In vitro co-culture of cardiomyocytes and myoblasts

Since skeletal myoblasts (SM) are highly proliferative in vitro, co-culture conditions were optimized to avoid overgrowth of the rat cardiomyocytes (RCM) by myoblast populations during the 8 days of perfusion culture on PGS scaffolds. Control CD8 myoblasts were used for these experiments. Three different media compositions (cardiac medium, Cm; myoblast differentiation medium, Diff; 1:1 mixture of the two, Mix) and three different SM/RCM ratios were investigated. In all groups, the seeding density of RCM has been maintained at 1.48×10⁸ RCM/cm³ (corresponding to 3.3×10⁶ cells per scaffold).

The ratio of SM/RCM in the mixture was 0, 1:6 or 1:2, corresponding to 0, 0.55×10⁶ or 1.65×10⁶ SM/scaffold, respectively. As shown in Figure 1A, cardiac medium Cm and Mix medium supported the cultivation of RCM significantly better than Diff medium, and allowed the maintenance of SM without excessive growth. Co-cultures with a SM/RCM ratio of 1:6 in cardiac medium yielded the most reproducible maintenance of both cell types and this condition was adopted for subsequent studies. In all experimental groups, all cell types exhibited spatially uniform distributions throughout the scaffold volume, in agreement with our previous studies using the same method for scaffold seeding [17].

Four experimental groups were formed for animal implantation studies: (1) RCM alone (RCM), (2) 1:6 CD8-SM/RCM (CD8), (3) 1:6 VEGF-SM/RCM (VEGF), and (4) cell-free scaffold (PGS). VEGF-SM were purified by a flow cytometry-based technology we recently developed [27] to generate a population expressing a homogeneous level of VEGF that we have previously shown to be safe and effective (20.2±1.0 ng of VEGF/day/10⁶ cells). After 7 days in culture, the DNA content of the constructs from the three cell-loaded groups (RCM, CD8 and VEGF) was comparable (Figure 1B).

In vivo cardiac function

High-resolution myocardial elastography allowed myocardium strain to be measured in the region of the infarction, both in the short and the long axes. Strain measures the deformation of cardiac muscle and was quantified at baseline before LAD ligation, 1 week after LAD ligation right before patch implantation, and 4 weeks after patch implantation (Figure 2A). Before surgery, systolic strain values were within the normal range in all mice and very similar between short and long axis (0.50±0.10 and 0.52±0.08, respectively). One week after infarction, the heart function was significantly compromised in all mice while short- and long-axis strain values were in good agreement (0.16±0.06 and 0.09±0.08, respectively; at p<0.001 and p<0.0001 compared to the respective pre-infarction values).

To assess the functional effect of treatment, the ratio between the strain values 4 weeks after treatment and right before patch implantation (1 week after infarction) was calculated for each animal. These values therefore represent the fold-change in cardiac function caused by implantation of the different patches. As shown in Figure 2B–C, patches made of scaffold alone or seeded only with RCM did not improve cardiac function (strain ratio of about 1 both in the short and long axes). Addition of control SM caused a slight, but not significant improvement in the strain ratio to approximately 2. In contrast, treatment with patches generated by co-culturing RCM with the SM expressing a specific VEGF level significantly improved cardiac function, by 6- to 10-fold compared to all other groups (short- and long-axis, p<0.05 and p<0.01 respectively).

In vivo patch engraftment

In order to determine the mechanism by which VEGF-expressing patches significantly improved cardiac function, we analyzed the effects of the different conditions on patch engraftment, cell survival and differentiation, infarct size and tissue vascularization. At four weeks after implantation, all PGS scaffolds were completely degraded. Acellular patches made of acellular PGS contained loose fibrous tissue surrounded by a fibrous capsule (Figure 3A–B, Figure 4A). The patches formed from RCM alone or from RCM and control SM (not expressing VEGF) generated fibrous tissues with a high cellular content, but without organized muscle structure (Figure 3C–F, Figure 4B–C). In contrast, implantation of patches containing RCM and VEGF-expressing SM led to the robust generation of dense tissue with compact cellular structure (Figure 3G-H), and without fibrosis (Figure 4D). These tissues also showed best integration with the underlying myocardium, as evidenced by seamless transitions and connections between the tissue layers (Figure 3G).

The thicknesses of the infarct, the patch-derived tissue and the underlying myocardium were quantified on sections stained with Masson trichrome. The thickness of the infracted region underneath cardiac patches was similar in all groups, except for the RCM alone group, in which the infarct was slightly but significantly thinner (Figure 4E, p < 0.01). The improvement in cardiac function following implantation of VEGF-expressing patches was therefore not due to the change in the ventricle geometry in the infarct region.

The implantation of patches based on RCM, control SM and co-culture of RCM and VEGF-expressing SMs respectively decreased, maintained and increased the thickness of the tissue in the patch region, defined as the area above the infarction including the associated capsule, compared with the acellular elastomer patch alone (Figure 4F). Notably, the thickness of the VEGF-expressing patches was significantly higher as compared to all other conditions (p<0.01 vs CD8, p<0.001 vs PGS, p<0.001 vs RCM). Consistently, VEGF expression had a major effect on the survival of the underlying myocardium, leading to about 3-fold greater thickness compared to all other groups (p <0.001; Figure 4G).

Cell survival and differentiation in vivo

In order to determine the effects of controlled VEGF expression on the survival and differentiation of implanted cells, RCM and SM were tracked by expression of cardiac-specific Troponin-I and of the retrovirally encoded β-galactosidase enzyme, respectively. In the absence of VEGF expression, neither RCM nor SM could be found 4 weeks after implantation (Figure 5A–B and Figure 5D, respectively). However, in the VEGF-expressing patches not only both RCM and SM survived, but successfully differentiated into morphologically interconnected cardiomyofibers (Figure 5C) and mature skeletal myotubes (Figure 5E–F), respectively.

Vascularization in vivo

Vessel density was quantified in different areas of the implantation sites: the fibrous capsule, the patch-derived tissue and the underlying myocardium. As shown in Figure 6, the vascularization of the fibrous capsule was similar in all groups (Figure 6A). Since the patches made of PGS alone were completely resorbed and left no recognizable tissue except for the fibrous capsule, vessel density of the patch area could not be measured in this condition. The disorganized tissue generated by patches made of RCM alone or together with control SM had similar vessel densities, whereas VEGF-expressing patches had significantly increased vascular in-growth into the patch area (Figure 6B, p<0.01).

Interestingly, the VEGF-expressing patches induced angiogenesis in the underlying myocardium and caused a greater than 3-fold increase in its vessel density compared to the condition implanted with a PGS scaffold alone (Figure 6C, p<0.001). In contrast, RCM alone or RCM with control SM, did not affect vascularization of the underlying myocardium (Figure 6C). Controlled VEGF expression induced the formation of vessels resembling mature capillaries, of regular size and surrounded by NG2-positive pericytes (Figure 6D).