Robust cardiomyocyte-specific gene expression following systemic injection of AAV: in vivo gene delivery follows a Poisson distribution

Abstract

Newly isolated serotypes of AAV readily cross the endothelial barrier to provide efficient transgene delivery throughout the body. However, tissue-specific expression is preferred in most experimental studies and gene therapy protocols. Previous efforts to restrict gene expression to the myocardium often relied on direct injection into heart muscle or intracoronary perfusion. Here, we report an AAV vector system employing the cardiac troponin T (cTnT) promoter. Using luciferase and enhanced green fluorescence protein (eGFP), the efficiency and specificity of cardiac reporter gene expression using AAV serotype capsids: AAV-1, 2, 6, 8 or 9 were tested after systemic administration to 1-week-old mice. Luciferase assays showed that the cTnT promoter worked in combination with each of the AAV serotype capsids to provide cardiomyocyte-specific gene expression, but AAV-9 followed closely by AAV-8 was the most efficient. AAV9-mediated gene expression from the cTnT promoter was 640-fold greater in the heart compared with the next highest tissue (liver). eGFP fluorescence indicated a transduction efficiency of 96% using AAV-9 at a dose of only 3.15 × 10(10) viral particles per mouse. Moreover, the intensity of cardiomyocyte eGFP fluorescence measured on a cell-by-cell basis revealed that AAV-mediated gene expression in the heart can be modeled as a Poisson distribution, requiring an average of nearly two vector genomes per cell to attain an 85% transduction efficiency.

Fig. 1

(A) Schematic representation of wild type AAV and AAV vectors: ACMVLuc and AcTnTLuc carry the firefly luciferase cDNA driven by CMV and cTnT promoters, respectively. AcTnTeGFP encodes eGFP protein driven by the cTnT promoter. AAV inverted terminal repeats (ITR), and SV40 poly-adenylation (pA) are also indicated. B, Bioluminescence imaging illustrates cardiomyocyte-specific gene expression directed by the cTnT promoter. The AAV vectors [ACMVLuc (CMV) and AcTnTLuc (cTnT)] were packaged into AAV-6 capsids. One week old mice (n=4 per group) were injected with the indicated AAV vector (1×10¹¹ viral genomes/mouse) via jugular vein. (B) In vivo bioluminescence images obtained on 28^th day following vector administration. (C) Ex vivo bioluminescence images of the indicated tissues (heart; H, liver; L, thymus; T, kidney; K, spleen; S, intercostal muscle; Im, brain; B, Gastrocnemius; Gn,) from mice that received ACMVLuc (CMV) or AcTnTLuc (cTnT). Organs were immersed in D-luciferin solution for one minute and bioluminescence was imaged immediately thereafter using the IVIS 100 system. (D) Bar graph showing luciferase activities in protein extracts of the indicated tissues from the mice that received ACMVLuc (CMV) or AcTnTLuc (cTnT). Luciferase activity is expressed as relative light units per mg tissue (RLUs/mg tissue).

Fig. 2

Bioluminescence imaging comparing the efficiencies of five AAV serotypes for cardiomyocyte-specific gene delivery. The AAV vector AcTnTLuc was packaged into the indicated AAV serotype capsids (AAV-2, AAV-1, AAV-6, AAV-8 and AAV-9). One week old mice (n=4 per group) were injected with 1×10¹¹ viral genomes/mouse via jugular vein. (A) In vivo bioluminescence images obtained on day 28 after vector administration. (B) Bar graph showing luciferase activities in protein extracts from various tissues (brain; B, kidney; K, spleen; S, thymus; T, intercostal muscle; Im, Gastrocnemius; Gn, liver; L, heart; H) collected 28 days after vector administration. Luciferase activities are expressed as relative light units per mg tissue (RLUs/mg tissue).

Fig. 3

Fluorescence microscopy of heart cryosections from mice treated with increasing doses of AAV vector packaged in AAV capsid serotypes 2, 1, 6, 8, and 9. The AAV vector AcTnTeGFP was packaged into AAV capsids from the indicated serotypes (AAV-2, AAV-1, AAV-6, AAV-8 or AAV-9). One week old mice were injected with the indicated dose (n=2 per dosage) of viral genomes via jugular vein. Four weeks following vector administration, 6 μm cryosections of heart were prepared for analysis by fluorescence microscopy. All images shown here were captured at 40x magnification with a constant 0.5 sec exposure.

Fig. 4

AAV vector genome copy numbers in heart following systemic administration of increasing doses of AAV vector packaged in AAV capsid serotypes 2, 1, 6, 8 and 9. (A) Bar graph illustrating the increase in % transduction as a function of administered dose for each of the five capsid serotypes. (B) Graph illustrating the linear correlation between vector dose and mean vector genome copy number per μg of genomic DNA for each of the five capsid serotypes. (C) Graph of percent transduction versus mean vector copy number per μg of genomic DNA illustrates the artificial plateau imposed by defining a “transduced” cell as a cell containing one of more vector genomes.

Fig. 5

Fluorescence microscopy of heart cryosections illustrating eGFP expression. Shown are representative fluorescence images of heart cryosections from mice injected with AcTnTeGFP (AAV-9, 3.15×10¹⁰ viral genomes/mouse). (A) short-axis cross-section illustrating the homogeneity of expression throughout the extent of the left and right ventricles. (B) medium magnification images with the location of C indicated by white box. (C) 200x magnification image illustrating the lack of eGFP expression in the endothelium (open arrow) and smooth muscle cells (solid arrow) located within arterioles. (D and E) Comparison of eGFP protein detection by fluorescence imaging and by immunohistochemical staining. Immunohistochemical analysis was performed to confirm the eGFP expression observed by fluorescence microscopy. First, eGFP expression in the mouse heart was documented by fluorescence microscopy of heart cryosections (D). Next, after removing the coverslips, the same sections were processed for eGFP detection by immunohistochemistry (E).

Fig. 6

In vivo cardiac gene delivery follows a Poisson distribution. The intensity of cellular eGFP fluorescence in heart cryosections was used to assess the number of expressed viral genomes in individual cardiomyocytes. (A) Fluorescence image from the heart of a mouse injected with AcTnTeGFP (AAV-9, 1×10¹⁰ viral genomes/mouse). The tip of each arrowhead identifies a cell that was subsequently identified by K-means cluster analysis to contain n = 0 (red), 1 (green), 2 (blue), 3 (light blue), 4 (purple) and 5 or greater (yellow) AAV genomes per cell. (B) Three-dimensional bar graph showing the video intensities (y-axis) of individual cardiomyocytes (x-axis) from each group containing 0, 1, 2, 3, 4 or ≥5 vector genomes (vg) per cell (z-axis) as determined by the clustering algorithm. Each bar represents a single cardiomyocyte from the FOV shown in Panel A. The colors red, green, blue, light blue, purple and yellow represent clusters containing 0, 1, 2, 3, 4 or ≥5 AAV genomes per cardiomyocyte, respectively. (C) Colorized image showing the classification of each cardiomyocyte into clusters containing n = 0, 1, 2, 3, 4 or ≥5 AAV genomes per cell, as indicated by the colors red, green, blue, light blue, purple and yellow, respectively. (D) Bar graph showing the percentage of cardiomyocytes expressing 0, 1, 2, 3, 4 or ≥5 viral genomes (vg) per cell determined as described in the Methods (Actual) compared to the corresponding theoretical distribution (Poisson Mean) obtained by using the mean of the actual viral genomes per cell (2.0) as the Poisson parameter (λ ) = 2.0 to generate the corresponding Poisson distribution. (E) Correlation between the actual percentages of cells expressing “n” viral genomes (vg) as determined by image analysis and the theoretical percentages predicted by Poisson distribution. The percentages of cardiomyocytes expressing 0, 1, 2, 3, 4 or ≥5 viral genomes (vg) per cell were determined as described above (for AAV-9 at 1×10¹⁰ vg/mouse) as well as for AAV-9 at 3.15×10⁹ vg/mouse and AAV-8 at 3.15×10⁹ and 1×10¹⁰ vg/mouse (see Supplementary Information). The mean numbers of viral genomes per cell determined from image analysis were then used as the Poisson parameters (λ ) to generate the theoretical distributions of percentages (as in Panel D) for both AAV-8 and -9 at both doses (3.15×10⁹ and 1×10¹⁰ vg/mouse). Linear regression analysis of the correlation between the actual and theoretical percentage values yielded an R² = 0.85 with a slope near unity and a y-intercept near zero, indicating a strong correlation between the empirical data and the theoretical distribution predicted by Poisson.