Effective in vivo and ex vivo gene transfer to intestinal mucosa by VSV-G-pseudotyped lentiviral vectors

Background Gene transfer to the gastrointestinal (GI) mucosa is a therapeutic strategy which could prove particularly advantageous for treatment of various hereditary and acquired intestinal disorders, including inflammatory bowel disease (IBD), GI infections, and cancer. Methods We evaluated vesicular stomatitis virus glycoprotein envelope (VSV-G)-pseudotyped lentiviral vectors (LV) for efficacy of gene transfer to both murine rectosigmoid colon in vivo and human colon explants ex vivo. LV encoding beta-galactosidase (LV-β-Gal) or firefly-luciferase (LV-fLuc) reporter genes were administered by intrarectal instillation in mice, or applied topically for ex vivo transduction of human colorectal explant tissues from normal individuals. Macroscopic and histological evaluations were performed to assess any tissue damage or inflammation. Transduction efficiency and systemic biodistribution were evaluated by real-time quantitative PCR. LV-fLuc expression was evaluated by ex vivo bioluminescence imaging. LV-β-Gal expression and identity of transduced cell types were examined by histochemical and immunofluorescence staining. Results Imaging studies showed positive fLuc signals in murine distal colon; β-Gal-positive cells were found in both murine and human intestinal tissue. In the murine model, β-Gal-positive epithelial and lamina propria cells were found to express cytokeratin, CD45, and CD4. LV-transduced β-Gal-positive cells were also seen in human colorectal explants, consisting mainly of CD45, CD4, and CD11c-positive cells confined to the LP. Conclusions We have demonstrated the feasibility of LV-mediated gene transfer into colonic mucosa. We also identified differential patterns of mucosal gene transfer dependent on whether murine or human tissue was used. Within the limitations of the study, the LV did not appear to induce mucosal damage and were not distributed beyond the distal colon.


Background
Gene transfer to the gastrointestinal (GI) mucosa is a therapeutic strategy which could prove particularly advantageous for treatment of various hereditary and acquired intestinal disorders, including inflammatory bowel disease (IBD), GI infections, and cancer [1][2][3][4][5].
Non-viral vectors for delivery of exogenous DNA are limited by low efficiency of transduction in vivo, and do not provide long-term expression [6]. First-generation retroviral vectors can achieve long-term expression through their ability to integrate permanently in the genome of target cells, but gene transfer to the GI tract by these vectors was also found to be inefficient [7][8][9][10]. Conversely, adenoviral vectors can infect a wide range of cells, including intestinal epithelial cells, and show high levels of transgene expression [11,12], but these are non-integrating vectors and so expression is transient [13]. Furthermore, adenoviral vectors induce robust cellular and humoral immune responses in vivo, resulting in elimination of transduced cells and neutralization of the vector upon repeat administration [14]. In contrast, adeno-associated virus (AAV) vectors lack all viral genes and have limited capacity to induce cell mediated immune responses. In addition, AAV may have the potential to transduce intestinal crypt progenitor cells resulting in extended transgene expression [15].
LV, such as those derived from human immunodeficiency virus (HIV), are distinct from earlier generation retroviral vectors in their ability to infect quiescent cells through active import of the viral preintegration complex across the intact nuclear membrane, even in post-mitotic cells [16,17]. LV pseudotyped with the vesicular stomatitis virus envelope glycoprotein (VSV-G) exhibit an expanded host range that allows entry into most cell types in a wide variety of species ranging from zebrafish to man [18]. Highly efficient gene delivery by VSV-Gpseudotyped LV has been reported in various types of terminally differentiated primary cells in vitro and in vivo, including neurons, hepatocytes, cardiomyocytes, vascular endothelium, alveolar pneumocytes, and keratinocytes [19][20][21][22][23][24][25][26][27][28][29]. With regard to intestinal cells, these vectors have been shown to be capable of transducing human and canine colonic epithelial cell lines via the apical membrane in polarized monolayer cultures in vitro [30]. However, to date there have been no studies evaluating VSV-G-pseudotyped LV for gene delivery to primary intestinal cells, especially those subjacent to the mucosal epithelia, particularly in the context of the architecturally complex native tissue.
Therefore, in these studies we tested the ability of VSV-G-pseudotyped LV to deliver reporter genes to colonic mucosa via the apical surface in vivo, after intraluminal instillation per rectum in a murine model, and ex vivo in a human intestinal explant system [31].

In vivo studies
All in vivo studies were performed according to institutional guidelines under protocols approved by the UCLA Animal Research Committee. Briefly, 6 to 8-week-old female BALB/c mice (Charles River Laboratories Inc., Wilmington, MA, USA) were divided into a non-treated control (NC) group, and two groups that received intrarectal instillation of either LV or vehicle solution (DMEM). Prior to instillation, anesthetized mice were given an intrarectal enema of 50% ethanol (vol/vol in distilled H 2 O). Pretreatment with ethanol enemas has been shown to increase intestinal transduction with other vectors [36]. Two hours after the enema, 100 μL of vehicle or viral vector solution containing a total of approximately 1000 ng p24 equivalent units was instilled intrarectally. The mice were inverted for 30 sec after administration of intrarectal products to prevent leakage.

Macroscopic assessment and ex vivo bioluminescence imaging
Health status and body weight was monitored throughout the study. On days 2, 7, and 21 after vector or vehicle instillation, cohorts of mice were sacrificed, and the entire colon was removed. Colon length from cecum to anus, weight, and a macroscopic colonic damage score were recorded The macroscopic colonic damage score was based on the degree of tissue adhesion, mucosal ulceration, and intestinal wall thickness (Table 1) [37].
Ex vivo bioluminescence imaging was performed 2 days after rectal instillation of LV-fLuc using the Xenogen IVIS system (Caliper Life Sciences, Alameda, CA, USA). Bioluminescent signal intensity was expressed as photons per second per cm 2 per steridian (p/s/cm 2 /sr), and color images were processed with Living Image and IGOR-PRO analysis software (Wave Metrics, Portland, OR, USA).

Histological evaluation and X-Gal histochemical staining
For routine histology, colon and other tissues (spleen, liver, lung, kidney) were fixed in 4% paraformaldehyde overnight, placed in 30% sucrose/PBS for 2 hr, and embedded in OCT compound. Serial 5-μm cryosections were stained with hematoxylin/eosin (H&E) for evaluation of histopathology score. Mucosal inflammation was scored using a semi quantitative technique (Table 2) [38].
For X-Gal histochemistry, tissues were fixed in 2% glutaraldehyde for 2 hr, embedded in OCT compound, and 10-μm cryosections were stained in 1 mg/ml X-Gal solution at 37°C for 24sr, rehydrated, and counterstained with 0.1% Nuclear Fast Red (Sigma, St. Louis, MO, USA). The number of positive staining cells was counted in five independent fields in random areas on two non-consecutive slides at 200× magnification.

Real-time quantitative PCR (qPCR) analysis
Quantification of vector copy number was performed at each time point by TaqMan qPCR assay to detect the HIV-1 packaging signal sequence, using 300 ng genomic DNA (equivalent to 5 × 10 4 genomes) isolated from murine colon and other tissues, including stomach, small intestine, liver, kidney, spleen, lung, heart, brain, and bone marrow [39]. A reference curve was prepared by amplifying serial dilutions of LV-encoding plasmid in a background of genomic DNA isolated from untransduced murine colon. Genomic DNA from a PC3 cell line previously confirmed by flow cytometry to be 100% transduced by LV-GFP vector was used as a positive assay control.

Human colorectal tissue explant culture and ex vivo gene transfer
All human endoscopic biopsies were collected from healthy HIV-negative volunteers at UCLA Medical Center, Los Angeles, USA. The protocol for the use of human endoscopic biopsies was approved by Institutional Review Board of the David Geffen School of Medicine at UCLA (#02-05-001-13).
Explant cultures were established as previously described [31]. Briefly, explants were placed on presoaked Gelfoam ® (Pharmacia and Upjohn Company, Kalamazoo, MI, USA) rafts with the epithelium uppermost. Tissues were transduced by pipetting LV solution and polybrene (8 μg/mL) onto the top of each Gelfoam ® -supported explant. After 2 hr incubation, tissues were washed three times, placed on fresh Gelfoam ® rafts, and incubated at   37°C for a total of 24 hr. Explants were then either fixed in glutaraldehyde, embedded in OCT, and 10-μm cryosections prepared for X-Gal as above, or were fixed in formaldehyde, embedded in OCT, and 7-μm sections prepared for immunofluorescence staining.

Immunofluorescence staining and quantitative morphometry
The identity of transgene-expressing cells in harvested murine colon tissues or human colorectal explant tissues was examined by immunofluorescence double-staining with antibodies against β-Gal and cell-specific phenotypic markers. Briefly, serial 7-μm cryosections were stained for E. coli β-Gal (Promega, Madison, WI, USA avoid the high background staining caused by antibody binding to endogenous murine immunoglobulin when secondary antibodies were used to amplify primary murine antibodies. Slides were mounted and nuclei counterstained using VECTASHIELD ® and DAPI (4',6-diamidino-2-phenylindole; Vector Laboratories, Burlingame, CA, USA). In order to quantify LV-mediated β-Gal transduction efficiency, the total number of cells showing colocalization of β-Gal-positive and CD45 or CD4 positive signals in 3 randomly selected fields per section at 200 × magnification were identified, and expressed as percentages of the total number of β-Gal-positive cells or the total number of CD45 +/CD4+ cells.

Statistical analysis
All values were expressed as means ± SD. Comparisons between groups were made using the Student t-test and the Mann-Whitney U test, and p values <0.05 were considered statistically significant.

LV-mediated gene transfer to human colorectal cancer cell lines in vitro
LV-mediated transduction efficiency was first assessed in vitro using human colorectal adenocarcinoma cell lines Caco-2, WiDr, and LoVo. In each cell line, a dose-dependent increase in the number of cells showing positive signals upon X-Gal staining was observed with increasing LV concentration (Figure 2A, B). The highest dose tested was 10 ng p24 equivalent units of LV, which corresponds to a biological infectious titer of 10 6 transducing units (TU) on a standardized cell line such as 293T [40]. At this dose, transduction rates per 1 × 10 5 target cells were 40.6% for WiDr, 13.7% for Caco-2, and 13.3% for LoVo, indicating that although these colon adenocarcinoma cell lines are less permissive for LV compared to 293T embryonic kidney cells, significant gene transfer can be achieved with higher multiplicities of infection. Seppen et al. have previously reported higher rates of transduction in Caco-2 cells but used a transwell system with GFP expressing first and third generation lentiviral vectors [41].

Administration of LV for transduction of murine colorectal mucosa in vivo
The cant differences among these three groups in macroscopic damage assessment or histopathological evaluation scores (Additional File 1). More specifically, the use of ethanol enemas did not appear to induce significant mucosal damage. No pathological findings could be observed in kidney, spleen, lung, and liver harvested from control and LV-treated mice at each time point (data not shown).

Ex vivo bioluminescence murine imaging following rectal instillation of LV-fLuc
Strong positive signals were observed in the distal colon adjacent to the rectum in 8/8 (100%) of mice exposed to VSV-G-pseudotyped LV encoding the fLuc reporter gene ( Figure 3A). Quantitative measurement of bioluminescent photon emission from LV-transduced distal colon showed a significantly higher level of transduction than that of mock-treated colon (57560 ± 28960 p/s/cm 2 /sr for the LV instillation group vs. 6260 ± 813.3 p/s/cm 2 /sr for the mock instillation group; p < 0.001) ( Figure 3B). Additional studies are needed to determine whether the physical distribution of the ethanol enemas and/or the LV influenced the location of the bioluminescence signal.

Real-time qPCR analysis of LV transduction in murine colon tissue
Real-time qPCR was employed to quantify the vector copy number of LV integrated into genomic DNA from murine colon tissues after in vivo administration. Using spiked samples to obtain a reference curve, this method was determined to be sensitive enough to detect 50 copies of LV per 5 × 10 4 cellular genomes. Genomic DNA from both positive control cells (data not shown) and transduced rectosigmoid colon showed amplification of proviral LV sequences. The average copy number per 300 ng genomic DNA from LV-transduced colon was 231.3 ± 183.7 ( Figure 3C). As expected, no detectable qPCR signals were observed in genomic DNA from colon tissues of non-treated or mock-treated controls. Importantly, even in LV-treated animals at 2 days post-vector instillation, no detectable qPCR signals were found in any extraintestinal tissues examined, including stomach, small intestine, liver, kidney, spleen, lung, heart, brain, and bone marrow ( Figure 3C). As the LV was not treated with DNAse prior to injection, we cannot exclude the possibility that the data in Fig 3C in part reflect plasmid contamination.

X-Gal histochemistry and double immunofluorescence staining of LV-βGal transduction in murine colon tissues
Histochemistry was performed using optimized X-Gal concentrations and pH conditions to minimize background staining due to endogenous mammalian β-galactosidase in the GI tract. As expected, under these conditions, no staining was observed in mouse colon tissues from non-treated and mock-treated control groups, nor in the non-intestinal tissues (kidney, spleen, lung, and liver) from any animals including the LV-treated group. In contrast, positive-staining cells were observed in colon tissues from animals treated with LV expressing the E. coli β-Gal reporter gene. The initial number of positive-staining cells in the LV-treated group was 62.3 ± 18.5 per microscopic field at 200× magnification on Day 2 post-vector instillation (Figure 4). However, the number of positive-staining cells was then observed to decrease, to 19.1 ± 5.3 per 200 × field on Day 7, and 15.7 ± 5.1 per 200 × field on Day 21. These positive-staining cells were found predominantly toward the luminal surface of the colon, but were identified in both mucosal epithelium and, importantly, the lamina propria (LP) ( Figure 5A).
To further characterize the identity of the transduced cells, immunofluorescence double-staining was performed using an E. coli β-Gal-specific antibody in combination with various cellular immunophenotypic antibodies. Quantification of double-positive staining in colon tissues harvested on Day 2 post-LV instillation demonstrated that 27 ± 5% of the β-Gal-positive cells were also positive for cytokeratin. Consistent with their histological origin, β-Gal/cytokeratin double-positive cells were observed solely in the epithelial layers of colon tissues from transduced animals. Cells exhibiting doublepositive staining for both β-Gal and the common leukocyte antigen CD45 were found in both mucosal epithelium and lamina propria (LP) regions. These β-Gal/CD45 double-positive cells represented 70 ± 11% of the total population of β-Gal-positive cells.
A subset of LV-transduced white blood cells was further identified to consist of T lymphocytes, dendritic cells, or macrophages, as 27 ± 2% of the total population of β-Gal-positive cells were also positive for CD4. In fact, of the total population of CD4-positive cells, 48 ± 13% were β-Gal-positive, indicating a significant transduction of half of this cell population. These β-Gal/CD4 doublepositive cells were observed only within the LP of transduced murine colon. Notably, there were no cells showing co-expression of both β-Gal and CD8 or CD11c ( Figure  5B-F).

LV-β Gal-mediated ex vivo gene transfer to human colorectal explant tissues
Human colon tissue specimens were incubated with 1000 ng p24 units of LV-β-Gal, and examined within 24 hr of explant culture. The number of X-Gal-positive cells after transduction was 12.4 ± 1.5 per low power (×200) field. In contrast to the murine studies, the transduced cells in human samples were primarily found within the LP (Figure 6A). These data might reflect the availability of both apical and basal aspects of the explant to LV transduction, and/or the use of polybrene. By double immunofluorescence staining using human cellular immunophenotypic antibodies, 84 ± 16% of β-Gal-positive cells were also found to be positive for CD45. Of these, 41 ± 12% of β-Gal-positive cells were found to be CD4-positive, and 35 ± 8% were CD11c-positive. Notably, no β-Gal-positive cells expressed CD8 or cytokeratin ( Figure 6B-F). Non-specific background staining was not observed using this technique (Additional File 2). This identifies transduced cells as mucosal lamina propria CD4+ lymphocytes, macrophages, dendritic cells, and other leukocytes. Consistent with the results obtained after in vivo LV instillation in mice, 54 ± 15% of the total population of CD4-positive cells in ex vivo LV-transduced human explant tissues were also positive for β-Gal. We did not perform immunohistochemical examination of mesenteric lymph node tissue and so we do not know whether LV transduced cells could be identified at this site.

Discussion
Using a combination of molecular, imaging, and histological techniques, we have demonstrated for the first time that VSV-G-pseudotyped LV can mediate detectable gene transfer to architecturally intact human and murine intestinal mucosa after topical application. The overall transduction level was observed to decrease over time, with a 3-fold reduction between Day 2 and Day 7, but was then relatively consistent between Day 7 and Day 21. This may reflect both initial transgene expression from unintegrated episomal reverse-transcribed circular forms of LV provirus, which is transient in duration, as well as reduced numbers of transduced non-adherent cells (such as lymphocytes) as they migrate or circulate out of colon tissue in vivo, or reflect cells lost through apoptosis or karyolysis in the ex vivo explant model.
An additional finding is the demonstration that in human intestinal tissue, the transduced cells are predominantly in the lamina propria, an observation differing from previous beliefs that it is predominantly epithelial lineage cells that are transduced. Co-localization of phenotypic cell surface markers and transgene expression showed notable differences in the types of cells transduced between experiments involving LV-mediated transduction of murine vs. human colon tissue. Significant transduction of cytokeratin-positive cells was observed after in vivo transduction of murine colon, but not after ex vivo transduction of human explants colon tissue. Conversely, after ex vivo transduction of human explant colon tissue, but not after in vivo transduction of murine colon, we observed significant transduction of cells positive for CD11c, a cellular marker of mac-rophages and dendritic cells, but which is also weakly expressed on B cells, NK cells, and T cell subsets.
A number of parameters may have contributed to these differences. For example, to avoid degradation of the tissue architecture, LV-transduced human colon tissues had to be analyzed after no more than 24 hours of explant culture, and the time course for accumulating a detectable level transgene product may require a longer time interval for epithelial cells. Of course, LV transduction after rectal instillation in vivo also necessitates penetration from the intestinal lumen, and the mucosal epithelium represents the only tissue surface directly in contact with the virus solution. In contrast, virus applied ex vivo to the surface of endoscopically acquired human tissue explant samples has access to sub-epithelial cell layers that would not normally be available to virus administered intraluminally to intact intestine in vivo. Another difference between the in vivo and ex vivo experiments was the use of ethanol in the former, and polybrene in the latter, to facilitate lentiviral transduction. We did not explore the extent to which these different experimental parameters might influence the results of our experiments.
We did observe a difference between the present study and previous studies using explant tissues, in terms of the specific localization of transduced cells within the LP. In the present study, LV-transduced cells in human explant samples were found primarily in the subepithelial region of the LP whereas in other colorectal explant studies using adenoviral vectors [42] transduction was commonly observed in the basolateral region. However, it is difficult to compare our data with other published studies as the cell tropism may differ between vectors. In addition, differences in the histological distribution of vectorspecific cellular receptors, such as the coxsackievirus/ adenovirus receptor (CAR) required for efficient binding of adenovirus to target cells may influence experimental results.
There were also similarities in cell transduction results from the murine rectal in vivo transduction and human explant ex vivo transduction models. In both cases, VSV-G-pseudotyped LV exhibited the ability to transduce CD4-positive cell population. Notably, LV pseudotyped with the VSV-G envelope, which only requires binding to phospholipid constituents of most mammalian cells [43], achieved much more efficient infection of T cells than is possible with identical vectors pseudotyped with amphotropic retrovirus envelope [44]. However, CD4 is not only expressed by helper T lymphocytes and T regulatory cells, but also by macrophages and dendritic cells at lower levels. All of these cell types generally reside in the LP, which is where the predominant staining was observed in humans. It is not clear why the LV was unable to transduce CD8 positive cells and this finding warrants additional in vitro studies with purified T cell populations.
As the vectors used are replication-defective and can only mediate a single cycle of infection, this finding suggests that LV may be efficiently transported intact across the mucosal epithelium as has been suggested as a natural route of HIV infection. It has also been reported that dendritic cells may form projections into the intestinal lumen to sample incoming antigens, which may permit HIV infection as well as LV transduction of this cell population.

Conclusions
In summary, these studies have demonstrated the feasibility of using VSV-G-pseudotyped LV to safely achieve appreciable levels of localized gene transfer into architecturally intact primary murine and human intestinal tissues.