Gene Delivery to Human Limbal Stem Cells Using Viral Vectors

Abstract

Limbal stem cell (LSC) transplantation is a promising treatment for ocular surface diseases especially LSC deficiency. Genetic engineering represents an attractive strategy to increase the potential for success in LSC transplantations either by correcting autologous diseased LSCs or by decreasing the immunogenicity of allogeneic LSCs. Therefore, two popular viral vectors, adeno-associated viral (AAV) vector and lentiviral (LV) vector, were compared for gene delivery in human LSCs. Transduction efficiency was evaluated by flow cytometry, quantitation of viral genomes, and fluorescence microscopy after introducing eight self-complementary AAV serotypes or LV carrying a green fluorescent protein (GFP) cassette to fresh limbal epithelial cells, cultivated LSC colonies, or after corneal intrastromal injection into human explant tissue. For fresh limbal epithelial cells, AAV6 showed the highest transduction efficiency, followed by LV and AAV4 at 24 h after vector incubation, which did not directly correlate with internalized genome copy number. The colony formation efficiency, as well as colony size over time, showed no significant differences among AAV serotypes, LV, and nontreated controls. The percentage of GFP+ colonies at 14 days post-seeding was significantly higher in the LV group, which plateaued at 50% GFP+ upon serial passages. Interestingly, AAV6-treated colonies initially showed a variegated transduction phenotype with no GFP+ colonies in serial passages. Quantitative polymerase chain reaction and AAV6 capsid staining revealed that transduction was restricted to differentiated cells of LSC colonies at a post-entry step. Following central intrastromal injection of human corneas, both LV and AAV6 transduced the stroma and endothelial cells, and AAV6 also transduced cells of the epithelia. However, no transduction was observed in derived LSC colonies. The collective results demonstrate the effectiveness of LV for stable human LSC genetic engineering and an unreported phenomenon of AAV6 transduction restriction in multipotent cells derived from the human limbus.

Introduction

Continuous repopulation of the four to six layers of the transparent corneal epithelia relies on an apparent differentiation continuum initiating with a multipotent cell population, limbal stem cells (LSCs), found in the corneoscleral junction (limbus). Despite decades of research, the identification of LSCs and their mechanism in regulating corneal epithelial maintenance and/or corneal wound repair are not well understood. The human limbus exhibits radially oriented vascular ridges between which are the limbal epithelial crypts where LSCs normally reside. The identification of LSCs themselves is complicated by an apparent lack of universally accepted marker(s); however, p63α,^1,2 the ATP-binding cassette subfamily G member 2 (ABCG2),^3,4 cytokeratin 14 (CK14),² CK15,⁵ antigen Ki67, frizzled class receptor 7 (FZD7),⁶ cadherin 3 (CDH3), CDH2, PAX6,⁷ and ABCB5⁸ are among those suggested.⁶

Although somewhat controversial,⁹ studies have postulated the centripetal migration model of epithelial maintenance that relies on daughter cells of LSCs to migrate centripetally from the limbus toward the central cornea residing in the basal epithelia.¹⁰ These cells have the capacity for transient amplification and migrate anteriorly generating differentiated epithelial cells. As the anterior layer of the corneal epithelia is lost approximately daily, replacement cells are continually needed, thus representing normal corneal epithelial homeostasis. Additionally, it is thought that LSCs and/or daughter cells also provide barrier function to prevent infiltration by epithelial cells of the conjunctiva.^10,11

Limbal stem cell deficiency (LSCD) is the absence or loss of normal function of LSCs resulting in the inability to repopulate the corneal epithelia, inflammation, and corneal vascularization. This deficiency disrupts the corneo-conjunctival equilibrium and results in the encroachment of the conjunctiva into the cornea occluding vision.¹¹ LSCDs can be trauma-induced or genetic and include ectrodactyly-ectodermal dysplasia, corneal dermoids, Stevens–Johnson syndrome, ocular cicatricial pemphigoid, aniridia, and ocular surface squamous cell carcinoma. Currently, there is no effective cure for LSCD, although autologous or allogeneic corneal LSC transplantation has been clinically performed in Europe with some success.^12,13 However, major hurdles remain including correcting the disease state in autologous transplants and tissue rejection in allogeneic procedures.¹²

Genetic engineering represents an attractive strategy to increase the potential for success in LSC transplantations by correcting autologous diseased LSCs or perhaps by decreasing the immunogenicity of allogeneic LSCs. For example, genetic modification of epidermal stem cells and hematopoietic stem cells has generated optimism for the treatment of epidermolysis bullosa and severe combined immune deficiency, respectively.^11,14–16 In the case of LSCDs, ex vivo or in vivo stem cell genetic manipulations have yet to be reported in a human context, and therefore, the therapeutic potential remains unrealized. In fact, only a handful of reports have investigated gene delivery in LSCs, the majority of which rely on viral vectors.^17–19 Oliveira et al. successfully transduced 14% of cultivated rabbit corneal epithelial cells using a lentiviral (LV) vector¹⁸; however, transduction of true LSCs in that population was only suggestive.¹⁸ More recently, Basche et al. reported LV-mediated gene delivery to limbal epithelial stem cells following corneal injections in mice.¹⁹ In that instance, gene expression was noted in corneal epithelial cells for 1 year, highly suggestive of permanent LSC genetic modification.¹⁹

Adeno-associated viral (AAV) vector^20,21 transduction of corneal epithelial cells following intrastromal injection was found to be transient, perhaps highlighting the nonreplicative nature of transgenic episomal genomes without chromosomal integration in resident corneal stem cells.¹⁹ Currently, there are no studies investigating AAV and LV gene delivery to human LSCs in situ in harvested primary limbal epithelial cells or in cultivated LSC colonies.

In the present study, the efficiency of gene delivery of eight natural AAV serotypes and LV vectors was investigated in human LSCs and in viable human corneas. For fresh limbal epithelial cells, AAV6 showed the highest transduction efficiency, followed by LV and AAV4 at 24 h post-transduction, which did not directly correlate with intracellular genome copy number. Green fluorescent protein (GFP) expression was relatively stable after in vitro propagation in the LV group, whereas the loss of GFP was observed in AAV-treated cells over time. Size and density analyses of cultivated LSCs demonstrated a small-sized cell population responsible for colony formation and, distinctly, larger more differentiated cells that do not continually divide. While LV vectors transduced both these cell populations, AAV6 transduction was biased for large/differentiated cells, with minimal to no transduction in the less differentiated small cell population, despite vector entry. Following AAV6 or LV intrastromal injection, GFP fluorescence was noted in the stroma, corneal endothelial cells, and central epithelium. However, only AAV6 resulted in the minimal transgene expression in the limbal epithelium by histology and flow cytometry. Importantly, intrastromal vector injections of either viral vector did not result in transgenic expression in derived LSC colonies or influence colony formation efficiency (CFE). These observations for the first time report that both AAV6 and LV successfully deliver genes to human primary limbal epithelial cells or to LSC colonies. Stable transgene expression was noted following LV transduction, highlighting its utility for treatment of LSCDs. Additionally, an unexplained phenomenon of AAV vectors is reported in which a post-entry step restricts AAV6 transduction of human LSCs with no apparent influence on cell viability.

Materials and Methods

Tissue procurement and cell dissociation

Seventy-eight human cadaver corneoscleral rims not suitable for use as transplants were selected based on death-to-preservation time (<25 h), death to experimental time (<14 days), and slit lamp and specular evaluations. Corneal rims were transported in Optisol GS from the eye bank, Miracles in Sight, to the laboratory. Under a dissecting microscope, the Tenon's capsule, conjunctival overhang, corneal endothelium, iris residue, and sclera were removed/trimmed off with sterile dissection tools and cotton swabs. The central cornea was recovered using an 8.0 mm trephine. The corneal limbus or central corneas were then incubated with 2.4 U/mL dispase II (neutral protease, grade II; Roche, Indianapolis, IN) at 37°C for 2 h. Under aseptic conditions, the epithelium cell sheaths were then gently peeled/scraped off under a dissecting microscope. Single cells were obtained by digestion with 0.25% trypsin-EDTA solution (Invitrogen, Thermo Fisher Scientific, Waltham, MA) for 10 min at 37°C. The limbal epithelium cells or the central corneal epithelium cells were then collected and counted for the next step.

Cultivation of human limbal and central epithelial cells

Cells recovered from the limbus or central epithelia were cultivated for CFE, colony immunostaining, and serial propagation on NIH/3T3 feeder layer cells as described by Mei et al. with minor modifications.⁶ In short, NIH/3T3 cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Life Technologies, Carlsbad, CA) containing 10% bovine calf serum and penicillin–streptomycin. One day before use, NIH/3T3 cells were treated with 4 μg/mL mitomycin C (Sigma–Aldrich, St. Louis, MO) at 37°C for 2 h and seeded at a density of 3 × 10⁴ cells/cm². Limbal or central epithelial cells were then seeded in the prepared 3T3 monolayer at a density of 300 cells/cm². Cells were fed with supplemental hormonal epithelial medium, a combination of DMEM/Ham's F-12 nutrient mixture (DMEM/F12, 1:1; Invitrogen, Gibco Cell Culture, Portland, OR), containing 2 ng/mL recombinant human epidermal growth factor (Thermo Fisher Scientific), 8.4 ng/mL cholera-toxin (Millipore, Sigma–Aldrich, St. Louis, MO), 0.5% dimethyl sulfoxide (Sigma–Aldrich), 0.5 μg/mL hydrocortisone (Sigma, St. Louis, MO), 1X N-2 supplement (Thermo Fisher Scientific), penicillin–streptomycin (Invitrogen, Life Technologies, Carlsbad, CA), gentamicin/amphotericin B (Lifeline, LS-1008), and 5% fetal bovine serum. The medium was changed approximately every 3 days. Colony growth was monitored under microscopy using phase contrast at different time points. The colony area was determined by the cellSens Dimension software (Olympus, Hamburg, Germany) using two-point circle measurement, and the data were plotted and analyzed with the GraphPad Prism software. The CFE assay was performed following rhodamine B staining (Thermo Fisher Scientific) and calculated as the colony number divided by the total number of cells seeded per dish multiplied by 100. For serial cultivation, the LSC colonies were treated with 0.05% trypsin and gently washed to remove the 3T3 feeder layer cells. The colonies were then dissociated for about 10 min with 0.25% trypsin at 37°C, and cells were then plated back on mitomycin C-treated 3T3 feeder layers.

Viral vector administration and transduction

Self-complementary AAV-cytomegalovirus (CMV)-GFP vectors, provided by the Vector Core at the University of North Carolina, were used in this study and characterized by quantitative polymerase chain reaction (qPCR), silver staining, and alkaline gel electrophoresis. An LV vector carrying the GFP reporter gene under the control of a CMV promoter (pTK945) was described earlier.²² Glycoprotein G from vesicular stomatitis virus (VSV-G) pseudotyped vector particles were generated by transient three plasmid transfection in 293T cells as described earlier.²³ Vector titers (3.5 × 10⁸ IU/mL) were determined by quantifying GFP+ cells following serial dilutions on 293T cells. Viral vectors were either directly added to freshly isolated cells or incubated with growing colonies on day 8 post-seeding. In some experiments, a single corneal intrastromal injection using a 31-gauge needle was performed on cadaver human corneas at a dose of 1 × 10¹¹ viral genome (vg) (AAV) or 1.75 × 10⁶ IU (LV) in a volume of 50 μL. Flow cytometry was used to determine the transduction efficiency at the following time points: 24 h post-transduction of freshly isolated primary cells, 24 h and 6 days post-transduction of the cultured colonies, or 48 h following intrastromal injection. Propidium iodide (PI) served as an indicator of the cell viability in the flow analyses. GFP fluorescence was also monitored by microscopy. To analyze the GFP expression pattern, cells were sorted by cell size using an FACSAria III machine.

Reverse transcription and qPCR

To detect the vg copy number, total DNA was isolated using the DNeasy Blood and Tissue Kit (Qiagen, Inc., Valencia, CA) and then subjected to qPCR using the Roche Universal probe no. 67 (Roche) and GFP primers (forward primer 5′-ccatgccgagagtgatcc-3′; reverse primer 5′-gaagcgcgatcacatggt-3′). Expression of specific markers was analyzed by quantitative reverse transcription PCR (qRT-PCR), as previously described.²⁴ Total RNA was isolated using the Qiagen RNeasy Mini Kit (Qiagen, Inc.) and was subjected to DNase I treatment (Ambion, Life Technologies) before reverse transcription. cDNA was then synthesized with the Second Strand Synthesis Kit (Invitrogen) in the presence and absence of reverse transcriptase. qPCR was conducted using the LightCycler^® SYBR Green I Mastermix. Primers are listed in Supplementary Table S1. The expression of each marker was normalized to that of recovered cDNA glyceraldehyde-3-phosphate dehydrogenase (GAPDH). For data analysis, 2^−ΔΔCt method was used to calculate the relative fold-change between the indicated experimental groups.

Immunofluorescence staining

For the immunostaining of histological cross sections of colonies, freshly isolated limbal epithelium cells were cultured using the methods described above, but on a transwell membrane instead of on standard tissue culture dishes. Colonies were fixed with neutral buffered 10% formaldehyde for 1 h at room temperature. For the primary human cornea, samples were fixed with 10% formaldehyde for ∼16 h. The fixed colonies and human corneas were then processed and embedded in paraffin. Five micron sections were prepared by the UNC CGIBD histology core. Immunofluorescence staining was performed following a previously described method.²⁴ Briefly, after treatment with blocking solution consisting of normal goat serum (10%) and Triton X-100 (0.4%) in phosphate-buffered saline (PBS) for 1 h at room temperature, antibodies specific for p63α (D2K8X, 1:200 dilution, no. 13109; Cell Signaling Technology, Danvers, MA), CK12 (1:1,000 dilution, Ab185627; Abcam, Eugene, OR), CK13 (1:250 dilution, EPR3671; Abcam), CK14 (LL002, 1:1,000 dilution, MA5-11599; Thermo Fisher Scientific), and GFP (1:500 dilution; AVES Labs, Inc., Davis, CA) were diluted in blocking buffer and incubated with samples overnight at 4°C. After washing three times with PBS containing 0.025% Triton X-100, samples were incubated with one of the following secondary antibodies for 1 h at room temperature: goat anti-mouse IgG (Alexa Fluor^® 488-conjugated, 1:1,000 dilution; Life Technologies), goat anti-rabbit IgG (Alexa Fluor 594, 1:1,000 dilution; Abcam), or goat anti-chicken (Alexa Fluor 488, 1:1,000; Abcam). Slides were mounted with coverslips using mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI) (p36971; Invitrogen) and observed using a fluorescence microscope.

Intact AAV6 capsid, p63α, CK12, and CK14 in different cell populations were detected by immunofluorescence after fluorescence-activated cell sorting (FACS). In brief, 2–10 μL of cell suspension was placed in the center of a poly-l-lysine–coated glass slide (Thermo Fisher Scientific) and spread to form a thin film using a pipette tip. After the cell smear dried, cells were fixed with 1% paraformaldehyde in PBS for 15 min, rinsed with PBS for three times, and then blocked for 1 h at room temperature with 10% normal goat serum and 0.25% Triton X-100 in PBS. Cells were then incubated at room temperature in blocking buffer containing one of the following antibodies: ADK1a antibody (mouse IgG, 1:50; BM5093; Acris Antibodies GmbH, Herford, Germany) (for detection of intact AAV6 capsids), or a CK14 antibody (described above), p63a antibody (described above), or CK12 antibody (described above). Following washes, slides were incubated with an appropriate secondary antibody (described above), washed, and counterstained with Hoechst 33342 (Thermo Fisher Scientific) and then covered with a coverslip with antifade mountant (P36930; Thermo Fisher Scientific). The positive stained cells were visualized and counted using microscopy (Olympus IX31 or Olympus FV3000RS).

Statistical analyses

Statistical analyses were performed using the GraphPad Prism 8 (GraphPad Software, San Diego, CA) and also by using the unpaired Mann–Whitney test for nonnormally distributed data, otherwise, t-test. Significant differences were defined as a p value <0.05. For the size distribution and growth rate analyses, a two-way analysis of variance (ANOVA) was applied.

Results

Isolation and characterization of corneal epithelial cells

To identify the population of corneal epithelial stem cells for harvest and subsequent investigations, human corneas were stained for p63α,^1,25 CK14,² and CK13,²⁶ which are reported to be LSC or LSC-niche markers. The results showed that CK14 or p63α-strong nuclear staining was primarily restricted to the limbus with minimal to no signal in the central cornea, as previously reported (Fig. 1A). CK13²⁷ showed the expected expression in the superficial and suprabasal limbal epithelial cells and minimal to no expression in basal limbal epithelial cells or corneal epithelial cells (Fig. 1A). CK12, a cytokeratin abundant in terminal differentially corneal epithelial cells, exhibited staining in the central corneal epithelia and in the superficial limbal epithelial cells and was absent in basal limbal epithelial cells where LSCs reside (Fig. 1A).

Figure 1.

Isolation and characterization of corneal epithelial stem cells. (A) Immunofluorescence staining for p63α, CK12, CK13, and CK14 in limbal (upper panel) and central (lower panel) epithelial layers of the human cornea (bar size = 50 μm). (B) Isolation of limbal and central epithelial cells. Approximately a 20 mm diameter sclerocorneal biopsy (a) containing cornea, limbus, and conjunctiva was obtained from the eye bank. Central epithelial cells were isolated from the central 8 mm diameter portion (b), whereas limbal epithelial cells were isolated from about a 2 mm diameter region defined as the limbus (c, the part inside the broken black circle). (C) A comparison of the indicated gene expression in limbal and central corneal epithelia by qRT-PCR. Data were collected from four individual donor corneas and presented as the fold-change between epithelial cells harvested from limbal or central areas. (D) Representative photomicrographs of a colony formed from limbal epithelium cells at day 12 after seeding. Scale bars: 400 μm (left panel) and 200 μm (right panel). (E) The CFE (defined as the colony number divided by the total number of cells seeded per dish and multiplied by 100) of epithelial cells freshly isolated from limbus or central corneas, the bar shows the median with range. Significantly higher CFE was demonstrated from cells isolated from limbal epithelium (n = 10, ****p < 0.0001, Mann–Whitney test). CFE, colony formation efficiency; CK, cytokeratin; qRT-PCR, quantitative reverse transcription polymerase chain reaction.

Next, the epithelial cells were isolated from the central cornea (8 mm diameter) or limbal region (2 mm diameter) (Fig. 1B). Flow analysis indicated that the majority of cells harvested from the limbus were smaller in size compared with those from the central corneal epithelium, consistent with a previous report (Supplementary Fig. S1A).²⁸ Next, epithelial cell cDNA from the limbal or central cornea was characterized by qPCR using primers listed in Supplementary Table S1 to amplify markers associated with stemness, including CK14, CK15, Δp63α, PAX6, Ki67, FZD7, CDH2, CDH3, ABCG2, and CK12 (Fig. 1C). CK15, ABCG2, and CDH3 expressions were >19-fold higher in limbal cells compared with central epithelial cells. The p63α, FZD7, and CDH2 cDNA abundance was moderately greater in the limbal epithelium, whereas CK12 was higher in the central cornea epithelium as predicted (Fig. 1C). Harvested limbal and central epithelial cells were then cultured on mitomycin C-treated 3T3 feeder cells and colony formation was tallied (Fig. 1D, E, Supplementary Fig. S1B). Under the tested conditions, the limbal epithelial cells gave rise to macroscopic colonies 5–7 days after inoculation and a typical colony is depicted (Fig. 1D, Supplementary Fig. S1B). The CFE for limbal epithelial cells (3.57 ± 4.04%) was significantly different from the central epithelial cell (0.26 ± 0.31%) group (n = 10, p < 0.0001) (Fig. 1E).

Transduction efficiency of AAV serotypes and LV in fresh LSCs

As the limbal epithelial cells demonstrated stem cell markers and clonogenic capacity, gene delivery to this cell population was investigated using self-complementary AAV and an LV vector harboring a CMV-driven GFP reporter cassette. Different AAV serotypes (1–6, 8, and 9) and a VSV-G pseudotyped LV vector were added to freshly isolated primary epithelium cells at a dose of 40,000 vg per cell for AAV and 10 IU/cell for LV. The transduction efficiency was determined 24 h post-incubation of limbal epithelial cells and vector. Viral vector transduction of 293 cells served as an additional control (Supplementary Fig. S2A).

Compared with 293 cells, microscopy and flow cytometry analyses, in all cases, demonstrated much lower transduction of limbal epithelial cells (Fig. 2, Supplementary Fig. S2A). The quantitative transduction detected by flow cytometry was from independent experiments using cells collected from 12 individual human donors. Although some variability existed among donors, AAV6 (8.3 ± 6.2%) and LV (4.6 ± 4.4%) showed the highest transduction for limbal epithelium cells, followed by AAV4 (4.1 ± 4.4%) 24 h following exposure (Fig. 2A, B), whereas none of the other AAV serotypes showed >2% GFP+ cells (Fig. 2B). PI staining demonstrated no significant difference in the cell viability following limbal epithelial cell exposure to the viral vectors under the conditions tested (Supplementary Fig. S2B). Since the transduction is often determined by whether the vectors get into the cells, internalized viral genomes were measured by qPCR. In contrast to induced GFP expression, AAV4 was the most efficient for cell entry followed by AAV2, which itself demonstrated little to no transduction (GFP+ cells), especially when compared with AAV6 (Fig. 2). Similar vector transduction efficiencies were observed for central corneal epithelial cells (Supplementary Fig. S2C).

Figure 2.

Transduction efficiency of primary limbal epithelium cells. The indicated AAV serotypes and LV vectors harboring a GFP reporter were incubated with freshly isolated human limbal epithelial cells. Transduction efficiency was examined by fluorescent microscopy and flow cytometry at 24 h post-incubation. (A) Representative fluorescence images. (B) Percentage of GFP+ cells detected by flow analysis from independent experiments, data (n = 12) are plotted as mean with SD, t-test, *p < 0.05, AAV6 versus NC, AAV1, AAV2, AAV3, AAV4, AAV5, AAV8, AAV9. (C) Viral genome copy number in treated cells detected by probe-specific qPCR. Data (n = 3) are plotted as mean with SD, t-test, *p < 0.05, AAV4 versus NC, AAV1, AAV3, AAV5, AAV6, AAV8, AAV9, LV. AAV, adeno-associated viral; GFP, green fluorescent protein; LV, lentiviral; NC, negative control; SD, standard deviation.

Viral vector–mediated transgene expression during clonal outgrowth

To observe the kinetics of transgene expression in cells capable of clonal expansion, and potential adverse effects of transduction on this process, limbal epithelium cells were incubated with AAV6 or LV vectors for ∼24 h and then seeded on inactivated 3T3 feeder cells. The colony morphology, size (defined as the colony area in square micrometers), and numbers were monitored daily and analyzed using image-analysis software (cellSens; Olympus). The CFE of LSCs treated by vehicle, AAV6, or LV ranged from 0.3% to 10.2%, and there were no significant differences in any of the observed metrics between the groups (n = 10; Fig. 3A, Supplementary Fig. S3). No significant difference between any of the treatments was noted for the colony growth rate, defined as the mean colony size from day 8 to day 20 using the two-way ANOVA test (Supplementary Fig. S3B).

Figure 3.

Transgene expression during clonal outgrowth. (A) The CFE of LSCs following the indicated treatments (NC, vehicle control), data (n = 10) are plotted as mean with SD. (B) Percentage of GFP+ colonies, defined by at least one GFP+ cell in the colony following the indicated treatment at P1 and P2. (C) Representative full GFP+ colony growth over time, bar size = 2 mm. (D) Variegated GFP phenotype was observed for AAV6-treated cells, bar size = 100 μm. LSC, limbal stem cell; P1, passage 1; P2, passage 2.

Next, the GFP+ colony phenotype, defined as at least one observable cell in the colony being GFP+, was evaluated following AAV6 or LV exposure in cultivated LSC colonies. For serial propagation, the clonal cells were passaged after 14–21 days of culture, seeded at a density of 1 × 10³ cells/cm² on a 3T3 feeder layer, and cultured for another 14–21 days. In passage 1, AAV6 and LV groups showed 10.9 ± 6.0% and 88.4 ± 11.1% of colonies were GFP+, respectively (Fig. 3B). However, in passage 2, no GFP+ cells were observed in the AAV6 group, whereas the LV group maintained 55.5 ± 16.3% GFP+ colonies (Fig. 3B).

Compared with the uniform GFP distribution in colonies following LV transduction, variegated GFP+ colony phenotypes for the AAV6 group were observed (Fig. 3C, D). Generally, the majority of the AAV-induced GFP+ LSC colonies exhibited only central or partial transduction. In these instances, the GFP+ cells appeared large in size and located on the surface of the multilayered colonies (Fig. 3D).

AAV6 transduction restriction to differentiated cells of an LSC colony

From the whole colony images, it was not possible to discern the precise localization of the GFP+ cells in the AAV6 treatment group (Fig. 3C, D). Therefore, LSC colony cross sections were stained for microscopic analyses. Hematoxylin and eosin (H&E) staining and immunofluorescence staining of p63α, CK14, and GFP were performed on processed paraffin-embedded colonies, as described in the Materials and Methods section. From the H&E and the nuclei staining (DAPI) (Fig. 4, Supplementary Fig. S4A), multiple layered colonies were observed, reminiscent of the human corneal epithelium (Figs. 1A and 4, Supplementary Fig. S4). Similar to actual corneal histology (Fig. 1A), p63α staining in the LSC colonies was primarily located in the basal layer (Fig. 4), whereas CK14 was observed throughout the colony section (Supplementary Fig. S4B–D).

Figure 4.

Localization of the GFP expression in the cultivated colonies. Immunofluorescence staining on cross sections of LSC colonies given the indicated treatments. p63α (red), GFP (green), and nuclei counterstaining (4′,6-diamidino-2-phenylindole [DAPI], blue), bar size = 50 μm.

As for viral vector–induced GFP expression, the majority of GFP+ colonies in the LV group showed positive staining in all cell types throughout the colony layers (Fig. 4). In stark contrast, and consistent with the earlier observation (Fig. 3D), the majority of colonies in the AAV6 group demonstrated transduction restriction to the superficial cell layer(s) that stained negative for the stem cell marker p63α (Fig. 4).

AAV6 LSC transduction restriction is post-entry

The observation of AAV6 vector restriction to differentiated cells within LSC colonies following primary limbal epithelia stem cell transduction is perhaps a consequence of the episomal nonreplicative nature of AAV transgenic genomes. In this division/dilution model, LSCs lose AAV genomes during division while nondividing differentiated cells of the same colony (located minimally in the superficial layer) persistently display GFP (Figs. 3 and 4). To investigate this possibility, the transduction efficiency of AAV6 or LV was determined following incubation with already established LSC colonies. As shown in Fig. 5A, both viral vector formats resulted in about 20% GFP+ cells 6 days post-treatment. Although the LSC colony presumably is clonal, at least two different cell populations were identified based on size/density and general morphology (termed large and small cells hereafter; Fig. 5B). Interestingly, it was apparent that LV transduced both large and small cell types, whereas AAV6 elicited a GFP+ phenotype in predominantly the larger cells of the overall population (Fig. 5C). The ratio between the percentage of GFP+ in the small and large cell populations was plotted in Fig. 5C, with a mean sixfold bias for large cells for AAV6 while only a twofold bias for the same cells using LV (Fig. 5C). To eliminate concerns of viral vector access to cells of the different colony layers, this experiment was repeated using dissociated cells from established colonies with very similar results, AAV6, not LV, was biased for large cell transduction (Supplementary Fig. S5).

Figure 5.

Transduction pattern following direct addition to cultivated LSC colonies. (A) Percentage of GFP+ cells at the indicated time point following incubation with AAV6 or LV vectors, t-test. (B) Representative dot plot of the FSC (size) and SSC (granularity) and the gating strategy for small and large cell populations (left graph) and representative small and large cell images stained with CK14 (right panel, bar size = 20 μm). (C) The ratio of GFP+ % in large and small cells following incubation with AAV6 or LV vectors on day 6. (D) Quantitation of Δp63α and CK12 cDNA in the large and small LSCs by RT-qPCR (data presented as the cDNA value in large cells divided by that determined in small cells). (E) Percentage of CFE of untreated small and large cells derived from LSC colonies. (F) Absolute quantitative analysis of intracellular viral genome copy number in large and small cells. (G) Representative immunofluorescence images of intact AAV6 capsid and p63α staining in large and small cells, bar size = 10 μm. *p < 0.05, t-test. FSC, forward scatter; SSC, side scatter.

The large and small cell populations observed in human LSC colonies were characterized by RT-qPCR detection of Δp63α and CK12 (Fig. 5D). The results show that Δp63α cDNA is approximately threefold higher in small cells compared with large cells, whereas CK12 cDNA is approximately ninefold higher in the large cells. Immunofluorescence detection of p63α, CK12, and CK14 demonstrated that the smaller cell population constituted about 80% of CK14+ cells and 30% of cells with strong nuclear p63α signal. In contrast, the larger cell population contained about 60% of CK14+ cells and less than 2% were positive for bright nuclear p63α (Supplementary Fig. S6). CK12+ cells in large and small populations were 3.4% and 1.3%, respectively (Supplementary Fig. S7). Importantly, only the small cell population was able to generate LSC colonies (Fig. 5E).

Next, to determine if the observed AAV6 transduction restriction is a consequence following particle uptake, internalized viral genomes from the large and small sorted populations were subjected to qPCR. The results demonstrate that the small population does internalize the AAV6 vector (around 10⁴ vg per host genome; Fig. 5F). Consistently, AAV6 intact particle staining was obvious in both large and small cell populations (Supplementary Fig. S7). The co-staining of p63α and AAV6 capsid results (Fig. 5G) indicate that the AAV6 did get into the p63α+ cells. Collectively, these results indicate that the LSCs colonies contain cells at different states of differentiation and that AAV6 transduction (not LV) is prevented at a post-entry step in the small cell population, which is solely responsible for colony formation.

LSC transduction was not observed following intrastromal injection

Recently, Ali et al. reported that AAV and LV vectors transduced limbal epithelial stem cells following intrastromal injection in mice, resulting in transgene-positive daughter corneal epithelia cells. As this would be of importance for anterior eye gene therapy, as well as of interest for ex vivo genetic manipulations, AAV6 or LV LSC transduction was analyzed following a 50 μL central intrastromal injection in viable human corneas ex vivo. Both AAV6 and LV resulted in corneal stroma and endothelial cell transduction 2 days post-injection (Fig. 6A, Supplementary Fig. S8). AAV6 also resulted in minimal central and limbal epithelial cell transduction (Fig. 6A, Supplementary Fig. S8). Flow cytometry of harvested central and limbal epithelial cells revealed 4.3 ± 0.6% or 0.5 ± 0.1% GFP+ cells, respectively, following AAV6 injection (Fig. 6B). No GFP+ epithelial cells (limbal or central) were detected following intrastromal injection of LV. Regardless, limbal epithelial cells recovered from corneas injected with either vector were then cultured for LSC colony formation. In no instances were any GFP+ cells detected, and the CFE was similar to the PBS injection control.

Figure 6.

Central intrastromal injection of viral vectors fails to induce stable LSC transduction. (A) The histology results showed that although stromal layer has strong GFP expression 2 days post-injection of the indicated vector, only minimal GFP was expressed in the epithelial cells. No GFP expression was colocalized with p63α+ cells in either groups, bar size = 20 μm. (B) Flow cytometric detection of percentage of GFP+ cells isolated 2 days post-injection of the indicated viral vector from the limbal and central epithelia (NC, negative control treated by vehicle; ND, not detected). ***p < 0.0005, t-test. AAV6 versus NC, LV. (C) CFE and GFP phenotype using limbal epithelial cells harvested from corneas injected with AAV6 or LV vectors.

Discussion

LSCD results from both intrinsic and extrinsic factors resulting in a compromised LSC niche and subsequent loss of the LSC population. In the absence of LSCs, the conjunctival epithelium invades and repopulates causing conjunctivalization of the previously transparent cornea. Additionally, corneal vascularization, scarring, and chronic inflammation also contribute to LSCD-induced vision loss. Genetic and trauma-induced causes include Stevens–Johnson syndrome, epidermal dysplasia, chemical/thermal burns, corneal infection, and contact lens. Treatment of LSCD is somewhat difficult as corneal clarity cannot be restored by a traditional corneal transplantation as the limbal region, which includes that LSCs is not included in the incoming tissue. Therefore, both autologous and allogeneic LSC transplantations have been performed in humans depending on the nature of the disease and its unilateral or bilateral nature. If genetic, the LSCD predominantly manifests bilaterally and requires treatment using allogeneic healthy cells that do not contain the disease mutation.

Although prolonged systemic immune-suppressive regimens are effective at preventing allogeneic graft rejection, they produce adverse side effects including other vision-threatening manifestations including glaucoma. Advancements in nucleic acid delivery have assisted cellular genetic engineering including high efficiency gene addition strategies and/or more complex yet less efficient approaches including precise gene editing.^29,30 These and other techniques present concerns in vivo, however, are ideal for ex vivo cell engineering as expanded cells can be well characterized before the transplantation. In such cases, AAV and LV vectors are the leading viral delivery vectors each with attractive attributes based on the mechanism of transgenic DNA persistence.

The data herein highlight LV as the most efficient format for the permanent modification of human LSCs, which can be greater than 50% upon serial passages. This result is not necessarily surprising as LV genomes integrate into host chromosome(s) and are replicated and therefore maintained upon cell division. Consistently, LV modification of stem cells is popular in the clinic with trials completed or underway for applications, including, HIV, β-thalassemia, metachromatic leukodystrophy, cerebral adrenoleukodystrophy, and sickle cell disease. Other popular LV applications include the genetic modification of T cells for cancer immunotherapy. Clinical trials underway using LV for chimeric antigen receptors delivery to autologous T cells ex vivo, followed by patient infusion, show promise as a therapy for multiple myeloma and metastatic melanoma. Considering LSC transplantation for LSCD, LV vectors have the potential to serve two different functions: (1) autologous cell correction for genetic disorders such as Stevens–Johnson syndrome and (2) to reduce the immunogenicity and overall induced inflammation of healthy allogeneic transplanted cells. The concept of a well-characterized universal donor LSCs for the treatment of LSCD is an attractive and perhaps soon to be realized therapeutic reagent.³¹

The AAV transduction results of fresh limbal epithelial cells, which include LSCs, demonstrated interesting results. First, AAV6 was deemed best for transduction at 24 h, which is consistent with a lot of reports demonstrating that to date, AAV6 appears best, in general, for AAV gene delivery to most types stems cells.^32,33 However, this result is not simply a measure of cell entry as AAV2 internalized more efficiently but was hindered in some capacity for transgene expression (Fig. 2). This phenomenon of AAV transduction restriction has been observed in other cell types including human dendritic cells.³⁴

Regarding GFP colony formation following LSC transduction, derived colonies from AAV6-treated cells demonstrated variegated phenotypes that were not maintained during serial passages (Fig. 3). This observation may be explained by the primarily episomal nature of AAV transgenic genomes, which are not replicated and therefore diluted during cell division. This understanding could rationalize the observation of minimal GFP expression following AAV6 treatment in p63α+ cells in colony cross sections, with indirect support from observed GFP+ cells in all layers when using integrating LV (Fig. 4, Supplementary Fig. S3). To determine if this hypothesis was correct, already formed LSC colonies (intact or dissociated before vector addition) were treated with AAV6, and surprisingly, transduction was still restricted to the larger differentiated cells of the LSC colony (Figs. 4 and 5). These larger, as well as the smaller, cells demonstrated AAV6 entry by intracellular copy number and intact capsid immunofluorescence staining; however, the smaller cells did not express GFP (Fig. 5, Supplementary Table S1). The precise mechanism for this restriction remains under investigation and may rely on multiple levels of action, one of which was recently reported for AAV transduction of dendritic cells.³⁴ In that study, deficiencies in AAV capsid uncoating was highlighted as a post-entry defect in dendritic cells and the authors were able to identify a mutant capsid that overcame this restriction.³⁴

Considering AAV transduction of stem cells in general, controversy in the literature exists with reports from our laboratory and others demonstrating that the AAV-inverted terminal repeats induce apoptosis in a p53-dependent manner in human embryonic stem cells.^35,36 Therefore, it remained possible in this work that LSCs also underwent apoptosis upon AAV transduction. However, in every context investigated in the current study, no evidence of AAV vector–induced toxicity was observed (Supplementary Figs. S2 and S3).

The collective results highlight that ex vivo LV gene delivery to LSCs represents an efficient approach to modulate LSCs for therapeutic applications. Both disease correction and immunomodulation using a molecule such as human leukocyte antigen G to prevent allogeneic LSC transplant rejection are potential therapeutic strategies.³⁷ In the case of AAV gene delivery to LSCs, this work uncovered a multipotent cell type that demonstrates transduction restriction at a post-entry step; however, this restriction is eliminated upon further differentiation.³⁴ Additional studies are underway to offer a better understanding of the relationship between AAV transduction and cellular “stemness.”

Author Disclosure

M.L.H. is an inventor on technology that has been licensed to AskBio. AskBio maintains a separate license on scAAV, which was evaluated as part of this study.

Funding Information

This study was supported by grants from the NIH RO1AI072176-06A1 (M.L.H.) and RO1AR064369-01A1 (M.L.H.). Pfizer-NCBiotech Distinguished Postdoctoral Fellowship (L.S.). A portion of the imaging was performed using the Neuroscience Center Microscopy Core Facility equipment, which is supported by funding from the NIH-NINDS Neuroscience Center Support Grant P30 NS045892 and the NIH-NICHD Intellectual and Developmental Disabilities Research Center Support Grant U54HD079124. The flow data reported in this publication were obtained using LSRFortessa and FACSAria II in the UNC Flow Cytometry Core Facility, which is supported in part by P30 CA016086 Cancer Center Core Support Grant to the UNC Lineberger Comprehensive Cancer Center, AIDS Research award number 5P30AI050410, and North Carolina Biotech Center Institutional Support Grant 2012-IDG-1006. This study was also supported by the Miracles in Sight Eye Bank.

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4

Supplementary Figure S5

Supplementary Figure S6

Supplementary Figure S7

Supplementary Figure S8

Supplementary Table S1