Non-syndromic genetic diseases affecting enamel formation have been broadly classified as amelogenesis imperfecta (AI). The most prevalent autosomal dominant (AD) form of AI is commonly caused by the ENAM gene mutations, while the X-linked form of AI is caused by alterations in the AMELX gene [14]. Several treatment options have been described to rehabilitate AI patients, ranging from preventive intervention to interim composite restorations, orthodontic treatment, orthognathic surgery and placement of cast crowns [15].

The complex secretory pattern of ameloblasts and narrow developmental window for tooth formation makes this process highly susceptible to disruption resulting in irreparable tooth malformation early in an affected individual's life. The restricted temporal pattern of amelogenesis suggests feasibility of ameloblast-targeted gene replacement strategy aimed at rescuing the AI-associated phenotype by transiently replacing defective gene products in the ameloblast cellular layer within this time window. A gene delivery vector, which could be injected into ameloblasts to accomplish this aim, offers a conceptually plausible approach to restoring tooth formation in AI-afflicted individuals.

A gene therapy vector capable of efficient delivery and expression of genes in ameloblasts is pivotal for such a strategy. Human adenovirus type 5 (Ad5)-based vectors offer the strongest in vivo transgene expression of limited duration without host DNA integration [16], [17], thereby minimizing human safety concerns. Further, advanced-generation “gutless” Ad vectors with reduced immunogenic properties [18], [19] could potentially be employed to minimize immune response in patients and prolong expression of the therapeutic/replacement gene if necessary. Those factors make Ad an attractive system for localized time-limited gene expression in ameloblasts. Improving the efficiency of target cell transduction to minimize viral load and the associated Ad immune response is pivotal for clinical applications of the above gene therapy strategies.

With this unique research tool in hand we next sought to evaluate the utility of Ad-based vectors for gene delivery to AI-ameloblasts as a potential strategy for gene replacement therapy of AI.

Efficiency of gene transfer by an Ad5 vector depends mainly upon the efficiency of the virus attachment to cellular receptors [49]–[53] and viral internalization [34], [35], [41]. Our initial experiments indicated that AI-WAm cells are about 23 times more resistant to Ad5 infection than A549 cells (see Fig. 2A), which was generally consistent with the lower levels of CAR expression observed in AI-ameloblasts on both mRNA (see Fig. 2B and C) and protein levels (see Fig. 2D and E). Although intracellular localization of hCAR is typically confined to tight junctions at the sites of intercellular contacts [54], the IHC staining for hCAR in AI-WAm revealed mostly diffuse cytoplasmic distribution of this protein (see Fig. 2E). Surprisingly, a similar hCAR staining pattern was observed also for hCAR-positive A549 cells, whereas in HEK293-T cells, used as another positive control in the same experiments, the protein showed a distinct localization to tight junctions (see Fig. 2E) in full agreement with the literature [54]. This difference between HEK293-T and A549 cells was most likely due to a substantially higher expression of hCAR in HEK293-T cells as suggested by a 15-fold higher expression of its mRNA in these cells as compared to A549 cells (data not shown). This agreed with the notion that A549 cells were more resistant to transduction with unmodified Ad5 as compared to HEK293-T or HEK293 cells (our unpublished observations). Other possible reasons for the observed difference in hCAR staining patterns between A549 and HEK293-T cells could be related to differences in physiological state (confluence, passage number) or growing conditions of those cells.

In addition to lower expression of hCAR in AI-WAm cells relative to A549 or HEK293-T control cells evidenced by our RT-qPCR and flow cytometry data, the IHC staining suggested paucity in cell surface-localization of the CAR protein, which could in part account for the observed resistance of AI-ameloblasts to Ad5 transduction.

To overcome the intrinsic CAR deficiency of AI-WAm cells we chose an approach of Ad5 infectivity enhancement using an array of Ad5 vectors with various genetic modifications of the capsid protein fiber that allow targeting of the vectors to alternate cell-surface molecules. We hypothesized that incorporation of a small cyclic peptide (RGD-4C) with affinity to cellular integrins (α_vβ3 and α_vβ5) [23], [36], [38], [39] or/and heptalysine peptide (pK7), targeting polysaccharide moieties of glycosylated cell surface proteins (HSPGs) [23], [43], could substantially increase Ad5 binding efficiency to cells of epithelial origin such as AI-WAm.

Expression of HSPG and α_vβ3/α_vβ5 integrins in AI-WAm and RD cells was confirmed by IHC staining and flow cytometry analysis (Fig. 3). The flow cytometry data were generally consistent with IHC staining of cells for HSPGs and integrins. The staining patterns for both integrins (α_vβ3 and α_vβ5) were similar in all cell types with both diffuse cytosolic distribution and cell membrane-localized signals (Fig. 3B). In contrast, staining patterns for total HSPGs based on detection of GAG moieties was very distinct from that for syndecan 4, which showed a highly polarized localization throughout all analyzed cell types (Fig. 3D), particularly conspicuous in A549 cells (Fig. 3F). This observation was consistent with specific intracellular localization of syndecan 4 to complex cytoskeletal adhesion sites, i.e., focal adhesions [55]. Of note, expression of syndecan 4 in ameloblasts was higher than in A549 cells relative to total HSPG in those cells (Fig. 3C and D).

Adequate comparison of Ad vectors with various tropisms in the context of the established dental cell population was an important, but challenging task due to the problem of proper dose normalization of vectors carrying different tropism modifications. Gene transfer assay with reporter gene expression as readout is typically employed to compare transduction efficiencies of viral vectors using either “physical” or “infectious” titers for vector dose normalization. Vector dose normalization for gene transfer experiments has been a subject of controversy in the field because the level of transgene expression in infected cells may not be reflective of the number of viral particles (VP) used for transduction or infectious (pfu/TCID₅₀) titer i.e. viral ability to form infectious progeny and produce plaques on infected cell monolayer. Moreover, Ad infectious titers are typically determined in helper (HEK293 or 911) cell lines regardless of tropism of analyzed vectors, which potentially leads to under- or overestimation of vector infectivity in other cell types with different repertoire of cell surface receptors.

To minimize potential errors in vector dose normalization due to the aforementioned factors we sought to employ a different approach based on empirical adjustment of reporter gene (Luc) expression levels for each virus in CAR-positive A549 cells as an equivalent of Ad5 infectious dose or “transgene expression dose”. We reasoned that the efficiency of vector-encoded transgene expression in a cell line with high levels of the native Ad5 receptor (CAR) represents an “equivalent” of gene transfer (cell binding and internalization) capability of fiber-modified Ad5 vectors that retain their natural ability to bind CAR (non CAR-ablated). Our rationale was based on the assumption that in CAR-expressing cells transduction by infectivity-enhanced vectors would occur predominantly via the native cognate receptor (CAR) pathway with relatively lesser contribution of other receptors provided small ligand modifications of the fiber do not significantly compromise CAR tropism of the vectors. On the contrary, in CAR-deficient or CAR-negative cells lines the cell binding mechanisms mediated by Ad5 fiber modifications would become predominant and determine transduction efficiency of the modified vectors.

The validity of our approach for viral dose normalization in A549 cells has been supported by the results of direct quantification of internalized genomic DNA (E4 copy number) for each modified virus in A549 (data not shown) and AI-WAm cells (see Fig. 5D). This analysis showed direct correlation between the Luc expression readouts following vector dose normalization and the actual Ad5 gene transfer levels assessed by intracellular quantification of Ad5 genomic DNA using the E4 region-specific qPCR (see Fig. 5D).

This study demonstrated that each tested fiber modification drastically enhanced Ad5 infectivity in CAR-deficient cells such as RD and AI-ameloblasts (see Fig. 5). The observed synergistic effect of the two small ligand modifications in the Ad5-pK7/RGD (G/L) vector was consistent with the original report [23] using RD as CAR-negative cells for evaluation of pK7 and/or RGD modification and suggested that pK7/RGD double modification represents an optimal fiber modification also for gene transfer to AI-ameloblasts at therapeutically relevant doses.

Our reciprocal blocking experiments with heparin or a mixture of recombinant α_vβ3 and α_vβ5 integrins demonstrated that the mechanism for the observed infectivity enhancement of Ad5-RGD (G/L) and Ad5-pK7 (G/L) vectors indeed involves their binding to α_vβ3/α_vβ5 integrins and HSPGs, respectively. Our experiments also suggest transduction through both types of molecules for double-modified Ad5-pK7/RGD (G/L) virus (Fig. 6). The same mechanism apparently mediated augmentation of transduction of control RD cells with negligible expression of CAR but substantial expression of HSPGs, including syndecan 4, since transduction of these cells was blocked by heparin with similar efficiency (data not shown). Although our blocking data generally agreed with the earlier study [23], we observed a substantially more robust inhibition of gene transfer by both integrins and heparin (Fig. 6). Despite higher doses of heparin used in our experiments, which could account for stronger blocking effects (58-fold and 11.2-fold for Ad5-pK7 and Ad5-pK7/RGD, respectively, versus 3.3-fold and 1.5-fold inhibition reported for the respective vectors in RD cells previously), it had no inhibitory effect on Ad5-RGD vector transduction in AI-WAm cells (Fig. 6B), in contrast to the slight effect in RD cells reported previously [23]. The reason for this discrepancy is unclear and could reflect subtle differences in the cell-binding mechanism of the HSPG-targeted Ads in AI-WAm versus RD cells.

Integrins α_vβ3 and α_vβ5 have been implicated in internalization of the group C adenoviruses (Ad2/Ad5), which involves their interaction with an RGD motif of the Ad capsid's penton base [34]–[38], [41] following the fiber knob-mediated step of the viral attachment to CAR [20], [33], [49]–[53]. The affinity of RGD interaction with α_vβ3 and α_vβ5 integrins is relatively lower than that of the fiber-CAR interaction, although it is essential for triggering formation of endosomes as well as for Ad5 endosome escape. The latter step requires α_vβ5 (β5 cytolasmic tail) but not α_vβ3 for endosome membrane permeabilization critical for Ad5 entering the cytoplasm. Of note, differential role of α_vβ3 and α_vβ5 has been suggested more recently for transduction of RGD-modified Ad (Ad5-RGD), implicating primarily α_vβ3 in binding to linear RGD peptide of the penton base as well as to the cyclic peptide RGD-4C in the fiber knob [56]. In this regard, robust expression of α_vβ3 on AI-WAm is consistent with strong augmentation of the Ad5-RGD (G/L) vector transduction of those cells observed in this study, whereas higher overall expression levels of HSPGs and integrins in AI-ameloblasts relative to RD cells was consistent with more efficient transduction of AI-WAm cell population with the RGD and/or pK7-modified viruses.

Because both replicative and replication-deficient Ad5/3 vectors carrying a chimera fiber modification demonstrated superior infectivity enhancement in different types of cancer cells both in vitro and in vivo [57]–[62], it was of interest to evaluate efficacy of this modification for transduction of non-cancer epithelial cells such as AI-WAm relative to the small modification ligands. However, Ad5/3 dose normalization to the panel of CAR-binding vectors presented a problem due to ablated CAR tropism of the virus resulting from the replacement of the entire Ad5 fiber knob domain with that of the serotype 3 Ad (Ad3) known to target a different set of cellular receptors [24], [63]. Several groups have identified the membrane cofactor CD46 as an attachment receptor for human Ad group B serotypes, including Ad11 [64], Ad35 [65] and Ad3 [47]. Furthermore, high-throughput receptor screening approach identified HSPGs as low-affinity Ad3 co-receptor interacting with its fiber knob domain [46]. It has thus been suggested that more than one receptor exists for species B Ads [46], [47], [64]–[67]. It remains controversial whether CD46 functions as attachment receptor for all species B serotypes. Most recently, the major Ad3 receptor was identified with desmoglein 2 (DSG2) [68], making the relevance of CD46 to the mechanism of Ad3 transduction highly questionable.

Our strategy of Ad5/3 infectious dose normalization to those of the fiber-modified CAR-binding Ads was originally based on the assumption that Ad5/3 and Ad5 possess similar abilities to transduce A549 cells resulting from comparable levels of CAR and Ad3 receptor(s). Due to the fact that this study was carried out prior to the discovery of DSG2 as a primary Ad3 receptor, in this study we analyzed expression of CD46 as a tentative Ad3 receptor [51] on A549 cells and found comparable expression of this molecule to CAR at least on mRNA level (Fig. 2C). We also found high levels of CD46 and CAR proteins in A549 cells (Fig. 2D and E), in line with earlier reports [28], but were unable to determine the relative ratio of those molecules quantitatively using flow cytometry approach since anti-CAR and anti-CD46 antibodies could have different affinities and cell labeling with different antibodies might not reflect the actual level of each receptor. Furthermore, after DSG2 was identified as the major Ad3 receptor, the analysis of CD46 expression in A549 cells became no longer relevant to the assessment of the bona fide Ad3 receptor levels and justification of our Ad5/3 normalization approach. In this regard, to validate our strategy for Ad5/3 normalization in A549 cells we employed a novel quantitative approach developed in a separate study to determining relative efficiency of A549 cell transduction by Ad5 and Ad5/3 vectors. Briefly, we used Ad5 and Ad5/3 viruses with EGFP-labeled capsids [68] as fluorescent tags in flow cytometry assay to probe A549 cells for the corresponding cognate receptors under conditions of receptor saturation and viral internalization block (4°C). Considering that the viruses had comparable infectious titers and capsid labeling efficiencies, the difference in Mean Fluorescent Intensities (MFI) of the cells bound to viral particles of each type (peak positions) relative to MFI of unlabeled cells (background) was reflective of the difference in A549 cell binding capacity of each vector (Ad5 or Ad5/3) under receptor saturation conditions. The above-mentioned approach revealed that A549 cells were capable of binding somewhat larger number of Ad5/3 particles than Ad5 particles (MFI=109.4 versus MFI=38.2, with 97% and 94% labeled cells, respectively). In light of these findings transduction efficiency of Ad5/3 vector (normalized to other vectors in A549 cells) observed for AI-WAm cells is likely to be underestimated and could potentially be higher than observed in our experiments (see Fig. 5). This could account for the relatively lower gene transfer augmentation demonstrated by the Ad5/3 vectors (L or G) in AI-ameloblasts as compared to the other fiber-modified vectors (Fig. 5B and C). Although expression of DSG2 in pre-ameloblasts was evidenced by gene expression arrays (our unpublished observations), expression of this protein in AI-WAm cell population may be down-regulated relative to HSPG and integrins, which could also account for the relatively lower augmentation of their transduction by the chimera Ad5/3 vectors.

This initial study thus defines an optimal gene delivery strategy to ameloblast-like cell population derived from a human patient with AI. Extrapolating in vitro results to clinical situation is a common challenge of the gene therapy field, although some general principles that have been derived from studies in cell lines have been supported by observations from clinical trials [69]. Ad-mediated gene transfer is dose-dependent, but increasing Ad doses inevitably leads to enhanced host immune response to Ad vectors and systemic toxicities. Therefore, the superior efficiency of gene transfer to AI-ameloblast cell population demonstrated by some capsid-modified vectors in this study along with the observed longevity of transgene expression (up to 4 weeks) in the target cells offers a potential utility of fiber-modified Ad5 vectors for local administration in dental tissues.