AAVrh32.33 capsid Demonstrates Unexpected Dermal Tropism Regardless of Immunodominant Epitope

Abstract

Cutaneous gene therapy has the potential to treat a wide range of skin disorders, but effective delivery remains limited by the skin’s barrier properties and immune surveillance. Here, we identify AAVrh32.33 as a potent vector for targeting dermal stromal compartments. Following systemic administration in mice, AAVrh32.33 mediated robust and durable transgene expression, with preferential targeting of dermal fibroblasts and hair follicle bulge cells. Expression peaked at one month and persisted for up to two years, highlighting its suitability for chronic conditions. To reduce immunogenicity, a dominant CD8⁺ T cell epitope was disrupted, generating the IDPΔ variant. This modification attenuated peptide-specific T cell responses while preserving stromal transduction. In human skin explants, IDPΔ achieved high levels of gene expression, primarily in dermal fibroblasts and precursors, confirming translational relevance. Finally, vectors encoding CCL17, CCL20, and CCL22 demonstrated localized targeted therapeutic gene delivery in both healthy and inflamed skin, underscoring the feasibility of using this platform to reshape local immune responses. Together, these findings establish AAVrh32.33 and IDPΔ as promising platforms for durable cutaneous gene therapy, with direct applications in diseases such as vitiligo where long-term modulation of the dermal microenvironment is essential.

Graphical Abstract

AAVrh32.33 and its IDPΔ variant show durable transgene expression in mouse and human skin, revealing unexpected dermal tropism and supporting their use in cutaneous gene therapy.

Introduction

Skin disorders encompass a diverse group of over 3,000 conditions, ranging from minor skin irritations to severe chronic diseases that can profoundly impair both physical health and psychological well-being. Globally, these disorders represent one of the most common categories of human disease, affecting individuals across all ages and demographics.¹ Approximately 700 skin conditions have a confirmed genetic basis, including severe monogenic diseases such as epidermolysis bullosa (EB), lamellar ichthyosis, and xeroderma pigmentosum. In addition to inherited conditions, autoimmune and inflammatory diseases such as vitiligo, psoriasis, and alopecia areata contribute substantially to clinical burden and unmet therapeutic needs.² Importantly, skin diseases differ in the primary compartment underlying pathogenesis, with some arising from defects in the epidermis and others involving dermal cells or dermal-immune interactions.³ Compared to the rapidly renewing epidermis, the dermal compartment is relatively slow-cycling, making it an attractive target for durable gene delivery.⁴ Therapeutic gene expression within dermal fibroblasts and adnexal structures may be particularly well suited for diseases in which benefit can be achieved through sustained local expression. This is especially relevant for immune mediated skin diseases, where dermal expression of chemokines or immunomodulatory factors can shape the local immune cell profile.

Gene therapy has emerged as a promising approach for addressing both genetic and immune mediated skin diseases. The skin offers distinct advantages as a target tissue for gene delivery, including easy accessibility, the potential for localized treatment, and non-invasive monitoring of therapeutic efficacy.⁵ Furthermore, in the context of immune mediated disorders, the skin’s immunological richness provides a potential gateway for targeted immune modulation.⁶ Despite these opportunities, effective gene delivery to skin tissue remains a major challenge. Both viral and non-viral delivery platforms have been evaluated, yet each comes with notable limitations in terms of efficiency, specificity, and durability.⁷ Among viral vectors, adeno-associated viruses (AAVs) have garnered particular interest due to their favorable safety profile, low immunogenicity, and episomal persistence, which minimizes the risk of insertional mutagenesis.⁸ The approval of seven AAV-based gene therapies by the U.S. Food and Drug Administration (FDA) has firmly established AAVs as a clinically validated therapeutic platform, underscoring their broad potential beyond current indications.⁹ Several AAV capsid variants such as AAV2 and AAV8 have been explored for skin gene therapy applications; however, they have demonstrated limited efficiency in transducing skin cells, especially following systemic administration.^10–12.

These limitations have prompted ongoing efforts to identify or engineer AAV capsids with improved tropism for skin. Less conventional or synthetic AAV capsid variants are increasingly being evaluated as potential solutions to the poor transduction efficiency observed with traditional vectors. In parallel, strategies to mitigate immune recognition and enhance transgene durability are being actively pursued to improve therapeutic outcomes.^13,14

In this study, we investigate the potential of AAVrh32.33, a less common AAV capsid variant previously shown to elicit immune responses in murine models,^15,16 to target skin. We assess its ability to transduce various skin resident cell types in both murine and human models and evaluate strategies to reduce its immunogenicity without compromising efficacy. By focusing on dermal-targeted gene delivery, our goal is to determine whether AAVrh32.33 and its engineered variant can support long-lasting, cell type specific expression suitable for immune modulation and therapeutic protein delivery in immune mediated skin diseases.

Results

Systemic Delivery of AAVrh32.33 Efficiently Targets Dermal Fibroblasts and Skin-Resident Cells

AAV capsids differ markedly in their tissue tropism, and biodistribution studies can reveal unanticipated sites of transduction. While working with AAVrh32.33, we observed an unexpected surface associated bioluminescence signal two weeks after intramuscular delivery of AAVrh32.33 expressing luciferase, in contrast to AAV1-injected mice (5 × 10¹⁰ vector genome (vg) per mouse) (Figure 1A). This finding suggested potential skin involvement, prompting us to investigate whether AAVrh32.33 preferentially transduces skin resident cells. To investigate this potential tropism more directly, we next examined the spatial distribution of transgene expression within the skin following systemic delivery (1 × 10¹¹ vg per mouse). Tail skin samples were collected two weeks post injection, enzymatically separated into dermal and epidermal layers and each compartment analyzed by flow cytometry (Figure 1B). GFP positive cells were found predominantly in the dermis, indicating that AAVrh32.33 preferentially transduces dermal rather than epidermal cells following systemic delivery (Figure 1C and Figure S1A). We then compared the skin transduction efficiency of multiple AAV capsid variants, including AAV1, AAV2, AAV5, AAV6, AAV8 and AAVrh32.33, following intravenous delivery of GFP expressing vectors (1 × 10¹¹ vg per mouse). Flow cytometric analysis of tail skin post injection revealed detectable GFP expression only in mice treated with AAV1, AAV6, AAV8 or AAVrh32.33. Among these, AAVrh32.33 resulted in significantly higher frequencies of GFP positive cells compared to AAV1, AAV6 and AAV8, further supporting its superior ability to transduce skin cells in vivo (Figure 2A and Figure S1B).

Figure 1. In Vivo Distribution and Dermal Preference of AAVrh32.33.

(A) Images of full-body bioluminescence in NU/J mice two weeks after intramuscular injection with AAVrh32.33- or AAV1-luciferase (5 × 10¹⁰ vg per mouse). Representative images are shown (n = 2 for PBS, n = 3 for AAVrh32.33, and n = 4 for AAV1). (B) Experimental schematic. (C) Percentage of GFP⁺ cells among live dermal and epidermal populations following systemic injection of AAVrh32.33-GFP (1 × 10¹¹ vg per mouse) in C57BL/6 mice two weeks post injection, as measured by flow cytometry (n=6). Error bars are mean ± SEM; **** p ≤ 0.0001 by paired Student's t test.

Figure 2. Enhanced GFP Expression in Dermal Cells Following Systemic Delivery of AAVrh32.33 Compared to AAV1, AAV6 and AAV8.

(A) Percentage of GFP-positive cells among live cells in the dermis one month following i.v. injections of AAVrh32.33, AAV1, AAV6, or AAV8-GFP in C57BL/6 mice, as measured by flow cytometry. One-way ANOVA; * p ≤ 0.0332, **** p ≤ 0.0001. Error bars are mean ± SEM (n = 6 for PBS, AAV1, AAV6, and AAV8 groups; n = 10 for AAVrh32.33 group). (B) Quantification of the percentage of GFP-positive cells within each cell type in the dermis one month following i.v. injections of AAVrh32.33, AAV1, AAV6, or AAV8-GFP in C57BL/6 mice, as measured by flow cytometry. One-way ANOVA; * p ≤ 0.0332, ** p ≤ 0.0021, *** p ≤ 0.0002, **** p ≤ 0.0001. Error bars are mean ± SEM (n = 6 for AAV1, AAV6, and AAV8 groups; n = 10 for AAVrh32.33 group). (C) Pie charts depicting the distribution of GFP-positive cell types following AAVrh32.33-GFP injections at the one-month time point, as measured by flow cytometry (n = 10). ISM = isthmus; IFD = infundibulum. (D) Representative images of mouse ear one month post i.v. injections of AAVrh32.33-GFP or PBS in C57BL/6 mice, stained for GFP (green), DAPI (blue) and vimentin (red). scale bar, 200 μm. Arrows indicate areas of colocalization between vimentin and GFP. Asterisks mark the location of hair follicles. (E) In vivo intravital imaging of ear one month following i.v. injections of AAVrh32.33-GFP or PBS in C57BL/6 mice. Vessels are labeled with an anti-CD31 antibody conjugated with the Alexa 647 fluorophore (red). Blue and white signals reflect intrinsic tissue autofluorescence, primarily from dermal collagen and hair follicle structures. All vectors were administered intravenously at a dose of 1 × 10¹¹ vg per mouse.

To determine whether enhanced skin transduction by AAVrh32.33 was associated with altered systemic biodistribution, we quantified vector genome copy numbers across multiple non skin organs by droplet digital PCT (ddPCR) following intravenous delivery (Figure S2A). AAVrh32.33 exhibited a biodistribution profile that was largely comparable to the tested AAV capsids across examined organs. Higher vector genome copy numbers were observed in skeletal muscle with AAVrh32.22 compared to the other capsids. These data indicate that the enhanced skin transduction observed with AAVrh32.33 occurs in the context of broadly similar systemic biodistribution.

To identify the specific skin cell populations transduced by AAVrh32.33, we performed flow cytometric analysis on tail dermis one month post injection and characterized GFP expression across various cell types using two flow cytometry panels (Figure S2B and C).¹⁷ Following AAVrh32.33 injection, 40% of all dermal fibroblasts (CD200⁻ Sca-1⁺) and 40% of all bulge stem cells (CD200⁺ CD34⁺) were GFP positive, indicating efficient gene expression within both stromal and epithelial compartments. Lower levels of GFP were observed in ~15% of infundibulum cells (CD200⁺ CD34⁻ Sca-1⁺) and ~5% of isthmus cells (CD200⁺ CD34⁻ Sca-1⁻) within the hair follicle. GFP-labeled cells were also observed within immune cell populations in the dermis, including approximately 35% of monocytes and 30% of macrophages (Figure 2B). In stromal and epithelial compartments, transduction by AAV1, AAV6 and AAV8 was markedly lower, with fibroblast and bulge stem cell transduction rates not exceeding 15%. The proportion of GFP⁺ monocytes (Lin⁻ CD11b⁺ Ly6c⁺) and macrophages (Lin⁻ CD11b⁺ Ly6c⁻ MHCII⁺ CD64⁺) with AAVrh32.33 was comparable to AAV1 and AAV8, and only higher than AAV6. This pattern indicates that AAVrh32.33 preferentially targets cutaneous stromal cells rather than immune cell populations (Figure 2B).

When examining the distribution of skin cell types among the total GFP positive population, dermal fibroblasts represented the majority (35%), followed by bulge cells (20%). Approximately 30% of GFP-positive cells could not be assigned to a defined population based on available markers, suggesting the involvement of additional, uncharacterized dermal cell types. These findings highlight the broad tropism of AAVrh32.33 and its preferential targeting of dermal fibroblasts and skin resident progenitor cells (Figure 2C).

The tropism of AAVrh32.33 for dermal fibroblasts was further confirmed by immunofluorescence (IF) staining of ear sections one month post injection, which showed strong GFP expression co-localizing with vimentin, a marker of dermal fibroblasts (Figure 2D). Consistent with this, in vivo intravital imaging (IVIM) of the ear at one month post injection revealed GFP positive cells with spindle-shaped morphology characteristic of dermal fibroblasts (Figure 2E). Together, these complementary approaches validate dermal fibroblasts as a major target of AAVrh32.33-mediated gene delivery following intravenous injection.

Sustained GFP Expression in Skin Following AAVrh32.33 Delivery

AAVrh32.33 is known to be immunogenic in mice, with prior studies from our lab and others demonstrating that transgene expression is typically lost around 3–4 weeks post injection due to immune mediated clearance.^15,18 To determine whether transgene expression could persist, we intravenously injected mice with AAVrh32.33-GFP (1 × 10¹¹ vg per mouse) and analyzed expression in the skin over a time course extending to two years post injection. Flow cytometry demonstrated durable GFP⁺ stromal populations up to two years post injection. Expression peaked at one month and was significantly higher compared to all time points except one year post-injection. From three months onward, levels remained stable, with no significant differences through two years (Figure 3A and S3A). IVIM imaging further confirmed this persistence, revealing elongated GFP-labeled cells with a morphology and distribution consistent with dermal fibroblasts, underscoring the long-term stability of transgene-positive cells despite the capsid’s immunogenic profile (Figure 3B). Vector genomes were similarly maintained at comparable levels across the later time points (Figure 3C). To track how transgene expression changed in different skin cell types over time, we analyzed GFP⁺ populations by flow cytometry. GFP expression in dermal fibroblasts varied over time, accounting for much of the overall decline in signal (Figure 3D). In contrast, GFP expression in bulge stem cells remained relatively stable, with no significant differences across time points. Among immune populations, both monocytes and macrophages showed significant reductions in GFP expression at 3 months and 2 year (Figure 3D). To determine whether this trend was unique to AAVrh32.33 or more broadly applicable to other AAV capsid variants, we also evaluated mice injected with AAV1- and AAV6-GFP, euthanized at one and three months post injection. As expected, overall GFP expression in the skin of AAV1- and AAV6-treated animals was markedly lower than that observed with AAVrh32.33. Similarly, these groups also showed a decrease in GFP⁺ monocytes and macrophages by three months, suggesting that early GFP positivity in myeloid cells could reflect transient antigen uptake by phagocytic cells rather than stable transduction (Figure 3D and Figure S3B). These findings show that AAVrh32.33 can support long-term gene expression in the skin, with GFP⁺ cells remaining detectable two years after a single administration.

Figure 3. Two-year Persistence of GFP-labeled Cells in Mouse Skin.

(A) Percentage of GFP-positive cells among live dermal cells isolated from mouse tails following i.v. injections of AAVrh32.33 in C57BL/6 mice, as measured by flow cytometry at two weeks, 1 month, 3 months, 6 months, 1 year, and 2 years post-injection. Two-way ANOVA; * p ≤ 0.0332, ** p ≤ 0.0021. Error bars represent mean ± SEM (n = 5, 10, 6, 3, 3, 4 for each respective time point). (B) In vivo intravital imaging of the ear two years post i.v. injections of AAVrh32.33-GFP or PBS in C57BL/6 mice. (C) ddPCR analysis of ears collected from C57BL/6 mice following i.v. injections of AAVrh32.33-GFP or PBS, measured at 1 month, 1 year, and 2 years post-injection. One-way ANOVA; p ≤ 0.0332. Error bars represent mean ± SEM (n = 3, 3, 4 for 1 month, 1 year, and 2 years, respectively; PBS: n = 6). (D) Quantification of percentage of GFP-positive cells within each cell types in the dermis following i.v. injections of AAVrh32.33-GFP in C57BL/6 mice, as measured by flow cytometry. Two-way ANOVA; * p ≤ 0.0332, ** p ≤ 0.0021. Error bars represent mean ± SEM (n=10, 6, 3, 4 for 1 and 3 months, 1- and 2-years post-injection, respectively). All vectors were administered intravenously at a dose of 1 × 10¹¹ vg per mouse.

Rational Engineering of AAVrh32.33 Attenuates CD8⁺ T Cell Responses Without Compromising Skin Transduction

Previous work by Mays et al. identified an immunodominant CD8⁺ T cell epitope in the AAVrh32.33 capsid in C57BL/6 mice, located after β-sheet E, between variable regions II and III (Figure 4A).^16,19 To confirm these findings, we performed peptide-specific flow cytometry using a peptide MHC dextramer, which specifically detects CD8⁺ T cells recognizing the immunodominant AAVrh32.33 peptide. As expected, robust CD8⁺ T cell responses were observed against the reported epitope in intramuscularly injected mice (Figure 4B). Multiple sequence alignment of AAVrh32.33 with AAV1, AAV2, AAV4, AAV5, AAV6, AAV8 and AAV9, revealed that this immunodominant region differs by only three amino acids across these capsid variants, suggesting that minor sequence variation in this region may underlie the differential immunogenicity of AAVrh32.33 (Figure 4C). To test this, we engineered a modified AAVrh32.33 capsid in which the three divergent amino acids within the immunodominant region were substituted with those from less immunogenic AAV capsid variants (AAV1, AAV2 and AAV6) hence named IDPΔ. To evaluate CD8⁺ T cell responses against AAVrh32.33 and the engineered IDPΔ capsid, IFN-γ ELISpot assays were performed using splenocytes isolated one month post AAV injection. Peptide #68, which contains the minimal sequence required for recognition of the immunodominant CD8⁺ T cell epitope of AAVrh32.33 and is contained within pool 2 of the capsid peptide library. As expected, splenocytes from AAVrh32.33-injected mice produced robust IFNγ spot formation following stimulation with peptide #68 and pool 2, reflecting strong antigen-specific activation. In contrast, IDPΔ treated mice showed significantly reduced ELISpot responses compared to AAVrh32.33, indicating attenuated recognition of this epitope. PBS controls did not exhibit peptide-specific responses (Figure 4D). Furthermore, flow cytometric analysis using the peptide-MHC dextramer for the immunodominant AAVrh32.33 peptide revealed a complete absence of such cells in mice treated with the IDPΔ capsid (Figure 4E), confirming successful epitope silencing. These results demonstrate that targeted mutation of a single immunodominant epitope in AAVrh32.33 can effectively blunt peptide-specific CD8⁺ T cell responses, offering a strategy to mitigate adaptive immune responses in gene therapy applications.

Figure 4. Reduced CD8⁺ T Cell Response and Sustained GFP Expression in the Dermis Following IDPΔ Injection.

(A) 3D structural representation of vp3, with the immunodominant peptide labeled in purple, β-sheets in blue, α-helices in red, and loops/turns in yellow, visualized using PyMOL software. (B) Quantification of dextramer staining for AAVrh32.33 immunodominant peptide-specific CD8⁺ T cells in isolated splenocytes from mice one month post-injection with AAVrh32.33 or PBS via i.m. or i.v. administration, measured by flow cytometry. Two-way ANOVA; ** p ≤ 0.0021, *** p ≤ 0.0002. Error bars represent mean ± SEM (n=7, 12 and 10 for i.v., i.m. AAVrh32.33 and PBS groups respectively). (C) Multiple sequence alignment of the AAVrh32.33 immunodominant peptide with AAV1, AAV2, AAV4, AAV5, AAV6, AAV8 and AAV9 using ESPript, highlighting sequence conservation and variability (Colors indicate sequence conservation: red background with white letters = identical residues across all sequences; red letters = similar residues (shared physicochemical properties); black letters = variable residues). (D) ELISpot analysis of splenocytes from mice one month post i.m. injection with AAVrh32.33, IDPΔ, or PBS, assessing IFNγ secretion in response to AAVrh32.33 capsid peptide pools, the immunodominant peptide (peptide # 68), positive control (ConA), or unstimulated conditions (n=6, 4 and 2 for AAVrh32.33, IDPΔ and PBS respectively). (E) Quantification of dextramer staining for AAVrh32.33 immunodominant peptide-specific CD8⁺ T cells in isolated splenocytes from mice one month post i.m. injection with AAVrh32.33, IDPΔ, or PBS, as measured by flow cytometry. Two-way ANOVA; ** p ≤ 0.0021, *** p ≤ 0.0002. Error bars represent mean ± SEM (n = 6, 12, and 10 for IDPΔ, AAVrh32.33, and PBS groups, respectively). (F) Percentage of GFP-positive cells among live dermal cells isolated from mouse tails one month post i.v. injections of PBS or AAVrh32.33 or IDPΔ in C57BL/6 mice, as measured by flow cytometry. Two-way ANOVA; *** p ≤ 0.0002, **** p ≤ 0.0001. Error bars represent mean ± SEM (n = 6 for IDPΔ and n = 10 for AAVrh32.33 and PBS groups). (G) In vivo intravital imaging of ear following i.v. injection of IDPΔ in C57BL/6 mice. Vessels are labeled with an anti-CD31 antibody conjugated with the Alexa 647 fluorophore (red). (H) Quantification of percentage of GFP-positive cells within each cell types in the dermis tails one month post i.v. injections of AAVrh32.33 or IDPΔ in C57BL/6 mice, as measured by flow cytometry. Two-way ANOVA; * p ≤ 0.0332, ** p ≤ 0.0021. Error bars represent mean ± SEM (n=6 and 10 for IDPΔ and AAVrh32.33 groups respectively). ISM = isthmus; IFD = infundibulum. (I) Pie charts depicting the distribution of GFP-positive cell types following i.v. injection of IDPΔ-GFP at one month time point, as measured by Flow cytometry (n=6). Intravenous injections were performed at a dose of 1 × 10¹¹ vg per mouse, and intramuscular injections were performed at a dose of 5 × 10¹¹ vg per mouse.

To evaluate whether modification of the immunodominant epitope in AAVrh32.33 affected transduction efficiency, we compared GFP expression in the skin of mice injected with either wild-type (WT) AAVrh32.33-GFP or the modified capsid (IDPΔ). Flow cytometry and IVIM imaging both showed that overall GFP expression in the skin was not significantly reduced with the IDPΔ capsid (Figure 4F and G), indicating that epitope silencing did not compromise transduction efficiency at the tissue level. However, cell type specific analysis revealed reductions in GFP-labeled dermal fibroblasts, bulge cells, isthmus and infundibulum, following IDPΔ capsid delivery (Figure 4H). Notably, GFP⁺ monocytes and macrophages were also reduced in animals injected with the IDPΔ capsid, further supporting the conclusion that modification of the immunodominant epitope attenuated immune cell activation and antigen uptake. To complement this data, proportional analysis of the total GFP⁺ pool revealed a redistribution of transduced populations (Figure 4I). In IDPΔ-treated animals, dermal fibroblasts represented a greater share of all GFP⁺ cells (66% vs. 35.7% in WT), whereas bulge cells (4.8% vs. 20% in WT) and other dermal populations (17% vs. 33% in WT) accounted for a smaller fraction (Figure 2C and 4I). This indicates that, although the frequency of GFP positivity within individual cell types declined, fibroblasts became the predominant contributor to overall GFP expression following IDPΔ delivery.

To assess off target biodistribution following systemic administration, vector genome copy numbers were quantified by ddPCR in liver, lung, and skeletal muscle. Comparable vector genome levels were detected in liver and lung following delivery of either AAVrh32.33 or IDPΔ. In contrast, vector genome levels in skeletal muscle were lower in mice receiving IDPΔ compared with the parental capsid (Figure S4A). Together, these data show that epitope deletion did not increase vector genome accumulation in liver and lung and was associated with reduced muscle biodistribution.

To further assess vector durability, we extended the analysis to a 3-month timepoint. Compared to the one month measurement, overall GFP expression in the skin was reduced for the IDPΔ capsid (Figure S4B), with a statistically significant decline observed only in dermal fibroblasts. In contrast, the wild-type capsid showed a broader reduction in transgene expression at this timepoint, affecting not only dermal fibroblasts but also monocytes, macrophages (Figure 3A and D). Importantly, monocytes and macrophages, which exhibited low GFP expression early on in the IDPΔ group, did not decline further. These findings suggest that the IDPΔ capsid is taken up by immune cells less efficiently than the wild type, thereby reducing immune-mediated clearance and supporting more stable transgene expression in non-immune skin cell populations over time.

Together, these findings demonstrate that rational epitope editing of AAVrh32.33 reduces immunogenicity without substantially impairing gene delivery to the skin, offering a promising strategy for engineering immune-evading vectors for cutaneous gene therapy.

IDPΔ Efficiently Transduces Human Skin Explants

To evaluate the transduction potential and cellular tropism of engineered and natural AAV capsids in human skin, we intradermally injected discarded human abdominal skin biopsies with AAVrh32.33, IDPΔ, AAV1, AAV6, AAV8, or PBS and cultured explants ex vivo for three days (Figure 5A). Flow cytometry revealed robust GFP expression in samples treated with IDPΔ, whereas GFP⁺ cells were detected at markedly lower frequencies with AAV1, AAV6, AAV8, and AAVrh32.33 and were not significantly different compared to PBS controls (Figure 5B, Figure S5A).

Figure 5. Enhanced Skin Tropism of IDPΔ Targeting Dermal Fibroblasts and Precursors.

(A) Experimental schematic. (B) Quantification of GFP-positive cells among live cells in human skin biopsies after intradermal injection with AAV1, AAV6, AAV8, AAVrh32.33, IDPΔ-GFP, or PBS. One-way ANOVA; *** p ≤ 0.0002, **** p ≤ 0.0001. Error bars represent mean ± SEM (3 subjects; Subject 1: 1 replicate for AAVrh32.33 and PBS, Subject 2: 3 replicates for AAVrh32.33 and PBS, Subject 3: 3 replicates for AAV1, AAV6, AAV8, AAVrh32.33, IDPΔ, and 1 replicate for PBS). (C) Quantification of GFP-positive percentages within individual cell types in human skin biopsies after intradermal injection. One-way ANOVA; ** p ≤ 0.0021, *** p ≤ 0.0002. Error bars represent mean ± SEM (same replicate breakdown as in B). (D) Pie charts showing the distribution of GFP⁺ cell types (proportion of total GFP⁺ cells) following intradermal IDPΔ-GFP injection (1 subject, 3 replicates). Biopsies were injected intradermally with 9 × 10⁹ vector genomes per biopsy.

To assess cell-type specificity, GFP⁺ frequencies were first quantified within defined dermal populations by flow cytometry. IDPΔ transduced ~21% of dermal fibroblasts (CD34⁺CD90⁺) and ~17% of dermal precursors (CD34⁺CD90⁻), representing a substantial enrichment compared to other capsids (Figure 5C). To further define the biodistribution of transduced cells, the composition of the total GFP⁺ pool was analyzed. Among all GFP⁺ cells, 28% corresponded to dermal fibroblasts, 30% to dermal precursors, and ~41% remained unassigned by the marker panel used (Figure 5D). GFP expression was not detected in EpCAM⁺ keratinocytes, c-Kit⁺ melanocytes, or CD200⁺ hair follicle cells (Figure S5B).

These results parallel our murine data, in which dermal fibroblasts represented the predominant transduced population. However, unlike murine studies where hair follicle populations were transduced, GFP⁺ hair follicle cells were not observed in human explants. This discrepancy is likely attributable to the use of discarded abdominal skin from panniculectomy specimens, which are largely devoid of hair follicles due to low follicular density at that anatomical site.

Localized Cutaneous Expression of Therapeutic Chemokines Following Systemic AAV Delivery in Wild-Type and Inflamed Skin

The selection of appropriate immunomodulatory transgenes is a critical step for skin-directed gene therapy. To explore the potential of AAVrh32.33 and IDPΔ as a therapeutic platform, we evaluated chemokines with well-established roles in regulating cutaneous immune responses. CCL17, CCL20, and CCL22 were selected based on their ability to mediate immune cell trafficking and their relevance to inflammatory skin conditions.^20,21 Among these, CCL22 has been most extensively studied, with murine models of vitiligo demonstrating that its local expression reduces depigmentation by recruiting CCR4⁺ regulatory T cells (Tregs) to sites of inflammation.²⁰ CCL17, another CCR4 ligand, was included as a complementary factor that may act in concert with CCL22 to strengthen CCR4-mediated signaling.^20,22,23 To further diversify the chemokine profile, CCL20 was incorporated, as it signals through CCR6, a receptor required for the skin homing of a distinct subset of immune cells implicated in autoimmune pathogenesis.²¹ Together, these candidates represent a rational chemokine panel for altering the inflammatory milieu of the skin. To enable simultaneous expression of all three chemokines from a single vector, CCL17, CCL20, and CCL22 were arranged in a polycistronic cassette separated by self-cleaving 2A domains. C57BL/6J mice were intravenously injected with IDPΔ- or AAVrh32.33-based vectors encoding this cassette (chemokine) (5 × 10¹¹ vg per mouse), alongside AAVrh32.33-GFP control. One month post injection, ear skin was analyzed by anti-CCL22 HRP staining (Figure 6A). Quantification revealed significantly higher CCL22 staining in the ears of animals treated with chemokine vectors compared to the AAVrh32.33-GFP group, which showed only minimal signal consistent with basal expression (Figure 6 B and C). These findings demonstrate that chemokine vector delivery markedly enhances CCL22 expression in the skin. Serum chemokine levels were analyzed using cytometric bead array (CBA). A trend toward higher levels was observed in the chemokine-treated groups (IDPΔ-Chemokine and AAVrh32.33-Chemokine) compared to control (AAVrh32.33-GFP), although differences did not reach statistical significance (Figure S6A). These results show that systemic delivery of chemokine vectors resulted in effective expression within the skin, which is the primary therapeutic target, while only modest increases were observed in circulation.

Figure 6. Intravenous Delivery of AAV Expressing Chemokines Leads to Localized Expression in Mouse Ear Skin.

(A) Experimental schematic. (B) DAB staining was quantified in ear sections from mice injected systemically with AAVrh32.33-Chemokine, IDPΔ-Chemokine, or AAVrh32.33-GFP control vectors. For each animal, three tissue sections were analyzed and averaged to generate a single biological replicate. One-way ANOVA; * p ≤ 0.0332. Error bars are mean ± SEM (n=3 for AAVrh32.33-Chemokine and IDPΔ-Chemokine, and n=2 for the AAVrh32.33-GFP control). Quantification was performed using ImageJ color deconvolution and thresholding, with the percentage of DAB-positive area calculated relative to total tissue area. (C) Representative ear sections from C57BL/6 mice collected one month after i.v. injection with IDPΔ-Chemokine, AAVrh32.33-Chemokine, or AAVrh32.33-GFP. Sections were stained with anti-CCL22, shown in brown. scale bar, 200 μm. Intravenous injections were performed at a dose of 5 × 10¹¹ vg per mouse.

To determine whether AAV mediated therapeutic gene expression is maintained in inflamed skin, we employed the MC903 induced murine model of atopic dermatitis (AD), a Th2 driven inflammatory skin disease characterized by epidermal barrier dysfunction and immune dysregulation. Mice received topical calcipotriol (MC903) applied to both ears to induce dermatitis and were systemically injected with the IDPΔ vector encoding a CCL17-CCL20-CCL22 chemokine cassette (1 × 10¹² vg per mouse) or PBS control three days after the first calcipotriol application. Ear tissue was harvested one month after disease induction to assess local transgene expression (Figure 7A). Immunohistochemical analysis revealed increased CCL22 staining in the skin of diseased mice receiving the chemokine expressing vector compared to PBS group, confirming successful AAV mediated gene delivery within an inflammatory microenvironment. Notably, chemokine expression levels in inflamed skin trended higher than those observed in non inflamed skin receiving the same vector, indicating that cutaneous inflammation did not impair AAV driven transgene expression (Figure 7B and C).

Figure 7. AAV-Mediated Chemokine Expression Is Maintained in Inflamed Skin in the Atopic Dermatitis Mouse Model.

(A) Experimental schematic of the MC903-induced atopic dermatitis (AD) model. (B) Representative ear skin sections collected one month after the first application of calcipotriol (MC903). CCL22 expression was visualized by HRP-mediated immunostaining (brown). Scale bar: 200 μm. (C) Quantification of CCL22 DAB staining in ear sections from mice injected intravenously with IDPΔ-chemokine (1 × 10¹² vg per mouse) or PBS, with or without AD induction. For each mouse, ten sections were analyzed and averaged to generate one biological replicate. DAB-positive area was quantified using ImageJ using color deconvolution and thresholding and expressed as a percentage of total tissue area. Data are shown as mean ± SEM (n = 4). Statistical analysis was performed using one-way ANOVA (*p ≤ 0.0332).

Discussion

Skin gene delivery has faced challenges due to the skin’s barrier properties and immune surveillance.²⁴ While several AAV capsid variants have been evaluated for skin targeting, most display limited penetration or restricted tropism. For example, AAV2 and AAV8 have shown some ability to transduce fibroblasts and keratinocytes in vitro but underperform in vivo. In our study, AAV2 did not show detectable transduction in mouse skin, and AAV8 exhibited only modest activity, consistent with previous reports of inefficient penetration and uptake within complex skin architecture.^11,12,25 In contrast, AAVrh32.33 demonstrated markedly superior tropism following systemic delivery. It robustly and consistently transduced dermal fibroblasts and hair follicle bulge cells, a dual targeting profile not observed with AAV8.^10,14,26 CD164 has recently been identified as a critical entry receptor for AAVrh32.33 in both human and mouse systems.²⁷ CD164 is expressed in multiple tissues, including basal epidermal cells and endothelium, and has been associated with progenitor cell populations.^28,29 Notably, hair follicle bulge cells analyzed in this study were defined in part by CD34 expression, raising the possibility that CD164 may contribute to their preferential transduction, although its distribution across specific skin-resident cell populations remains to be defined. Importantly, this targeting was achieved following intravenous administration, underscoring the capsid’s ability to exit the vasculature and transduce deeper dermal structures. This property is essential for therapeutic applications in disorders that affect the dermis, such as recessive dystrophic epidermolysis bullosa (RDEB) or localized scleroderma, as well as for transient modulation of regenerative epithelial progenitors in chronic wounds, vitiligo, or chemotherapy-induced skin damage. To assess translational relevance, we compared several AAV capsids in human skin explants. Our modified capsid, IDPΔ achieved the highest levels of transduction, outperforming AAVrh32.33, AAV1, AAV6, and AAV8. IDPΔ preferentially targeted stromal compartments, with dermal fibroblasts and precursors as the main transduced populations, consistent with murine data. The lack of GFP⁺ follicular cells likely reflect the limited hair follicles in panniculectomy tissue rather than a true absence of tropism, underscoring the need for follicle-rich samples in future studies. Notably, IDPΔ demonstrated enhanced transduction compared to AAVrh32.33 in human skin explants, a pattern not observed in mouse studies. While delivery route differed between systems, intravenous in mice vs intradermal in human explants, this alone may not fully explain the improved performance of IDPΔ in human tissue. The recently identified CD164 receptor for rh32.33 was characterized in human and mouse, however the IDPΔ may increase the efficiency in human thorough an unknown cause.²⁷ Additionally, neutralizing antibodies against AAVrh32.33 are rarely detected in human sera.³⁰ The low seroprevalence of AAVrh32.33 suggests that variants such as IDPΔ may offer a clinical advantage by expanding the pool of eligible patients. Moreover, when delivered intradermally, pre-existing neutralizing antibodies may have limited impact on vector performance. Together, these findings highlight IDPΔ as a strong candidate for cutaneous gene delivery, with conserved stromal tropism underscoring its therapeutic potential.

The capacity to transduce bulge stem cells is particularly noteworthy. These cells represent a pool of long-lived progenitors capable of regenerating the epidermis and hair follicles over time. Their strategic location within the hair follicle makes them difficult to access, especially through systemic routes.³¹ Our finding that AAVrh32.33 and IDPΔ can reach and transduce these cells after intravenous injection suggests a unique ability of this vector to navigate both vascular and tissue barriers, potentially facilitated by CD164 expression on endothelial cells, which may support efficient vascular interaction and tissue access.^27–29 Therapeutically, this opens the door to durable gene modification of self-renewing skin progenitors, with potential applications not only in inherited skin diseases but also in disorders of hair growth, such as alopecia areata or androgenetic alopecia.

Another central advance in this study is the demonstration of sustained transgene expression in the skin, despite the known immunogenicity of AAVrh32.33. In previous work, this capsid was associated with rapid CD8⁺ T cell responses and clearance of transgene expression in skeletal muscle, limiting its utility for gene therapy.^15,18 Here, we applied a rational capsid engineering approach to mutate a dominant CD8⁺ T cell epitope, as previously defined in murine models.¹⁶ Mice receiving the modified vector displayed significantly attenuated immune responses, including a complete absence of peptide-specific CD8⁺ T cells and a marked reduction in GFP⁺ monocytes and macrophages compared to those injected with the wild-type vector, suggesting reduced uptake or phagocytosis of transduced cells. Importantly, this immune silencing did not compromise the vector’s ability to transduce skin cells. Although a reduction in transduction efficiency was observed in dermal fibroblasts with the IDPΔ variant, these cells remained the predominant transduced population and continued to contribute the majority of the sustained GFP signal. Consistent with this, overall skin fluorescence remained comparable to that observed with the wild-type vector. Further, engineering of the IDPΔ variant largely preserved the biodistribution profile of the parental capsid, with comparable vector genome levels in liver and lung and reduced skeletal muscle biodistribution. Although immunodominant epitopes are expected to differ across species due to MHC restriction, raising concerns about the generalizability of this strategy, our results demonstrate that a mutation originally designed to eliminate a murine CD8⁺ T cell epitope also enhanced gene transfer in human skin grafts. The successful removal of an immunodominant epitope without compromising skin tropism indicates that such approaches can extend beyond mice and may be applicable in humans, provided the modified residues do not disrupt receptor interactions or tissue targeting. Together, these findings highlight the translational potential of epitope-guided capsid engineering as a means of generating immune-evading AAV vectors tailored to diverse host immune contexts. Our time course analysis further revealed that AAVrh32.33 supports durable transgene expression in the skin, with GFP⁺ cells detectable for up to two years following a single administration despite containing the immunodominant epitope. Notably, the persistence observed here may in part reflect differences in route and dose compared to prior studies that reported stronger T cell responses following intramuscular administration.^15,18 In this study, intravenous delivery and relatively low systemic dosing differ from prior reports and may contribute to the observed differences in transgene persistence. Although expression peaked at one month, declined modestly by three months, and then stabilized at sustained levels, these experiments were conducted using independent cohorts of animals which is a limitation to this study and may reflect experimental variability rather than true changes in transduction efficiency over time, including the apparent early reduction and modest increase in GFP⁺ signal at later time points. The initial decline was most pronounced in dermal fibroblasts, which likely reflects their greater contribution to the overall GFP signal as the dominant transduced population. Dermal fibroblasts are relatively long-lived, with subsets persisting for months to years, yet they also undergo turnover during normal skin homeostasis and remodeling.³² In contrast, GFP signals in myeloid cells decreased rapidly, consistent with transient antigen uptake or phagocytosis of transduced material rather than stable genomic transduction. Together, these results highlight the interplay between vector biology, dosing parameters, animal to animal variation and immune responses in shaping long-term expression.

Importantly, the durability observed with AAVrh32.33 is highly relevant for chronic dermatologic conditions such as RDEB, where long-term gene correction in skin fibroblasts can restore collagen VII and improve skin integrity since adult dermal fibroblasts are largely quiescent and long-lived in vivo.³³ Current gene therapies for EB often require repeated applications due to short-lived expression or loss of corrected cells.^34,35 For example, Vyjuvek, a topical HSV-1-based product, requires ongoing application to maintain efficacy, while Zevaskyn, an ex vivo gammaretroviral-based therapy, necessitates labor-intensive and invasive autologous cell grafting.^34,35 In contrast, systemic delivery of AAVrh32.33 enables broad skin targeting across the body, including inflamed tissue, without the need for repeated localized administration, providing a more scalable approach, particularly for generalized or deep dermal diseases where topical or graft-based methods are less feasible. AAVrh32.33 uniquely combines high initial efficacy with long-term persistence, critical features for durable gene therapy. Moreover, its ability to stably transduce long-lived dermal fibroblasts makes it an attractive platform for genome editing strategies such as CRISPR-based correction, which could achieve permanent restoration of defective genes in situ. Finally, within the broader landscape of cutaneous gene delivery platforms, non-viral approaches such as lipid nanoparticles (LNPs), electroporation, and topical formulations offer advantages in redosability and reduced immunogenicity. However, they are often limited by shallow tissue penetration and transient expression. Recent studies using LNPs for CRISPR-Cas9 delivery in skin achieved only modest editing rates (~5–12%) and limited depth,³⁶ while electroporation requires specialized equipment and yields short-lived, localized expression.³⁷ By contrast, AAVrh32.33 achieves stable, widespread expression following a single dose, with particular strength in targeting dermal and follicular structures that remain largely inaccessible to non-viral platforms.

Building on the demonstrated tropism, durability, and translational potential of IDPΔ, this study further highlights its capacity to deliver therapeutic chemokines relevant to vitiligo. CCL17, CCL20, and CCL22 were selected for their roles in shaping cutaneous immune responses, particularly through CCR4 and CCR6 dependent pathways implicated in Treg skin homing and immune regulation.^21,22 Importantly, chemokine expression was maintained in inflamed skin, demonstrating that AAV-mediated transgene expression is preserved in a pathological inflammatory microenvironment. Although functional immune outcomes were not directly assessed, the ability to establish cutaneous chemokine expression provides a foundation for future therapeutic strategies aimed at restoring tolerance and preventing melanocyte destruction in vitiligo.

While this study provides strong evidence for the utility of AAVrh32.33 and IDPΔ in cutaneous gene delivery, several considerations remain for translation. First, the immunodominant epitope targeted in this study was defined in a murine context, and species-specific differences in MHC restriction may influence the translatability of this approach in humans. Although enhanced transduction was observed in human skin explants, further studies in models with human-relevant immune systems will be important to fully evaluate immune responses. Additionally, while multiple skin sites were analyzed using complementary approaches, a systematic comparison across distinct anatomical regions was not performed. Differences in route of administration between mouse (intravenous) and human (intradermal) studies may also influence biodistribution and transduction patterns and will require further investigation. AAVrh32.33 does have a broad tropism with vector genomes found in liver, kidney and lung similar to other commonly used AAV capsids. Although skeletal muscle shows higher vector genome levels, this pattern is consistent with systemic AAV biodistribution observed in clinical settings, where muscle transduction is generally well tolerated.^38,39 Importantly, potential off-target transgene expression can be further minimized through the use of tissue-specific or cell type-restricted promoters, providing an additional layer of control over transgene activity.¹⁴ Further, although major organs were evaluated, broader assessment of vector distribution in additional tissues, including spleen, intestine, and vasculature, will be important to fully define off-target exposure and support clinical translation. Future studies will evaluate disease-relevant therapeutic payloads and optimized dosing strategies to determine whether sustained expression achieved with these capsids can translate into meaningful clinical benefit in inflammatory skin disorders.

In summary, our findings position AAVrh32.33 and IDPΔ as promising vectors for dermatologic gene therapy. These capsids efficiently and durably transduce key cutaneous cell types, including fibroblasts and follicular stem cells. Immune evasion through capsid engineering enables long-term expression without compromising efficacy, and successful transduction of human skin supports translational relevance. Importantly, the ability to drive localized expression of therapeutic chemokines demonstrates their utility beyond reporter genes, establishing a foundation for targeted immune modulation in diseases such as vitiligo. Together, these features make IDPΔ compelling platforms for cutaneous gene therapy applications, ranging from rare inherited disorders to acquired autoimmune conditions.

Material and Methods

AAV production

AAVs were manufactured by University of Massachusetts Chan Medical School viral vector core as previously described (Worcester, MA).^40,41 All constructs were designed as self-complementary vectors under the control of CB promoter to drive strong and durable expression. For therapeutic vectors, CCL17, CCL20, and CCL22 transgenes were linked by E2A self-cleaving peptides to enable coordinated expression of multiple chemokines from a single transcript.

Flow cytometry

Single-cell suspensions from spleen were prepared and stained as previously described.¹⁸ Cells from mouse tail skin were isolated and stained according to established protocols.⁴² Mice were sacrificed at predefined time points, and tissues were collected from independent cohorts for each time point.

For detection of antigen-specific CD8⁺ T cells, splenocytes collected at one month post-injection from mice injected either intravenously (tail vein, 1 × 10¹¹ vg per mouse) or intramuscularly (right hind limb, 5 × 10¹⁰ vg per mouse) were stained with MHC class I dextramers specific for the AAVrh32.33 immunodominant peptide. Dextramer staining was performed according to the manufacturer’s instructions (Immudex), followed by surface antibody staining.

Dead cells were excluded from all analyses using fixable LIVE/DEAD cell stain (BioLegend, San Diego, CA, USA). Data were acquired using the Cytek Aurora spectral flow cytometer (Cytek Biosciences, Fremont, CA, USA) and analyzed using FlowJo software v10.7.1 (FlowJo, LLC, Ashland, OR, USA). The antibodies used for the myeloid and non-myeloid panels are listed in Tables S1 and S2.

Animals and in vivo models

All procedures involving animals were carried out in accordance with the guidelines of the University of Massachusetts Chan Medical School Animal Care and Use Committee (IACUC). Male and female 6–12-week-old inbred C57BL/6J mice (Jackson laboratories, Bar Harbor, ME) and NU/J mice (Jackson laboratories) were used. NU/J mice were used for in vivo bioluminescence imaging studies, while all other experiments were performed in C57BL/6J mice. In vivo studies were conducted using either intravenous or intramuscular delivery, depending on the study design, as described below.

Intramuscular delivery:

Intramuscular injections were administered in the right hind limb at a volume of 50 μL per mouse (5 × 10¹⁰ vg per mouse) and were used for IVIS imaging and immune response studies, including flow cytometry-based dextramer staining and ELISpot assays.

Intravenous delivery:

Intravenous injections were performed via the tail vein at a volume of 200 μL per mouse. Mice received either 1 × 10¹¹ vg per mouse for transduction, biodistribution, and durability studies, or 5 × 10¹¹ vg per mouse for therapeutic transgene studies. For early assessment of skin compartment transduction, mice were analyzed at 2 weeks post-injection, with dermal and epidermal fractions isolated from mouse tail skin and analyzed by flow cytometry. For detailed characterization of transduced dermal cell populations, mice were analyzed at one month post-injection by flow cytometry.

To evaluate persistence of transgene expression, mice were euthanized at multiple time points (2 weeks, 1 month, 3 months, 6 months, 1 year, and 2 years post-injection). Dermal cells from tail skin were analyzed by flow cytometry, and ear tissues were collected for vector genome quantification by ddPCR at selected time points.

For therapeutic transgene studies, mice were analyzed at one month post-injection; dermal cells from tail skin were used for flow cytometric analysis, and ear tissues were collected for histological staining.

Atopic dermatitis like inflammation was induced in male and female C57BL/6J mice using an established MC903 (calcipotriol) sensitization protocol with minor modifications.⁴³ Mice were randomized into experimental groups and received daily topical MC903 application to both ears for 14 consecutive days, with ethanol-treated mice serving as controls. On day 3 after disease induction, animals received a single systemic injection of AAVrh32.33-IDPΔ (1 × 10¹² vg/mouse) encoding a chemokine cassette or PBS. One month after disease induction, ear was harvested for histological analysis and immunostaining.

IFNγ ELISpot

At one month post-injection, spleens were collected from mice injected intramuscularly (right hind limb, 5 × 10¹⁰ vg per mouse) with AAVrh32.33, IDPΔ, or PBS. Splenocytes were isolated via mechanical dissociation followed by red blood cell lysis. Cells were stimulated for 24 hours with overlapping AAVrh32.33 capsid peptide pools, the immunodominant peptide (peptide #68), a negative control (media only), or a positive control Concanavalin A (from Canavalia ensiformis (Jack Bean) Millipore Sigma) and then IFNγ cytokine secretion was assessed. Spot numbers were analyzed and counted using the Mabtech IRIS reader. For the cellular immune response to be considered positive, the number of spot-forming units per 10⁶ cells must be both greater than 50 and 3-fold higher than the unstimulated control. Cells were plated in technical triplicates, and results are reported as mean spot-forming units (SFU) per million cells ± standard deviation (SD).

Tissue Collection, DNA Extraction, and ddPCR Analysis

At the study endpoint, frozen tissues, including muscle, liver, kidney, heart, ear, dorsal skin, and lung, were collected to assess vector biodistribution. Genomic DNA was extracted using the Gentra Puregene Tissue Kit (Qiagen, Hilden, Germany), following the manufacturer’s protocol. Digital droplet PCR (ddPCR) was performed to quantify vector genomes, using 50 ng of genomic DNA per reaction. Assays were carried out according to the manufacturer’s recommendations, with vector-specific primers and probes. Results were normalized to a reference gene to calculate vector genomes per diploid genome (vg/dg).

Intravital Imaging of the Ear

In vivo intravital microscopy was performed to assess transgene expression following intravenous injection of AAVrh32.33-GFP, IDPΔ-GFP or PBS in C57BL/6J mice. Mice were anesthetized with isoflurane, and vasculature was labeled by i.v. injection of an IVI Tag^™ for Endothelial Cell Labeling Kit (IVI^™−991–0001, IVIM Technology, inc., Korea Republic) of anti-CD31 antibody conjugated to a Cy5 spectrum-compatible fluorophore. The ear pinna was immobilized on a stage, and intravital imaging was performed using IVM-CMS3(IVIM Technology, Inc., Korea Republic), integrated with appropriate filter sets for GFP and Cy5. Images were acquired and analyzed to evaluate transgene expression and vascular architecture.

IVIS imaging

Inbred NU/J mice (Jackson Laboratory) were injected with AAV vectors expressing luciferase intramuscularly (5 × 10¹⁰ vg/mouse) into the right hindlimb muscle. Two weeks post-injection, mice were administered 100 mg/kg D-luciferin (Gold Biotechnology) via intraperitoneal injection and anesthetized with isoflurane inhalation. Bioluminescence imaging was performed 10 minutes later, at the peak of luciferase emission, using the IVIS Spectrum CT system (Revvity, Waltham, MA, USA). Imaging conditions were optimized with an emission filter set at 660 nm.

Histology and Immunofluorescence

Tissues were fixed in 10% neutral-buffered formalin for 24 hours at room temperature (Fisher Scientific), followed by paraffin embedding performed by the Morphology Core at the University of Massachusetts Chan Medical School (Worcester, MA). Immunohistochemistry (IHC) or immunofluorescence (IF) staining was performed on tissue sections. Primary antibodies included anti-CCL22 (LSBio), anti-GFP (Abcam, Cambridge, UK), and anti-vimentin (Abcam). Images were acquired using a Leica DM5500 B microscope (Leica Microsystems, Wetzlar, Germany).

Preparation and AAVrh32.33 Transduction of Human Skin Explants

De-identified, surgically discarded fresh abdominal skin samples were collected from the UMass Chan Biospecimen, Tissue, and Tumor Bank. These samples were obtained with approval from the Institutional Review Board (IRB) at the University of Massachusetts Chan Medical School, and all patients provided written informed consent before surgery. Subcutaneous fat was completely removed, and 8-mm punch biopsies were taken from the abdominal tissue. Prior to culture, skin biopsies were incubated in antibiotic-antimycotic solution (Gibco, Cat# 15240–062; 100× stock) diluted to 4× in PBS (1:25 dilution from stock; final concentrations: 400 U/mL penicillin, 400 μg/mL streptomycin, and 1 μg/mL amphotericin B) for 10–15 minutes at room temperature. Biopsies were then washed three times in sterile PBS by sequential transfer through fresh PBS-containing wells to remove residual antibiotic-antimycotic. The skin biopsies were then infected by intradermally injecting 50 μL of AAVrh32.33-GFP at a concentration of 1.8 × 10¹¹ vg/mL (equivalent to ~9 × 10⁹ vector genomes per biopsy), or were mock-injected with 50 μL of PBS. All vectors used in this experiment were derived from the same viral preparation and quantified by ddPCR. Injected skin punches were cultured at the air-liquid interface in Iscove’s Modified Dulbecco’s Medium (Sigma-Aldrich; #I3390), supplemented with 10% fetal bovine serum, 100 U/mL penicillin-streptomycin, and 50μM 2-mercaptoethanol. Biopsies were placed epidermis-side up on transwell inserts (compatible with 24-well plates), and each insert was positioned in a well containing 700 μL of culture medium, such that the dermal side remained in contact with the medium while the epidermal surface was exposed to air. Each biopsy was cultured in a single well of a 24-well plate (Thermo Scientific^™; #142475) for 72 hours at 37°C in a humidified incubator with 5% CO₂. After culture, skin biopsies were harvested and minced into roughly 1 mm² fragments. The tissue was transferred into gentleMACS C Tubes (Miltenyi Biotec; #130-093-237) containing 1 mL of digestion buffer, consisting of RPMI with 0.5 mg/mL DNase I (Sigma; #EN0521) and 1.6 mg/mL collagenase IV (Worthington Biochemical; #LS004186). Tubes were incubated at 37°C on an orbital shaker at 200 rpm for 3–4 hours. One milliliter of cold RPMI was added to each tube, and samples were mechanically dissociated using the h_skin_01 program on the gentleMACS Octo Dissociator (Miltenyi Biotec). The resulting cell suspensions were filtered through a 70 μm mesh, washed, and stained for flow cytometry antibody panel.

Quantification of DAB Staining

Immunohistochemical DAB staining was quantified using ImageJ software (National Institutes of Health, USA). Brightfield images were first processed with the Color Deconvolution (H-DAB) plugin to separate the DAB channel from hematoxylin. The resulting DAB image was converted to grayscale, and thresholds were applied using manually to identify positively stained regions. To restrict measurements to tissue only, a tissue mask was generated from the original image by color thresholding and applied to the DAB channel. For each section, the total tissue area and the DAB-positive area were measured, and the percentage of DAB-positive area was calculated as:

$$\%\text{DAB-positive}\text{area}=(\text{DAB-positive}\text{area}/\text{total}\text{tissue}\text{area})\times 100$$

Measurements were obtained from three sections per animal, and values were averaged to represent a single biological replicate. Group data are presented as mean ± SEM across animals.

Statistical analysis

Statistical analysis was performed using GraphPad Prism (GraphPad Software Inc, La Jolla, CA) version 10. Results were reported as mean ± SEM with statistically significant differences determined by tests as indicated in figure legends.

Data availability statement

All data associated with this study are present in the paper or the supplemental materials.