Development of an ex vivo skin model for quantitative recovery and preservation of donor-specific human microbiota

Abstract

Background

Through multi-omics approaches, our understanding of skin microbiota has substantially advanced. However, the development of reliable and physiologically relevant experimental models would greatly accelerate progress in microbiome-targeted science. Most investigations of skin microbiota rely on in vitro 2D models or reconstructed epidermis colonized with a limited number of strains, which do not reflect the complexity and the spatial microbial organization observed on in vivo skin. To overcome these limitations, we developed a model based on perfused human skin explants (Perfex model), that enables the transfer of complex bacterial communities from human volunteers onto ex vivo skin.

Results

After quantitative validation of recovered microbial DNA by qPCR from volunteers (V1-V2-V3) and corresponding inoculated explants (MV1–MV3), 16S rRNA gene sequencing was performed to assess whether microbiota transfer influenced community structure and diversity. No statistically significant differences in alpha-diversity were observed between original and transferred microbiota at the genus level, suggesting that the transfer procedure did not substantially alter overall community diversity. Clustering analysis revealed distinct grouping of samples according to donor origin, indicating that donor-specific microbial signatures were largely maintained on the explants after 48 hours. Control samples formed separate clusters, suggesting minimal background signal and supporting the absence of cross-contamination. These observations indicate that the model enables clear discrimination between the endogenous microbiota of the explants and the transferred microbiota, supporting the integrity of the experimental conditions. Beta-diversity analysis based on Bray-Curtis supported these findings. Across samples, microbial communities were characterized by the recurrent presence of common skin-associated genera such as Cutibacterium, Staphylococcus, and Corynebacterium. Histological analyses demonstrated preserved tissue architecture, while immunostaining revealed modulation of selected innate immune markers consistent with a localized skin response to microbial transfer.

Conclusions

The Perfex model enables transfer of donor-specific skin microbiota to ex vivo human skin while preserving tissue integrity over short-term culture. This model provides a human-relevant platform for studying host-microbiota interactions under controlled conditions and for preclinical evaluation of microbiome-targeted therapies and dermo-cosmetic treatments.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12866-026-05316-w.

Introduction

The human skin is not only the largest organ of the body but also one of its most intricate ecosystems [1, 2]. The coexistence of a wide range of skin microorganisms is believed to have resulted from millions of years of mutual adaptation and functional integration [3], with the skin being colonized by bacteria, fungi, viruses, and mites. This microbiota contributes to essential physiological functions such as maintaining skin barrier integrity [4–6], modulating innate and adaptive immune responses, and protecting against pathogens. Far from being a passive inhabitant, the skin microbiome engages in continuous crosstalk with host cells, influencing inflammation, repair processes, and even systemic health [7]. With the advent of next-generation sequencing and multi-omics approaches [8], our understanding of the skin microbiota has substantially advanced, revealing a highly diverse and individual-specific microbial landscape shaped by genetic, environmental, nutritional, anatomical, and lifestyle factors that together define unique microhabitats for skin microorganisms [9, 10]. Disruptions in this microbial balance - referred to as dysbiosis - have been associated with a growing list of dermatological disorders such as atopic dermatitis [11–14], acne vulgaris [15], psoriasis [16, 17], and seborrheic dermatitis [18, 19]. Consequently, the skin microbiota has become a promising target for novel therapeutic strategies, including the development of topical bacteriophages [20], probiotics, prebiotics, postbiotics [21–23] and microbiome transplants [24, 25].

However, progress in microbiome-targeted dermatological research is hampered by the lack of reliable and physiologically relevant experimental models [26, 27]. Most traditional in vitro skin models, such as keratinocyte monolayers or even more advanced reconstructed human epidermis (RHE), are typically sterile or colonized with limited reference strains [28–30]. While these models have provided valuable insights into skin-microbe interactions, they generally fail to maintain viable, complex microbial communities that reproduce the spatial organization characteristic of on in vivo skin [8, 31].

Animal models have played an important role in advancing our understanding of host–microbiota relationships, but they come with notable drawbacks. Differences in skin anatomy, immune function, and microbial communities between species limit how well these models reflect human biology [32–34]. Their use is also becoming increasingly constrained by ethical concerns and evolving regulations, especially in the cosmetic and pharmaceutical sectors, which encourage the adoption of alternative methods in line with the 3Rs principles (Replacement, Reduction, and Refinement). Human skin explants serve as a physiologically relevant ex vivo model, preserving the structural, cellular, and immunological complexity of native human skin as well as their communication with external environment including the microbiota. They have been used in past studies to analyze the cutaneous microbiota [35–37] but standard culture conditions, typically ~ 37 °C and 5% CO₂ in incubators, do not recapitulate the true physicochemical environment of the skin surface.

To address these limitations, we propose a novel approach that involves the direct transfer of complex, intact bacterial communities from healthy human volunteers onto an ex vivo skin model cultured under physiological atmospheric conditions that we have developed recently and that we named PERFused EXplant (Perfex) skin model [38]. This strategy offers several advantages: it preserves the donor-specific features of the microbiota, maintains microbial diversity, and allows for controlled study of colonization dynamics in a reproducible, human-relevant system. Such a colonized ex vivo model has broad applicability. It enables fundamental research into microbial ecology, competitive interactions, and host signaling; it supports the screening of new active ingredients or formulations for microbiome compatibility; and it provides a scalable platform for studying disease-associated dysbiosis or testing microbiota-modulating interventions. The facial microbiota, specifically the cheek skin, was selected for this study mainly because cheek skin offers a suitable microbiota donor site, providing relatively high microbial biomass and reproducible collection conditions. While the cheek skin was ideal for microbiota sampling, it was not used as a tissue donor for aesthetic reasons. Instead, abdominal skin explants were chosen, as they are among the most commonly available surgical skin samples for ex vivo studies. This design enabled an evaluation of whether donor-specific microbial communities could be transferred and maintained in a physiologically distinct skin environment.

The aim of this study was to establish and validate a reproducible protocol for the quantitative transfer of complex human skin microbiota onto an ex vivo human skin explant model maintained under physiologically relevant conditions. We hypothesized that donor-specific microbial communities could be transferred and maintained over short-term culture while preserving tissue integrity and inducing a specific microbiota-trigged skin response. By comparing microbiota composition before and after the transfer using quantitative and sequencing-based approaches, this work evaluates the suitability of this model as a human-relevant platform for microbiota research, bacteria-host interaction and preclinical testing.

Materials and methods

Volunteer recruitment and sampling

To validate the protocol for in vivo microbiota collection, skin microbiota was sampled from ten healthy volunteers using a standardized swabbing protocol. The cheek was chosen as the sampling site because of its higher microbial biomass and reproducibility of sampling. Briefly, one each volunteer an area of interest per cheek was delimited using a square layer recovering 16 cm² per cheek (in total 32 cm²). Microbiota from each cheek were collected using a swab (E-swab^® 480 C, Copan, Italy) and transferred into two sterile (DNAseRNAse free) tubes containing 500 µl of maintenance buffer. The two suspensions were subsequently pooled into a single sterile tube to obtain 1 mL of microbiota suspension, which was processed within 24 h. Microbiota samples from ten donors were initially screened to assess microbial recovery, viability, and DNA yield. Due to logistical constraints associated with obtaining fresh surgical skin explants and the technical complexity of implementing the Perfex model, subsequent experiments were conducted on a subset of donors for which sufficient microbiota material and compatible skin samples were available. This selection was therefore based on practical feasibility and sample availability rather than experimental outcome.

DNA extraction and quantification

DNAs were extracted from skin microbiota of the volunteers and optimized using an extraction method for efficiently recovering DNA from low mass samples (P-209B-SpinColumns, DNAGenotek (TM)). Bacterial DNAs (ADNmb) were quantified via qPCR using the Femto™ Bacterial DNA Quantification Kit (ZE2006, OZYME) and 16 S rRNA copy numbers were estimated. Similarly, DNAs were extracted from the swab collected on the Perfex vivo explants at 48 h using the same optimized protocol. Quantification of extracted DNAs was performed by Qubit 4.0 using DNA HS kit (ThermoFisherScientific).

Transfer onto human Perfex system

Full-thickness human skin was obtained from the abdomen of female donors aged between 25 and 65 years old who had undergone plastic surgery. Circular skin explants were successively sterilized using a cocktail of antibiotics and prepared under sterile conditions as indicated by Peno-Mazzarino et al. [38]. Circular skin explants (3 cm in diameter) were incubated overnight in BEM culture medium supplemented with fungicidal amphotericin B (2.5 µg/mL; Fisher Scientific, ref. 15290) and bactericidal gentamicin (50 µg/mL). After treatment, explants were transferred to antibiotic-free BEM medium for 48–72 h to allow clearance of residual antibiotics prior to installation in the Perfex system. This continuous perfusion step minimizes potential carryover effects of antibiotics on microbial survival while preserving tissue viability. The microbiota collected from three healthy volunteers (I-III) was collected as previously described and transferred to Perfex model. The microbiota suspension was applied directly onto the epidermal surface of each explant using a pipette and gently spread with a sterile inoculating loop over a volume of 150 µL, corresponding to an estimated bacterial load of 2.10⁵ CFU. The suspension was evenly distributed across the surface and incubated under controlled conditions for 48 h. Earlier time points (e.g., 24 h) were evaluated in preliminary experiments, but low bacterial recovery and DNA yield precluded robust microbiome analysis. After 48 h of incubation, microbial recovery from the explants was performed by gently swabbing the skin surface with sterile swabs (E-swab^® 480 C, Copan, Italy), thereby sampling surface-associated communities without homogenizing the tissue. The collected material was subsequently processed for 16 S rRNA gene sequencing. For each microbiota obtained from in vivo human skin, the transfer procedure was repeated in two independent experiments. At the end of incubation with the bacteria, skin explants were collected and cut in two halves: half was frozen directly at -80 °C for immunofluorescence, and half was fixed in formalin solution for evaluation of the cellular viability and immune-histochemistry.

Details of the experimental design and characteristics of the skin donors are provided in Fig. 1.

Fig. 1.

A Schematic representation of the protocol used to transfer the cheek microbiota from healthy volunteers to the Perfex skin explant model. The microbiome was first sampled from the cheek of healthy volunteers (1) and analyzed by 16 S rRNA sequencing (2). In parallel, circular skin explants were prepared from abdominal skin obtained from healthy donors (3) and mounted in the Perfex setup (4). The collected microbiota was then transferred onto the surface of the Perfex explants (150 µl per explant of the E-Swab solution), cultured for 48 h, and subsequently analyzed by 16 S rRNA sequencing (5). B Summary of the characteristics of the volunteers and skin donors used in this study. Volunteers V1, V2, and V3 represent the sources from which the microbiota was collected, whereas the skin donors represent the individuals from whom the abdominoplasty samples were obtained. Volunteers and skin donors were grouped according to the experimental transfer procedure performed. Two independent experiments were performed after sampling microbiota from same volunteers (V1 or V2 or V3) which were transferred onto skin explants. The microbiome samples from volunteers 1 and 3 were transferred onto explants prepared from the same skin donor. For each experimental condition, three explants were processed

BIO-EC Laboratory possesses an authorization from the Bioethics group of the general director services of the French research and innovation ministry (registered under the number DC-2022-5373) to use human skin from surgical waste. The study was performed in accordance with the Declaration of Helsinki after the patients had given informed consent to use their skin samples by BIO-EC Laboratory.

Evaluation of skin tissue viability and protein immunolocalization

The analysis of cellular and tissular viability was carried out after staining of formalin-fixed paraffin-embedded (FFPE) skin sections according to Masson’s trichrome protocol, Goldner variant. For immunohistochemistry, a pre-diluted horse serum (Vector laboratories, ref. PK7200) and a universal horse secondary antibody were used (Vector laboratories, ref. PK7200). For immunofluorescence, AF488 secondary antibody (A11078, ThermoFisher Scientific, USA) was diluted 1:1000 in PBS–0.3% BSA. Nuclei were counterstained using propidium iodide. The immunostainings were realized on 5 μm-thick FFPE or on 7 μm-thick frozen skin sections. The list of the antibodies is shown in Table 1.

Table 1.

List of the primary antibodies used for protein immunolocalization on skin sections

ANTIBODY	REFERENCE	SUPPLIER	CLONALITY	HOST	DILUTION	TYPE OF IMMUNOSTAINING
TLR2	MA5-32787	Invitrogen	Monoclonal	rabbit	1: 50	IMMUNOHISTOCHEMISTRY
hBD2	AHP849	BIORAD	Polyclonal	goat	1: 50	IMMUNOHISTOCHEMISTRY
TNF-alpha	AF-410-NA	R&D System	Polyclonal	goat	1: 50	IMMUNOFLUORESCENCE

The microscopical observations were performed using a BX63 Olympus microscope. Pictures were digitized with an Olympus DP74 camera and cellSens data storing software. The stainings were semi-quantified by image analysis using cellSens software by calculating the surface percentage positive to the staining inside a defined region of interest. To allow for semi-quantitative comparison across experiments, all datasets were subsequently normalized to 100 before statistical analysis. Nine images per condition were analyzed. First, the Shapiro-Wilk normality test allowed to determine which data followed a normal distribution (https://www.statskingdom.com/). In case of a Gaussian distribution, the data were analyzed statistically by unpaired t-test using Excel software. In case of a non-Gaussian distribution, the data were analyzed using the Mann-Whitney test (https://www.statskingdom.com).

Innate immune response was assessed by immunohistochemical detection of selected markers (hBD2, TLR2, TNF-α). This analysis provides a semi-quantitative evaluation of localized host responses but does not represent a comprehensive assessment of immune function. Data were considered statistically significant when *p < 0.05 and ** p < 0.01. All data were shown as mean + standard deviation (SD).

Bacterial viability assessment

For flow cytometry viability analysis, swab samples were collected and placed in Liquid Amies Medium and stored at 4 °C for up to 8 h were processed by vortexing for 10 s to homogenize the suspension. The swab was then removed, and 500 µL of the Amies medium was transferred into a sterile tube. Bacterial viability was assessed using the LIVE/DEAD BacLight Bacterial Viability Kit (Thermo Fisher Scientific, L7012). 1.5 µL of staining solution was added to each sample, followed by a 20 min incubation at room temperature in the dark. Samples were then analyzed on a BD Accuri™ C6 Plus flow cytometer (BD Biosciences). The bacterial population was first isolated using a forward/side scatter gate (R1), and viable or non-viable cells were subsequently classified according to fluorescence profiles: SYTO™ 9-positive events in gate P4 (viable bacteria) and PI-positive events in gate P3 (non-viable bacteria). Quantification of bacterial concentrations was achieved through volumetric event counting, allowing determination of viable and non-viable bacteria per mL of sample.

Cultivable bacteria enumeration was performed on cheek-swab eSwab™ samples processed using the same procedure described above. After vortexing, serial tenfold dilutions were prepared in sterile physiological saline. For aerobic culture, 100 µL of the appropriate dilutions were spread onto Tryptic Soy Agar (TSA, BD Difco™) plates and incubated aerobically at 37 °C for 48 h. For anaerobic culture, 100 µL of the same dilution series were plated onto Columbia Blood Agar (BD Difco™) and incubated under anaerobic conditions (BD GasPak™ EZ) at 37 °C for 72 h. After incubation, colonies were counted on plates containing 30–300 CFU, and bacterial loads were expressed as CFU/mL of the original sample, considering the dilution factor and plated volume.

Bacterial DNA detection and quantification were performed using a primer mix targeting the 16 S rRNA (Femto™ Bacterial DNA Quantification Kit, ZymoResearch E2006) with a real-time and quantitative PCR instrument (CFX96, BIO-RAD). Bacterial DNA standards were measured in duplicates comprise a 10-fold dilution series ranging from 20 ng to 20 fg (femtograms) as well as no template control and extracted DNAs from eSwab ™ samplings. The Bacterial DNA Standards provided are purified from E. coli strain JM109. The genomic copy number of the 16 S rRNA may vary between bacterial species. This kit provides an exact quantity of bacterial DNA presents and gives an approximate number of bacterial cells based on E. coli strain JM109 copy number.

16 S rRNA sequencing and analysis

The V1–V3 region of the 16 S rRNA gene was amplified (primers 27 F–534R), and libraries were sequenced using the Element AVITI system (150 bp paired-end). The V1–V3 hypervariable region of the 16 S rRNA gene was amplified to maximize taxonomic resolution for skin-associated bacteria, although species-level classification remains uncertain for some taxa. The FASTQ data have been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB105206 (https://www.ebi.ac.uk/ena/browser/view/ PRJEB105206). Quality control (base quality scores, GC content, and detection of overrepresented k-mers) was conducted. Taxonomic assignment was performed with Kraken2 and RefSeq databases. OTU tables were imported and processed using the phyloseq package in R., filtered (≥ 3 counts in ≥ 5% samples), and analyzed at the genus level.

The sequencing data were also processed using QIIME 2 version 2024.10 bioinformatics platform (https: Qiime2//qiime2.org/) [39]. Metrics for alpha diversity (Shannon index, and Chao1 estimator) were generated based on rarefaction at 153 541 reads per sample. Similarly, beta diversity was evaluated by Bray-Curtis method and a principal coordinates analysis (PCoA) plots generated using QIIME 2 pipeline [40]. Differences in the microbial composition between evaluated conditions were tested using permutational multivariate analysis of variance (PERMANOVA). Additionally, hierarchical clustering using Bray-Curtis dissimilarity and Ward’s linkage method was performed using R tool which measured compositional dissimilarity between samples based on abundance data.

Results

Successful colonization of ex vivo model by donor-derived microbiota

a-Quantification of bacteria recovered from the skin surface

To evaluate the efficiency of our skin microbiota recovery protocol, both bacterial load and DNA quantification were assessed across samples collected from human volunteers (Volunteers V1-V10, Supplementary S1). On average, 1.89 × 10⁷ bacteria per mL were recovered, including 7.14 × 10⁵ viable cells and 1.82 × 10⁷ non-viable cells. These numbers are in line with previous findings reporting approximately 1 × 10⁶ bacteria per cm² of skin, validating the robustness of our sampling method. Although only about 3.8% of the recovered bacteria were viable, this is consistent with known limitations related to environmental exposure (e.g., desiccation, oxygen, UV treatment) that may impact microbial viability during handling. Nevertheless, the total microbial load was sufficient to support downstream applications such as culturing, flow cytometry, and high-throughput sequencing.

To evaluate the quality and quantity of the recovered microbial DNA, we quantified total microbial DNA (ADNmb) and estimated the number of 16 S rRNA gene copies via quantitative PCR (Supplementary S2). DNA concentrations in the samples ranged from 0.03 to 0.37 ng/µL, which corresponded to estimated 16 S rRNA gene copy numbers between 5,207 and 260,027 copies/µL. The positive control produced 203.7 ng of DNA in 45 µL and roughly 913,114 copies/µL, supporting the sensitivity and reliability of the quantification method. These estimates were derived using the DNA mass, genome size of E. coli (4.6 Mbp), and Avogadro’s number, assuming an average molecular weight of 649 Da per base pair. The variation in DNA yields and 16 S copy numbers reflect individual differences in skin microbiota density and sample handling. Nevertheless, most samples yielded sufficient material for downstream applications such as 16 S rRNA gene sequencing.

Taken together, these results indicate that the protocol efficiently recovers both viable and non-viable bacteria of the skin microbiota, and that the DNA obtained is of sufficient quality and quantity for molecular analyses.

b-Viability and recovery of cutaneous microbiota on an ex vivo skin explant model

To evaluate the viability and persistence of human skin microbiota following transfer onto ex vivo skin explants, viable bacterial concentrations were quantified at baseline (day 0) and after 48 h of incubation on the explant surface (day 2), using both flow cytometry and culture-based enumeration. Results are presented in Fig. 2.

Fig. 2.

Quantification of viable bacteria from representative volunteer at sampling and from bacteria recovered from explants. Bacterial viability was assessed using flow cytometry (V1 cyto, MV1 cyto) and confirmed by CFU counting on agar plates (V1 cultivable, MV1 cultivable). Analyses were performed on the initial sample obtained from volunteer 1 at sampling (V1) and after 48 h from bacteria recovered from skin explants (MV1). Bars represent mean ± standard deviation (SD)

As a result, 48 h of contact with the skin microbiota was evaluated on our model. Bacterial viability was quantified using both culture-based colony enumeration and flow cytometry, and expressed in colony-forming units per milliliter (CFU/mL). Culture-based quantification showed a progressive decline in culturable bacteria over time. In the original suspension, total bacterial load reached 2.30 × 10⁶ CFU/mL, composed of 1.59 × 10⁶ CFU/mL aerobes and 7.13 × 10⁵ CFU/mL anaerobes (Data not shown). After 48 h of incubation, only 1.42 × 10³ CFU/mL were recovered, with 5.00 × 10² CFU/mL aerobes and 9.20 × 10² CFU/mL anaerobes (data not shown).

Flow cytometry analysis revealed a similar decline in total viable bacterial counts. The initial suspension contained 1.46 × 10⁶ viable bacteria/mL After 48 h, 3.37 × 10⁴ CFU/mL were recovered, corresponding to 2.3% viability relative to the inoculated dose. Although bacterial viability was decreased during incubation, a measurable fraction of viable and culturable bacteria remained detectable after 48 h.

These data show that the Perfex model is able to maintain a portion of the viable and culturable skin microbiota over a 48-hour period, including both aerobic and anaerobic populations. The similar decline observed in viability and culturable underscores the need for complementary analytical approaches to accurately assess microbial survival in skin explant systems. The two methods indicate that the model preserves a complex microbial community that includes both culturable and non-culturable bacteria.

c-preserved tissue viability and controlled immune marker response following microbiota inoculation

To evaluate the impact of microbiota inoculation on skin explants, both tissue integrity and innate immune markers immunolocalization were assessed after 48 h of culture. Comparisons were made between uninoculated control explants and explants inoculated with human-derived skin microbiota. Immunological assessments were performed directly on tissue sections, allowing spatial and semi-quantitative evaluation of specific immune markers (Fig. 3).

Fig. 3.

Tissue morphology and immunolocalization of hBD2, TLR2, and TNF-α following human-derived skin microbiota transfer to Perfex model and relative signal intensity semi-quantification by image analysis versus uninoculated control explants. Bar graphs represent the quantitative analysis of immunostaining intensity for HBD2, TLR2, and TNF-α. Scale bar: 50 μm

Microscopy-based analysis confirmed preserved tissue morphology in both conditions. No signs of cellular and tissular alteration were observed (Fig. 3). Histological imaging revealed intact epidermal stratification and dermal structure in both control and inoculated explants (Fig. 3a). In inoculated samples, qualitative differences in immune marker localization were noted. Human β-defensin 2 (hBD2) was visibly increased in the upper layers of the epidermis in the inoculated group, showing a strong signal in granular keratinocytes. Toll-like receptor 2 (TLR2) displayed enhanced staining across suprabasal layers in inoculated explants, suggesting active microbial sensing. In contrast, Tumor necrosis factor-α (TNF-α) staining appeared comparable between inoculated and control samples, except for the V1 for which a decrease in TNF-α expression was observed surprisingly. Semi-quantitative image analysis of signal intensity further confirmed these observations (Fig. 3b). hBD2 levels were significantly higher in microbiota-exposed explants than in controls, supporting an antimicrobial response triggered by the applied microbial communities. TLR2 expression was also moderately elevated following inoculation, consistent with the recognition of microbial-associated molecular patterns. By contrast, TNF-α levels did not significantly differ in 2 out 3 conditions, highlighting the absence of a marked inflammatory response.

The results demonstrate that the Perfex model maintains tissue viability and structural integrity in the presence of inoculated skin microbiota, while supporting a localized, balanced innate immune activation. The tissue responded with upregulation of barrier-defense and microbial recognition markers, without triggering pro-inflammatory cytokine production.

Profiling and statistical analysis of transferred skin microbiota

a-alpha diversity analysis of transferred microbiota

Following the assessment of bacterial load and viability, we investigated whether the transfer of microbiota onto the Perfex influenced community structure and diversity (Fig. 4). To this end, alpha diversity metrics, including the Shannon index and the Chao1 estimator, were calculated for each group: controls (C), explants after microbiota transfer (MV1, MV2, MV3), and microbiota recovered from corresponding volunteers (V1, V2, V3).

Fig. 4.

Indices of Alpha diversity, Shannon index (a) and the Chao1 (b) estimator, were calculated for each group from first experiment: controls (C), explants after microbiota transfer (MV1, MV2, MV3), and microbiota recovered from corresponding volunteers (V1, V2, V3)

No statistically significant differences in alpha diversity were observed between original and transferred microbiota at the genus level. The Shannon index, which reflects both richness and evenness, showed limited variation between groups. The overall diversity of the transferred microbiota (MV1–MV3) remained comparable to that of the corresponding in vivo samples (V1–V3), suggesting that the transfer process did not substantially alter the community diversity. The Kruskal-Wallis test did not detect statistically significant differences between groups (H = 10.50, p = 0.162; Fig. 4a) (Table 2).

Table 2.

The table summarizes alpha diversity indices (Shannon and Chao1) for microbiota samples before and after transfer onto skin explants

Similarly, the Chao1 estimator, which emphasizes rare taxa and estimates species richness, showed overlapping distributions between volunteer samples and their corresponding explants. The slight fluctuations observed between V and MV groups were not statistically significant (H = 9.44, p = 0.222; Fig. 4b) (Table 2), indicating that the ex vivo system preserved donor-specific richness over the 48-hour incubation period.

b-clustering and beta diversity evaluation of skin microbiota before and after transfer

To further investigate the global structure of microbial communities before and after transfer onto skin explants, beta diversity analyses were conducted. Hierarchical clustering using Bray-Curtis dissimilarity and Ward’s linkage method was performed on two independent experiments (Fig. 5). Samples included in vivo microbiota from volunteers (V1, V2, V3), post-transfer explant samples (MV1, MV2, MV3), and uninoculated control explants (C).

Fig. 5.

Clustering of skin microbiota collected from 3 volunteers before (V1-V2-V3) and after transfer on the Perfex model (MV1-MV2-MV3) compared to control skin explants (C). (a) and (b) represent two independent experiments performed at different time points to evaluate the reproducibility of the microbiota transfer. The table including identification of samples and their corresponding numbers is available in Supplementary S3

Two independent experiments were performed at different time points to assess the reproducibility of the microbiota transfer procedure. In both experiments, the dendrograms showed a clear grouping of samples by donor, with each volunteer’s sample (V1, V2, V3) clustering closely with its corresponding explant sample (MV1, MV2, MV3). This pattern reflects a strong similarity between the microbiota before and after transfer, indicating that the individual-specific microbial signature is largely preserved on explants over the 48-hour period. The control samples formed separate clusters, showing a profile clearly distinct from that of inoculated explants and supporting the absence of cross-contamination or any significant impact of potential background signal.

Principal coordinates analysis (PCoA) based on Bray-Curtis dissimilarity confirmed these findings (Fig. 6).

Fig. 6.

Principal coordinates analysis (PCoA) based on beta diversity using Bray-Curtis method. a PCoA was performed using the volunteer samples before transfer (V1-V2-V3) and post transfer (MV1-MV2-MV3) on human skin explants as well as control conditions b Beta diversity comparison between post-transfer samples (MV1–MV3) and control explants was conducted using PERMANOVA test

Principal coordinate analysis (PCoA) based on Bray–Curtis dissimilarity was used to visualize differences in microbial community composition among samples. Samples clustered into distinct groups according to their donor origin along the principal axes, with Axis 1 explaining 40.62% of the variance, followed by Axis 2 (20.74%) and Axis 3 (13.90%). These principal coordinates represent the main gradients of community dissimilarity derived from the Bray–Curtis distance matrix. Post-transfer samples (MV) remained closely associated with their original in vivo counterparts (V), indicating that donor-specific microbial signatures were largely maintained following transfer onto the explants. Dendrogram analysis showed a clear grouping of samples by donor, with each volunteer sample (V1, V2, V3) clustering closely with its corresponding explant sample (MV1, MV2, MV3). Control samples formed separate clusters, suggesting minimal cross-contamination. This pattern indicates that the model maintains the inter-individual variability of the skin microbiota while preserving its overall compositional structure following transfer during the 48-hour incubation period.

A second comparison of beta diversity based on Bray–Curtis dissimilarity between post-transfer microbiota (MV1, MV2, and MV3) and control samples was performed using PERMANOVA (Permutational Multivariate Analysis of Variance) (Fig. 6b). The results showed a pseudo-F statistic of 5.61 with a p-value of 0.001 based on 999 permutations, indicating that the observed differences between groups exceeded what would be expected by chance. These results indicated that microbial community composition differed significantly between post-transfer samples and control explants and reflect shifts in microbial community structure rather than random variation among samples. Together, these findings support the ability of the model to distinguish transferred microbiota from control conditions and to track donor-specific microbial dynamics under controlled experimental conditions.

Preservation of donor-specific microbiota structure in the Perfex model

a-taxonomic composition of the skin microbiota before and after transfer on human skin explants

To evaluate the taxonomic structure of the microbiota before and after transfer to skin explants, we analyzed the Top 30 most abundant bacterial genera across all samples, based on 16 S rRNA gene sequencing from both independent experiments. For each volunteer, the 30 genera with the highest sequence counts were compiled, with duplicates removed using OTU identifiers as well as the unassigned genus, and their relative abundances calculated as percentages (Fig. 7).

Fig. 7.

Conservation of the more relevant genus according to relative abundance of sequenced bacterial taxa at the genus level from each volunteer (V1-V2-V3) in experiment 1 (a) or 2 (b) before and after transfer on Perfex system (MV1-MV2-MV3) as well as in control conditions (C). (a) and (b) correspond to two independent experiments at different time points

Across both experiments, the microbial communities were largely dominated by typical skin-associated genera such as Cutibacterium, Staphylococcus, Streptococcus and Corynebacterium. These genera remained highly represented in both the original in vivo samples (V1-V3) and their corresponding post-transfer explant samples (MV1–MV3), suggesting that key taxonomic features of the native microbiota were largely maintained after transfer to the Perfex model during the 48-hour incubation period.

In experiment 2, the Top 30 list included a broader diversity of genera Lactobacillus, Fusobacterium, Moraxella, Micrococcus, Pseudomonas, Rothia, Finegoldia, Anaerococcus, and Paracoccus reflecting greater inter-individual and inter-sample variations (Fig. 7b). Despite this, Cutibacterium, Staphylococcus and Corynebacterium remained among the dominant genera across most conditions, indicating their relative stability under ex vivo culture conditions. Overall, the taxonomic profiles of the explant samples broadly comparable to those of the original donor microbiota, supporting the successful microbial transfer and short-term maintenance of donor-specific microbial communities during the 48-hour incubation period.

b-taxonomic composition following suppression of resident microbiota on the untreated human skin explants

To complement analyses of microbial diversity and taxonomic structure, we investigated the relative abundance of dominant bacterial genera across samples before and after microbiota transfer onto the Perfex model. Sequence data were filtered to exclude low-abundance OTUs (≤ 100 reads) from control samples which can correspond to residual microbiota, in order to minimize overestimation of diversity. Thus, the most abundant genera were identified per volunteer across two independent experiments. Genera were compiled and de-duplicated based on OTU identifiers to allow for cross-sample comparisons for each experiment (Fig. 8). While Fig. 7 highlights the overall taxonomic structure across samples, Fig. 8 focuses on the most abundant genera within each donor-specific community, allowing visualization of individual variability in microbial composition.

Fig. 8.

Conservation of the most relevant genera according to relative abundance of sequenced bacterial taxa at the genus level from each volunteer (V1-V2-V3) in experiments 1 (a) or 2 (b) before and after transfer on Perfex model (MV1-MV2-MV3) following the suppression of resident microbiota from control conditions. (a) and (b) correspond to two independent experiments at different time points

In Experiment 1, clear donor-specific taxonomic patterns were observed (Fig. 8a). For volunteer V1, both the in vivo microbiota sample and corresponding explants (MV1) were dominated by Roseomonas, Streptococcus, Pseudomonas and Staphylococcus. V2 and MV2 samples showed enrichment in Corynebacterium, Nitrosomonas and Neisseria while V3 and MV3 exhibited a more complex composition, with higher levels of Staphylococcus, Lactococcus and Rothia. These differences indicate that donor-specific microbial signatures were largely maintained after transfer onto explant. Notably, dominant genera detected in the original samples remained strongly represented in the corresponding explant samples compared with uninoculated controls.

In Experiment 2, which involved the same volunteers sampled in an independent experimental run, similar taxonomic trends were observed (Fig. 8b). Notably, the presence of the genera Neisseria and Staphyloccus was preserved in both experiments independently of the volunteers. V1-derived explants again exhibited predominance of Facklamia, Fenollaria and Corynebacterium, while V2 remained enriched in Corynebacterium, Neisseria, Porphyromonas, Fusobacterium and Streptococcus. V3 explants consistently demonstrated a complex signature with notable contributions from Fusobacterium, Veillonella, Paracoccus, and Neisseria. These patterns broadly recapitulated the donor-specific profiles observed in Experiment 1, indicating reproducibility of the model across independent experimental runs.

Control explants exhibited uniformly low relative abundance according to the filtering criteria, suggesting limited background signal in the sequencing dataset. While minor variation was observed in the abundance of lower-ranked genera between experiments, likely due to biological variability or sampling noise, dominant genera remained consistently represented across samples.

Importantly, the overall taxonomic structure of the microbiota remained broadly comparable between volunteers’ samples and their corresponding explant-associated communities, supporting the preservation of donor-specific microbial signatures after transfer. Together, these results indicate that the Perfex explant model supports the short-term maintenance of complex, donor-specific microbial communities during the 48-hour incubation period, providing a suitable platform for controlled microbiome studies.

Discussion

In this study, we established and validated an ex vivo human skin explant model, termed Perfex, that enables the transfer and short-term maintenance of complex donor-specific skin microbiota. The present study demonstrates that these communities can be transferred to human skin explants while preserving key features of community structure at the genus level and maintaining tissue integrity. Although microbial viability decreased during incubation, a measurable fraction of viable bacteria remained detectable, and only moderate modulation of selected innate immune markers was observed, consistent with a controlled host response rather than overt inflammation. Taken together, these findings support the Perfex model as a physiologically relevant and ethically acceptable platform for studying host–microbiome interactions under controlled conditions. Such validation is consistent with recent studies showing that complex human skin explant models preserve tissue viability and structural integrity under controlled conditions, while maintaining donor-specific microbial communities and only a balanced activation of innate immune responses [38, 41, 42]. By bridging this gap, the Perfex model may also help standardize protocols for skin microbiome studies, a current limitation in translational dermatological research [9].

One of the most notable outcomes is the preservation of donor-specific microbial communities following transfer to the explants. Both alpha and beta diversity analyses showed that richness, evenness, and global community structure were maintained, while hierarchical clustering confirmed that each explant-associated microbiota clustered tightly with its corresponding in vivo sample. This ability to retain inter-individual variability is crucial, given the highly personalized nature of the human skin microbiome [43, 44]. Personal microbial profiles are increasingly recognized as determinants of susceptibility to skin diseases such as atopic dermatitis, acne, and psoriasis, as well as modulators of therapeutic outcomes [4, 18]. Importantly, taxonomic analyses revealed that dominant skin-associated genera, including Staphylococcus, Cutibacterium, and Corynebacterium, remained stable post-transfer, consistent with their well-established roles in barrier function and host defense [5, 6]. Although beta-diversity analyses demonstrated preservation of donor-specific community structure and taxonomic composition at the genus level, the use of 16 S rRNA gene sequencing-even targeting the V1-V3 region-limits reliable species- and strain-level identification for several taxa. Beyond these dominant taxa, donor-specific differences such as enrichment in Streptococcus, Lactococcus, or anaerobic genera like Prevotella and Fusobacterium were also maintained, highlighting the ability ot the model to reproduce complex and individualized microbial ecologies. This distinguishes Perfex model from reconstructed human epidermis models, which are typically colonized with only a few reference strains and thus fail to capture the ecological and functional diversity of the native skin microbiome [26]. Preserving such individuality is essential, since even subtle microbial shifts can influence cutaneous immune responses [7].

Equally important is the observation that microbiota inoculation elicited a controlled tissue response without compromising structural integrity. Histological and immunostainings analyses confirmed intact epidermal and dermal organization, while innate immune markers such as hBD2 and TLR2 were upregulated in inoculated explants. This indicates that the tissue mounted antimicrobial and microbial-recognition responses, while avoiding an overt inflammatory reaction, as reflected by unchanged TNF-α levels. Such balanced activation is consistent with the nature of commensal host–microbe interactions in vivo, where the microbiota typically promotes barrier defense and immune homeostasis rather than inflammation [24]. This feature greatly enhances the relevance of the Perfex model compared with both sterile in vitro systems and animal models, the latter being limited by interspecies differences in immunity and microbial composition as well as by increasing ethical restrictions [38]. By contrast, ex vivo explants derived from surgical waste comply with the 3Rs principle and provide a human-relevant system [25]. The ability to recapitulate protective host responses while minimizing inflammatory artifacts makes this model suitable for long-term evaluation of cosmetic and therapeutic formulations [45, 46]. Moreover, it may help clarify mechanisms by which commensals exert protective effects against opportunistic pathogens, a major focus of current microbiome research [27].

Several studies have recently sought to overcome the limitations of traditional in vitro systems by developing reconstructed or standardized skin models that can host complex microbial communities. For instance, Wang et al. designed a microecological skin model (SCmic) capable of reproducing native-like community structures with minimal taxonomic drift [47]. Similarly, Maloney et al. introduced the SURFACE model (Skin µbiome Reconstruction For Assessment of Cutaneous Effects), which supports stable colonization of multi-strain consortia, including anaerobic Cutibacterium acnes, on human skin equivalents maintained at the air–liquid interface for up to seven days [48]. Other approaches such as SkinCom [49] or organotypic culture models [50] provide standardized environments that facilitate host–microbiome interaction studies. Furthermore, Loomis et al. demonstrated that reintroduction of a mixed microbial community onto 3D human skin equivalents altered epidermal morphology and gene expression, underscoring the functional consequences of microbial colonization [51]. More recently, an alternative ex vivo culture strategies demonstrated tissue viability even under non-CO₂ conditions, further broadening the accessibility of microbiome research platforms [52]. Together, these models highlight the growing interest in reconstructing and sustaining complex skin microbiota in controlled systems. Within this context, the Perfex model provides a complementary strategy that uniquely preserves donor-specificity, an essential feature for translational and personalized applications. These comparative advances reinforce the notion that combining reconstructed systems with donor-derived microbiota could maximize both control and clinical relevance [53].

The ability of this system to sustain viable, donor-specific communities while triggering host responses without compromising tissue integrity suggests a wide range of potential applications. It provides a useful tool for fundamental microbiome research, investigations into colonization dynamics, ecological competition, and host signaling pathways in a donor-specific context. The model also offers a promising platform for preclinical evaluation of microbiome-targeted therapies, including probiotics, postbiotics [54, 55], bacteriophages [20], and microbiome transplantation strategies [56–58]. By preserving inter-individual variability, it further supports personalized dermatology, enabling the investigation of subject-specific responses to cosmetic or therapeutic treatments. Given its adaptability, this model may also be useful for investigating antimicrobial resistance gene transfer within skin communities [59]. Moreover, the model provides a scalable testing environment for industrial applications, particularly as microbiome-friendly claims increasingly require rigorous experimental validation [60].

Nevertheless, several limitations of this study should be acknowledged. The number of microbiota from volunteers and skin donors was limited due to logistical constraints associated with obtaining fresh surgical samples, and microbial composition was analyzed using 16 S rRNA gene sequencing at the genus level, which does not provide species-level resolution or direct information on bacterial viability.

Microbial viability decreased markedly over the 48-hours incubation period, with rentention of culturable bacteria below 1% of the inoculated dose. Although this aligns with previous challenges in maintaining anaerobes and fastidious taxa outside their natural environment [61], optimizing culture conditions such as nutrient supplementation, including artificial sebum deposition on skin explant surface, humidity, or the addition of host-derived factors could prolong microbial survival. Moreover, the present work focused primarily on bacterial communities, without assessing the fungal, viral, or arthropod components of the skin ecosystem, which may also play important roles in host-microbiota interactions. Integrating multi-omics approaches could provide a more comprehensive view of microbial community dynamics [62, 63]. Finally, while innate immune responses were evaluated with a limited set of markers, broader transcriptomic or proteomic profiling would allow a deeper characterization of host-microbe crosstalk. Extending the duration of culture beyond 48 h and exploring multi-donor inoculations will also be important for future steps to study long-term stability, dysbiosis, and microbial competition. Future refinements of the model may include longer culture durations, incorporation of additional microbiome components such as fungi and viruses, and evaluation of disease-associated microbiota, which would further enhance the translational value of this system.

In conclusion, this study shows that the Perfex model supports the transfer and short-term maintenance of complex, donor-specific skin microbiota while preserving both tissue integrity and a balanced immune response. Compared with conventional in vitro systems or animal models, the Perfex model provides a more physiologically relevant and ethically sustainable platform for dermatological and microbiome research. The system could also be adapted to investigate microbiota from individuals with skin disorders such as atopic dermatitis, enabling mechanistic studies of host-microbiome interactions under pathological conditions. By bridging the gap between simplified models and the complexity of human skin, the Perfex model holds considerable promise for advancing basic research as well as the development of microbiome-based therapeutic and cosmetic strategies.

Supplementary Information

Supplementary Material 1. ^{(50.3KB, docx)}

Authors’ contributions

Contributions: Contributions to the conception: AP, GP, CG, CN, DS. Performed the experiments: CD, DR, GP, AP, DS. Design of the work : AP, GP, CN, DS. Acquisition and analysis : GP, AP, DS. Interpretation of data: GP, AP, CN, DS. Have drafted the work: GP, AP, CN, DS. Review and editing: CG, VP, KV. All authors reviewed the manuscript. Each author is expected to have approved the submitted version.

Data availability

The datasets generated during this study are available in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB105206. ( [https://www.ebi.ac.uk/ena/browser/view/PRJEB105206](https:/www.ebi.ac.uk/ena/browser/view/PRJEB105206).

Declarations

Ethics approval and consent to participate

Skin microbiota samples were collected from healthy adult volunteers by non-invasive skin swabbing after obtaining written informed consent. Participation was entirely voluntary. Samples were anonymized prior to use, and no personal, demographic, clinical, or health-related data were collected or used in the study. Donors were not involved in any subsequent study procedures or follow-up.

The collection and use of human-derived samples were conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. BIO-EC Laboratory is authorized by the Bioethics Department of the French Ministry of Higher Education and Research (Direction Générale de la Recherche et de l’Innovation) to collect, store and use human biological tissues for research purposes (authorization number DC-2022-5373).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.