The development of an in vitro human hair follicle organoid with a complexity similar to that in vivo

In vitro hair follicle (HF) models are currently limited to ex vivo HF organ cultures (HFOCs) or 2D models that are of low availability and do not reproduce the architecture or behavior of the hair, leading to poor screening systems. To resolve this issue, we developed a technology for the construction of a human in vitro hair construct based on the assemblage of different types of cells present in the hair organ. First, we demonstrated that epithelial cells, when isolated in vitro, have similar genetic signatures regardless of their dissection site, and their trichogenic potential is dependent on the culture conditions. Then, using cell aggregation techniques, 3D spheres of dermal papilla (DP) were constructed, and subsequently, epithelial cells were added, enabling the production and organization of keratins in hair, similar to what is seen in vivo. These reconstructed tissues resulted in the following hair compartments: K71 (inner root-sheath), K85 (matrix region), K75 (companion layer), and vimentin (DP). Furthermore, the new hair model was able to elongate similarly to ex vivo HFOC, resulting in a shaft-like shape several hundred micrometers in length. As expected, when the model was exposed to hair growth enhancers, such as ginseng extract, or inhibitors, such as TGF-B-1, significant effects similar to those in vivo were observed. Moreover, when transplanted into skin biopsies, the new constructs showed signs of integration and hair bud generation. Owing to its simplicity and scalability, this model fully enables high throughput screening of molecules, which allows understanding of the mechanism by which new actives treat hair loss, finding optimal concentrations, and determining the synergy and antagonism among different raw materials. Therefore, this model could be a starting point for applying regenerative medicine approaches to treat hair loss.


Introduction
The human hair follicle (HF) is an appendage of the skin known for its wide range of functions, from external protection and thermoregulation to social interactions [1].The diversity of its roles is also well presented on the cellular level, where a single follicle is composed of several cell types [2] with very distinct cell-to-cell interactions [3] and multifaceted signaling [4].Furthermore, this organ behaves dynamically, passing through repetitive cycles of growth, regression, and quiescence [5].These cycles are well aligned and described in mice [6,7], but are rather stochastic in humans [8] increasing the difficulties in understanding the biology of the human hair [9].The biological complexity of the HF reflects on several fields of research, including pigmentation [10], microbiome [11], aging [12], stem cell therapeutics [13], plasticity [14], differentiation [15], and others [16][17][18].
The HF contains an epithelial compartment of outer and inner root sheaths, a matrix, and a hair shaft.It also contains dermal papilla (DP) and dermal sheath (DS) cells derived from the mesenchyme.DP lies adjacent to a population of transit-amplifying matrix cells and secret diffusible proteins that regulate their proliferation and differentiation into the hair shaft, thereby playing a key role in the regulation of hair cycling as well as regenerating hair growth [5,19,20].DP and DS cells, which constitute the dermal part of the HF, can induce HF formation by interacting with the epithelial cells [21].
More than 30 years ago, Philpott et al [22] developed the HF organ culture (HFOC), an ex vivo model that enabled the survival of hair units inside a petri dish after microdissection from surgery tissue waste.Nowadays, this is considered the gold standard in vitro model when working with human hair [16].Thus, researchers in the field rely on this model not only to study the biology of the hair [23] but also to assess the effect of pharmaceutical [24,25] and cosmetics compounds [26,27].However, the availability of this model is decreasing because surgeries are now less invasive, and only small tissue samples can be recovered [28].Philpott himself, who developed the HFOC model, urged for the development of next-generation in vitro hair models to fulfill the demand for them [29].Another method for evaluating hair loss therapeutics, which involves the use of specific types of hair cells, such as the DP cells, in 2D models, has been used mostly due to its simplicity and commercial accessibility [30][31][32][33].However, the hair-inductive capacity of these cells is lost during subculture [34].In this respect, alternatives have been proposed and the hair-inducing capacity of human DP cells can be restored in the form of 3D spheroid cultures [35].The 3D architecture of the DP is capable of substantially changing the expression profile when cultured while attached to the flasks [36] and, just switching the DP from a 2D monolayer to a 3D spheroid could restore the transcriptome by more than 20% when compared to fresh intact DP [35].Even though the use of the DP in a 3D structure has only recently been adopted by the research community, major key interactions, such as the epithelial-mesenchymal interaction (EMI), cannot be mimicked using the DP alone [37].The EMI plays a central role in organogenesis in many tissues including HFs.The development of HFs is initiated by a signaling cascade of EMIs.Extensive interactions between these two different HF compartments have led to the formation of the hair shaft-producing miniorgan that shows a cyclic activity during postnatal life with periods of active growth and hair shaft formation (anagen), apoptosis-driven involution (catagen), resting and hair shedding (telogen) [38,39].Studying the human hair-shaft formation or elongation without the use of the HFOC model, necessitated the use of animal models in which human DP cells [36], DP spheroids [35], or pluripotent stem cells [40] were grafted.Besides the intrinsic difficulties and ethical questions concerning the use of animal models, the efficiency of generating new human HFs tends to be low when grafting human cells in them [41].This is a notable difference from the works performed with mouse cells [42].
Remarkably, Abaci et al [43] were able to mimic hair development and EMI in vitro by co-culturing epithelial cells with DP cells that were transfected with certain key genes, such as Lef-1, to upregulate their expression and then deposited inside dermal equivalent microcylinders generated by casting 3D microprinted molds in a skin equivalent.Recently, another group showed that activating pluripotent stem cells, through specific modulation of transforming growth factor β and fibroblast growth factor, over 4 months could generate hair-bearing skin organoids [44].Even though both studies can be considered breakthroughs for in vitro development, their methods are not expected to be widely applied because the experiments are time-consuming and employ sophisticated techniques, such as printing, genetic modification, and pluripotent stem cell technology.
Recently, various models that replace human HF tissue and facilitate the evaluation of treatments have been studied.In that respect, other alternatives such as the microfollicle model [45] and the two-cell assemblage model [46] for evaluating hair loss therapeutics in vitro have been proposed.However, at present, the HF growth rate and the expression of specific HF proteins are insufficient to replace human HFs.
In this study, by using simple tissue-engineering techniques with high-quality primary cell isolation, we developed for the first time an organoid that mimics the follicle in vivo and can be easily constructed.Desired hair properties were assessed such as the capacity of epithelial elongation, the development of tubular structure, and keratin formation.The results were correlated with ex vivo data from HFOC models for the evaluation of screening for substances with activity, and a hint on the.This study fills the gap between the ex vivo gold standards and the limited in vitro options and shows the potential of our model in regenerative medicine applications.

Stem cell optimum conditions (SCOCs) of epithelial cells
For ORS cell isolation, the HFs with the lower part of the bulb and the upper part of the sebaceous gland removed were placed in a 60 mm dish in which 3T3-L1 cells were inactivated.For matrix cell isolation, the top of the DP at one side of the follicle was cut to make space for separating the matrix region.Using a small syringe, the matrix region was pushed up at the bottom of the bulb and taken out of the follicle.The isolated matrix cells were also cultured on inactivated 3T3-L1 feeder cells.Both cell types were cultured in EpiLife media (GIBCO, MEPI500CA) supplemented with 1% EpiLife Defined Growth Supplement (EDGS, GIBCO, S0125), 20% FBS (Hyclone, SH30084.03),0.5% Penicillin-Streptomycin (Gibco, 17-602E), 0.2% amphotericin B (Gibco, 15290026) and 10 µM ROCK inhibitors (Y27632, Merck, US, SCM075) for 5 d without media change after which FBS was reduced to 10%.

Standard culture conditions (SSCs) of epithelial cells
The same region as the HF for SCOC was used for SSC.Collagen type I-coated 60 mm dishes were used for isolation and culture (Corning, US, 354401).Cells were cultured in EpiLife media supplemented with 1% EDGS, 20% FBS, 0.5% Penicillin-Streptomycin, and 0.2% amphotericin B for 5 d without changing media after which FBS was removed.

Generation of IVHF
DP cells were seeded into a 3D cell culture plate (Thermo Scientific, 174 925) at a density of 2500 cells/well with mild centrifugation and incubated for sphere formation.Epithelial cells were cultured on inactivated 3T3-L1 feeder cells to 70%-80% confluency and detached with a trypsin-EDTA solution.Detached epithelial cells were resuspended in HFOC medium with 30 µM of Rock inhibitor and seeded on the DP sphere cultured for 5 d at a density of 3000 cells/well with mild centrifugation.The HFOC medium was changed every 2-3 d, and the Rock inhibitor concentration was reduced to 5, 2.5, and 0 µM.

Histology
IVHF cultured for 7 d were used for histology analysis.After washing with phosphate-buffered saline (PBS), IVHF samples were fixed in 4% formaldehyde overnight at room temperature.After fixation, IVHF permeabilized with 0.3% Triton X-100 in PBS and incubated with primary antibodies for 1.5 d at 37 • C.Then, the secondary antibody and DAPI were added and incubated for 1 d at 37 • C. The samples were then treated with Mounting and storage solution (Binaree, Korea, SHMS-060) and incubated overnight at 37 • C. For specific protein detection and visualization, confocal microscopy was used (LSM 800, ZEISS).
Anagen phase HFs and ex vivo skin were embedded with OCT compound in dry ice.Frozen blocks were sectioned 10 µm thickness and made slides.Slides fixed in 4% formaldehyde and permeabilized by 0.3% Triton X-100 in PBS and blocked for 1 h (Abcam, UK, ab126587).Primary antibodies were applied overnight at 4 • C. The secondary antibody was applied for one hour at room temperature.The staining pattern of each protein was detected by confocal microscopy (LSM 800, ZEISS).

Ex vivo HFOC
The materials for the experiment were diluted to their specified concentrations in HFOC medium and were used to treat anagen phase HFs.The HFOC medium was changed once every 2-3 d, and hair shaft elongation was analyzed using a stereomicroscope.

Preparation for sequencing
RNA was extracted from cells under three conditions and transcriptome analysis was performed.Total RNA was obtained by treatment with Trizol reagent, and only mRNA was isolated with TruSeq Stranded mRNA LT Sample Prep Kit (Illumina).For short read sequencing, the purified RNA was randomly fragmented, and the RNA fragments were converted to cDNA by reverse transcription.After amplification, an insert of size 200-400 bp was obtained by the size selection.

Sequencing and analysis
Sequencing was performed using an Illumina Novaseq 6000 sequencer.The raw reads were obtained, and artifacts such as adapter sequence, contaminant DNA, and PCR duplicates or poor quality were removed.Aligned reads were generated after mapping to the reference genome using the HISAT2 program.Transcript assembly was performed using the ordered read information of the StringTie program.Transcript quantification was applied to each sample to extract its expression profiles.Differentially expressed genes (DEGs) were selected, and functional annotation and gene set enrichment analyses based on GO and KEGG databases were performed.

Statistical analysis
The experimental data are displayed as the mean ± standard error of the mean (SEM).We applied the Student's t-test to examine variations between two groups, with statistical significance defined as p-values below 0.05 ( * p < 0.05, * * p < 0.01, * * * p < 0.001).

Culture conditions alter the morphology and function of epithelial cells
Our initial hypothesis was that isolating a primary epithelial stem cell population with higher differentiation ability, such as the one located in the matrix region, would facilitate the creation of a cell bank that would be highly trichogenic.After optimizing the dissection procedure to obtain the epithelial cells located in the bulb region (supplementary video 1), several conditions were applied for the successful isolation of these cells.SCCs for culturing ORS were applied, however, the cells failed to attach and proliferate (figure 1(E)).Interestingly, a combination of the standard conditions with an initial supplementation of FBS, Rho-kinase inhibitor, and 3T3i feeders, enabled the attachment and further proliferation of cells from the matrix region (figure 1(F)); these will be referred to as SCOCs.The ORS counterpart was successfully isolated in standard conditions and SCOC (figures 1(B) and (C)).
The epithelial cells isolated and cultured under SCOC were morphologically distinguished from those cultured under SCC, even when they were isolated from the ORS region.They featured a high nucleus-to-cytoplasm ratio and increased 3D colony numbers.The use of IVHF media on ORS or matrix cells derived from SCOC, resulted in fiber-like shapes that were hundreds of micrometers in length (figures 1(D) and (G)), however, when IVHF media was used on cells derived from SCC, cells failed to continue to proliferate or differentiate in any distinct pattern (data not shown).
In addition to evidence from microscopy, transcriptome analysis also showed differences in culture conditions.For the characterization of DEGs under the three different conditions, we analyzed the transcripts using fold change and p-value.DEGs from the comparisons SCC ORS vs SCOC ORS, SCC ORS vs SCOC Matrix and SCOC ORS vs SCOC Matrix identified 1970, 1576, and 110 genes respectively (figure 1(J)).In the volcano plot, the yellow dots represent up-regulated genes, and the blue dots and bars represent down-regulated genes with a p-value lower than 0.05 (figures 1(K)-(M)).This RNA sequencing corroborates with light microscopic observations, demonstrating that culture conditions rather than the isolation region of epithelial cells play a decisive role in cell function and fate.
For the mesenchymal part, the DP cells were isolated from HFs (figure 1(H)) as described by Higgins et al [35] and expanded as described in Material and Methods (figure 1(I)).

Generation of IVHF by interaction between DP spheroid and epithelial cells mimicking in vivo markers and behavior
To generate IVHF, we aimed to mimic the interaction of mesenchymal and epithelial cells in the HF.DP cells were transformed into spheroids using a nonadherent surface plate.After 5 d of spheroid maturation, epithelial cells, cultured under SCOC, were added.Mild centrifugation was performed to allow the epithelial cells to come closer to the DP spheroids and wrap them; this was recorded as day 0 for the IVHF formation (figure 2(B)).This process plays an important role in IVHF formation and development.If centrifugation is not performed, IVHF elongation starts slowly and results in high variation (supplementary figure 1).To promote differentiation of structure, the concentration of ROCK inhibitor was gradually decreased.The sample was kept in nonadherent well plates to limit the potential interactions with the plastic surface.After a 48 h incubation, the association of epithelial cells with DP spheroids was observed using a phase-contrast microscope, and both cell types were polarized.On the subsequent days, elongation through the differentiation and proliferation of epithelial cells began, reaching its plateau between the 6th and 10th days (figure 2(C)).The elongation rate in vitro was similar to that in ex vivo culture from plucked HFs, having an average elongation of 100-200 µm d −1 (figure 2(D)).
During the telogen-to-anagen transition, epithelial cells interact with DPs and differentiate into layers of growing HFs such as hair shaft, inner root sheath, and basement membrane.These newly differentiated compartments are classified as site-specific keratins.To identify the structural similarity with human HFs, we checked whether IVHF expresses these special keratins.On day 7 of the IVHF culture, we performed immunofluorescence for specific hair keratins.Figure 2(E) shows IVHF express keratin 5, 71 and forms slightly different layers.Two keratins identify ORS and IRS, respectively, in human HFs.Figures 2(F)-(H) show the hair shaft markers keratin 75 (medulla), 82 (cuticle), and 85 (precortex), respectively, that were expressed in IVHF, and their expression appeared at the interface between DP and epithelial cells and disappeared in the olddifferentiated area.As expected, proliferation markers, such as Ki67, were mainly present in the epithelial compartment of the IVHF.Additionally, the basement membrane marker laminin 5 was expressed at the interface of differentiation, showing good in vitro to in vivo congruence.Vimentin, a specific marker for the mesenchymal compartment in HFs, was specifically found at the spherical part of the IVHF, providing proof of its compartmentalization.To validate that the keratin antibodies could differentiate between the different layers of the HFs, all layer-specific keratins were validated for their specificity using plucked HFs as a control (supplementary figure 2).
Next, we determined that IVHF can show hair cycle-specific changes with the passage of culture time.During anagen to catagen transition, the HF undergoes structural and physiological changes, and several markers can be used to specify its stage.As the catagen period progresses, the number of DAPI positive dots below the Auber's line decreases and the expression of TCHH and HPSE1, known as anagen-specific markers of the IRS layer, is reduced.Figure 2(I) shows that these specific changes are the same in IVHF.Hematoxylin positive dot, which indicates a nucleus in a location similar to the HF Auber's  line, decreased as the culture continued.And the expression of TCHH and HPSE1 also decreased with culture time.

IVHF shows similar responses with positive materials of the ex vivo culture model
Next, we investigated whether IVHF shows similar drug responses with the ex vivo organ culture model.Two well-known drugs and a herbal extract were tested for their effects on elongation.Minoxidil is one of the FDA-approved drugs and is widely used as a positive control for in vitro, ex vivo, and clinical studies on anti-hair loss treatment.TGFβ has the opposite effect of minoxidil, as it induces hair loss and hair catagen phase acceleration in ex vivo.Ginseng extract, which contains a variety of saponins, is known to alleviate hair loss.Their detailed mechanisms of efficacy are well elucidated [48][49][50].We treated the model with these drugs for 7 d with epithelial cell seeding to DP spheroid.During the first 48 h of culture, there were no structural changes after drug treatment.After 7 d of treatment, minoxidil and ginseng extract enhanced the elongation of IVHF, and TGFβ inhibited it (figures 3(A) and (C)).Ex vivo results showed the response of IVHF to drugs is similar to that of the organ culture model (figures 3(B) and (D)).

IVHF integration into ex-vivo skin explants
To infer the trichogenic potential of IVHF in-vivo and speculate on its regenerative medicine potential, skin explants were injected with IVHF.Firstly, skin explants containing epidermis, dermis, and around 5 mm of hypodermis (Hyposkin ® ) were purchased from Genoskin (France) and put into culture following the manufacturer's guidelines for one day.On the following day, IVHF cultured for less than 48 h (figure 2(B)) were collected, and injected around the reticular dermis close to the subcutaneous tissue of the skin.Each skin explant received eight injections containing six IVHF constructs per injection.To be retrieved with ease, the injections were done in parallel with a line that was drawn on the center of each skin explant.After 7 d of culture, the explants were sliced serially around the injection region, and at least 40 slides were recovered for each skin for histological analysis.Surprisingly, the group which received IVHF injections had a 70 ± 4.2% (p < 0.05) increase in HFlike structures (figure 4(A)).Additionally, most of the additional HF structures were found exactly in the reticular dermis and had smaller structures, as if they were not mature HFs, (figure 4(B)), supporting the evidence that the injection of IVHF contributed directly or indirectly to the data.Additionally, the lack of pyknotic nucleus within these miniaturized structures suggests that the IVHF injection did not cause immediate cell death or destruction of the structures into the skin explant environment suggesting potential to a further integration into the host tissue.

Discussion
In 2014, the European Union (EU) banned the sale of cosmetics tested on animals, and similar bans were imposed in countries such as Israel, India, and the United States.Furthermore, since 2022, the FDA has no longer required animal tests before human drug trials.Both governmental decisions are boosting the need for better alternatives to animal models.In this study, we aimed to establish an evaluation method that can replace animal experiments and ex vivo human HF models and to establish a model that can be used in the field of HF regeneration.
The limitations of existing models are that many characteristic markers of the human follicle are quickly lost in vitro.To overcome this problem, various methods have been combined, including the use of a fibroblast feeder layer [51,52], ROCK inhibitors [53], growth factors [54,55], biocompatible scaffolds or non-adherent 3D cultures [56,57], and cell isolation from different anatomical sites.Surprisingly, we noticed that once in vitro, the culture conditions had a greater impact on the trichogenic potential of human HF epithelial cells than the anatomical location of the cells, suggesting that under the right conditions, these cells are capable of differentiating into several different and distinct hair layers.
Once we established the SCOC, human hair epithelial cells exhibited high nucleus-to-cytoplasmic ratios and increased 3D colony numbers, and the final IVHF maintained a structure similar to human HFs.The immunostaining results of DP, ORS, and IRS markers indicate that the IVHF was composed of several layers and defined compartments.Additionally, the growth rate of the in vitro model was maintained for 5-7 d while growing at a rate of 100-200 µm d −1 .This indicates that the growth mechanism of the IVHF is similar to that of HFs in the body.Finally, after reaching the plateau, hair growth of up to 1.5 mm in length was observed, which is a level that previous studies [58] have not achieved.Overall, these results demonstrate that the SCOC we established has trichogenic potential similar to that of existing hair epithelial cell cultivation conditions.To this point, when we tried to generate IVHF using immortalized cells or primary cells typically isolated from the skin's anatomical compartment rather than the hair compartment, elongation and compartment organization failed to occur (supplementary figure 3), confirming that SCOC is associated with good cell selection and is pivotal for the success of HF mimicking in vitro.
We used a 2-step method in which DP spheroid formation and epithelial cell attachment were separated.Although a direct comparison with the 1step method is difficult, the use of the 2-step fabrication has the advantage of more closely mimicking the in vivo environment.The incubation time of DP spheroids alone is expected to reach a genetic signature profile similar to that of in vivo [35].
In addition, we have found that the elongation is more natural and reproducible when the attachment of DP spheroid and epithelial cell can interact quickly and evenly (supplementary video 2).Therefore, the quality control of IVHF formation can be ensured when the constructs are formed by a 2-step method.
Even though the trichogenic potential was undoubtedly demonstrated, this model should be used carefully, as the IVHF does not contain all the surrounding environments in vivo; for example, they lack blood vessels and fat cells.In the case of certain actives, such as the FDA-approved drug minoxidil, differences between the elongation performance of ex vivo follicles and the IVHF could be explained by the mechanism of action of minoxidil, which is dependent either on vascularization improvement or modulation through adipose stem cell tissue.Even though we substantially made advances in terms of in vitro model creation for hair research, bio-printing and organ-on-a-chip are some of the technologies that could enhance the complexities of in vivo models, and these approaches have been attempted recently [59][60][61].
The entire human hair cycle takes place over many years.Due to technological limitations, we have only been able to demonstrate the anagen-catagen transition in IVHF by protein expression.However, technological advances that allow culturing for months or years will be needed to confirm the link to the full hair cycle at the in vivo level.Once the IVHFs were injected into skin explants, hair tissues were retrieved for histological analysis without being fully developed into new mature hairs.This could be due to several reasons.However, the most likely one is the failure to keep the skin explant alive for several days, limiting the integration of the engineered implants and their development into the host tissue.For a better proof of concept, further studies with animal models or preclinical trials are needed to fully demonstrate the regenerative medicine potential of the IVHF made with cells isolated under SCOC.This can eventually lead to sustainable and autologous therapies to treat hair baldness.

Conclusion
This study demonstrated that the newly developed 3D hair organoid model closely mimics the structural and functional properties of in vivo human HFs.This model is expected to be useful for large-scale screening of active molecules, especially when the mechanism of action does not rely on other cell compartments which are absent in this model.Nonetheless, the IVHF seems to be a good tool that supports studies on the molecular biology of the hair and holds big promise for autologous regenerative medicine approaches.

Figure 1 .
Figure 1.Characteristics of cells isolated and cultured from the human hair follicle.(A) Hair follicles plucked from the ex vivo scalp.(B), (C) Cells isolated from ORS and (E), (F) matrix regions were primarily cultured using SCC or SCOC.(D), (G) Cells isolated under SCOC differentiated into fiber-like structures.(H) Dermal papilla cells were isolated and (I) cultured.(J) DEGs number from three different groups comparison analysis (fold change < 2, p-value < 0.05) (K) The log2 fold change of the expression value and the p-value derived from the average comparison between the two groups are shown in the Volcano plot.SCC: standard culture conditions SCOC: stem cell optimum condition (all scale bars: 200 µm).