A Guide to Studying Human Hair Follicle Cycling In Vivo.

Introduction

Limitations of the murine hair follicle model
Human and murine hair follicles (HFs) share the same essential features of organization and function, and basic hair research in the mouse has long been both the foundation and at the forefront of our understanding of hair biology (Dry, 1926; Hsu et al., 2014; Montagna and Ellis, 1958; Plikus and Chuong, 2014; Schneider et al., 2009; Sundberg et al., 2005). In both species, HFs contain the same principal cell types and undergo repetitive cycling, alternating between phases of active growth (anagen), regression (catagen), and relative “quiescence” (telogen) (Geyfman et al., 2014; Paus and Cotsarelis, 1999; Schneider et al., 2009).
However, significant interspecies differences exist, limiting the translational potential of the murine HF model. Critically, anagen in the human scalp lasts for several years, whereas murine dorsal skin anagen is only 2–3 weeks long (Garza et al., 2012; Halloy et al., 2000; Müller-Röver et al., 2001), and epithelial HF stem cells differ in their markers and characteristics (Cotsarelis, 2006; Kloepper et al., 2008; Purba et al., 2014). Furthermore, while murine pelage HFs synchronize their cycles and grow in coordinated domains (Plikus et al., 2011; Plikus et al., 2008), human scalp HFs cycle asynchronously (mosaic, stochastically-driven hair cycle) (Dawber, 1997; Halloy et al., 2000) (see also Supplementary Text S1).
Although both human and murine HFs are exquisitely responsive to hormonal stimulation, their responses differ. For example, while estrogens and prolactin inhibit murine HF growth and cycling, both hormones prolong anagen duration in human female temporofrontal scalp HFs (Langan et al., 2010; Ohnemus et al., 2006). Thus, the response of murine HFs to stimulation with candidate hair growth-modulating agents does not necessarily predict how human HFs will respond, and may actually be misleading. Finally, the characteristic phenomenon of androgen-dependent HF miniaturization, seen in androgenetic alopecia (Dawber, 1997; Lattanand and Johnson, 1975), is not reproducible in currently available mouse strains (Crabtree et al., 2010; Nakamura et al., 2013; Sundberg et al., 1999).

The clinical importance of standardized human hair cycle staging
Considering that scalp skin harbors ca. 100,000 terminal HFs, even minor variations in their cycling have major clinical effects (Dawber, 1997). Thus, a small increase in the percentage of telogen scalp HFs by just a few percent can cause substantial effluvium, e.g. due to premature catagen induction by hormones, inflammatory mediators, neuropeptides, autoimmune reactions, cytotoxic drugs, psychoemotional stress, or malnutrition (reviewed in Atanaskova Mesinkovska and Bergfeld, 2013; Dawber, 1997; Paus, 2006; Paus and Cotsarelis, 1999; Paus and Foitzik, 2004; Paus et al., 2013; Shapiro, 2007). Moreover, establishing an accurate anagen-to-catagen-to-telogen HF ratio is important for diagnosing the kind of alopecia at hand and for assessing its severity and progression. While the telogen-to-anagen ratio can be determined non-invasively via a phototrichogram, skin biopsies and histological staging are required to identify catagen HFs and to distinguish defined anagen sub-stages (Van Neste, 2002). Additionally, accurate histological hair cycle stage assessment is essential for quantitative preclinical and clinical hair research.
Therefore, an easy-to-follow, objective guide for the precise, standardized, and reproducible identification of human HF cycle stages is needed, ideally on the basis of routine histochemistry alone, without having to examine stage-specific molecular markers by immunohistochemistry, unless the latter provides crucial, otherwise unobtainable insights. While a comprehensive guide for murine hair cycle staging has long been available (Müller-Röver et al., 2001), the only major review on the human hair cycle dates back to 1959 (Kligman, 1959), yet it provides insufficient detail to guide accurate hair cycle staging. While this review has since been complemented by excellent atlases (e.g., Sperling et al., 2012; Whiting, 2004), and by a guide for evaluating the anagen-catagen transition of microdissected, organ-cultured human HFs ex vivo (Kloepper et al., 2010), a standardized, comprehensive, user-friendly, and electronically accessible human hair cycle guide in vivo is missing. The current study strives to provide this.

Standardized assessment of human HF cycling in the xenograft mouse model
HF xenotransplantation is currently the only preclinical assay that permits complete human HF cycling and supports long-lasting human anagen studies in vivo and is therefore a uniquely instructive and indispensable human hair research tool. However, despite several early reports (De Brouwer et al., 1997; Gilhar et al., 1988; Gilhar et al., 1998; Hashimoto et al., 2000, 2001; Jahoda et al., 1996; Krajcik et al., 2003; Lyle et al., 1999; Tang et al., 2002; Van Neste et al., 1989), and more recent uses for the experimental induction of alopecia areata (Gilhar et al., 2013), post-grafting human scalp hair cycle dynamics remain poorly characterized, hindering broader adaptation of this model. Furthermore, as xenografting is inevitably associated with surgery-, wound healing-, reinnervation-, and reperfusion-related phenomena that are absent during normal scalp HF cycling in vivo (see below), a detailed morphological comparison between xenografted and freshly biopsied human scalp HFs is needed. Because such a comparison has previously been unavailable, there is limited understanding of the extent to which human hair cycle events seen in host mice are representative of normal human hair cycle progression in vivo.
Therefore, this human hair cycle guide is complemented with a systematic analysis of HF cycling in xenografted human scalp skin, noting major similarities alongside minor differences and specific transplantation-related phenomena that one needs to be aware of. Finally, we report statistically validated, practical recommendations for designing human-on-mouse HF xenotransplantation experiments.

Results

Human hair cycle staging
HF cycle stages were evaluated based on the following histological characteristics (Supplementary Table S1): (i) size and shape of the dermal papilla (DP) and hair matrix, (ii) epithelial outer root sheath (ORS) morphology, (iii) connective sheath and vitreous membrane morphology, (iv) hair shaft characteristics, such as length and the presence of club, (v) the presence of the inner root sheath (IRS), (vi) pigment distribution, and (vii) the presence of apoptotic and/or proliferating cells, following the example of murine hair cycle staging (Müller-Röver et al., 2001). Additional markers can be assessed immunohistologically to demarcate selected cell populations or structures, such as epithelial stem cells or HF-associated keratins, but are dispensable for hair cycle staging (Supplementary Table S2).
Similar to routine hair transplantation in humans (Unger, 2005) or in chemotherapy-induced alopecia (Paus et al., 2013), xenografted anagen HFs (HFs-XG) predominantly enter catagen, thereby inducing a new hair cycle and allowing for quick recovery from surgery-associated damages. While these HFs-XG often shows signs of dystrophy (“dystrophic catagen”), less damaged HFs-XG enter into the “dystrophic anagen” damage-response pathway, with retarded progression into a new hair cycle (Paus et al., 2013). In the following sections, we first describe HF morphology in human scalp skin in situ (HF-IS) and subsequently explain the extent to which the hair cycle stages of HFs-XG recapitulate HFs-IS. Importantly, when staging HFs-IS, HF size and position relative to neighboring follicles and to epidermal/dermal or dermal/adipose tissue boundaries can be used as morphological landmarks. However, these landmarks cannot be recruited for hair cycle staging of HFs-XG.

Early catagen
This guide covers catagen first because after human HFs have completed their fetal morphogenesis (Montagna and Ellis, 1958), their life-long cycling activity begins with the first catagen entry in utero. For practical reasons, the eight distinct stages of catagen development in mice (Müller-Röver et al., 2001) are best subdivided into three, relatively easily recognizable stages (Kloepper et al., 2010): early catagen, equivalent to murine catagen phases I–IV; mid catagen (i.e. murine catagen V–VI); and late catagen (i.e. murine catagen VII–VIII) (Müller-Röver et al., 2001).
In HFs-IS, matrix and DP volume reduction, together with a complete cessation of HF pigmentation, are the earliest signs of catagen development that can be positively distinguished from anagen stage VI. Characteristically, the DP becomes more condensed and almond-shaped. Termination of melanogenesis (Bodo et al., 2007; Slominski et al., 2005; Tobin, 2011) results in the proximal end of the hair shaft becoming notably less pigmented than in anagen VI HFs (Figure 1a). Some melanin incontinence into the DP can also be seen, as the normal transfer of melanosomes into precortical hair matrix keratinocytes is interrupted (Tobin, 2011) (Figure 1b, feature #5). Importantly, morphology of the bulge region and the overall follicle length remain largely unchanged compared to anagen VI HFs-IS, and the lower HF portion rests below the dermal/adipose junction. Positive staining for apoptotic cells (e.g. by caspase-3 or TUNEL immunofluorescence) in the regressing epithelium above the DP can be used as a definitive immunohistological marker of early catagen, since apoptotic cells are essentially undetectable in healthy anagen VI HFs in vivo (Botchkareva et al., 2006; Botchkareva et al., 2007; Sharova et al., 2014). Furthermore, downregulation of IRS and DP immunohistological markers can be used to differentiate early catagen HFs from anagen VI HFs (see Supplementary Table S2) (Commo and Bernard, 1997; Malgouries et al., 2008a; Malgouries et al., 2008b).
In HFs-XG, anagen VI progresses into catagen unusually rapidly so that on day two post-grafting, follicles that closely correspond to murine catagen stage IV can already be found (Müller-Röver et al., 2001) (Figure 1d–h). In catagen HFs-XG, the matrix is reduced down to just two-three cell layers, yet still envelops a small, almond-shaped DP (Figure 1f, 1g, feature #2). The newly forming club hair is located a short distance above the condensed DP (Figure 1f, 1g, feature #6). A significant portion (76.4%) of HFs-XG undergo “dystrophic catagen” (Paus et al., 2013), during which a normal, serrated club hair shaft fails to form, and the regressing hair matrix above the DP commonly contains ectopic melanin deposits (Supplementary Figure S1a–e, S3).

Mid-catagen
In HFs-IS and HFs-XG, the matrix and DP further decrease in volume – residual matrix is only 1–2 cell layers thick and only partially wraps around the condensed, almond-shaped DP (Figure 1i, features #1, 2). A brush-like club hair becomes prominent at this stage, and it resides above the dermal/adipose boundary (Figure 1i, 1p, feature #4). The newly formed epithelial strand (the remnant of the regressing hair matrix and proximal ORS) between the club hair and the DP is thin, generally lacks pigment, and can have a ruffled, zipper-like appearance (Figure 1i, 1o, 1p, feature #3). Compared to early catagen, mid-catagen HFs acquire visible thickening of the vitreous membrane of the connective sheath, which prominently stains for the glycoprotein, biglycan (Figure 1i, 1o, 1q, feature #5). Because the IRS regresses and disappears during catagen, its absence can be used to differentiate mid- to late catagen HFs from early anagen III HFs upon H&E (Commo and Bernard, 1997). Dystrophic mid-catagen HFs-XG either lack or have incompletely formed club hairs, and melanin clumps and vitreous membrane thickening are prominent (Supplementary Figure S1f–j, S3).

Late catagen
In both HFs-IS and HFs-XG, the matrix disappears, and the DP becomes condensed and ball-shaped (Figure 1r, 1y, feature #1). The club hair is now prominently visible (Figure 1r, 1z, feature #4), and the epithelial strand has shortened (about half the length of that of mid-catagen HFs) (Figure 1r, 1x, 1y, feature #3). The thickened connective sheath, which characteristically trails below the DP into the adipose tissue and can contain melanin clumps (in HFs-XG), becomes prominent at this stage (“dermal streamer”) (Figure 1r, feature #5). A few apoptotic cells can still be detected in the epithelial strand (Figure 1t, feature #6). Importantly, in late catagen, apoptotic cells can also be found in the shrinking sebaceous gland (Figure 1t, feature #7), like in mice (Lindner et al., 1997). Dystrophic late catagen HFs-XG display ectopic melanin deposition in the epithelium (Supplementary Figure S1k–o, S3) and prominent pleats in the bulge region, which co-localize with CD200-positive epithelial progenitors (Figure 1x, 1z, feature #8).

Telogen
HFs with typical telogen morphology can be seen in situ, but are generally absent in xenografts. Their defining characteristics are: (i) positioning of the HF entirely above the dermal/adipose boundary (Figure 2a, feature #4), (ii) prominent unpigmented, serrated club hair (Figure 2a, feature #3), (iii) very compact, well-rounded DP separated from the club hair by a maximally shortened, unpigmented epithelial strand, the “secondary hair germ” (SHG) (Figure 2a, features #1, 2). Apoptotic cells are generally lacking (Figure 2c, feature #5).
However, consistent with previous reports (reviewed in Geyfman et al., 2014), a few dispersed (not clustered) proliferating cells can often be seen in the SHG and the distal epithelium of telogen HFs-IS (Figure 2c, feature #6). Thus, telogen HFs are not really “resting”; unfortunately, the functionally crucial distinction between “refractory” and “permissive” telogen HFs is not possible by histology, and the corresponding molecular signatures have only been characterized for murine telogen (see Geyfman et al., 2014). Importantly, human telogen HFs-IS can undergo exogen, the phase of active club hair shedding (Higgins et al., 2009; Stenn, 2005). Following exogen, HFs-IS enter kenogen, the telogen phase without club hair (Rebora and Guarrera, 2002), which can last for several months (Courtois et al., 1994).

Anagen I
Due to their relatively short duration, early stages of anagen can be quite difficult to identify in situ. One also needs to keep in mind that hair cycle staging describes a continuous and dynamic morphogenetic process in a discontinuous manner (only anagen VI and telogen are relatively stable stages; for detailed discussion see Bernard (2012)). Unlike in situ, anagen I is relatively common in xenografts, making this the model of choice for investigating the human telogen-anagen transformation. Anagen I HFs-XG display a hybrid morphology: (i) similar to late catagen, their bulge region's epithelium retains a pleated appearance (Figure 2g, feature #8); (ii) the SHG becomes triangular or crescent-shaped and wraps around the DP (Figure 2g–i, feature #7), which remains condensed and ball-like, may contain melanin clumps (i.e., pigment residue from the preceding anagen VI stage) (Figure 2i, feature #9), and still shows a trailing connective sheath (Figure 2h, feature #1).

Anagen II
In HFs-IS and HFs-XG, the SHG undergoes proliferation-driven thickening and elongation (Figure 2k–t). Its proximal end develops into a new hair matrix, which at this stage is still unpigmented, crescent-shaped, and only partially encloses a small, yet slightly larger, less densely packed, ball-shaped DP (Figure 2k, 2q–s, features #1, 2). Proliferation markers reveal localized clusters of proliferating cells in the thickening hair germ (Figure 2m, feature #4), while apoptotic cells are lacking. The entire length of stage II anagen HFs-IS resides above the dermal/adipose boundary. In HFs-XG, the bulge region's epithelium retains its pleated appearance (Figure 2r, 2t, feature #6), and the DP still contains melanin deposits (Figure 2s, feature #5).

Anagen III
In both HFs-IS and HFs-XG, the hair matrix has now formed and is 4–5 cell layers thick. It encloses at least 60% of the DP, which becomes enlarged and oval-shaped (Figure 3a, 3g–k, feature #2). Prominently, at this stage, HFs develop a hair shaft and IRS, both of which are easily identifiable in routine H&E stains (Figure 3a, 3b, 3g, 3i, 3k, feature #3). Immunostaining for proliferation markers reveals actively dividing cells both in the hair matrix and in the ORS (Figure 3c, 3j, feature #5). In situ, the hair bulb now reaches and extends into the adipose layer. In both HFs-IS and HFs-XG, three anagen III sub-stages can be differentiated based on hair shaft appearance. Anagen IIIa shafts lack a visible cortex (Figure 3g). Anagen IIIb and IIIc shafts have a visible cortex, while anagen IIIc hair shafts are long, reaching approximately twice the length of the hair matrix (Figure 3k). Importantly, throughout anagen III, hair shafts still lack visible pigmentation, even though HF melanogenesis in the HF pigmentary unit commences in anagen IIIc (Slominski et al., 2005). Lastly, the bulge epithelium of HFs-XG retains a pleated appearance (Figure 3i, 3k, feature #6), and the DP still contains occasional melanin clumps.

Anagen IV
At this stage, the hair shaft is fully mature, with a distinct medulla (in terminal HFs), cortex, and cuticle easily identifiable on H&E, and the hair tip reaches the level of the sebaceous gland duct (Figure 4a–h, feature #5). Importantly, melanin production and transfer are now fully reactivated, and hair shafts become visibly pigmented. In situ, the hair bulb now reaches down to the upper dermal adipose layer (Figure 4a, feature #2), and a distinct connective sheath trail is visible proximal to the bulb, which guides further HF downgrowth (Figure 4a, feature #7).

Anagen V
In situ, the hair bulb extends further into the adipose layer, and the connective sheath trail disappears at this stage (Figure 4i, feature #2). In both HFs-IS and HFs-XG, the tip of the hair shaft enters the hair canal (Figure 4i, 4j, 4k, feature #5). The DP is now onion-shaped, and in the hair matrix, pigmentation reaches down to Auber's line (Figure 4l, 4m, features #1, 2, 6). Additionally, in HFs-XG, bulge epithelium contours begin to smoothen (Figure 4l, feature #8).

Anagen VI
The vast majority of HFs in situ are in anagen stage VI. The hair bulb is located deep in the dermal adipose layer, while the hair shaft emerges above the skin level (Figure 4p–w). In pigmented HFs, the hair matrix contains the maximum amount of melanin, which now reaches below Auber's line. In HFs-XG, bulge epithelium smoothens, but residual undulations, which can be homologous to the “follicular trochanter” in HFs-IS (Tiede et al., 2007), can persist (Figure 4u, feature #8). Compared to anagen V, the DP is maximally enriched in extracellular matrix.

Practical recommendations for the xenograft model
Long-term survival of individually grafted human scalp HFs is much more consistent in SCID mice, averaging between 55–67% (Supplementary Figure S6d). In nude mice, it was extremely variable, ranging from 0% to 82%, likely reflecting mouse-to-mouse variability in graft rejection (Supplementary Figure S4). Also, among actively cycling HFs-XG, average hair growth rates are faster and more consistent in SCID than in nude mice (Supplementary Figure S6c). This confirms that SCID mice are the host of choice for xenografted human scalp HFs (Gilhar et al., 2013; Gilhar et al., 1998).
Xenograft transplantation provides a strong stimulus for catagen induction, thereby partially synchronizing hair cycling behavior (Figure 5). However, significant hair cycle stage heterogeneity is retained during all post-grafting time points (Figure 5b, 5c, 5d), demonstrating that the mosaicism of human HF cycling is partially maintained even after transplantation. We recommend using statistically adjusted peak time points generated here (see Figure 5e; Supplementary Figure S5) to evaluate the post-grafting human hair cycle. Moreover, because the majority of HFs-XG enter anagen stage VI on day 92, studies on anagen should be performed after this time point.

Discussion
Here, we provide a guide for staging terminal human scalp HFs in situ and in xenografts (Supplementary Figure S2) on the basis of a minimal set of characteristics, identifiable on routine histology. Depending on the specific hair research question(s) asked, additional standard read-out parameters can be employed that make the analysis of human HFs even more instructive, and Supplementary Table S2 lists selected examples for further guidance (Purba et al., 2014; Purba et al., 2015).
The mouse xenotransplant model remains indispensable for studying and experimentally manipulating human HF cycling in vivo. Besides follicular unit transplantation, as in the current study, one can also transplant carefully trimmed full-thickness scalp skin (Gilhar et al., 2013; Gilhar et al., 1998; Sintov et al., 2000; Van Neste et al., 1989). This greatly reduces the level of surgery-related damage suffered by HFs located away from the transplant edge and has the added advantage of permitting one to study the cycling behavior of an entire HF field as well as terminal HFs alongside vellus HFs, complete with associated sebaceous and sweat glands. However, perfusion, oxygenation, and re-innervation can be precarious in the center of such full-thickness transplants.
When interpreting data obtained with the xenotransplant model, one must keep in mind a number of confounding factors that may influence the results profoundly. Namely, xenotransplanted human HFs are re-perfused and re-innervated by cells and structures derived from an alien host and are shock-exposed to and must rapidly adjust to the foreign endocrine, innate immune, and metabolic system of SCID mice. In addition, the murine host launches a stress response to the trauma of surgery (note that perceived stress in mice triggers substantial perifollicular neurogenic inflammation, which is NGF-, substance P- and mast cell-dependent, centers around the bulge, and prematurely induces catagen in murine anagen HFs (Arck et al., 2005)). Coupled with the fact that human scalp HFs also respond to key stress-mediators (reviewed in Paus et al., 2014), all of these confounding factors are expected to impact greatly on human HF cycling, growth, immune status, pigmentation, and metabolism in vivo after xenotransplantation. Therefore, caution is advised in extrapolating from observations made with human HF xenotransplants in mice to the response of healthy human scalp skin.
At any given time, the vast majority of asynchronously cycling HFs in healthy human scalp are considered to be in anagen (80–90%), between 10–20% in telogen, and only 1–5 % in catagen (Dawber, 1997; Shapiro, 2007; Sperling et al., 2012; Whiting, 2004). However, our current histological analysis of HFs-IS suggests that the number of catagen HFs can exceed that of telogen HFs (catagen: 5–10%, telogen: 1–2%). This discrepancy likely reflects differences in assessment methodologies, since phototrichograms cannot distinguish between telogen and catagen and are thus less accurate compared to histology-based hair cycle staging (Hoffmann, 2001; Van Neste and Trueb, 2006). Additional histomorphometric hair cycle staging will be required to refine the true anagen:catagen:telogen scalp HF ratio. Due to the relatively short duration of anagen I to V, these anagen stages are rarely found in situ, with the notable exception of the weeks following extensive telogen effluvium, when a surge in premature anagen termination is followed by semi-synchronous anagen reactivation (Hadshiew et al., 2004; Harrison and Sinclair, 2002; Katz et al., 2006). Thus, an unusually high percentage of anagen stage I to V HFs points towards a preceding telogen effluvium.
Unlike in situ, anagen I-V HFs can be readily identified in xenografts due to a telogen effluvium-like resetting effect from the traumatic transplantation procedure (Gilhar et al., 1988; Hashimoto et al., 2000, 2001; Jahoda et al., 1996; Van Neste et al., 1989), complicated by various degrees of HF dystrophy, just as after chemotherapy (Paus et al., 1994; Paus et al., 2013). This resetting, however, is incomplete. While individual xenotransplanted anagen HFs rapidly enter catagen by day 3, their progression through catagen is variable, and late catagen HFs can still be found on day 50. This likely reflects variable response to trauma, when some HFs enter into normal, but premature catagen or the “dystrophic catagen”, while others undergo a “dystrophic anagen”, which protracts catagen development (Hendrix et al., 2005; Paus et al., 2013). This variable timing of the catagen program leads to incomplete hair cycle synchronization, heralding the reestablishment of cycling mosaicism. Additionally, grafted HFs do not appear to enter long-lasting telogen, suggesting that the normal HF stem cell quiescence mechanisms (Geyfman et al., 2014; Mardaryev et al., 2011) may be altered, perhaps as a result of the confounding, host-derived factors summarized above.

SCID mouse xenotransplantation model optimization for studying human anagen
Despite limitations of the xenograft model, HFs-XG closely resemble cycling HFs-IS and are able to enter long-lasting anagen. Therefore, the SCID mouse xenograft model (see also Gilhar et al., 2013; Gilhar et al., 1998) provides an extremely valuable experimental system for investigating multiple, otherwise difficult-to-study aspects of human HF biology, and instructively complements in vitro human HF and scalp skin organ culture (Al-Nuaimi et al., 2014; Hardman et al., 2015; Kloepper et al., 2010; Lu et al., 2007; Oh et al., 2013; Philpott et al., 1990; Poeggeler et al., 2010). We recommend using at least three post-grafting time points to study catagen-to-anagen progression, and waiting until after post-grafting 92 days for studying anagen VI HFs. This is substantially later than the post-grafting days 60–70 reported previously (Hashimoto et al., 2000, 2001). Future studies wishing to investigate human HF responses to hormonal stimulation, e.g. in the context of androgenetic alopecia, also need to consider the intricate hormone-sensitivity of human HFs (Paus et al., 2014) and their keratin expression patterns (Ramot and Paus, 2014); therefore, imitating donor-like hormone levels in host mice (e.g. testosterone) is important (De Brouwer et al., 1997; Krajcik et al., 2003; Sintov et al., 2000; Van Neste et al., 1991).
In summary, while Mus musculus remains unrivaled in the insights it has helped to generate into basic HF biology, murine HF physiology is quite different from that of human HF. The xenotransplant model characterized above provides an indispensable tool for human preclinical hair research in vivo, if employed together with the comprehensive guide for human hair cycle staging developed here.

Materials and Methods

Human scalp hair follicles and xenografting
Institutional approval and written informed patient consent were received for all studies using human tissue samples, and institutional approval was received for all animal studies. Human scalp skin in situ studies were performed on normal occipital and temporal scalp skin samples following previously published protocols (Harries et al., 2013; Harries and Paus, 2010; Kloepper et al., 2010). For xenografting, non-balding occipital scalp skin specimens were used. The method for human HF xenografting was adapted after Hashimoto et al. (2000). Briefly, 15 to 40 (on average 25) microdissected anagen VI follicular units were transplanted onto 6–8 weeks old female nude or SCID mice (Jackson Laboratory, Bar Harbor, Maine, USA). A total of 1,164 HFs were transplanted and then biopsied and analyzed at 45 consecutive time points (see Supplementary Table S3 and Supplementary Materials and Methods).

Histological tissue analysis
Paraffin embedded HF samples were sectioned at 3um thickness, and O.C.T compound embedded follicles were sectioned at 8um under −20°C. Sections were processed either for routine histology (H&E staining) or for immunofluorescence staining (see Supplementary Materials and Methods).

Computational analysis and statistical tests
Hair cycle stage's mean date was determined by averaging the time points when biopsied HFs were at the corresponding stage. To estimate the time point with the greatest probability of selecting a HF in the desired stage, the naïve Bayes classifier (Mitchell, 1997) was used. Additional computer simulations were employed to derive probability values for each hair cycle stage, and two-sample Kolmogorov-Smirnov (K-S) test (Conover, 1999) was utilized to compare the speed of HF-XG hair cycle progression between nude and SCID host mice (see Supplementary Materials and Methods for details).

Supplementary Material
01