Axin2 marks quiescent hair follicle bulge stemcells that are maintained by autocrineWnt/β-catenin signalingXinhong Lima,b,c,d,e,1, Si Hui Tana,c,f, Ka Lou Yua, Sophia Beng Hui Lima, and Roeland Nussec,d,1

aInstitute of Medical Biology, Agency for Science, Technology, and Research, Singapore 138648; bLee Kong Chian School of Medicine, NanyangTechnological University, Singapore 639798; cHoward Hughes Medical Institute, Stanford University School of Medicine, Stanford, CA 94305-5329;dDepartment of Developmental Biology, Stanford University School of Medicine, Stanford, CA 94305-5329; eProgram in Cancer and Stem Cell Biology, Duke-NUS Graduate Medical School, Singapore 169857; and fCancer Biology Program, Stanford University School of Medicine, Stanford, CA 94305-5329

Contributed by Roeland Nusse, January 28, 2016 (sent for review December 16, 2015; reviewed by Salvador Aznar-Benitah and Cheng-Ming Chuong)

How stem cells maintain their identity and potency as tissueschange during growth is not well understood. In mammalian hair,it is unclear how hair follicle stem cells can enter an extendedperiod of quiescence during the resting phase but retain stem cellpotential and be subsequently activated for growth. Here, we uselineage tracing and gene expression mapping to show that theWnt target gene Axin2 is constantly expressed throughout thehair cycle quiescent phase in outer bulge stem cells that producetheir own Wnt signals. Ablating Wnt signaling in the bulge cellscauses them to lose their stem cell potency to contribute to hairgrowth and undergo premature differentiation instead. Bulge cellsexpress secreted Wnt inhibitors, including Dickkopf (Dkk) and se-creted frizzled-related protein 1 (Sfrp1). However, the Dickkopf 3(Dkk3) protein becomes localized to the Wnt-inactive inner bulgethat contains differentiated cells. We find that Axin2 expressionremains confined to the outer bulge, whereas Dkk3 continues tobe localized to the inner bulge during the hair cycle growth phase.Our data suggest that autocrine Wnt signaling in the outer bulgemaintains stem cell potency throughout hair cycle quiescence andgrowth, whereas paracrine Wnt inhibition of inner bulge cellsreinforces differentiation.

Wnt signaling | hair follicle | stem cells | homeostasis | regeneration

The hair follicle is a complex miniorgan that repeatedly cyclesthrough stages of rest (telogen), growth (anagen), and de-

struction (catagen) throughout life (1). During anagen, growinghair follicles emerge adjacent to the old telogen hair follicles thatremain there throughout the cycle and create an epithelial pro-trusion known as the “bulge.” At the end of the hair cycle, incatagen, cells from the follicle migrate along the retracting epi-thelial strand and join the two epithelial layers of the telogenbulge—the inner and outer bulge layers—surrounding the clubhair shaft (2).Several studies have established that stem cells residing in the

outer bulge are the source of the regenerative capacity of thecycling hair follicle (3–5). During telogen, these stem cells arethought to be generally quiescent (6). In response to signals fromtheir microenvironment during anagen, the stem cells divide andproduce proliferative progeny that participate in the growth ofthe new follicle (7). Some of these activated stem cells and theirprogeny are believed to migrate away from the bulge, but aresubsequently able to rejoin it after anagen is complete (2, 5).Cells that return to the outer bulge take on a follicular stem cellidentity, ready to divide and participate in the next hair cycle (2,8). Conversely, cells returning to the inner bulge do not divideand, instead, form an inner bulge niche of differentiated cells forthe outer bulge cells (2). Stem cells remain quiescent duringtelogen for an extended period, and the identity of signals thatmaintain stem cell identity during this time are poorly understood.In the hair, Wnt/β-catenin signaling is required right from the

earliest stages of development, for the initiation of hair placode

formation (9). Wnt signals are needed later during postnatalhomeostasis as well, for the initiation of anagen in postnatal hair(10). Therefore, in view of their well-established importance forstem cell maintenance in multiple adult tissues, including theskin (11), Wnts are candidate hair follicle stem cell (HFSC)-maintaining signals. However, Wnt signaling is generally be-lieved to be inactive in the telogen bulge (8, 10, 12), which isthought to be quiescent. Wnt signaling becomes strongly elevatedwhen bulge cells are “activated” to undergo the transition fromtelogen to anagen (13, 14). During anagen, Wnt signaling hasbeen described to primarily specify differentiated cell fates in theanagen follicle (12, 15). As anagen proceeds and the follicleenters catagen and telogen again, the bulge is thought to revertto a Wnt-inhibited state (12, 13, 16, 17).Conversely, there is evidence for a functional requirement of

Wnt/β-catenin signaling in the bulge other than initiating anagenand specifying differentiation during anagen. For instance, post-natal deletion of β-catenin in outer bulge cells results in the loss oflabel-retention and HFSC markers, suggesting that β-catenin isrequired for maintenance of HFSC identity (10).Here, beyond its role in hair differentiation and anagen ini-

tiation, we sought to determine whether Wnt/β-catenin signalingis also involved in HFSC maintenance during telogen. We foundthat Axin2 expression persists in HFSCs in the outer bulgethroughout telogen and anagen, suggesting that active Wnt

Significance

Hair follicle stem cells (HFSCs) remain quiescent for long pe-riods of time during the resting phase of the hair cycle. Howthey maintain their stemness and identity during quiescencewhile being responsive to growth-inducing cues remains poorlyunderstood. Here, we identify Axin2 as a previously unidentifiedmarker of HFSCs and use it to show that quiescent HFSCs un-dergo and require active Wnt/β-catenin signaling. By mappingWnt and its inhibitors with high sensitivity, we show that HFSCssecrete their own self-renewing Wnt signals and inhibitors thatpromote differentiation outside of the stem cell compartment.Our findings suggest that careful modulation of Wnt signalingmay be important for the derivation and maintenance of HFSCsfor alopecia treatment and drug screens.

Author contributions: X.L. and R.N. designed research; X.L., S.H.T., K.L.Y., and S.B.H.L.performed research; X.L. and S.H.T. contributed new reagents/analytic tools; X.L., S.H.T.,K.L.Y., S.B.H.L., and R.N. analyzed data; and X.L., S.H.T., K.L.Y., and R.N. wrote the paper.

Reviewers: S.A.-B., Institute for Research in Biomedicine, Barcelona; and C.-M.C., Univer-sity of Southern California.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. Email: xinhong.lim@imb.a-star.edu.sg orrnusse@stanford.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1601599113/-/DCSupplemental.

E1498–E1505 | PNAS | Published online February 22, 2016 www.pnas.org/cgi/doi/10.1073/pnas.1601599113

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

0

signaling is a consistent feature of bulge stem cells. Furthermore,these hair outer bulge stem cells produce autocrine Wnts andparacrine-acting Wnt inhibitors that may specify the positionalidentity of cells residing within the bulge niche.

ResultsTo determine whether Wnt/β-catenin signaling is active duringthe telogen stage, we examined telogen follicles for the expres-sion of Axin2, a well-established Wnt/β-catenin target gene (18,19), using RNA in situ hybridization. We found that Axin2 wasexpressed mostly in telogen outer bulge cells (Fig. 1A). Similarly,when we examined telogen hair from Axin2–lacZ Wnt reportermice, we observed clear reporter activity, specifically in the outerbulge (Fig. 1B). To determine whether Wnt signaling activityvaries during telogen, we looked at Axin2 mRNA expressionduring early [postnatal day 43 (P43)], mid (P56), and late (P69)telogen using RNA in situ hybridization. We found that Axin2mRNA is expressed in the bulge throughout telogen (Fig. S1A),suggesting that Wnt signaling persists through all of thetelogen phases.Because HFSCs are known to reside in the outer bulge (3, 5,

20), we asked whether Axin2-expressing telogen bulge cells havestem cell properties. We permanently labeled Axin2-expressingcells in the telogen bulge by exposing Axin2–CreERT2 Rosa26–mTmGflox mice (21, 22) to tamoxifen (Tam) during early second

telogen (P49) and then tracked the fate of these cells and theirprogeny for various lengths of time (Fig. 1C). Consistent with ourobservations described above (Fig. 1 A and B), Axin2–CreERT2-labeled cells in the bulge persisted from early to mid-telogen(Fig. 1D). As the hair follicles entered anagen, the labeled cellsappeared to accumulate in the growing secondary hair germ andmigrate along the elongating outer root sheath and into thedeveloping matrix, where they lay adjacent to the dermal papillaand generated progeny spanning all of the layers of the anagenhair follicle (Fig. 1E). These cells were able to participate insuccessive cycles of hair follicle growth up to a year after theinitial labeling (Fig. 1F). We observed the same patterns of Axin2expression and long-term, self-renewing potential of outer bulgecells labeled during the first telogen (P21) occurring immediatelyafter morphogenesis (Fig. S1 B–K), suggesting that Axin2-expressing outer bulge cells are present during successive telogencycles of hair growth. These data suggest that Axin2–CreERT2labels long-lived HFSCs in the outer bulge.Hair bulge cells are thought to be quiescent during early sec-

ond telogen (23). To determine whether Axin2-expressing bulgecells were indeed quiescent cells, we performed cell-cycle analysison Axin2-expressing bulge cells harvested by flow cytometry fromAxin2–lacZ reporter mice. Almost all of these cells exhibited aG0/G1 nuclear profile (>90%; Fig. 1G), suggesting that Axin2-expressing telogen bulge cells are quiescent.

CA B

P56 P414

D

F

P49

Chaseperiod

P88

E

Taminjection

D EP56

FP414 P414P88 P88 P414P88

Axin2-lacZAxin2 mRNA

G

Fig. 1. Axin2-expressing cells in the telogen bulge are quiescent HFSCs. (A and B) Axin2 mRNA (A) and Axin2–lacZ (B) expression in the early telogen hairbulge (P43). Dotted lines denote approximate outer boundary of the hair bulge. [Scale bars: 10 μm (Axin2mRNA); 20 μm (Axin2lacZ).] (C) Schematic depictingpulse-chase experiment to track the fate of Axin2+ cells and their progeny. (D–F) Histological sections of hair follicles from Axin2–CreERT2/Rosa26–mTmGflox

mice chased from P49 for 1 wk (P56, mid-telogen; D), 39 d (P88, anagen; E), and 1 y (P414; F). Dotted lines denote approximate outer boundary of the hairbulge. (Scale bars: 10 μm.) (G) Cell-cycle analysis of Axin2+ bulge cells by propidium iodide staining using flow cytometry.

Lim et al. PNAS | Published online February 22, 2016 | E1499

DEV

ELOPM

ENTA

LBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

0

To test whether Axin2 is indeed a Wnt/β-catenin signalingtarget gene in the hair bulge, we conditionally inactivated theβ-catenin gene in Axin2-expressing cells during telogen by crossingAxin2–CreERT2 mice with β-catΔex2-6 floxed mice. Axin2mRNA expression that was present in heterozygous Axin2–CreERT2/β-catΔex2-6-fl/+ control hair bulge cells was abrogated inAxin2–CreERT2/β-catΔex2-6-fl/fl mutant bulge cells (Fig. 2 A andB) cells. This result was consistent with the observation thatAxin2 expression is also greatly decreased when β-catenin isdeleted in Keratin-15–expressing bulge cells (Fig. S2B) (14).Thus, Axin2 expression in the outer bulge cells is dependenton the presence of β-catenin. Furthermore, other Wnt targetgenes like Lgr5 are also expressed in the outer bulge (Fig. S2A)(5) and similarly require β-catenin for their expression (Fig. S2B)(14). The expression of multiple Wnt target genes suggests thatWnt/β-catenin signaling is active in the telogen bulge and thatAxin2 expression is a reliable indicator of Wnt signaling activity inhair follicles.The ability to produce differentiated cells is a defining prop-

erty of stem cells. We next asked whether Axin2-expressing HFSCsrequire Wnt/β-catenin signaling to function as stem cells by pro-ducing differentiated hair lineages during the anagen stage ofhair growth. In heterozygous Axin2–CreERT2/β-catΔex2-6-fl/+ con-trol littermates, hair follicles entered anagen normally (Fig. 2C). Incontrast, Axin2–CreERT2/β-catΔex2-6-fl/fl mutant hair follicles did notgrow, and most exhibited an abnormal telogen-like morphology(Fig. 2D). This result resembles the arrested hair growth seen inanimals where β-catenin is mutated in all outer bulge cells (10, 24).The loss of hair follicle growth suggests that Axin2-expressing bulgeHFSCs require β-catenin to maintain their stem cell potency/state.To determine the source of Wnt signals for these Wnt-depen-

dent HFSCs, we screened for the expression of Wnt genes atvarious stages of the hair cycle using RNA in situ hybridization(Fig. 3 A–C). We found that Wnt1, Wnt4, and Wnt7b were ex-pressed in the bulge throughout telogen. The highest levels ofexpression appeared to be in outer bulge cells. Consistent withour Axin2 mRNA expression data, these Wnts remain expressedthroughout telogen. In comparison, we did not detect any Wnt5aexpression in the bulge throughout telogen, whereas other Wnts,includingWnt3,Wnt6,Wnt10a, andWnt10b, are expressed, but atmuch lower levels (Fig. S3). These data suggest that the outer

bulge HFSCs themselves are a source of Wnt signals in thetelogen bulge.How is Wnt gene expression in the bulge regulated? When we

looked at Wnt gene expression using RNA in situ hybridizationin Axin2–CreERT2/β-catΔex2-6-fl/fl mutant bulges, we found a lossof Wnt4 mRNA expression compared with Axin2–CreERT2/β-catΔex2-6-fl/+ control bulges (Fig. 3D). This is similar to otherreports where β-catenin deletion in keratin-15–expressing bulgecells also resulted in a great reduction in Wnt7b expression (Fig.S2B) (14). Together, these data suggest that production of atleast some Wnts is regulated by and requires autocrine Wnt/β-catenin signaling in the outer telogen bulge.Is Wnt secretion required for hair growth? We observed that

Wntless, a Wnt cargo-binding protein required for Wnt secretion,is expressed in bulge cells (Fig. 4A), consistent with previousreports (25, 26). Wntless is also expressed during hair de-velopment (27), and our lineage tracing experiments showthat embryonic Axin2-expressing cells become HFSCs (Fig.S1L). Conditional knockout of Wntless from this stage in mutantAxin2–rtTA/tetO–Cre/Wntlessfl/fl mice resulted in much less hairand thickened epidermis in several regions compared with theirheterozygous littermate controls (Fig. 4 B and C), suggesting thatWntless is required by Axin2-expressing cells for hair growth. Thisfinding is consistent with other studies where hair is lost andremaining hairs are arrested in telogen when Wntless is deletedin the entire epidermis (25–27) and also specifically in the hairbulge (24, 25). Together, these data and the literature suggest thatautocrine Wnt signaling is required by HFSCs for hair growthand maintenance.To determine whether Wnt production and β-catenin signaling

change as the hair enters the anagen growth phase, we examinedfollicles at the telogen-to-anagen transition and found thatAxin2, Wnt1, Wnt4, and Wnt7b continued to be expressed in thebulge (Fig. 3C). However, the expression of several Wnts, in-cluding Wnt6, Wnt7b,Wnt10a, and Wnt10b, became up-regulatedin cells at the boundary of the secondary hair germ and thedermal papilla (Fig. S3C), consistent with the strong expressionof Axin2 in the secondary hair germ and previous reports (7, 28).This result suggests that specific Wnts may be involved in theepithelial–mesenchymal cross-talk in secondary hair germ cellsneeded to initiate anagen, as has previously been proposed (28).

A BAxin2 mRNA Axin2 mRNA

C D

het cKO

het cKO

Fig. 2. Bulge cells functionally require β-catenin for Axin2 expression and to maintain hair-forming potency. (A and B) Representative images of RNA in situhybridization for Axin2 gene expression as performed on Axin2–CreERT2/β-cateninΔex2-6-fl/+ (het; A) and mutant Axin2–CreERT2/β-cateninΔex2-6-fl/fl (cKO; B)hair follicles. Dotted lines denote approximate outer boundary of the hair bulge. (Scale bars: 20 μm.) (C and D) Representative images of eosin-stained sectionsof control Axin2–CreERT2/β-cateninΔex2-6-fl/+ (het; C) and mutant Axin2–CreERT2/β-cateninΔex2-6-fl/fl (cKO; D) dorsal skins. (Scale bars: 20 μm.)

E1500 | www.pnas.org/cgi/doi/10.1073/pnas.1601599113 Lim et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

0

We next asked whether the bulge becomes Wnt-inhibitedduring anagen, as is widely believed (12, 13, 16, 17). Contrary toexpectations, we observed that Axin2 mRNA and the Axin2–lacZreporter gene continued to be expressed in the anagen bulge(Fig. 5 A–C and F). At the same time, we found the persistent

expression of Wnt4 and Wnt7b in the anagen bulge (Fig. 5 D andF). This finding is in contrast to other Wnt genes like Wnt5a,Wnt6, Wnt10a, and Wnt10b that are expressed elsewhere in theanagen follicle (Fig. S4C), but not in the anagen bulge (Fig. 5E).By double-labeling RNA in situ hybridization, we found that

Wnt1 mRNA Wnt4 mRNA Wnt7b mRNANegative Control

P43 P43 P43 P43

AWnt5a mRNA

P43

P57 P57 P57 P57

Negative Control

Wnt1 mRNA Wnt4 mRNA Wnt7b mRNA

B

P57

Wnt5a mRNA

C

P69 P69 P69 P69 P69

Negative Control Wnt5a mRNA Wnt1 mRNA Wnt4 mRNA Wnt7b mRNA

het

D

cKO

Wnt4 mRNA

Fig. 3. Bulge cells express Wnts throughout telogen. (A–C) Representative images of RNA in situ hybridization for Wnt gene expression as performed on hairfollicles during early (P43; A), mid (P57; B), and late (P69; C) second telogen. (D) Representative images of RNA in situ hybridization for Wnt4 gene expressionas performed on Axin2–CreERT2/β-cateninΔex2-6-fl/+ (het) and mutant Axin2–CreERT2/β-cateninΔex2-6-fl/fl (cKO) hair follicle bulges. Dotted lines denote ap-proximate outer boundary of the hair bulge. (Scale bars: 20 μm.)

Wls mRNA

A B CWls-fl/flWls-fl/+ Wls-fl/+

Wls-fl/fl

n = 4 n = 4

Fig. 4. Axin2-expressing cells require Wnt secretion for hair growth. (A) Representative image of RNA in situ hybridization for Wntless gene expression asperformed on telogen hair follicles. Dotted line denotes approximate outer boundary of the hair bulge. (Scale bar: 20 μm.) (B and C) Representative images ofthe whole body (B) or H&E-stained sections of the skins (C) of Axin2–rtTA/tetO–Cre/Wls fl/+ or fl/fl mice. (Scale bars: 1 cm.)

Lim et al. PNAS | Published online February 22, 2016 | E1501

DEV

ELOPM

ENTA

LBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

0

Axin2-expressing bulge cells also expressedWnt4 andWnt7b (Fig.5F). These data suggest that the anagen bulge is not Wnt-inhibited and that Wnt production and β-catenin signaling persistin the anagen bulge. In addition, we found Axin2 (Fig. S4A) anddistinct Wnt gene expression patterns (Fig. S4 B and C) inmultiple differentiated lineages of the anagen hair follicle, con-sistent with previous reports (28). Other Wnts like Wnt7a andWnt16 were present either at very low levels or were not expressed(Fig. S4D). Specific Wnts may thus play distinct functional roles inthe different compartments of the anagen hair follicle.Our data show that Wnt signaling, as reported by Axin2 ex-

pression, is restricted to the outer bulge (Fig. 1 A and B). If Wntsignals are available in the bulge, then why do inner bulge cellsnot also respond to Wnt and proliferate? In seeking to addressthis question, we found evidence for active Wnt inhibition ofthe inner bulge. Using RNA in situ hybridization, we saw thatseveral secreted Wnt inhibitors, including Dkks and Sfrp1, areexpressed in the outer bulge during telogen (Fig. 6A). Expres-sion of some of these Wnt inhibitors persists into the anagen

phase (Fig. 6B). This finding is consistent with other studies ofgene expression in isolated mouse bulge cells (3, 16) and is alsosimilar to human hair follicle bulges, where both Dkk3 andSFRP1 are expressed in the outer bulge, but are either at verylow levels or not present in the inner bulge (Fig. 6C) (29).When we performed antibody staining for the Dkk3 protein inthe telogen bulge, however, we found that the Dkk3 protein islocalized to the inner bulge (Fig. 6D). This distribution of Dkk3mRNA and protein in the bulge persists even through anagen(Fig. 6E), suggesting that HFSC stemness is restricted to theouter bulge by constant inhibition of Wnt signaling in the innerbulge cells.

DiscussionOur data lead us to propose a model in which Wnt/β-cateninsignaling is a constant and persistent feature of the HFSC niche(Fig. 7). HFSCs located in the outer bulge are Wnt-dependentand produce Wnt signals that act locally, while simultaneouslyinhibiting Wnt/β-catenin signaling in the inner follicle bulge by

A

FNegative Control

Axin2 mRNAWnt4 mRNA

Axin2 mRNAWnt7b mRNA

Negative Control

Wnt7b mRNANegative control Wnt4 mRNA

Axin2 mRNA

Anagen Bulge

outer root sheath

inner

outerbulge

D

Axin2-lacZ

EWnt10a mRNA Wnt10b mRNAWnt5a mRNA Wnt6 mRNA

B C

Fig. 5. Anagen hair bulge cells express Axin2 and Wnts. (A) Schematic depicting cellular organization of the anagen follicle bulge. (B–E) Representativeimages of Axin2–lacZ (B) and RNA in situ hybridization (C) for Axin2 and Wnt gene expression (D and E) as performed on anagen hair follicles. Dotted linesdenote approximate outer boundary of the hair bulge. Dotted lines denote approximate outer boundary of the hair bulge. [Scale bars: 20 μm (Axin2–lacZimage); 10 μm (RNA in situ hybridization images).] (F) Representative images of double-labeling RNA in situ hybridization in anagen hair follicles for Axin2(red spots) andWnt4 orWnt7b (turquoise spots). Arrows indicate bulge cells expressing both Axin2 and Wnts. Boxes show magnified view of individual bulgecells expressing both Axin2 and Wnts. Dotted lines denote approximate outer boundary of the hair bulge. (Scale bars: 20 μm.)

E1502 | www.pnas.org/cgi/doi/10.1073/pnas.1601599113 Lim et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

0

secreting Wnt inhibitors like Dkks and Sfrp1 (Fig. 7). Thisautocrine Wnt activation and paracrine Wnt inhibition may helpto maintain the discrete identity of cells located in each bulgelayer, even as cells traffic dynamically to and from the bulgeniche (2) throughout multiple hair cycles. Such a model predictsthat loss of Wnt/β-catenin signaling in the outer bulge would resultin disrupted compartmentalization of the inner vs. outer bulge.Consistent with this prediction, we find that Keratin-6, which istypically enriched in the inner bulge and is considered to be amarker of that compartment (Fig. S2 C and D) (2), becomes morestrongly expressed in outer bulge cells upon deletion of β-cateninin Axin2-expressing outer bulge cells (Fig. S2 B and E) (14).If HFSCs generate their own source of Wnt, and Wnt pro-

duction is regulated in an autocrine manner, what keeps hairfollicles in telogen and prevents them from undergoing autocrineamplification of Wnt signaling to initiate anagen? A likelycandidate is BMP signaling, which has been shown to inhibit Wntsignaling and hair growth (30). It is thought that the reciprocalinterplay between BMP and Wnt signaling controls HFSC ac-tivity and fate decisions and that BMP signaling suppresses Wntproduction (31). We found some evidence that Wnt signalingcontrols BMP pathway component expression. After deletion ofβ-catenin in Axin2-expressing bulge cells, we observed slightlyincreased expression of several BMP antagonists, including Noggin,Bambi, and Gremlin1 (Fig. S2B), suggesting that autocrine Wntsignaling suppresses BMP antagonist expression. This finding leadsus to speculate that a low level of autocrine Wnt signaling in HFSCsreduces BMP inhibition and thus permits BMP signaling, creating afeedback inhibition loop that moderates Wnt signaling levels andprevents autocrine entry into anagen. As environmental BMP levelsare reduced, autocrine amplification could then lead to elevated

Wnt signaling, which in turn promotes the transition from telogento anagen (23, 32).Our finding of Axin2–lacZ Wnt reporter gene expression in

the telogen follicle bulge contrasts with the lack of activity inother TCF-element-based Wnt reporter constructs such asTOPGAL (13). The more limited expression pattern of TOPGALin hair follicles may result from endogenous Wnt-responsivecis-regulatory elements being absent in the TOPGAL reporterconstruct (33). Consistent with this hypothesis, Axin2–lacZ isexpressed widely in the anagen follicle in domains of Wnt mRNAexpression, whereas TOPGAL is not. That Axin2 is a moresensitive and comprehensive Wnt reporter than TOPGAL is alsoobserved in other stem cell niches such as the intestine, whereAxin2 expression marks stem cells (22), even though TOPGALexpression is absent. The expression of other Wnt target genessuch as Lgr5 in the bulge further supports the notion that Wnt/β-catenin signaling is active in the bulge, and that Axin2 is areliable, sensitive, and comprehensive reporter of Wnt/β-cateninsignaling in the hair follicle.Whether or not Wnt/β-catenin signaling is required for stem

cell maintenance has been a matter of debate. Previous studiesshowed that deletion of β-catenin in the bulge results in loss ofhair stem cell identity, as determined by loss of both proliferationand expression of outer bulge cell markers like CD34 (10). Morerecently, however, others have shown that CD34 and Keratin-15continue to be expressed even after β-catenin deletion and haveproposed instead that Wnt/β-catenin signaling is not required formaintenance of stem cell identity but rather to induce anagen(14, 24). However, it is unclear that the expression of these genesis a reliable indicator of stem cell identity or function. Indeed,whenCD34 is knocked out, hair follicles continue to grow, suggesting

ADkk3 mRNA

telogen

B

telogen

Sfrp1 mRNA Dkk3 mRNA Sfrp1 mRNA

anagen anagen

C

DDkk3 mRNA

anagen

Axin2-CreERT2Dkk3 protein

Dkk3 Axin2 Dkk3 Axin2

Dkk3 mRNA

Axin2-CreERT2Dkk3 protein

E

telogen

telogen

Dkk4 mRNA

Fig. 6. Outer bulge cells express secreted Wnt inhibitors. (A and B) Representative images of RNA in situ hybridization for Dkk3, Dkk4, and Sfrp1 duringtelogen (A) and anagen (B). Dotted lines denote approximate outer boundary of the hair bulge. (Scale bars: 20 μm.) (C) Dkk3 and Sfrp1 expression in laser-capture microdissected primary outer vs. inner bulge cells from human hair follicles (error bars, SD). Expression values are from Gene Expression Omnibusaccession no. GSE3419. (D and E) Representative images of Dkk3 immunostaining in telogen (D) and anagen (E) bulges of Axin2–CreERT2/Rosa26–mTmGflox

mice. Representative images of RNA in situ hybridization for Dkk3 in telogen and anagen bulges are shown for comparison. Dotted lines denote approximateouter boundary of the hair bulge. [Scale bars: 20 μm (telogen Dkk3 RNA in situ hybridization image); 10 μm (anagen Dkk3 RNA in situ hybridization and Dkk3immunostaining images).]

Lim et al. PNAS | Published online February 22, 2016 | E1503

DEV

ELOPM

ENTA

LBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

0

that it is not functionally required for hair stem cell function (34).Additionally, Keratin-15 expression persists in follicles of baldingpatients, whereas the expression of another HFSCmarker, CD200, islost (35). As such, although CD34- and Keratin-15–expressing cellsmay persist in β-catenin–deleted bulges, it is unclear that they re-main as stem cells.Stem cells are functionally defined by their potency rather

than by marker expression (36–38), and a more reliable indicatorof HFSC identity is a test of its ability to contribute to hair folliclegrowth. Consistent with our own data (Fig. 2), all of the publishedstudies we know of involving conditional ablation of β-catenin inthe bulge result in the hair follicle being arrested in telogen and nolonger growing. Isolated bulge cells from these β-catenin mutantscannot form hair follicles, even when transplanted (14). We in-terpret these observations to mean that Wnt/β-catenin signalingis required to maintain HFSC potency and function.Our findings complement recent work which suggests that

attenuated Wnt signaling is part of the process of defining em-bryonic progenitors that become adult HFSCs (39). This studyshowed that hair progenitors start off as Axin2-expressing cells inthe placode, but gradually lose Axin2 expression as they committo becoming adult HFSCs during late embryogenesis. Our worksuggests that Axin2 becomes expressed again in the adult HFSCcompartment, once it is established.HFSCs have been proposed to actively organize their micro-

environment (16, 17). Here, we provide evidence that HFSCsmay influence the organization of their niche by acting as theirown source of self-renewing Wnt signals, while simultaneouslypromoting the differentiation of their progeny.

Materials and MethodsAnimals. Axin2–CreERT2 mice were described (22). Axin2-lacZ (40) were a giftfrom W. Birchmeier, Max-Delbrück-Centre for Molecular Medicine, Berlin.Rosa26–mTmGflox (21) and β-cateninΔex2-6-flox/flox (41) mice were obtainedfrom The Jackson Laboratory. All alleles were heterozygous, except wherestated. All experiments were approved by the Stanford University AnimalCare and Use Committee and the A*STAR Animal Care and Use Committeeand were performed according to NIH and Singapore Bioethics Councilguidelines.

Labeling and Tracing Experiments. All mice received a single i.p. injection of5–20 mg/mL stock solutions of Tam (Sigma, T5648) dissolved in corn oil/10%(vol/vol) ethanol, corresponding to specific doses per gram body weight as

specified below. Axin2–CreERT2/Rosa26–mTmGflox mice received a singledose of Tam at P21 or 49 (P49), totaling 1 mg per 25 g of body weight.

Conditional Knockout of β-Catenin and Wntless. Axin2–CreERT2/β-cateninΔex2-6-fl/+

andmutantAxin2–CreERT2/β-cateninΔex2-6-fl/fl animals were injectedwith a singledose of Tam corresponding to 4 mg per 25 g of body weight at 3 wk of age.Skins were harvested and processed for histology upon morbidity or death after10–11 d of chase. For Wntless ablation, mothers carrying Axin2–rtTA/tetO–Cre/Wls fl/+ and mutant Axin2–rtTA/tetO–Cre/Wls fl/fl pups were injected with a singledose of Tam corresponding to 1 mg per 25 g of body weight at embryonic day14.5 (E14.5). Skins were harvested and processed for histology when pups wereat postnatal age 1 mo.

X-Gal Staining. Tissues were harvested and fixed in 4% (vol/vol) para-formaldehyde (PFA) for 1 h at 4 °C, washed in PBS and detergent rinse (PBSwith 2mM MgCl2, 0.01% sodium deoxycholate, and 0.02% Nonidet P-40detergent), and then stained in staining solution (PBS with 2 mM MgCl2,0.01% sodium deoxycholate, and 0.02% Nonidet P-40, 5 mM potassiumferricyanide, 5 mM potassium ferrocyanate, and 1 mg/mL X-gal) in the darkat room temperature overnight. Tissues were then washed in PBS, postfixedin 4% PFA in PBS overnight at 4 °C, and processed for paraffin embeddingand histology.

Histology and Immunostaining. Animals were killed and skins were removedfrom the back (dorsal) and the tail regions. For cryosections, tissues were fixedfor 1–2 h in 4% PFA at 4 °C, washed in PBS, and incubated overnight in 30%sucrose/PBS at 4 °C. Tissues were then embedded in OCT medium and storedat −80 °C before cryosectioning. Sections of 5- to 6-μm thicknesses were cutusing a Leica CM3050S cryostat (Leica Microsystems).

For paraffin sections, tissues were fixed overnight in 4%PFA at 4 °C. Tissueswere washed in PBS, then dehydrated through an ethanol series into xylene,and embedded in paraffin wax. Sections of 5-μm thickness were cut by usinga Leica RM2255 microtome (Leica Microsystems), and then rehydrated andcounterstained with Eosin where specified.

All immunofluorescence staining was performed in the dark. Cryosectionswere washed in PBS and incubated in blocking buffer (2% normal donkeyserum and 0.2% Triton X in PBS) for 1 h at room temperature. Primary an-tibodies were diluted in blocking buffer and incubated with cryosectionsovernight at 4 °C for 1 h at room temperature. Sections were then washed inPBS, and incubated with secondary antibody diluted in blocking solution for1 h at room temperature. Sections were then washed in PBS and mounted inProlong Gold with DAPI mounting medium (Life Technologies). The fol-lowing primary antibodies were used: anti-GFP (Abcam), anti-Dkk3 (R&Dsystems), and anti–Keratin-6 (Covance). The following secondary antibodieswere used: anti-chicken conjugated to Alexa Fluor 488 (Life Technologies,Jackson ImmunoResearch) and anti-goat conjugated to Alexa Fluor 647(Jackson ImmunoResearch).

ATelogen

Wnt1,Wnt4, Wnt7b

Telogen-Anagentransition

Wnt6, Wnt7b,Wnt10a,Wnt10b

B CAnagen

Wnt4, Wnt7b

Dkk3Dkk4Sfrp1

= Wnt5a

= Wnt4, Wnt6, Wnt7b Wnt10a, Wnt10b

Dkk3Dkk4Sfrp1

Dkk3Dkk4Sfrp1

Hair germ-dermal papllaboundary

stemcellmaintenance

hairlineagespecification

Wnt-active (Axin2+)

Wnt-inhibited (Dkk protein+)

Wnt1,Wnt4, Wnt7b

Fig. 7. Model: Dual roles for Wnt/β-catenin signaling in HFSC maintenance and lineage determination. (A) During telogen, Wnt-responding stem cells in theouter bulge produce autocrine Wnt1, Wnt4, and Wnt7b, and paracrine-acting secreted Wnt inhibitors like Dkk3, Dkk4, and Sfrp1. This expression patternpersists throughout the hair cycle, potentially to maintain the niche signaling microenvironment. (B) At the telogen–anagen transition, Wnt6, Wnt10a, andWnt10b become strongly expressed in cells at the boundary of the hair germ and dermal papilla. (C) These and other Wnts continue to be expressed in varioushair layers proximal to the dermal papilla during anagen and are likely involved in hair lineage specification.

E1504 | www.pnas.org/cgi/doi/10.1073/pnas.1601599113 Lim et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

0

RNA in Situ Hybridization. Tissues were harvested and fixed in 4% PFA for 24 h,and then dehydrated and embedded in paraffin. Tissues sections were cut at5-μm thickness, air-dried at room temperature, and processed for RNA in situdetection by using the RNAscope 2.0 FFPE Reagent Kit according to themanufacturer’s instructions (Advanced Cell Diagnostics; ref. 42). RNAscopeprobes used were as follows: Axin2 (NM_015732, region 330–1287), Wnt1(NM_021279.4, region 1204–2325), Wnt3 (NM_009521.2, region 135–1577),Wnt4 (NM_009523, region 2147–3150),Wnt5a (NM_009524, region 200–1431),Wnt6 (NM_009526, region 780–2026), Wnt7a (NM_009527.3, region 1811–3013),Wnt7b (NM_009528, region 1597–2839), Wnt10a (NM_009518, region 479–1948), Wnt10b (NM_011718, region 989–2133), Wnt16 (NM_053116.4, re-gion 453–1635) Dkk3 (NM_015814, region 1513–2703), Dkk4 (NM_145592.2,region 22–1140), and SFRP1 (NM_013834.3, region 1810–2779), which weredetected using the Fast Red detection reagent.

Microscope Image Acquisition and Quantification. All sections were imaged byusing the Axioplan 2 and Imager.Z2 Microscopes, the AxioCam MRm (fluores-cence) and MRC5 (bright field) cameras, and using Axiovision AC software (Re-lease 4.1, Carl Zeiss) or Zen software (Carl Zeiss). Image acquisitions were done atroom temperature using ×10 NA 0.3, X20 NA 0.5, X20 NA 0.8, and ×40 NA 0.75EC Plan-Neofluar objectives (Carl Zeiss). For some images, contrast, color, anddynamic range were globally adjusted in Adobe Photoshop (Adobe Systems).

Cell Labeling and Flow Cytometry. For isolation of bulge cells, 6- to 8-wk-oldAxin2–LacZ animals were killed, hair was removed, and dorsal skins wereharvested and washed in ice-cold calcium and magnesium-free Dulbecco’sPBS (DPBS; GE Healthcare). After removing fat and s.c. tissues using a scalpel,the dorsal skins were then trypsinized in 0.25% Trypsin–EDTA (Gibco) at37 °C for 1 h. Epidermal cell suspension was neutralized with ice-cold 10%chelexed FBS (GE Healthcare) in DPBS and strained with 70 and 40 μM cellstrainers (Fisher Scientific). Cells were centrifuged at 300 × g for 30 min at

4 °C, washed with 10% chelexed FBS/DPBS, centrifuged at 300xg for 5–10 minat 4 °C and resuspended into 1 mL of single cell suspension with ice-cold 10%chelexed FBS/DPBS which was then incubated with primary antibodies eithercoupled directly with a fluorochrome or indirectly in chelexed FBS-precoatedpropylene tubes on ice for 60 min in the dark, washed two times before in-cubation with secondary antibodies in 10% chelexed FBS in DPBS for 20 minon ice, washed and analyzed with flow cytometry. The following primaryantibodies were used in our FACS analysis: Alexa 700-conjugated anti-mouseCD34 (Affymetrix), APC-conjugated anti-human/mouse CD49f (α6 Integrin)(eBioscience), Pacific Blue anti-mouse Ly-6A/E (Sca-1) (Biolegend), Polyclonalgoat antibody against mouse LRIG1 (R&D systems). The following secondaryantibodies, Donkey anti-goat IgG (H+L)-Phycoerythrin (R&D systems) was used.Cell viability was checked by staining with Zombie-Aqua Fixable Viability kit(Biolegend). Living cells were sorted positive for CD34 and α6 Integrin, andnegative for Lrig1 and Sca1. This CD34+ α6 Integrin+ bulge cell population wasfurther sorted into Axin2+ and Axin2− populations.

For detection of intracellular β-galactosidase activity of Axin2-expressingcells, a fluorescein di-β-D-galactopyranoside (FDG) staining kit (Thermo Fisher)was used. Keratinocytes stained with the antibodies were loaded with FDGusing hypotonic shock at 37 °C. Cell sorting was carried out using a FACSAriacell sorter using unstained control, single color compensation controls. Axin2-positive cells were defined by gating using fluorescence minus one withoutFDG. Axin2+ and Axin2− cell populations were fixed in ice-cold methanol for15 min, washed and centrifuged in DPBS. Cell cycle analysis was performed bystaining fixed cells with propidium iodide supplemented with RNase in thedark at room temperature and analyzed with a FACSAria cell sorter.

ACKNOWLEDGMENTS. We thank Poh Hui Chia for cell quantification assis-tance. This work was supported by the Howard Hughes Medical Institute; Cal-ifornia Institute of Regenerative Medicine Grant TR1-01249; and the Agency forScience, Technology, and Research in Singapore.

1. Schneider MR, Schmidt-Ullrich R, Paus R (2009) The hair follicle as a dynamic mini-organ. Curr Biol 19(3):R132–R142.

2. Hsu Y-C, Pasolli HA, Fuchs E (2011) Dynamics between stem cells, niche, and progenyin the hair follicle. Cell 144(1):92–105.

3. Morris RJ, et al. (2004) Capturing and profiling adult hair follicle stem cells. NatBiotechnol 22(4):411–417.

4. Levy V, Lindon C, Harfe BD, Morgan BA (2005) Distinct stem cell populations re-generate the follicle and interfollicular epidermis. Dev Cell 9(6):855–861.

5. Jaks V, et al. (2008) Lgr5 marks cycling, yet long-lived, hair follicle stem cells. NatGenet 40(11):1291–1299.

6. Cotsarelis G, Sun T-T, Lavker RM (1990) Label-retaining cells reside in the bulge area ofpilosebaceous unit: Implications for follicular stem cells, hair cycle, and skin carcino-genesis. Cell 61(7):1329–1337.

7. Greco V, et al. (2009) A two-step mechanism for stem cell activation during hair re-generation. Cell Stem Cell 4(2):155–169.

8. Hsu Y-C, Fuchs E (2012) A family business: Stem cell progeny join the niche to regulatehomeostasis. Nat Rev Mol Cell Biol 13(2):103–114.

9. Andl T, Reddy ST, Gaddapara T, Millar SE (2002) WNT signals are required for theinitiation of hair follicle development. Dev Cell 2(5):643–653.

10. Lowry WE, et al. (2005) Defining the impact of beta-catenin/Tcf transactivation onepithelial stem cells. Genes Dev 19(13):1596–1611.

11. Clevers H, Loh KM, Nusse R (2014) Stem cell signaling. An integral program for tissuerenewal and regeneration: Wnt signaling and stem cell control. Science 346(6205):1248012.

12. Blanpain C, Fuchs E (2006) Epidermal stem cells of the skin. Annu Rev Cell Dev Biol 22:339–373.

13. DasGupta R, Fuchs E (1999) Multiple roles for activated LEF/TCF transcription com-plexes during hair follicle development and differentiation. Development 126(20):4557–4568.

14. Lien W-H, et al. (2014) In vivo transcriptional governance of hair follicle stem cells bycanonical Wnt regulators. Nat Cell Biol 16(2):179–190.

15. Gat U, DasGupta R, Degenstein L, Fuchs E (1998) De novo hair follicle morphogenesisand hair tumors in mice expressing a truncated beta-catenin in skin. Cell 95(5):605–614.

16. Tumbar T, et al. (2004) Defining the epithelial stem cell niche in skin. Science303(5656):359–363.

17. Blanpain C, Lowry WE, Geoghegan A, Polak L, Fuchs E (2004) Self-renewal, multi-potency, and the existence of two cell populations within an epithelial stem cell ni-che. Cell 118(5):635–648.

18. Jho E-H, et al. (2002) Wnt/beta-catenin/Tcf signaling induces the transcription ofAxin2, a negative regulator of the signaling pathway. Mol Cell Biol 22(4):1172–1183.

19. Lim X, et al. (2013) Interfollicular epidermal stem cells self-renew via autocrine Wntsignaling. Science 342(6163):1226–1230.

20. Ito M, Kizawa K, Hamada K, Cotsarelis G (2004) Hair follicle stem cells in the lower bulgeform the secondary germ, a biochemically distinct but functionally equivalent pro-genitor cell population, at the termination of catagen. Differentiation 72(9-10):548–557.

21. Muzumdar MD, Tasic B, Miyamichi K, Li L, Luo L (2007) A global double-fluorescentCre reporter mouse. Genesis 45(9):593–605.

22. van Amerongen R, Bowman AN, Nusse R (2012) Developmental stage and time dic-tate the fate of Wnt/β-catenin-responsive stem cells in the mammary gland. Cell StemCell 11(3):387–400.

23. Plikus MV, Chuong CM (2014) Macroenvironmental regulation of hair cycling andcollective regenerative behavior. Cold Spring Harbor Perspect Med 4(1):a015198.

24. Choi YS, et al. (2013) Distinct functions for Wnt/β-catenin in hair follicle stem cellproliferation and survival and interfollicular epidermal homeostasis. Cell Stem Cell13(6):720–733.

25. Myung PS, Takeo M, Ito M, Atit RP (2013) Epithelial Wnt ligand secretion is requiredfor adult hair follicle growth and regeneration. J Invest Dermatol 133(1):31–41.

26. Augustin I, et al. (2013) Loss of epidermal Evi/Wls results in a phenotype resemblingpsoriasiform dermatitis. J Exp Med 210(9):1761–1777.

27. Fu J, Hsu W (2013) Epidermal Wnt controls hair follicle induction by orchestratingdynamic signaling crosstalk between the epidermis and dermis. J Invest Dermatol133(4):890–898.

28. Reddy S, et al. (2001) Characterization of Wnt gene expression in developing andpostnatal hair follicles and identification of Wnt5a as a target of Sonic hedgehog inhair follicle morphogenesis. Mech Dev 107(1-2):69–82.

29. Ohyama M, et al. (2006) Characterization and isolation of stem cell-enriched humanhair follicle bulge cells. J Clin Invest 116(1):249–260.

30. Kobielak K, Pasolli HA, Alonso L, Polak L, Fuchs E (2003) Defining BMP functions in thehair follicle by conditional ablation of BMP receptor IA. J Cell Biol 163(3):609–623.

31. Kandyba E, et al. (2013) Competitive balance of intrabulge BMP/Wnt signaling revealsa robust gene network ruling stem cell homeostasis and cyclic activation. Proc NatlAcad Sci USA 110(4):1351–1356.

32. Schlake T, Sick S (2007) Canonical WNT signalling controls hair follicle spacing. CellAdhes Migr 1(3):149–151.

33. Barolo S (2006) Transgenic Wnt/TCF pathway reporters: All you need is Lef? Oncogene25(57):7505–7511.

34. Trempus CS, et al. (2007) CD34 expression by hair follicle stem cells is required for skintumor development in mice. Cancer Res 67(9):4173–4181.

35. Garza LA, et al. (2011) Bald scalp in men with androgenetic alopecia retains hairfollicle stem cells but lacks CD200-rich and CD34-positive hair follicle progenitor cells.J Clin Invest 121(2):613–622.

36. Lander A (2009) The “stem cell” concept: Is it holding us back? J Biol 8:70.37. Snippert HJ, Clevers H (2011) Tracking adult stem cells. EMBO Rep 12(2):113–122.38. Blanpain C, Simons BD (2013) Unravelling stem cell dynamics by lineage tracing. Nat

Rev Mol Cell Biol 14(8):489–502.39. Xu Z, et al. (2015) Embryonic attenuated Wnt/β-catenin signaling defines niche lo-

cation and long-term stem cell fate in hair follicle. eLife 4:1–49.40. Lustig B, et al. (2002) Negative feedback loop of Wnt signaling through upregulation

of conductin/axin2 in colorectal and liver tumors. Mol Cell Biol 22(4):1184–1193.41. Brault V, et al. (2001) Inactivation of the beta-catenin gene by Wnt1-Cre-mediated

deletion results in dramatic brain malformation and failure of craniofacial develop-ment. Development 128(8):1253–1264.

42. Wang F, et al. (2012) RNAscope: A novel in situ RNA analysis platform for formalin-fixed, paraffin-embedded tissues. J Mol Diagn 14(1):22–29.

Lim et al. PNAS | Published online February 22, 2016 | E1505

DEV

ELOPM

ENTA

LBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

0