Primary Mouse Cultures Contain Follicle Progenitor Cells with Multiple Differentiation Potential

Jun Kamimura,*i" David Lee,* Howard P. Baden,* Janice Brissette,* and G. Paolo Dotto* *Cuwneous 13io logy ltese>1 rch Center, Massac husetts General Hospital and Harva rd Medica l School, C harl estown, M arylond, U .S. A.: and ·j·S hiseido R.esca rch Centre, Yokohama, Japan

The interfollicular contains a single type of reassociation of cells. Instead, the reconstituted inter­ tenninally differentiated , whereas hair follicular epidennis contained distinct colum.nar units, follicles are composed of a minimum of six or seven comprising all the overlying layers and most likely derived distinct types. Whether or not these various populations from a single progenitor . In contrast, hair follicles of terminally differentiated keratinocytes originate from were found to be composed of cells of multiple origin, one or more progenitor cells has not been established. A with each population showing a striking localization related and in1portant question is whether keratinocyte to a separate concentric region. The vast majority of reconstituted progenitor cells with a pluripotent potential, able to follicles appeared to derive from a n1initnutn of two or, in a significant fraction of cases, form. not only epidern1is but also hair follicles, can be three progenitor cells, one for the generation of the shaft maintained i11 11itro for any period of time. We have (, , and n1edulla), one for the inner root addressed these questions using reconstitution assays sheath, and the third for the outer . The with adrnixed populations of genetically labeled, cultured general implications of these findings for epidermis and keratinocytes. Examination of reconstituted epidennis fonnation and for keratinocyte sten1 cell and hair follicles showed that neither was composed of cultivation are discussed. Key words: dermal papilla cells/ a random mixture of differently labeled keratinocytes, as J•etJ•olliral geue tmnsfer I stem cells I tissue reco ustitution. J bwest would be predicted if they originated from a random Dermatol 109:534-540, 1997

ike o th er terminall y differentiating ti ss ues, the epidermis is (, cortex, and cuticle), three of th e inner r oot shea th (TR.S) likely to contain a compartment o F cells that are ca pable (H enlt! 's and Huxley's la yers, the c uticle), and the one or mo re of the of continuo us self-renewal and that can repl enish th e o mer root sheath (ORS). Whetb er or no t these va rious populations populations of keratinocytes that are los t dLI!;ng clifferenti­ of terminall y differentiated k eratinocytes Oti gin ate frotn one or more ati o n. Stem cells sho uld have two fundamental pro perties: progenitor cells has not been established. Finally, an important and as L(i) a n indefi nite life spa n, a11d (ii) t he ca pabili ty to give ri se to all o ther yet unsolved ques tion is w hether keratinocyte s tem ce!Js with a types of partia ll y committed and differentiated ·ells. The existence of pluripotent differentiation potential, able to form no t o nl y e pidermis ocyte subpopul ations with increased proliferative po tential has but also hair follicles, can be maintained i11 tlitm for any length of time, been well establi shed, rtven if their precise locali zation and functional thus a!J owin g additional and mo n:: direct exp erimental evaluatio ns. role is still a matter of so me debate (Cotsarelis ct a/, '1990; Ya ng et a/, As stem cells of th e bo ne marrow, k eratinocyte progenitor cells a re 1993; Kobayas hi et a/, 1993; Jahoda and R eynold s, 1993; R oc!Jat et nl, th o ught to be tightly regula ted and dependent on signals produced b y 1994; Moll, 1995; Jones et a/, 1995). By contrast, evidence fo r closely juxtaposed stromal cells (Sengel, 1990). This is particularly keratinocyte progenitor cell s with a broad differentiation potential, evident in hair follicl es, yvhere cycles of keratinocyte growth (a nagen) able to reconstitute o n their own interfo Lli cul ar epidermis as well as and regression (ca tagen and tdogen) appear to be determined by th e hair folli cles, is stil.l lack in g. underl ying dermal papilla cells (Stenn el a/, 1991; Hardy, 1992) . The interfo lli cular epidermis contains a sin gle type of terminally .Epitheli al- mesenchymal interactions also play an essen tial role in th e differentiated ke ratin ocyres, whereas hair fo llicl es are composed of a inducti on of hair follicle formation during development. Three distinct minimum of seven distinct t ypes (S tenn et a/, 199 1; R.eynolds and sigmls are involved in this process (H ard y, 1992). A ftrst "dermal Jahoda, 1993). T hese include the three cell types of th e hair shaft message" fr om relatively undifre rentiated mesenchymal cell s induces a condensation of th e overl ying e pidermis. The epidermal condensa ti o n in turn produces an "epitheli al message" leading to dermal papil.l a cell formati on. Dermal papilla cells then release a "second dermal message" Manuscript received 1\pril 30, 1997; revised June 20, 1997; accep ted fo r that instructs th e overl ying epidermal condensation (hair bud) to fo rm pu bli cation June 27, 1997. a hair fo llicle. T his sequence of events c an be reproduced by a very P...eprint requests to: G .l'. Dotto, Bl"tC, MG H Eas t, 13th St r c~ t , elegant hair reconstitution C harl es town, MA 02 129. assay that bas been r ecently developed T hi s paper is dedicated to the mem01y of Dr. Jun l

0022-202X/ 97/$ 10.50 · opyright © ·1997 by T he Society fo r Investi gative D ermatology, In c. 534 S VOL. 10'.1, NO. -1 OCT OUER 1997 KER.AT INOCYTE/ H A IR FO LLI C LE P I ~OGEN IT OR CELL 535 from a single cultured progenitor cell . [n contrast, hair fo Lli cl es appear in cubation w ith rh o daminL·-conjug3tcd goat anti- rabbit itnmunoglobulin G I 00 dilution in buffer A, Sigma, St. Louis, MO) and flu orescein -conjugated to derive fi·om a nunimum of two o r, in a significant fraction of cases, (I: strepra vidin ( I: I 00 in buffer A, Amershalll , Boston, MA) for I h in a li ght­ three progenitor cells, one for the generation of the shaft (cuticl e, right hlllnidificd box at room t e mp ~ ratur e. Fo ll owing in cubation with the a), o ne for the IRS, and th e third for the OR.S. cortex, and medull secondary antibodi es, secti o ns were washed exrensivcl y w ith PBS ~llld n1 0 U11ted fo lli cle T he general i.mplications of these findings for epidermis and hair in ·1% n-propylga ll are in glycerol. Comrol experim ems we re performed to formation and for keratinocyte cultiva tion are discussed . ve ri fY spec ificiry of recognition of 13A LB /c versus C57BL/ 6 and nude (Swiss) mouse . MATEI:tiALS AND METHOD .RESULTS Cells and Mouse prima1y keratinocytes we re iso lated fi·om 1-2-d­ old newborn Sencar, C57BL/6, or BALB/c mi ce and cultured in minimal Hair forming capability of cultured primary k eratinocytes essential n1 ediun1 at lovv ca lcium concentratio ns (0 .05 mN1; lo\v ca lciun1 The hai r reconstitution assay developed b y Li chti eta/ (1993; Weinberg ith 4% Chclex-treated feral calf serum and epid t!nnal medium) supplemented w et a/, 1993) is based on the grafting of hair buds (hair follicle precursors growth r.1ctor (10 ng per ml ; olbborative l:tesearch In c.) as previ ously present in newborn mo use skin as specif1 c ke ratinocyte aggregates) described (Hennings ct ril, 1980; Datto et nl, 1988) . Hair bud and interfolli cula r o nto n ude mice together with hair-inducing dermal papilla cell s keratin ocyte preparations were obtain ed as described by Weinberg el nl ( 1993) . mbinacion, Bri eAy, the epid ermis of '1 -2-d-old newborn mi ce was se parated fi·om th e (de ri ved fi·01n rat vibrissae) . These two components in co underl ying by overni ght trypsin or dispase treatmellt at -l °C, fo ll owed but not eithe r one se parately, result in the formation of full y differe nti­ by sepa ration of the hair bud and interfo l.li cular fra ctions by two consecmive ated a nd o rga ni zed hair fo lli cl es. lmeres ti ngly, th ese authors reported Fi co ll gradi ent purifica tion steps. Hair bud and inrerfollicular keratinocyte that hair tollicle formatio n is not affected by dissociation of the hair preparations were either used directl y for grafting, or cultured for 1 wk in low buds prior to grafti ng, suggesting that the keratinocytes pre e nt in these medium as described for th e unfi·acrionared prinmy ker;H in ocyte structures contain aU the necessaLy information for reassociating itt vivo cultures. (Weinberg et a/ , 1993) . W e in vestigated the mo re s pecifi c poss ibility Prim a1y dermal papilla cells and inrerfo lli cular dermal f1broblasrs were micro that the hai r bud structures, and the epidermis in general, contain dissected fi·om vibri ssae of -+-6-wk-old Fisher rats Oah oda et a/, 1984). Derm al l progenitor cell s capable f generating o u their own a hair papi lla cells were expanded three tim es (1 :3 s plits) , the first two pa ssages in individua Chang's medium (1-l ana 13iologics, Alm eda , CA) and th e last passage in foLl icle s tructure and/ or a distinct tract of the e pidermis. Dll lbecco's modified minimal essential med ium supplemented with ·10% fetal [n o ur initial studies, w e tested w he ther hair reconstitutio n assays calf serum . could be performed w ith culrured keratinocytes. In fact, we found C ultured kerarinocy res were infected with a hel per free recombinant retroviru s that p 1 ~ imaLy keratinocytes deLived fi·01n hair buds of newbo rn mouse ca rrying th e human alkaline phosphatase (AP) ge ne (Cepko et a/, '1993) 2 d skin and cultured for 1 wk, were as e ffi cient as fi·esh[y prepared hair et nl, after plating. Infection conditions were as previously desc ribed (Datto buds in forming hair fo lli cles w hen grafted together with instructing res were used for grafti ng experiments 4-5 dafter infecti on. 1988) . l1 tfecred cultu dermal papill a 'cell s (Fig 1A,B). When these sa me ceLl s were grafted The exact percentage of AP-positi ve and AP-negari ve keratinocy res was togethe r with fibrobL1Sts deri ved fi·o m the interfoll icubr dermis, rather determined for each experiment, by stain ing the same keratinocytc preparations no hair fonmtion was detected (F ig lC,D). that were used for grafti ng for AP producti on, as desc ribed by Cepk o ct nl (1993) . than dermal papill a cells, Intri guingly, simibr hair forming ability was also observed w ith prima1y the inte rfo lli cubr fi·a ction of th e e pidermis, Hair reconstitution assay Skin / luir reconstitutio n nss:Jys were perfo rm ed keratinocytes derived fi·om esse ntia iJ y as described by Weinberg et nl ( 1993) . We used eith er fi·eshly prepared and separated fi·01 n the hair bud li·action by double Fi coll gradie nt hai r buds (2 .5 skin equivalents per graft) or cultured keratinocy tes ( I X I 07 centrifugatio n. C ul tures of unfracti o nated k e ratinocyte preparations per graft) together with rat dernl:ll pap.ill a ce ll s or dermal fibroblasts (8 X :lOr, w ere just as e fl'i cient in hair fonnacion . Furthermo re, subsequent cell 6 per graft) plus, in aU cases, Swiss 3T3-J2 fibrob lasts (2 X 10 per graft, kindly lineage studies (see below) yielded sirniiJ r results, irrespective of provided by Dr. H. G reen). Immediatel y before grafting, cell s were trypsinized, whether they were perfo nned w ith kerarinocytes de1i ved fi· o m hair o ns, co n ce ntr~te d by low speed ce ntrifu g~ tion . mixed in v;uious con1binati and bud, interfollicular, or unfi:actionated epidermis preparations. Dense cel l suspensions were inj ected into sili con transplantation chambers Thus, the abi li ty of keratinocytes to form hair fo llicles is not lost implanted onto the do rsa l surf.1ce of nude mice (Datto eta/, 1988) . One week upon culturing, and is depe ndent on cells that are n ot n ecessarily an after grafting the transplantation chambers were removed. Mice were routinely s, even if they may still be associated with sacrifi ced 6 wk after graftin g. Grafts were f1xed in 4% paraformaldehyde at 4°C integral part of hair bud for 8 h, rin sed in phosphare-buflered saline (PBS), then pla ced in 30'){, sucrose these structures . in PBS containing 2 mM Mg lz overni ght at +°C. Samples were then embt!dded Cell lineage a nalysis of epidermis and h air follicle formation in OCT compound (Mi les, Elkhart, IN). with retrovirally labeled keratinocytes We next tested w hether the primary keratinocyte cultures contained individual cell s capable o f Histochemical staining of alkaJjnc phosphatase and citrulline Frozen and either a sections of grafts with rerroviraJi y labeled keratinocytes were fi xed in 4% generating on their own discrete tracts of the epidem1i s paraformaldehyde for 1Q- J 5 min and heated at 68°C for 30 min, to destroy wh ole hair foll icle or disc rete parts of it. For this purpose, cultured endogenous AP activity. AP colorimetric reactions were perfo rm ed fo ll owing prim ary keratinocytes deri ved fi:om albino nuce (Sencar) were infected the conditions o f Cepko 1'1 a/ ( 1993). BrieA y, sections were rinsed in X-P with a re trovirus carrying a human AP gene (Cepko eta/, 1993), under detection buffer (100 mM Tris-HC I, pH 9.5; JOO mM NaC I; 50 mM MgCI2) conditions w here o nly = 10-20% of cells expressed this marker. Fo ur for '10 min , and in cubated in X-P reac tion mix (0 . I mg 5-bromo-4-chl oro-3- o r five days after infection, cells w ere tryp sini zed, adm.ixed with dermal tetrazoli um per ml , in detection indolyl-phosphate per ml, I mg nitroblue papilla cells, and grafted. As w ith uni nfected cells, hair formation could bufrer) for 12 h in the dark. Slid es were rinsed in PBS plus 20 mM be detected by 3-4 wk of grafting. Mice were sacrificed at 6 wk, at a eth ylenediamine tetraacetic acid three rimes for 10 min pri or to mounting. time w hen reconstituted hair folli cles have unde rgone a t least two fi.1U Staining of the IRS for citrull ine-rich was perfonned by incubating (Lichti rl nl, 1993; Li chti, pers. paraformaldebyde-fixed fi·ozen secti ons with p-dimcthylaminoci nnamaldehyde hair cycles and have entered a third (0. 5% wr/ vol in 0.5 M HCI and 2 M NaCI) for 5 min at room remperawre, commun.). Grafts were analysed hi sto logically and after staining for as described by l:l...ogcrs and Harding {1 976). A P produccion. [ n all cases, histologic sections w ere heated at 68°C for 30 1nin, prior to the AP staining reaction. Control experiments h11111tmotluorescence Frozen sec ti ons of grafts of HALJ3 / c and C57BL/6 w ith sections of nonnal m ouse skin indicated that this he:1t treatmellt keratinocytes were incubated with normal goat serum ('I :5 in PBS for 30 min was sufticient to totall y destroy e ndogen ous AP activity that is at room temperature), and then treated with avidin/ biotin bl ocking solution specifi all y found in dermal papilbs (Handjiski cl a/, 1994). (Vector) according to th e nlanuf.'lcturer's instruc ti o ns. After washing twice When the reconstituted epidermis was analysed , no areas of randomly were in cubated overnight at 4°C with biotin-conjugated with PBS, sections mixed AP- positi ve (black) and AP-negative (white) cells were d etected. monoclonal aga inst the maj or histocompatib ili ty complex (M HC) Instead, distinct t racts of AP-positive ceLl s were found, w hich extended H2Kd hap lotype (1:10 di lution in 0.3% Triton X- 100, 1% BSA in P13S jbu Ae r units ti·om the basal to the uppermost differenciated layers AI, Pharmingen, San Di ego, CA) and affiniry pmifi ed antibodies as co lumnar (1 :500 dilution in buffer A, kindly provided by Dr. S. l-1. Yuspa) . Sections were (Fig 2). The average w idth of these units was 10-15 cells, but, on was hed fo ur tim es with buffer A fo ll owed by two washes with PBS , prior to rarer occasions, patches of 30 ceLl s were observed. 536 KAM IM U itA ET A L THE JOUI1..NAL OF INVESTIGAT IVE DERMATOLOGY

fresh H.B.+ D.P. cultured krtcs. + D.P.

F ig ure 1. Efficie nt hair forming ability of cultured primary keratinocytcs 11eo·s us freshly p repared hair buds. H air reconstitu­ ti on ass~ys wae pcrtonncd as desc ribed in Mnr erials a111l Methods, usin g a con•bination of / either fi·eshl y prepared hair buds ( 1-1. 13.) or cultured prinoaoy mouse keratinocytes (k rtcs.) together with either rat derm al papilla cells (D .P.) or derm al fibroblasts (D.F.). Similar results ' were obtained with cultun:d keratinocytes derived fro on hair bud and interfo ll icular frac­ tions, and fi·om total epidermis. Mice were sacrifi ced 6 w k after grafting. -?indica te grafted areas. Histologic analys is confirmed th e recon­ fresh H.B. + D.F. cultured krtcs. + D.F. stitution of a fu ll y orga ni zed epidermis and hair fo ll icl es . The spccifo city of these assays was demonstrated by th e f:1ct that no hair formation was obse rved after graftin g of hair buds or cultured ke ratinocytcs admi xed with no rmal dermal fib roblasts rather than dermal papi ll a ce lls. Also, no hair formati on was observed after graftin g of dermal papilla cells in th e absence of keratinocytes (no t shown).

A

F igure 2. C lonal pattern of epidermis formation as determined by skin reconstitution experiments with re trovirally labeled keratinocytes. Primary kerati nocytes fi ·om albino mi ce (Senca r) were infected with a helper fi·ee recombinant retrovirus ca rrying the human AP gene as described in Materials 11111/ Meth ods. Infection conditions were such that only I 0% of keratinocytes expressed th e transduced gene, as ve rifi ed b y staining th e cultured cells for AP producti on at the time of grafting. Grafts were analysed 6 wk later both histological ly a nd for AP production. (A , B) Two difrerent regions of th e reconsti tuted epidermis. Note that th e columnar arrange ment of the AP- positi ve patches some- B times does not fo ll ow a sharp verti cal lin e. For instance, in th e patch shown in (B), AP- positi vity at the left edge appears to extend more in th e upper epidermal la ye rs than in th e lower ones. The reverse situatio n was observed in oth er cases. Scale bars, (A) 60 ~l m , (B) 180 J.ln l.

When the reconstituted fo lli cles were examined, none were com­ (Fig 3A,Jollicle 4, 3B). A third type offolbcle was freq uently observed, posed of a random mixture of AP-positive and AP-negative celJ s, as whi ch showed positive staining in its innermost r egion, was AJ)_ would be expected if bair fo !J icles were generated by the reassociation nega tive in its middle portion, and was again positive in its o utermost of keratinocytes in bulk at the time of grafting. Instead, as shown in layers (Fig 3A, fo llicle 5). T he converse pattern of stain.in g, with Fig 3, so me reconstituted folli cles were negati ve fo r AP staining AP-positive cells restricted to the middle regio n, was also fo und. (Fig 3A, _follicle 1), whereas others were AP-positi ve in their external Interestingly, in a smalJ minority of cases, AP-positive cells were region, and AP-negati ve in their central region (Fig 3A, follicles 2 and confined to an arc of the fo llicle rather than a w hole concentric region 3). Foll icl es with th e converse sta ining pattern were ;dso present (Fig 3A, follicle 6). Very few if any of the fo JJicles that were observed ···- .. ------VC L. 109. NO. 4 0 TOilEl\. 1997 KER AT ! NOCYTE / 1-\ A I! ~ FOLLIC lE PROGENIT \1.. CELLS 537

Table I. Quantitative hair follicle distribution o f AP-positive versus AP-negative cells''

Experim ent I (AP-positi ve cells: 10.0'!\,) # of foll icles 13.t 32 2 1 17 19 2 2 %counted 59.0 l.t .l 9.3 7.5 8.4 0.8 0.8 % expected 72.9 '.!.0 9.0 8. 1 0.9 0. 1 NA

.Experiment 2 (Al'-negative cel ls: I '1.1 %) # of fo llicles 5-1 6 29 6 2 0 0 t}(, co unted 55 .7 6.2 29.9 6.2 2. 1 0 0 %1 expected 70.2 9.9 9.9 8.8 1.1 0. 1 NA

11 1-l ni r folli cle rcconstinttion assays with rcrro,·ir:tll y labeled kcratinocytcs wt.:rc performed as descri bed in .\ lrJtcricl[s tmd Nlt:tlwds. Histologic sections fi·onl rwo indcpendL'nt expe;: ri1n ents wert· used for qu:lntifi c:~ rion of ril e v:1 rious rypcs of n.:co nsrim red hair tOllidc:s. afrcr staining ror AP. Onlv c ross-s~c ri o n s w ith all three m:lj o r r~ gion s of the hair fo ll iclt.: (shaft, tRS. and O ltS) c.lea rly distinguishable, were included in th e couming. Conccnrri c bl:l ck and white circles n:: f~r to fOlli cles with v:wious combin:u-ions of AP-positive and AP-ncgative Figure 3. Clonal pattern of hair follicle formation as determined by rl'gions, ;1s they Wl'TL' described jn the text and shown in Fig 3A. T he p~ rcc nt:-.ge of slcin reconstitution experin1ents with retrovirally labeled kcratinocytcs. kcrarino cytcs expressin g th e rer.rovir::d AP m:1 rk cr w:ts dcrcnnincd experin1 entall y ar the. T he experiment was th e same as that desc ri bed in Fig 2. Similar results were beginning of each ex periment. The expcctl."d p c rc ~..·magl' of vario us types o f tO IIi cles was obtained in four independent experiments with cultured keratinocytes deri ved based on rhc known starting fi·acrion of AP-positive and AP-ncgarive cells, :t nd on rhl' assu n1pdon fi·om eitl1 er unfractionatcd epidermis o r separate hair follicle or interfo ll icular rh:1t rhn.:e distinct progt'lt imr cells were involved in :-. H cases. ln o ur C:licubrions, we took into ace unt that type 2 and 3 follicles could bt: cqu:-.lly dc:rivcd from rw fi·actions. (II) Cross-secti ons of reconstituted fo ll icles, as described in the text; AP­ positivt: and o ne AP- ncgarive cel l or, alt~.:rn:~rivc l y, fi·om o ne AP-positive and rwo 1\P­ (B) longicudi nal sections of reconstituted fo lli cl es . Cx/ Md, cortex/ medulla. The nl' g:ltivc cd ls. There coul d b~..· :1 signi fi cant 11 1:1rgin of error in counting spec ifi c hair follicle 1~/i pnuel in (B) corresponds to a follicle with AP-positive staining of th e cortex / combinati ons, panicubrly type I fo lli cles. which m ight be undcrcsri marcd bcc.msc of rht: IR S regions but not of th e O R.S. T he lower central region of AP positi vity difficulty ofide11 tifying these unstained follicles in rhc histolog-ic s~c tion s . NA. not a ppli c:~ b l c. fo und in this fo llicle does not correspond, by morphology and size, to th e dermal papilla. The main ax is of the folli cle is c ut at a bias, and, by c hangin g the focus on the sec ti on, even th is lower AP-positive region could be tenta tively identified with th e hair shaft. The riglu paucl co rr·esponds to a similar longitudinal section of:1 fo llicle composed entirely of AP-nega t.i w cells, next to an adjacent folli cle orientated at 90° and also AP-negati ve. The total lack of staining of th ese folli cles illu strates th e point that, as mentioned in the text, th e h ~a t treatment of secti ons prior to staining was sufftcient to destroy the endogenous Al' ac ti vity normally present in skin ti ssues. Scale bar, (A) 80 ~1 11 1 , (B) '160 l..llll .

in cross-secti on were uniformly composed of AP-posi ti ve cells. Longi­ tudinal secti o ns confirmed this d iscrete pattern of populatio n cell distributio n (Fig 3B, lift pa ueQ. Interestingly, unlike in no rnul skin, reconstituted hair fo lli cles were rand o mly orie ntated, so that it was frequently possible to identify follicles at 90° of each o ther (see, fo r instance, Fig 3B, rig/It paueQ . The vario us types of reconsti tuted hair fo lli cles that were obtained in tvvo independe nt experiments were quantified (Table I). T heir expected fi·equency was also calculated, based o n the known percentage of AP-positive and AP-negati ve cells that were used at the time of grafting, and based o n the ass umption that three distinct progenitor cells a re involved. As shown in Table I, there was a rather good agreem ent b etween observed and expected values for most hair fo lli cle combinations; however, even if suggestive, this concordance of va lues cannot be taken as strictly statisti cally s igni fi cant, given the relati vely low number o f fo lli cles that were counted and the likelihood of bias in counting particular hair fo llicle combinations (for instance, by underestimating the number of totall y unstained, AP-negative fo llicles). In their middle and upper parts, hair fo llicles are composed of three distinct concenni c regio ns: the O RS, the IRS, and th e shaft (cuticl e, cortex, m edulla). Keratinocytes in each of these regions share specific m o rphologic and biochemical characteristi cs , ~ nd it has been suggested that they may o ri ginate from distinct progenitor cells in the bulb (rteyn olds and Jahoda, 1993). A n important q ues tio n was w hether the alternating zones of AP-positive and AP-negative cells that w ere identifted in o ur analysis coincided with the classically defin ed concent­ Figure 4. Distinct clonal origin of the ORS, IRS, and shaft - c ri c regio ns of the hair fo lli cle. Cells of the ll<..S are c haracteri zed by ortex/ medulla (Cx/ Md) - of reconstituted h air follicles . Adjacent histologic the producti o n of proteins s uch as n·icho h yaJin w1th a high. citrulline. secti ons of the reconstituted fo ll icl es were sta ined fo r either AP-cxpression con tent, and a histologic stain specifi c fo r - bo und citrulline ca n (A, C) or protein-bound citru ll in e ( B, 0 ) as described in J'.J.u crinls 1111d Methods. be used to id entify these cells (R ogers and H arding, 1976) . Accordingly, Specitlcity of detection of the IRS layers by the citrullin-stai ning technique seri al sections of the g rafted tissues were analysed by e ither AP o r was confi rm ed by ana lys is of longitudinal hair follicle sec ti ons. Srnle b11r, 40 l..llll . 538 KAMIMUil..A ET A L T H E JOUI1..NAL OF INVESTIGATIVE DEI1..M ATOLOGY

Figure 5. Clonal pattern of epidermis and hair follicle formation as deterntined by skin reconstitution experiments with keratinocytes belonging to different MHC. Primary keratin ocytes deri ved fi·o m BALB /c mi ce were cultured for ·r wk and admixed at th e time of graftin g with similarl y prepared keratin ocytes fi ·o m C57BL/6 mi ce (in a 1:3 ratio). Cr:lfi:in g co nditions we r ~ as described in Ma rerials orrd M c: tlro rls. Reco nsti tuted epid ermis and fo!Jicles were analysed 6 wk after graftin g by doubl e immunoAu orescence with affi ni ty-purified antiserum against keratin 5 {A) and mo\1.oclon·.1l antibody against the MHC J-l 2Kd hapl otype (B). Small ~ indi cate the basa l laye r of th e epidermi s, and large ~ point to the hair follicl es. Noti ce th e uniform staining with KS anti serum of both epidermis and oute r Io yer of th e hair folli cles, and th e limi ted stainin g of l-l 2Kcl antibodi es to a subset of hair fo llicl es and a discrete region of th e epiderrnis. Contro l ex perim ents were perfor·med to veri fY that the H2Kd aJJtibod ies were specific fo r BALB /c keratinocyres and did not r ecog n~i ze C57B L/6 ce lls, nor ce lls of the host nu de mi ce (w hi ch we re on a Swi ss geneti c ba ckgro un d) . Antibodi es aga in st th e K2 Kb hapl otype (fo r C57BL/6 ce ll s) ge nerated a high background and co uld not be used for this kind of analys is. (C) Lo ngitudin al sec tion of a reco nsti tuted hair folli cle :mel neighborin g epid ermis stained with anti- J-l 2K'1 anti bodies. The ~ points to th e transition between H2K"-pos itive and H2K"- nega ti ve epidermis. The data in Fig 5 are representati ve of at kast f1 ve fr elds. Similar results wen: observed in two ind epe nd ent ex perim ents. Sw/c b11r, (A, B) 160 J.lm , (C) 80 ~1111.

ci trulline staining. As shown in Fig 4, the concentric layers detected DISCUSS ION by the two different techniques did overbp, suggestin g that their cell s of origin actuall y coin cid e. The noti on of a "keratinocyte stem cell " is of central releva nce ro our Thus, distinct colunlllar units were fo und in the reconstituted understanding of epidermal development and ho meostas is. A grea t epidermis tl1 at more likely ori gi nate fi:om a single AP-expressing am ount o f effo rt lJJs been dedi ca ted to identify such cells and defin e progenitor cell. In contrast, the vast majority o f reconstituted hair their properties . In l!ivo studies have sh own th e presence of slow cycling follicles are derived fro m a mjnimum of two or, in a signifi cant fra cti o n (l abel- retaining) and presumptive self-renewing cells at privil eged sites, o f cases, three retrovira.lly labeled cells, o ne for the hair shaft , one for such as the bulge regio n o f the hair follicle (Cotsa relis el a/, '1990) or the IRS, and the third for the O R S region. the deep pa lmar ridges (La vker and Sun, l 983). Keratinocyte popula­ ti ons w ith grea ter proli feration po tential ha ve been demonstrated in Lineage analysis of epidermal and hair follicle formation with cuLture (Barrando n and G reen, '198 7; Jo nes

sac epidermis (Oli ver, 1970) . As m entioned, the mature hair follicl e A 8 c consists of at least seven differem !)'pes of diFferentiated keratin ocytes (S tenn et a/, 199 ·1; R.eynolds and Jahoda, 1993). O ur data indicate that these ce ll s ori ginate, at leas t under certa in co nditions, fi·o m multiple progenitor cells rather than a sin gle o ne . Folli cl es that stai ned entirely for the AP cell marker would have been expected if these structures o ri gi nated 6·01n a single cell. I nstead, th ere was a striking locaJi z;nion of sta ining to distinct concenu·ic regions of the hair fo lli cles. ln the case of fo ll icles w ith di fle rential sta ining of three concentric regions, a minimum of three progenitor cells are in volved , one for the generati o n of the shaft (c uticle, cortex, and m edulla), one for the IRS, and a third fo r the ORS. The diR-e rential stai ning of two rather than three concennic regions found in the o th er fo lli cles coul d also be the resul t of three progeni tor cells, w ith the o nes givin g rise to two neighborin g regions being of the same ra ther than diflerent colo r. The in volvem ent o f onl y two progenitor cell s in these cases is equall y possible, however. Only very few reconstituted hair fo lli cles were ho mogeneously com ­ D E posed of AP-positi ve celJ s. Such fo lJi cles co uld be equally derived fi·o m a single totipotent stem cell , or fro m the rare combination of two or three retrovirally infected progenito r cells. Follicles in w hich an AP­ ORS IRS Shaft ORS IRS Shaft positive cell populati on was confined to an arc rather than a w hole concentric regio n were also o bserved , but ;1t similarly low freq uencies. t t t t t t This indicates that the participation of more than three progenitor cells b~t-@ is also a poss ibl e b ur unlikely event, either because the number of .., t l (t.~T~\~ progeni tor celJ s is very smalJ and thus limiting, or because there is a \ I I GFs , I I 'v~·~GFs mechanism that positively restricts their association to o nl y three. On.ly - DP occas io nall y woul d a fo u1·th competent cell p articipate in hair folli cle ~ formatio n, possibly jo ining at a later sta ge than the other three. An interesti ng ques tion raised by o ur results woul d be that of the Figure 6. Possible or1gu1 of partially committed kcratinocytc stem cells, and thei r specifi c association. We have demonstrated that the three populatio n cell distribution in areas such as the bulge region o r the concentric regions of the hair follicle (OI<.S, IRS, and sh3ft) ori ginate from germinative epithelium reservoir, which had been previously implicated distin ct pr o~cniror cdls. As disc ussed in tht" rexr. three poss ibilities ca n be as sites fo r putative s tem cells (Cotsa rdis ct a/, '1990; R eynolds and envisaged for the origin of th ese ce lls: (A) they may derive fi·om ;1 totipotent Jahoda, 1993). T he limit of o ur analys is is such that we were not able kerati nocyte ste1n ce ll that is still present and operati ve in the mature hair follicle; to address this ques tio n. In f<1ct, th e reconstituted hair fo llicles are (B) they ma y dcri 1·e fi·o 1n :1 totipotent 1-' rogenitor cell durin g devdol'nH:m, but randoml y o ri entated roward each o ther (see, for inst:m ce, Fig 3B), ex ist as indqJe ndent enti ties in the mature h;1ir fo ll icle; (C) they moy be and their main axis is probably sli ghtly slanted. For this reason, it is intrinsica ll y the sam e, but become conimitted ro ditre1·entiate along diflt!rent quite difficult to obtain good longitudinal sections of hair fo lli -Jes, and p:1 thw:1ys in respo nse ro stinwli fi·otn rh !..! local environment. For rhrct' progenitor cells to co to identify ~ p ec ifi c regions such as the bul ge in sufficient numbers to me together ond ori gi nate a hair follicl e, th ey n1ust either interact specifica ll y IVith cach other (D), or usc a spec ific anchoring mec make o ur analysis signifi ca nt. For this sam e reason , we could onl y hanism. whi ch co ul d be provided by dermal pap il b ce lls and/ or thei r extracellular matri x (E) . observe th e lower and upper porti ons of the ORS on the same plane T his latter mec hanis m could co ntribute to the hair- in duci ng ca pabilit)' of dermal in a vel)' limited numb er o f fo llicles . O ne such case is shown in Fig 5, papi lb cells (0 1'), rogethcr with the likel y. bur os yet elusive rel ease of hair­ w hi ch illustrates the continuity of the lower and upper ORS region indu cing growth f.1cto rs (G F). o f a reconstituted hair fo ll icle w ith a tract of the neighboring epidermis, w hich would be co nsistent w ith th em originating fi·om a single progenitOr cell. As fo r se b :~ceo u s glands, onl y very few if any of these co mmitted progenitors may be rehtively independent of each other in structures were found in our grafts, leaving open the q uestion of their the mature hair fo lli cle, and possibly reside at diFferent locatio ns population cell colllpositio n. (Fig 6B). Finall y, the progenitor cells givi ng ri se• to distinct parts of R econstituted hair fo lli cles were exam in ed at 6 wk after graftin g, at the hair fo lli cle may not be intrinsicall y difl:erent fi-o m each other, but a time w hen they have undergone two full hair cycles and therefore may becon1 e con11nitted to a specific cell lineage as a fun ct:i o n of their are not the di rect product of the original mixture of cell s used fo r micro environment (for instance, o nce they reach the bulb region at grafting ( Lichti cl n/, 1993; Li chri, pers. commun.). T he tact that these the beginning ofanagen) (Fig 6 C). foU icles remain composed of distin c tl y labeled cdl populations over M ore defini6 ve answers to the ori1,ri n of hair fo lli cle p rogenito r cells this time is consistent w ith di stinct cell s of o ri gin remaining operative could be provided by performing skin/ hair reconstitmion assays similar iu vi11v , even after th e ini tial reco nstitution b y graftin g and for at least to the ones described here w ith donally deri ved keratinocyte popula­ two consecutive hair cycles. ti o ns. Unfortunately, the present assays are based o n m o use primal)' Our fi n li ngs indi ca t~ that primary keratinocyte c ultures derived keratinocytes, and these cells are vet)' difficult to expand in large fi·om newborn mice are likely tO contai n multi ple progenito r cells w ith enough nu1n bers from individua l cl o nes. Thus, it wi ll be vei)' impo rtant a partially committed cli fl:e rentiati o n p otential. By analogy with the to test w hether human kerarin ocytes can also be employed in these bone marrow, these cells may o t~g in ate fro m a single undifre rentiated assays, as clonal growth of these cells ca n be rea cli.l y obtained , and stem cell that may exist even in th e newbo rn and/ or adult skin, and clones with diA:e rent proliferative potential have alre;1dy been well reside at some privileged site(s) , such as the bulge of the hair fo Lli cle. characterized (Barrandon and G reen, 1987; Jones and Watt, 1993). T his cell would then give 1i se to pa rtially co mmitted progenitor cells O ur skin reconstitution assays involved grafting of d isso iated cell.s resn-icted along a specifi c cell lineage, either generating them at a in suspension. U nder these conditions, for each hair follicle to o riginate constant rate, or p eriodica ll y, during the hair cycle (Fig 6A). Although fro m two or three distinct progeni tor cells, these cells must be able to we found li ttl e evid ence fo r totipotent stem cells present in our primal)' recognize and adhere to each other through som e specifi c m echanism (s). keratinocyte cultures, these cell s may still be present in very low Although these imeracrions could be direct (Fig 6D), it is tempting numbers, or they co uld be hard to in fe ct by the retrovi rus used for to speculate that they might be promoted by the presence of th e other our studies. Alternatively, n10re co111111i tted progenito r cells may cell rype essential for hair formation , the cl ermal papi Ll a cells (Fig 6E). ori ginate 11-o m a totipotent stem cell during development, fo r instance In f.1ct, whereas the capability of these latt.: r cells to produce hair­ as the result of th e first dermal mess;1ge. In this case, these m o re inducing growth f:1ctors is sti ll a matter o f specubtion, th eir tendency 540 KAM IM U I ~A ET AL THE J OURNAL OF INVESTIGATIVE DEI"tMATOLOGY

to form specific aggregates is well established (Jahoda and O li ver, 1984; Jahoda CAJ3 , R eynolds A.J : Denn al-cpidenml interactions- fo ll icle derived cell populatio ns Messenger et a/, 1986). Finall y, epidenn is and hair fo lli cle progenitor in the study of h ::~ir-growt h m echanisms. J fltPest Demwtol 10 I :33S-3SS, 1993 Jahoda CAO, H ornt: KA , O li ver JtF cells w ith broad differentiation potential ca n be maintai ned in culture : Inductio n of hair growth by impl:-~m:-~r i o n of cu ltured denm l papill a cell s. N11 1111r 3 11 :560-562, I 984 fo r at least 1 wk. This provides a very attractive sys tem f or future Jones Pl-1 . ·wntt FM: Sc p ar::~tio n of hun1:111 e p i d e rm::~] stem cells fi·ont rr:111 sit ;llltplify ing studies on control mechanisms - and agents - that inAu ence the cells o n the basis of difrercnces in inrcgrin fu nction :-~nd cxprcssiou. Cell 73:7"13- number of keratin ocyte stem cells, and their ba lance between growth 724, 1993 and differentiation. J o nes PH , H arper S, Watt FM: Stem cell p.:1ttc rning :md f.1tc in human epidermis. Cell 80:83-93, I 995 Kobayashi 1< , R och:J.t A, l3a rrandon Y: Scgrcg:nion of keratinocytc colo ny-forming cells in the bulge of the rat vibrissa. Pmr Nat! Amd Sci 90:7391-7395, 1993 La vker l ~M . Sun TT: Epidennal stem cells. ) l11vcst D cm"'tvi 81:12 JS-127S. 1983 vVe an: 11ery gmtiful w Drs K. ·. Stell// n11d Ste11e 11 P mut y for helpf111 advise for th e Lic h ti U, Weinberg WC, Goodm::t n L, Lcdbc[tcr S, Dooley T, N1organ D , Yuspa S!-1: In skirr /hair ruco 11 slit11tio11 assays, Dr: Sne111a ) eo 11 for help 111itlr tlr r initialretmt1iml it ifcclioll vivo regul ntio n ofnlllrinc hair growth: insights from gr:1ftin g defin ed ce ll popul:ttio ns e;\CperilltCIIt s, a11d Dts Jim Rlrei11111nfd arr d Pn11f Coe ti11ck fo r crilical tmdi11g ro liferaci ve potential of different keratinocytes of plucked human hair foll icles. j !t wes t Dcmwt,)/ 105 :.14-2 1, 1995 M o rris l"tj. Fischer SM. S l :~ga TJ: Evickncc that the centrally nnd peripherally located cells REFERENCES in the muri ne epidermal proliferative u nit are two distinct cell populatio ns. J l11 Pt..'Sf Denl/(uol 84:277-28'1, 1985 13nrra!1don Y. Green H : T hree.:: clolwl types of kcr:ltin ocyn: with difiCrcnt cnpnciri cs fo r O li ver lt F: T he induction of hnir fOll icle formati o n in rh c :~du l r h 111ultiplkation. Pmc Nat! Acari Sci 84:2302-2306, 1987 ooded r:1t by vibrissa dermal papill ae. J E11tlir)'ol Exp Mot)J!wl 23:219-236. '1970 Cepko L, R.yder EF, Austin C P. Wa lsh C. Fekete DM: Li11 e:~ge analysis using retrovirus Pottcn CS: T he cpicknti:JI pro liferative unit: the possible ro le of the ct• ntr:ll ba vectors. i\1/ctlwds t'11 E11 ;:inwl 225:933-960. 191J3 sa l cell. C(' /1 Tiss 11 e Ki11ct 7:77-78, 1974 o ts:1 relis G, Sun TT. Lavkcr R.M: L:Jbcl-rctaining cells reside in the bulgt: arc::~ of J ~cy n o l ds AJ , Jaho da CAll: H air fibre progen itor cell s: developm ental status anq interactive pi losebaceous unit: implic:1tions for fo ll icular stem cells. h :~ir cycle, and skin potenti al. Sc 111i11ars Dwl'l Bio i 4:24 1-250, carciJJ ogcncsis. Cell 6 1:1321)-1337, '1';)90 J 993 R och;,t AJ. K ob:-~yas hi K, lhrrandon Y: Loc:ttion of stem cc: lls of human follicles by clonal Datto G P, "\VeiJtbl!rg RA, Ariza A: Malig JJ :l l1t tr:t nsformation of mo use prirn:try kcrarinocytcs analys is. Cr/1 76: 1063-1073, 1994 by H nrvcy sa rco m:~ virm and its m odubrio n by su rrounding normal cells. Pmc J\Tm/ R ogers GE, H ard ing H WJ: In: 1<. Tod:1 t:t rd., eds, Biology (/lt d / lmd Sri USI J H5:6389-n393, 19HH Disease '!f f-lair. University of Tokyo l'ress, Tokyo, '1976, pp 4 11-435 Garlick J ATaichlll:ll1 Lil: A model r.o sw dy the f.1te of gcncrica ll y-mnrked kcr:ltinocyres in Sengcl P: Pattern formation in ski n development. ltu J Dev Bioi 34:33-50, 1990 culru re. J Demwrol 19:797-SO I, l 9Y2 Stcnn KS. M essenger AG, J3ad en I-ll' (eels) (199 1) The molecular and structur:Jl biology 1-l nndjisk i BK, Ei chmUll er S. H ofi-nann U . Cznrn ctzki DMOnus RP: Alka line phosph:na se ot· h;1ir. 11 111 wls q/ t/1 (~ Nt..w Yurk Awde111 y !!( Scicuccs 642, New York, 19Y I activity and localization during the murine hair cycle. Btj Dt•mJIItoiJ31 :303-31 U, 1994 Weinberg WC, Goodman LV. George C, M organ D L. Ledbetter S, Yuspa S J-1 , Li ch ri U: H ardy MJ-1: The secret life of the hair fol li cle. Jinu/s i11 Ce11 etirs 8 (2):55-61, 1992 l"tcconstitution of h :~ir folLicle dl.:!vc lopmcm in vivo: dcce rrnjnatio n of follicle Hennings H. Mi chae l D. C heng C. Steinert P, H o lbrook KYuspa Sl-1 : Calcium regld ation formation, hair gro\-vth, and hair quality by dermal ccUs. J fl we.s t Dcmwtol 100:229- of growth and diOCrentiat ion of m ouse epidermal cells in culture. Cell "llJ:2 45- 236. 1993 254, 1980 Y:~ng J-S, Lavkcr l"t.M. Sun TT: Upper hum:-~n hnir folli ck co nt:~in s a subpopulation o f j::d10da CAB, O li ve r RP: Vibrissa d e rm:-~1 pnp ill ;-~ cell ;,ggreg<~t i vc behaviour jn vivo :Hid keratinocytes w ith superi o r in vitro pro liferative potential.) bwc.H Dcmwrol '10 I :l152- in vi tro. Embq J ~o l Exp NIMph 79:2 11-224, 1984 659. 1993