Mouse digit tip regeneration is mediatedby fate-restricted progenitor cellsJessica A. Lehoczkya, Benoît Robertb, and Clifford J. Tabina,1

aDepartment of Genetics, Harvard Medical School, Boston, MA 02115; and bInstitut Pasteur, Génétique Moléculaire de la Morphogenèse, Centre National dela Recherche Scientifique, Unité de Recherche Associée 2578, F-75015 Paris, France

Contributed by Clifford J. Tabin, November 1, 2011 (sent for review September 19, 2011)

Regeneration of appendages is frequent among invertebrates aswell as some vertebrates. However, in mammals this has beenlargely relegated to digit tip regeneration, as found in mice andhumans. The regenerated structures are formed from a mound ofundifferentiated cells called a blastema, found just below the siteof amputation. The blastema ultimately gives rise to all of thetissues in the regenerate, excluding the epidermis, and has classi-cally been thought of as a homogenous pool of pluripotent stemcells derived by dedifferentiation of stump tissue, although thishas never been directly tested in the context ofmammalian digit tipregeneration. Successful digit tip regeneration requires that thelevel of amputation be within the nail bed and depends on ex-pression of Msx1. Because Msx1 is strongly expressed in thenail bed mesenchyme, it has been proposed that the Msx1-expressing cells represent a pluripotent cell population for theregenerating digit. In this report, we show thatMsx1 is dynamicallyexpressed during digit tip regeneration, and it does not marka pluripotent stem cell population. Moreover, we show that boththe ectoderm and mesoderm contain fate-restricted progenitorpopulations that work in concert to regenerate their own lineageswithin the digit tip, supporting the hypothesis that the blastema isa heterogeneous pool of progenitor cells.

epimorphic regeneration | lineage analysis | transdifferentiation |fibrin clot | mitogenic signal

Regeneration of appendages occurs widely throughout theanimal kingdom, although it is limited in higher vertebrates.

Although many fish regenerate fins and a number of urodele andlarval anuran amphibians regenerate entire limbs, in mammalssuch as humans and mice appendage regeneration is limited tothe distal digits. Regeneration begins with the formation of aspecialized regenerative epithelium across the plane of ampu-tation that is required for regeneration to proceed (1, 2). His-tological analysis of amphibian limb regenerates suggests that theregenerative epithelium forms by the migration of stump epi-dermal cells, although additional contributions from other cellpopulations have not been ruled out (3). The formation of theregenerative epithelium is followed by an accumulation of mes-enchymal cells distal to the stump, which proliferate to forma bud-like structure known as the blastema. These blastema cellssubsequently differentiate into the musculoskeletal and connec-tive tissues of the regenerated appendage.It is critical to understand the origin of blastema cells and the

degree to which they represent a multipotent cell population.The blastema could be formed from a preexisting multipotentstem cell as found in planaria; however, vertebrate researchsuggests, at least in amphibian limb regeneration, that the ma-jority of blastema cells are from dedifferentiation of the maturetissues of the stump (4, 5). In principle, blastema cells couldremember their origin and only differentiate into the cell typesfrom which they arose. Alternately, blastema cells could bemultipotent, capable of transdifferentiation into all cell types ofthe regenerate. Indeed, transdifferentiation has been describedin both axolotl tail regeneration (6) and amphibian limb re-generation experiments in which certain tissues were removed orirradiated (7, 8). However, recent lineage analyses after ampu-

tation of both the axolotl limb and the zebrafish fin stronglysuggest that transdifferentiation does not significantly contributeto the regenerates, and that instead the blastemas are made up oflineage-restricted cell populations (9–11).Digit tip regeneration has been reported in mammals including

mice and juvenile humans (12, 13). Amputations of the terminalphalanx through levels associated with the nail organ are capableof regeneration, whereas more proximal amputations are not.Intriguingly, mesenchymal nail bed cells in neonatal mice andhumans express the transcription factor Msx1 (14, 15), which isa transcription factor expressed in the proliferative, un-differentiated regions of regenerating newt limbs (16, 17). Thisexpression may reflect a functional importance of Msx1 in blas-tema formation andmaintenance. Msx1 induces dedifferentiationand proliferation of cultured myotubes, which are ultimately ca-pable of being redifferentiated down multiple lineages (18). Im-portantly, Msx1−/− digits do not regenerate in ex vivo embryoculture. Moreover, Msx1 is necessary for Bmp4 activity, whichcould rescue the ability of these Msx1−/− digits to regenerate (19).Msx1 expression is not observed in the digit tip blastema itself(20), suggesting that its critical function may be in the nail bedmesenchyme. Given the requirement of the nail bed for successfulregeneration and the association of Msx1 with undifferentiated,multipotent cells in both developmental and regenerative set-tings, Msx1-expressing nail bed mesenchyme could representa pluripotent stem cell population necessary for digit tip re-generation. Alternately, the Msx1-expressing nail bed mesen-chyme could serve as a signaling center, producing critical factorsrequired for regeneration.To determine whether mouse digit tip regeneration involves

the recruitment of multiple cell type determinant lineages or theformation of a pluripotent progenitor pool, and to answer therelated question of whether these blastema cells are derived froma preexisting stem cell population or from dedifferentiation ofmature tissues in the digit stump, we undertook a lineage analysisusing inducible alleles of Cre recombinase expressed in specifictissues within the neonatal limb.

ResultsDescriptive Analysis of the Process of Digit Tip Regeneration. Todevelop a context for our lineage studies, we established a timecourse of the early events of mouse digit tip regeneration. Inpreliminary experiments, we found that regeneration reliablyoccurs when the distal-most 400 μm are amputated on postnatalday 3 (PN3) CD1 neonatal mice (SI Text and Fig. S1). This rep-resents ≈50% of the length of the PN3 distal phalanx and trans-verses the proximal nail bed, although digit size and regeneration

Author contributions: J.A.L. and C.J.T. designed research; J.A.L. performed research; B.R.contributed new reagents/analytic tools; J.A.L. and C.J.T. analyzed data; and J.A.L. and C.J.T.wrote the paper.

The authors declare no conflict of interest.1To whom correspondence should be addressed. E-mail: tabin@receptor.med.harvard.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1118017108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1118017108 PNAS | December 20, 2011 | vol. 108 | no. 51 | 20609–20614

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

can vary with strain (Fig. S2). CD1 mothers were used in all sub-sequent experiments, except where specifically noted.Over the first few days after amputation of the distal 400 μm

of a PN3 digit, the epidermis closes across the wound. As haspreviously been noted for adult mouse digit tip regeneration, the

steps of epidermal closure are the same from digit to digit,whereas the timing is variable (21). It is not unusual to observe 1-to 2-d variability in each regenerative stage among littermates orbetween different genetic backgrounds or drug treatment groups.Fig. 1 shows an idealized course of events. Immediately afteramputation a superficial blood clot forms at the amputation site.One to three days after (post) amputation (dpa), the existingepidermis retracts and attaches to the terminal phalanx at a levelproximal to the amputation plane (Fig. 1 A and A’). The epi-dermal retraction leaves the distal tip of the stump bone pro-truding into the clot (Fig. 1 A and A’), and this bone sub-sequently fragments distally. Between 3 dpa and 7 dpa the clot ismaintained and integrates the residual bone fragment while theepidermis closes over the amputation, yet underneath the clot(Fig. 1 B–E and B’–E’). The regenerative blastema forms underthe new epithelium by 6 dpa (Fig. 1 D and D’). By 2 weeks afteramputation (wkPA) (Fig. 1 F and F’) the clot begins to sloughoff, and the majority of the bone regeneration is complete, andby 3 wkPA regeneration is complete (Fig. 1 G and G’). Thiscourse of regenerative events corroborates the postamputationevents of adult digit tip regeneration (21), underscoring thesimilarity in these two processes.

There Is No Transdifferentiation Between Ectodermal and MesodermalLineages During Digit Tip Regeneration. Transdifferentiation be-tween ectoderm and mesoderm has been described during am-phibian tail regeneration (6). To determine whether epidermallyderived cells contribute to the regenerating digit tip blastema, weused the Krt14-CreESR1 tamoxifen-inducible Cre allele (22) andeither the R26R-lacZ Cre reporter allele (23) or the R26R-CAG-tdTomato Cre reporter allele (24) to fate-map the stumpkeratinocytes. Keratin 14 is a well-established marker expressedin mitotically active keratinocytes in the basal epidermis. Ta-moxifen was administered to PN0 pups. PN3 400 μm digit tipamputations were carried out, and litters were analyzedthroughout regeneration to follow the fate of the keratinocytedescendants (Table S1).

Fig. 1. Lineage of Krt14-expressing keratinocytes during digit tip re-generation. (A–E) X-gal stained (blue) Krt14-CreESR;R26R-lacZ digit sectionsfrom 3 dpa to 1 wkPA. (F and G) TdTomato (red) fluorescence overlayingdifferential interference contrast brightfield (grayscale) images of Krt14-CreESR1;R26R-CAG-tdTomato digit sections at 2 wkPA and 3 wkPA. Epider-mal retraction and proximal attachment to terminal phalanx occurs between1 and 3 dpa (A and A’, arrows). Dotted line shows approximate plane ofamputation. Distally exposed bone becomes integrated into clot (B, B’, C,and C’) as epidermis covers the amputation site under the clot (C–E and C’–E’). Blastema formation occurs after epidermal closure (D and D’). bl, blas-tema; b, bone; c, clot; ct, connective tissue; e, epidermis. (Scale bars,100 μm.)

Fig. 2. Lineage of Sp7-expressing osteoblasts during digit tip regeneration.X-gal stained Sp7-tTA-tetO-EGFP::Cre;R26R-lacZ digit sections. (A and A’)Images at 2 wkPA show descendants in clot (arrow), bone, and blastema, butnot the epidermis. (B and B’) Images at 3 wkPA show descendants in boneand connective tissue but not epidermis. Staining resembles normal 3 wknonregenerative digit osteoblast contribution (C and C’). Insets in A’–C’ show40× magnification. bl, blastema; b, bone; ct, connective tissue; e, epidermis.(Scale bars, 100 μm.)

20610 | www.pnas.org/cgi/doi/10.1073/pnas.1118017108 Lehoczky et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

Because the Krt14-dependent Cre genetically causes re-combination and hence activation of the reporters exclusivelyin the PN0 epidermal cells and is not active during regeneration,x-gal staining allows for lineage analysis of stump keratinocytesthroughout regeneration. At 3 dpa, marked cells are seenthroughout the epidermis surrounding the lateral stump andextending to the retracted edge of the epidermis, where it reat-taches to the bone, just proximal to the amputation plane (Fig. 1A–C). By 6 to 7 dpa, stump-derived keratinocytes are clearlyincorporated into the newly formed epidermis (Fig. 1 D, E, D’,and E’). From these experiments we find that most, if not all, ofthe regenerated epithelium is descended from preamputationkeratinocytes. As seen at 3 dpa, there are a few unlabeled cells inthe regenerative epidermis, likely reflecting incomplete pene-trance of the Cre activity, although we cannot rule out a minor

contribution from a distinct cell pool. The epidermis continues toconsist of marked cells (tdTomato) at 2 wkPA and 3 wkPA (Fig.1 F, F’, G, and G’), indicating that the marked keratinocytes thatform the initial regenerative epithelium continue to maintain theepidermis as the digit tip regenerates. Hence, epidermal stemcells as well as mature keratinocytes of the regenerate derive fromthe ectodermal tissue overlying the stump. Most importantly, ouranalysis of postamputation digits up to 3 wkPA does not showsignificant keratinocyte descendants in any tissue type other thanthe epidermis. Krt14-expressing keratinocyte descendants are notfound in the clot, bone, or connective tissue, demonstrating that notransdifferentiation occurs from the ectodermal germ layer duringdigit tip regeneration. Conversely, we never see contribution to theregenerative epidermis from the various mesodermally derivedcell populations analyzed below.

Regenerated Bone Derives from Lineage-Restricted PreamputationOsteoblasts. The distal 400 μm of the PN3 mouse digit consistsprimarily of bony and connective tissues, with nomuscle or tendonpresent at this level. Moreover, there is no cartilage: the terminalphalanx forms during limb development by two discrete processes,endochondral ossification via chondrocytes limited to the growthplate of its proximal end, and direct ossification via osteoblasts atits distal tip (25). Although muscle, tendons, and cartilage are allabsent from the amputation plane, it was nonetheless possiblethat these proximal tissues could contribute pools of cells thatwould migrate into the blastema after amputation injury. To testthis, we addressed the possibility that chrondrocytes contributeto the regenerated bone with the Col2a1-CreERT allele (26)with the R26R-lacZ Cre reporter line and found that Col2a1-expressing chondrocytes do not contribute to any tissues of theregenerated digit (SI Text and Fig. S3).We turned our attention to the descendants of the pre-

amputation osteoblasts, to verify that they give rise to the regen-erating skeletal tissues and to test whether they remain lineage-restricted or dedifferentiate to give rise to multiple tissues in theregenerated digit. To address the contribution of preamputationosteoblasts to the regenerate, we used the Sp7-tTA-tetO-EGFP::Cre allele (27) with theR26R-lacZCre reporter line. Sp7 is knownto specifically mark osteoblasts in both endochondral and intra-membranous bones. In this cross, Cre is expressed in osteoblaststhroughout development until doxycycline is administered on PN0and throughout the rest of the experiment to repress Cre ex-pression. PN3 digits were amputated and regenerates were ana-lyzed 2 wkPA and 3 wkPA. (Mice developed and regeneratedsignificantly slower than animals from the other crosses in thisreport. In this set of experiments, all regenerative eventsare slightly delayed.) By 2 wkPA, Sp7-expressing osteoblastdescendants are found in the blastema, the newly forming bone,and the clot at the distal tip of the regenerate, but not the epi-dermis (Fig. 2 A and A’). The large concentration of osteoblastdescendants in the clot likely reflects the observed trapping of afragment of the distal bone into the clot as the epidermis closesunderneath it (Fig. 1). By 3 wkPA the clot has sloughed off, andthe digit is largely regenerated, with preamputation osteoblastscontributing exclusively to the regenerated bone and the adjacentperiosteum (Fig. 2 B and B’). In unamputated contralateral digits,Sp7-expressing osteoblasts are also found to contribute to thesetwo tissues by 3 wk (Fig. 2 C and C’), thus indicating that pre-amputation osteoblasts serve as fate-restricted progenitors, onlyrepopulating tissues in which they were normally expressed. Im-portantly, no x-gal stained cells were found in the epidermis, in-dicating that there is no transdifferentiation of mesodermallyderived osteoblasts into the ectodermal lineage.

Msx1 Is Expressed More Broadly than Previously Appreciated. Thenecessity of Msx1 for successful digit tip regeneration (19) led usto evaluate its expression pattern throughout the process, toensure that we have a complete understanding of the tissuesinvolved. Using the Msx1-nlacZ knockin line (28), we analyzed

Fig. 3. Expression of Msx1 during digit tip regeneration. Msx1-nlacZ x-galstained digit sections during regeneration. PN3 (A) and 3 wk (B) unampu-tated digits show expression in dorsal and ventral dermis and throughoutbone and sweat glands. (C and C’) Expression at 3 dpa is in normal stumpdomains and in the clot. (D and D’) Expression at 5 dpa is in the clot anda few bone cells where epidermis is closing under the clot (arrow). (E and E’)Expression at 1 wkPA remains in the clot but is absent from the blastema. (Fand F’) Expression at 3 wkPA returns to a nonregenerative expression pat-tern and resembles B. Epidermis is outlined in higher-magnification pictures.bl, blastema; b, bone; c, clot; e, epidermis; np, nail plate; sg, sweat gland.(Scale bars, 200 μm.)

Lehoczky et al. PNAS | December 20, 2011 | vol. 108 | no. 51 | 20611

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

preamputation PN3 digit tips and found the nail bed mesen-chymal expression to be the largest and most homogenous do-main of expression (Fig. 3A). Msx1 was also expressed in thelateral dermis (still contiguous with the nail) as well as, to a lesserdegree, the ventral dermis (not associated with the nail) (Fig.3A), domains that were not previously reported. Additionally,many lacZ-positive cells are found within the proximal articularcartilage, the growth plate, tendon attachment points, and theperiosteum (Fig. 3A), all of which have been previously noted forlong bones (29). Msx1 is also expressed within the sweat glandsof the ventral fat pad, blood vessels, and hair follicles, but thesetissues are not directly analyzed in this study.During regeneration there are dynamic changes to the ex-

pression ofMsx1 at the amputation plane. By 3 dpa the expressionof Msx1 in the stump seems unchanged, but a new expressiondomain forms in the clot surrounding the piece of trapped bone,in which there are also sparse Msx1-expressing cells (Fig. 3 C andC’). At 5 dpa Msx1 expression is still maintained in the clot, andadditional expression can now be found in the bone, perhapsosteoclasts (29), specifically where the epidermis is closing underthe clot (Fig. 3 D and D’). By 7 dpa the epidermis has closed, andthe blastema has formed. Msx1 expression remains in the clot andis absent from the blastema (Fig. 3 E and E’). By 2 wkPA the clothas sloughed off, and by 3 wkPA (Fig. 3 F and F’) regeneration iscomplete and the expression ofMsx1 once again resembles that ofan unamputated control digit (Fig. 3B).

Cells Expressing Msx1 in the Preamputation Digit Contribute Broadlyto the Regenerate. To explore the contribution of Msx1-expressingcells to the regenerate, we used the Msx1-CreERT2 tamoxifen-inducible Cre allele (30) with the R26R-CAG-tdTomato Cre re-porter line* (24). Intraperitoneal tamoxifen was given to PN0pups, followed by digit tip amputation on PN3. By 1 wkPA (Fig. 4A–A’’’), descendants of Msx1-expressing cells are found in the clot(Fig. 4 A’ and A’’’) as well as the blastema (Fig. 4A’’), but not theepidermis. The descendants found in the clot are consistent withthe Msx1 clot expression domain (Fig. 3 C–E) and may be thesame population of cells. The descendants found in the blastema,

however, represent a cell population that does not actively expressMsx1 (Fig. 3E), indicating a change in gene expression and per-haps cell type as the cells enter and proliferate in the blastema. By2 wkPA (Fig. 4 B–B’’) the clot is mostly lost, and Msx1 descend-ants are found within the bone and dermis, but not the epidermis.At 3 wkPA (Fig. 4 C–C’’) Msx1 descendants have heterogeneouslypopulated the bone and dermis of the regenerate, resemblingthe normal expression pattern of Msx1 at this stage (Fig. 3F).Importantly, Msx1 descendants were not found within the epi-dermis, once again indicating that there is no transdifferentiationbetween the mesodermal and ectodermal germ layers. BecauseMsx1 is expressed in a diverse set of cell types before amputation,we cannot conclude that Msx1-expressing cells do not trans-differentiate into another Msx1-expressing cell type. However, it isclear that the set of cell types expressing Msx1 before amputationis congruent to the cell types to which their derivatives contribute.Because Msx1 is expressed more broadly than just the nail bedmesenchyme, and because no alternative unambiguous markersexist for this tissue, we cannot determine the fate of these specificcells in the regenerate. Thus, it remains unclear whether the nailbed mesenchymal cells give rise exclusively to the regenerated nailbed mesenchyme and whether they represent a true cell type orsimply a dense accumulation of dermal cells.

Regeneration Is Impaired by Removal of Fibrous Clot. In our lineageanalyses we found that Sp7 and Msx1 descendants populate theclot after amputation (Fig. 2A and 4A). To address whether thefibrin clot at the distal tip of the regenerating digit serves a re-generative function in addition to classical wound healing, weremoved the clot at time points correlating with the closure ofthe regenerative epithelium. For technical reasons the clot couldnot be removed cleanly before 6 dpa. Regeneration, as assayedby bone outgrowth, is compromised compared with controlregenerates, when the clot was removed at 6 dpa or 7 dpa (P <2.205e−14 and P < 0.0001, respectively), but not 8 dpa (P < 0.416)(Fig. 5). Moreover, removal of the clot at 6 dpa often producessmall regenerates that near the size of a nonregenerated digit(Fig. 5 B and G, red line), consistent with the clot serving anessential regenerative function. Although the results of thesemanipulations were statistically significant, there was consider-able variability in the outcome (Fig. S5). The newly formedepidermis that lies just under the clot structure is fragile andeasily punctured or removed. Mechanical removal of the clot,although possible, is difficult and can result in fresh bleeding.These instances were documented, and the data were left in theanalysis. Thus, the data, as presented, may underemphasize the

Fig. 4. Lineage of Msx1-expressing cellsduring digit tip regeneration. TdTomatofluorescence from Msx1-CreERT2;R26R-CAG-tdTomato digit tips. (A–A’’’) Descend-ants at 1 wkPA are in blastema (A’’) and clot(A’’’) but not epidermis (A’–A’’’). (B–B’’)Descendants at 2 wkPA are found in boneand dermis but not epidermis. (C–C’’)Descendants at 3 wkPA have reestablishednormal Msx1 expression pattern of thedistal tip. (A–C) TdTomato expression withno overlay. (A’–A’’’, B’, B’’, C’, and C’’) Dif-ferential interference contrast brightfieldimages (grayscale) with tdTomato (red)fluorescent overlay. bl, blastema; b, bone;cl, clot; d, dermis; e, epidermis; np, nailplate. (Scale bar, 100 μm.)

*The R26R-lacZ reporter allele used in some of the previous experiments generally dis-plays robust expression when activated via Cre-mediated recombination and was usedinterchangeably with the R26R-CAG-tdTomato reporter. However, in some tissues re-porter expression is much weaker and hence the use of the R26R-CAG-tdTomato allelewas favored, as was the case here (Fig. S4). Although less dramatic, the staining from theR26R-lacZ reporter allele is consistent with the R26R-CAG-tdTomato resultsdescribed below.

20612 | www.pnas.org/cgi/doi/10.1073/pnas.1118017108 Lehoczky et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

impact of clot removal because puncturing the skin may lead tonew clot formation, possibly rescuing regeneration.

DiscussionFate-Restricted Progenitor Model of Mouse Digit Tip Regeneration.Using tissue-specific, inducible Cre alleles, we were able to showthat the neonatal digit tip regenerates via fate-restricted pro-

genitors derived from the preamputation tissue. First we estab-lished that there is no transdifferentiation between tissue layersduring regeneration. Following Krt14-marked keratinocytesduring regeneration, we were able to show that although theregenerative epithelium is derived from preexisting stump ker-atinocytes, these cells do not contribute to any other lineage inthe regenerate and remain fated to the epidermis. In comple-mentary experiments, we show that mesodermally derived tis-sues, marked by Sp7 and Msx1, contribute only to mesodermlineages, never ectoderm. Moreover, Sp7-expressing osteoblastsfrom the limb only contribute to skeletal tissues (bone andperiosteum) in the regenerate, indicating that at least theseblastema cells are lineage restricted. Altogether, the lineage datawe present are consistent with and support reports of fate-re-stricted progenitor cells contributing to the regenerates ofzebrafish fin, Xenopus tail, and axolotl limb (9–11, 31); and theemerging model that transdifferentiation between germ layers ortissue types does not occur during appendage regeneration.Additionally, these results indicate that digit tip regenerationmay be more analogous to other examples of appendage re-generation than previously believed.During the preparation of this manuscript, Rinkevich et al. (32)

published a report focusing on cell lineage restriction during adultmouse digit tip regeneration, using a similar approach to thattaken here. Their conclusion, that lineage-restricted progenitorcells repopulate the digit tip without transdifferentiation, cor-roborates the findings in our report. As in our study, Rinkevichet al. demonstrated that the ectoderm of the regenerate is derivedfrom preamputation progenitor cells and that these cells do notcontribute to any other lineage or germ layer. Moreover, theyshowed that no mesodermally derived tissue transdifferentiatesinto an ectodermlal fate and all cells remain faithful to theiroriginal lineage.

Insight into Human Adult Digit Tip Regeneration. Appendage re-generation occurs in teleost fish and urodele amphibiansthroughout life, yet only larval anurans are capable of limb re-generation. Similarly, mice can regenerate their digit tipsthroughout life, whereas humans lose the ability with age. Hy-potheses on the lack of human adult digit tip regeneration includethe following: preferential inflammation response over regener-ative response in adults, tumor suppressor gene-mediated in-hibition of progenitor cell hyperproliferation in postneonataltissues, age-dependent progenitor cell depletion, and neonataland adult digit tip regeneration occurring by two discrete pro-cesses (of which humans have lost the latter). Although the ma-jority of these hypotheses are beyond the scope of this work, ourdata combined with other mouse digit tip reports, can provideinsight into whether the neonate and adult processes, as exem-plified in the mouse, are the same.It is becoming increasingly clear that neonatal and adult mouse

digit tip regeneration are much more similar than previously ap-preciated. Fernando et al. (21) showed that during regenerationin adult mice the terminal phalanx regenerates via direct ossifi-cation, consistent with the study by Han et al. (20), in which thesame was found during neonatal regeneration. Our report buildsupon these commonalities and reveals that the regenerative epi-thelium of the neonatal postamputation digit tip forms at a vari-able rate and includes a fragmentation of the terminal bone at amore proximal level, highly similar to what is described for theadult. Moreover, our data provide evidence against a single plu-ripotent blastema cell type and support a mechanism of multiplefate-restricted progenitor populations participating in neonataldigit tip regeneration, which is consistent with what Rinkevichet al. (32) found during adult mouse digit tip regeneration. Col-lectively, the data support a single process used both duringneonatal and adult mouse digit tip regeneration, and the questionremains what the fundamental difference is in adult human digittips that prevents successful regeneration.

Fig. 5. Clot removal impairs bone regeneration. (A) Brightfield image offibrin clot at 6 dpa. Inset: PN3 amputated digit tip, with dotted line depictingamputated tissue. (B–F) Alizerin red/alcian blue stained digit tip skeletons at3 wkPA representing average extents of regeneration for the group. (B)Nonregenerate digit after nail removal. (C) Control digit after no clot re-moval. Digits after clot removal at 8 dpa (D), 7 dpa (E), and 6 dpa (F). (G)Boxplot analysis of clot removal experiment shows significant decrease inregeneration when the clot was removed at 6 dpa or 7 dpa, as measured by2D area in pixels. Open circles represent 1.5 interquartile range outliers.Average nonregenerate digit tip area is denoted with a red line.

Lehoczky et al. PNAS | December 20, 2011 | vol. 108 | no. 51 | 20613

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

Why Is Msx1 Necessary for Digit Tip Regeneration? The requirementof Msx1, coincident with its expression in the nail bed, led to thehypothesis that the Msx1-expressing nail bed mesenchymal cellscould serve as a stem cell population for the regenerate. Ourdata show that although Msx1 cells do broadly contribute to theblastema and their own lineages in the regenerate, they do notact in isolation, and multiple other progenitor populations ac-tively participate in the process. This suggests that the necessityof Msx1 during regeneration may be due to the signaling prop-erties of the cells rather than the strict necessity of the de-scendant population, although a combination of the tworemains possible.Msx1 has been shown to be an important mediator of Bone

Morphogenetic Protein (BMP) signaling during digit tip re-generation (19), and it follows that Msx1-expressing cells may benecessary to mediate BMP activity after amputation. In this re-port we show that Msx1 is expressed in the distal clot. The clothas not received much attention in regard to digit tip regenerationbecause it is a structure that forms routinely upon injury to themammalian skin, but by lineage analyses we show that Msx1 andSp7 descendants reside within the clot, suggesting a functionalrole during regeneration, possibly beyond typical wound healing.By removing the clot, we provide preliminary data in support ofthis hypothesis. In urodele amphibians, successful limb re-generation depends on mitogenic signals provided by the nerve.In contrast, mouse digit tip regeneration proceeds even in theabsence of innervation (33), suggesting that if these are analogousprocesses, the necessary signals are coming from another source.Agrawal et al. (34) provide support for this hypothesis by showingthat at nonregenerative levels in digits of adult mice, applicationof ECMdegradation products to the wound, as an artificial sourceof signals, led to the accumulation of stem-like cells at the planeof amputation. Collectively there is clearly a delicate balancebetween wound healing and regeneration, and although clot

formation is typically associated with a wound healing response, itwill be interesting to further explore the specific function(s) theclot serves during digit tip regeneration.

Materials and MethodsFull methods are provided in SI Materials and Methods.

Mouse Strains and Alleles. Msx1-nlacZmicewere described previously (28). TheMsx1-CreERT2 allele was generated by introducing the CreERT2 coding se-quence (35) (a kind gift of Pierre Chambon, Institut de Génétique et de Biol-ogie Moléculaire et Cellulaire, Illkirch, France) at the initiator ATG site of Msx1by homologous recombination (30). Additional inducible Cre mouse strainsSp7-tTA-tetO-EGFP::Cre (27), Krt14-CreESR1 (22), and Col2a1-CreERT (26) wereobtained from Jackson Laboratories (strains 006361, 005107, and 006774). Crereporter alleles R26R-CAG-tdTomato (24) and R26R-lacZ (23) were alsoobtained from Jackson Laboratories (strains 007905 and 003474). All crosseswere performed with compound heterozygous males carrying both the in-ducible Cre driver and Cre reporter alleles and a CD1wild-type female with theexception of the Sp7-tTA-tetO-EGFP::Cre allele, which was crossed directly tothe R26R-lacZ reporter.

Digit Manipulations. All mouse breeding and manipulations were done inaccordance with the Harvard Medical School Institutional Animal Care andUse Committee. Mouse digit tip amputations were done on PN3 as previouslyreported (20). A dissection microscope (Leica MZ6) fitted with an eyepiecereticule was used for digit amputations. Pups were cryoanesthetized before400 μm of hindlimb digits 2, 3, and 4 were amputated using microdissectionspring scissors (FST 15003-08). In all animals, right hindlimb digits were leftunamputated as controls. One dose of 0.05 mg/kg s.c. buprenorphine wasgiven as postsurgical analgesia.

ACKNOWLEDGMENTS. We thank Andrew McMahon and Susan Dymecki forhelpful conversations regarding expression from the R26R-lacZ allele. Thisresearch was funded by National Institutes of Health Grants F32 AR56149 (toJ.A.L.) and R01 ND045499 (to C.J.T.).

1. Thornton CS (1957) The effect of apical cap removal on limb regeneration in Am-blystoma larvae. J Exp Zool 134:357–381.

2. Mescher AL (1976) Effects on adult newt limb regeneration of partial and completeskin flaps over the amputation surface. J Exp Zool 195:117–128.

3. Repesh LA, Oberpriller JC (1978) Scanning electron microscopy of epidermal cell mi-gration in wound healing during limb regeneration in the adult newt, Notoph-thalmus viridescens. Am J Anat 151:539–555.

4. Hay ED, Fischman DA (1961) Origin of the blastema in regenerating limbs of the newtTriturus viridescens. An autoradiographic study using tritiated thymidine to followcell proliferation and migration. Dev Biol 3:26–59.

5. Butler EG, O’Brien JP (1942) Effects of localized x-radiation on regeneration of theurodele limb. Anat Rec 84:407–413.

6. Echeverri K, Tanaka EM (2002) Ectoderm to mesoderm lineage switching during ax-olotl tail regeneration. Science 298:1993–1996.

7. Namenwirth M (1974) The inheritance of cell differentiation during limb re-generation in the axolotl. Dev Biol 41:42–56.

8. Thornton CS (1938) The histogenesis of the regenerating for limb of larval Ambys-toma after exarticulation of the humerus. J Morphol 62:219–241.

9. Kragl M, et al. (2009) Cells keep a memory of their tissue origin during axolotl limbregeneration. Nature 460:60–65.

10. Tu S, Johnson SL (2011) Fate restriction in the growing and regenerating zebrafish fin.Dev Cell 20:725–732.

11. Sousa S, et al. (2011) Differentiated skeletal cells contribute to blastema formationduring zebrafish fin regeneration. Development 138:3897–3905.

12. Illingworth CM (1974) Trapped fingers and amputated finger tips in children. J Pe-diatr Surg 9:853–858.

13. Borgens RB (1982) Mice regrow the tips of their foretoes. Science 217:747–750.14. Reginelli AD, Wang YQ, Sassoon D, Muneoka K (1995) Digit tip regeneration corre-

lates with regions of Msx1 (Hox 7) expression in fetal and newborn mice. De-velopment 121:1065–1076.

15. Allan CH, et al. (2006) Tissue response and Msx1 expression after human fetal digit tipamputation in vitro. Wound Repair Regen 14:398–404.

16. Crews L, et al. (1995) Expression and activity of the newt Msx-1 gene in relation tolimb regeneration. Proc Biol Sci 259:161–171.

17. Simon HG, et al. (1995) Differential expression of myogenic regulatory genes andMsx-1 during dedifferentiation and redifferentiation of regenerating amphibianlimbs. Dev Dyn 202:1–12.

18. Odelberg SJ, Kollhoff A, Keating MT (2000) Dedifferentiation of mammalian my-otubes induced by msx1. Cell 103:1099–1109.

19. Han M, Yang X, Farrington JE, Muneoka K (2003) Digit regeneration is regulated byMsx1 and BMP4 in fetal mice. Development 130:5123–5132.

20. Han M, Yang X, Lee J, Allan CH, Muneoka K (2008) Development and regeneration ofthe neonatal digit tip in mice. Dev Biol 315:125–135.

21. Fernando WA, et al. (2011) Wound healing and blastema formation in regeneratingdigit tips of adult mice. Dev Biol 350:301–310.

22. Vasioukhin V, Degenstein L, Wise B, Fuchs E (1999) The magical touch: genome tar-geting in epidermal stem cells induced by tamoxifen application to mouse skin. ProcNatl Acad Sci USA 96:8551–8556.

23. Soriano P (1999) Generalized lacZ expression with the ROSA26 Cre reporter strain. NatGenet 21:70–71.

24. Madisen L, et al. (2010) A robust and high-throughput Cre reporting and character-ization system for the whole mouse brain. Nat Neurosci 13:133–140.

25. Casanova JC, Sanz-Ezquerro JJ (2007) Digit morphogenesis: is the tip different? DevGrowth Differ 49:479–491.

26. Nakamura E, Nguyen MT, Mackem S (2006) Kinetics of tamoxifen-regulated Cre ac-tivity in mice using a cartilage-specific CreER(T) to assay temporal activity windowsalong the proximodistal limb skeleton. Dev Dyn 235:2603–2612.

27. Rodda SJ, McMahon AP (2006) Distinct roles for Hedgehog and canonical Wnt sig-naling in specification, differentiation and maintenance of osteoblast progenitors.Development 133:3231–3244.

28. Houzelstein D, Cohen A, BuckinghamME, Robert B (1997) Insertional mutation of themouse Msx1 homeobox gene by an nlacZ reporter gene. Mech Dev 65:123–133.

29. Orestes-Cardoso SM, et al. (2001) Postnatal Msx1 expression pattern in craniofacial,axial, and appendicular skeleton of transgenic mice from the first week until thesecond year. Dev Dyn 221:1–13.

30. Lopes M, et al. (2011) Msx genes define a population of mural cell precursors requiredfor head blood vessel maturation. Development 138:3055–3066.

31. Gargioli C, Slack JM (2004) Cell lineage tracing during Xenopus tail regeneration.Development 131:2669–2679.

32. Rinkevich Y, Lindau P, Ueno H, Longaker MT, Weissman IL (2011) Germ-layer andlineage-restricted stem/progenitors regenerate the mouse digit tip. Nature 476:409–413.

33. Mohammad KS, Neufeld DA (2000) Denervation retards but does not prevent toetipregeneration. Wound Repair Regen 8:277–281.

34. Agrawal V, et al. (2010) Epimorphic regeneration approach to tissue replacement inadult mammals. Proc Natl Acad Sci USA 107:3351–3355.

35. Feil R, et al. (1996) Ligand-activated site-specific recombination in mice. Proc NatlAcad Sci USA 93:10887–10890.

20614 | www.pnas.org/cgi/doi/10.1073/pnas.1118017108 Lehoczky et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0