Direct Isolation of Satellite Cells
for Skeletal Muscle Regeneration
Didier Montarras,1* Jennifer Morgan,3,4Charlotte Collins,4
Fre ´de ´ric Relaix,1Ste ´phane Zaffran,1Ana Cumano,2
Terence Partridge,4Margaret Buckingham1*
Muscle satellitecells contribute to muscle regeneration. We have used a Pax3GFP/þ
mouse line to directly isolate (Pax3)(green fluorescent protein)–expressing
muscle satellite cells, by flow cytometry from adult skeletal muscles, as a
homogeneous population of small, nongranular, Pax7þ, CD34þ, CD45–, Sca1–
cells. The flow cytometry parameters thus established enabled us to isolate
satellite cells from wild-type muscles. Such cells, grafted into muscles of mdx
nu/nu mice, contributed both to fiber repair and to the muscle satellite cell
compartment. Expansion of these cells in culture before engraftment reduced
their regenerative capacity.
Satellite cells are skeletal muscle progenitor
cells responsible for postnatal growth and re-
pair (1). The difficulty of isolating pure popu-
lations of satellite cells in sufficient number
has precluded their use in cell-based tissue
repair assays. These assays have, therefore,
employed either muscle precursor cells that
correspond to the progeny of muscle satellite
cells, obtained after activation and proliferation
in culture (2–4), or mixtures of cells obtained
after enzymatic dissociation of skeletal mus-
cles (5, 6). In vivo, quiescent muscle satellite
cells are characterized by the expression of
surface markers such as M-cadherin (7, 8),
syndecan 3 and 4 (9), and CD34 (10); how-
ever, none of these permit unequivocal isolation
because of the lack of specificity or availability
of suitable reagents. Satellite cells also express
transcription factors, notably Pax7, a member of
the homeodomain/paired box family of Pax
proteins (11). Recently, we have shown that,
Pax3, the paralog of Pax7, is also expressed in
quiescent muscle satellite cells in a subset of
muscles (12, 13). The generation of a green
fluorescent protein (GFP)–tagged Pax3 mouse
line (Pax3GFP/þ) (14) permitted us to isolate
(Pax3)GFP-expressing cells from adult skeletal
muscles by flow cytometry.
Previous observations on Pax3nlacz/þmice
indicated that satellite cells expressing the
transcriptional regulator Pax3 were limited to
a subset of adult skeletal muscles, including
the diaphragm, most trunk muscles, and some
limb muscles (13). The Pax3GFP/þmouse line
shows similar expression. As illustrated in the
diaphragm (Fig. 1A), (Pax3)GFPþ cells are
found in a typical satellite cell position beneath
the layer of laminin that surrounds muscle
fibers. Most of these cells also express the
transcriptional regulator Pax7 (Fig. 1B), which
marks muscle satellite cells (11).
Flow cytometric analysis of the cells
prepared from diaphragm muscle of adult
Pax3GFP/þmice (Fig. 1C) indicated that
(Pax3)GFPþ cells constitute a population that
is negative for CD45 and Sca1 and positive for
CD34. These cells were also negative for the
endothelial markers CD31 and Flk1 (fig. S1).
Forward and side scatter gating (reflecting the
size and granularity of the cells, respectively)
(Fig. 1C, left) indicated that (Pax3)GFPþ cells
constitute a homogeneous population of small,
nongranular, mononucleated cells. Immediate-
ly after sorting, immunodetection showed the
presence of 93% Pax7þ cells and 8% MyoDþ
1CNRS Unite ´ de Recherche Associe ´e 2578, Department
of Developmental Biology,2Unite ´ du De ´veloppement
des Lymphocytes, Unite ´ 668, INSERM, Pasteur Insti-
tute, 75724 Paris Cedex 15, France.
Paediatrics, Imperial College London, The Dubowitz
Neuromuscular Centre, Hammersmith Hospital, Du
Cane Road, London W12 ONN, UK.
Biology Group, Medical Research Council Clinical
Sciences Centre, Imperial College, Du Cane Road,
London W12 ONN, UK.
*To whom correspondence should be addressed.
E-mail: firstname.lastname@example.org (D.M.); email@example.com
23 SEPTEMBER 2005 VOL 309 SCIENCEwww.sciencemag.org
cells. Pax7 marks both quiescent and activated
satellite cells (15), whereas MyoD marks acti-
vated satellite cells only (16, 17). This indi-
cates that the majority of the cells did not
undergo activation during the few (4 to 6)
hours that were required for dissociation and
sorting. Colony assays further established the
identity of these cells as muscle progenitors,
giving rise to 100% Pax7- and MyoD- express-
ing cells after 3 days in culture (fig. S2). The
(Pax3)GFPþ fraction that we isolated thus
constitutes a pure population of myogenic cells.
We functionally characterized (Pax3)GFPþ
cells, isolated by flow cytometry from adult
diaphragm muscle, by grafting them into ir-
radiated tibialis anterior (TA) muscles of im-
munodeficient nude mdx mice (mdx nu/nu).
These mice lack dystrophin, a structural protein
that is mutated in Duchenne muscular dystro-
phy patients (18). Satellite cells of the TA, like
those of other lower hindlimb muscles, do
not normally express Pax3. The contribution
of (Pax3)GFPþ cells to fiber repair was mea-
sured by the restoration of dystrophin expres-
sion in muscle fibers of host mice, 3 weeks
after grafting. Numerous dystrophin-positive
fibers werereadilydetectedin the grafted mus-
cles (Fig.1D, top). Only occasional dystrophin-
positive fibers, probably revertant fibers (19),
were found in the control contralateral, non-
grafted TA muscles (Fig. 1D, bottom). Grafting
of 2 ? 104cells led to dystrophin expression
in an average number of 587 fibers, and grafts
of as few as 103cells still resulted in
dystrophin expression in an average of 160
fibers (Fig. 1E, left). These yields are compa-
rable to those obtained after grafting 5 ? 105
cells isolated by enzymatic dissociation of
whole adult muscles (5, 6).
Most grafting experiments have employed
muscle precursor cells obtained after a phase
of amplification in culture (2, 3). To determine
whether such culturing procedures could alter
the capacity of cells to contribute to tissue
reconstitution, we grafted cultured and non-
cultured (Pax3)GFPþ cells. Grafting 104non-
cultured cells led to restoration of dystrophin
expression in an average number of 300 fibers,
whereas grafting the same number of cultured
cells resulted in significantly fewer dystrophin-
positive fibers (mean 0 88 fibers, P G 0.02)
(Fig. 1E, right). We also grafted 105cells,
corresponding to the progeny after 3 days in
culture of 104(Pax3)GFPþ cells. These cells
led to restoration of dystrophin expression in
an average number of 265 fibers, a figure that
is similar to that obtained when grafting 104
noncultured cells (Fig. 1E, right). These results
show that culturing muscle satellite cells for a
few days before grafting reduces their efficien-
cy in fiber repair, suggesting that in vitro
expansion is disadvantageous. Clonal assays
indicated that cultured cells display a lower
proliferation potential than freshly isolated
cells and a tendency to differentiate more
rapidly (table S1). These features may account
for their reduced regenerative capacity.
(Pax3)GFPþ CD34þ donor cells could be
recovered by flowcytometry fromgrafted mus-
cles (Fig. 2A). These cells displayed a myo-
genic phenotype in culture, expressing MyoD
and Pax7 and differentiating into TroponinT-
expressing myotubes (Fig. 2B). Single fibers
prepared from grafted muscles (Fig. 2C)
carried cells of donor origin in a muscle sat-
Fig. 1. Characterization of (Pax3)GFP-expressing cells from the diaphragm muscle of adult mice.
(A) Transverse section of a Pax3GFP/þmouse). Left: Immunodetection of laminin (red staining),
with 4¶,6-diamidino-2-phenylindole (DAPI) coloration. Right: Direct fluorescent detection of GFPþ
cells (green staining), together with immunodetection of laminin. Scale bar, 20 mm. (B) Transverse
section from a Pax3GFP/þ:Pax7lacz/þmouse. Left: Immunodetection of b-galactosidase (b-gal, red
staining) from the Pax7 allele, together with DAPI staining. Right: Detection of GFP fluorescence
(green staining) from the Pax3 allele and co-immunodetection of b-galactosidase (red staining)
from the Pax7 allele, resulting in yellow staining. Scale bar, 100 mm. Arrowheads indicate
candidate muscle cells. (C) Flow cytometry analysis of cells from Pax3GFP/þmice. CD34 and Sca1
expression on CD45þ cells (red, top panels) and on GFPþ cells (blue, bottom panels). Top and
bottom right panels correspond to forward scatter (FSC) and side scatter (SSC) gating of CD45þ
and GFPþ cells, respectively. (D) Detection of dystrophin-positive fibers in grafted muscles. Three
weeks after grafting with (Pax3)GFPþ cells, TA muscles of mdx nu/nu mice were processed for
detection of dystrophin. Top: Tranverse section of grafted muscle. Bottom: Control contralateral
nongrafted TA. Scale bar, 200 mm. (E) Quantitative analysis of experiments as shown in (D). Cells
were grafted immediately after sorting (left). The effect of cell culture was examined by injecting
cultured or noncultured cells (right). The numbers of injected mice were, from left to right: 4, 5, 4,
6, 5, and 4. Labeled error bars represent SD.
R E P O R T S
www.sciencemag.orgSCIENCE VOL 30923 SEPTEMBER 2005
ellite cell position, co-expressing (Pax3)GFP
and Pax7. Of 569 cells detected on the surface
of 120 single fibers from grafted TA muscles,
17% were satellite cells of donor origin co-
expressing Pax7 and (Pax3)GFP. These results
not only to muscle fiber repair but also to the
muscle satellite cell compartment. They also
show that (Pax3)GFPþ cells retain their Pax3þ
identity in the environment of the TA muscle,
where endogenous satellite cells do not express
Pax3. Injured, as well as intact, TA muscle from
Pax3GFP/þmice does not normally contain
(Pax3)GFPþ cells (fig. S3).
Flow cytometric analysis indicated that
(Pax 3)GFPþ cells express the surface marker
CD34. We used this surface marker and the
parameters defined by forward and side scatter
gating for (Pax3)GFPþ cells (Fig. 1C) to
determine whether muscle progenitor cells that
do not express (Pax3)GFP could also be
isolated from adult muscles. The GFPþ
CD34þ cells isolated from the diaphragm of
adult mice (Fig. 3A) represented 47% of the
cells analyzed by flow cytometry. Clonal anal-
ysis of the cells from each fraction showed that
all of the clones formed (78 out of 192 single
cells) were myogenic, as monitored by immu-
nodetection of MyoD and Pax7 and by
myotube formation. In contrast, the GFP–
CD34þ cells (Fig. 3A) displayed a cloning
efficiency of 6% and gave rise to only 2
myogenic clones out of 192 plated cells. The
same cell fractions from the lower hind leg
muscles (Fig. 3B) gave markedly different
results. GFPþ CD34þ cells, representing only
0.25% of the cells, gave rise to 33 clones (out
of 96 single cells), all of which were myo-
genic. The GFP– CD34þ cells, which now
represented 52% of the population, gave rise
only to myogenic clones, with a cloning
efficiency of 39% (76 out of 192 single cells).
These results confirm that adult muscle
progenitor cells belong to the (Pax3)GFPþ
CD34þ cell fraction in the diaphragm,
whereas, in lower hind leg muscles, they are
in the (Pax3)GFP– CD34þ fraction. Both cell
fractions express Pax7 (fig. S4). Thus, the
parameters of size and granularity defined for
(Pax3)GFPþ cells permit an equally efficient
isolation of muscle satellite cells by sorting on
the basis of CD34 expression. Skeletal muscle
repair assays confirm and extend these obser-
vations. Grafting of GFPþ CD34þ cells from
the diaphragm of adult mice or GFP– CD34þ
cells from lower hind leg muscles of the same
mice into TA muscles of mdx nu/nu recipients
produced comparable restoration of dystrophin
expression (Fig. 3, C and D). Thus both
preparations participate similarly in muscle
Flow cytometric analysis and characteriza-
tion of (Pax3)GFPþ cells present in skeletal
muscles of adult Pax3GFP/þmice have per-
mitted us to define parameters for isolating
adult muscle progenitor cells. These cells
comprise a population of small, nongranular,
CD34þ CD45– Sca1– cells expressing Pax7.
In accordance with this, a CD34þ cell fraction
from skeletal muscles has been shown to be
enriched in myogenic cells (20). In a recent
study (21), adult muscle-associated progenitor
cells were also shown to belong to a fraction of
CD45– Sca1– CD34þ cells. We have shown
recently that the progenitor cells of skeletal
muscle during late embryonic and fetal devel-
opment depend on both Pax3 and Pax7 (14). In
the adult, not all muscle satellite cells express
both genes. As shown here, although those in
the diaphragm are Pax3þ Pax7þ, satellite
cells in lower hindlimb muscles express only
Pax7. This distinction is maintained in re-
generating muscles as seen in the TA after
injury, where satellite cells remain (Pax3)GFP-
negative. This is in contrast to a previous report
on cells cultured from injured muscle (22).
When we extended this flow cytometric analy-
sis to a much larger gating window, we still did
not detect any (Pax3)GFPþ cells after TA
muscle injury (23). In muscles such as the dia-
phragm, Pax3 expression is cell-autonomous
because (Pax3)GFPþ cells engrafted into the
Pax3-negative TA muscle retain their initial
phenotype. The myogenic potential of Pax3-
expressing or -nonexpressing satellite cells is
indistinguishable in both in vitro and in vivo
in subpopulations of adult muscle satellite cells
awaits a conditional mutant, because embryos
do not survive in the absence of Pax3.
Assays for muscle repair that have been
developed to date are based on the injection of
5 ? 105to 106cells into the muscles of mdx
mice. Most have been performed with cells
either directly obtained by enzymatic dissoci-
ation of muscles (5, 6) or after a phase of
selection and amplification in culture (2, 3).
When 5 ? 105cells from freshly disaggregated
Fig. 2. Recoveryof(Pax3)GFPþ cellsfromgrafted
TA muscles. (A) Three weeks after grafting, TA
muscles were enzymatically dissociated and
(Pax3)GFPþ CD34þ cells were isolated by flow
cytometry. (B) Immediately after sorting, cells
were plated and their myogenic identity deter-
mined by immunodetection of MyoD, Pax7, and
Troponin T, after 3, 3, and 5 days of culture,
respectively (top panels). Bottom panels show
DAPI nuclear stain and phase contrast. Scale bar,
on single fibers of grafted TA muscles. Left:
Immunodetection of GFP with DAPI staining. Right:
Co-immunodetection of Pax7 and GFP with DAPI
marks both the cytoplasm and the nucleus of
Fig. 3. Isolation of muscle satel-
lite cells in the absence of
(Pax3)GFP expression. Flow cy-
tometric analysis of cells from
the diaphragm and lower hind
leg muscles of adult Pax3GFP/þ
mice. (A and B) Cells from the
diaphragm and from lower hind
leg muscles were analyzed for
both GFP and CD34 expression
as indicated in each panel. The
percentages shown correspond
to the fraction of positive cells
within a FSC/SSC gate as shown
in Fig. 1C. (C and D) Immunode-
tection of dystrophin in TA mus-
cles of mdx nu/nu mice 3 weeks
after grafting of 2 ? 104cells
from the fractions indicated by
the arrows. Scale bar, 50 mm.
R E P O R T S
23 SEPTEMBER 2005 VOL 309SCIENCE www.sciencemag.org
muscle were implanted into the TA of ir- Download full-text
radiated mdx nu/nu mice, they formed a mean
of 328 dystrophin-positive fibers (5). Similar
results were obtained after injecting one to two
million muscle-derived cultured cells into limb
muscle (2, 3). Our results now show that pu-
rified satellite cells are much more efficient
than these crude or cultured cell populations in
contributing to muscle repair.
The culture of muscle progenitor cells
before grafting markedly reduces their regen-
erative efficiency such that the culture expan-
sion itself is an Bempty[ process, yielding the
same amount of muscle as the number of cells
from which the culture was initiated. Culture-
induced modifications may affect survival or
engraftment capacity of the cells (24, 25). How-
ever, we did not detect a difference in survival
between cultured and freshly isolated cells 1
day after grafting (23). The activated state of
the grafted cells may diminish their regenera-
tive potential, because freshly isolated progen-
itor cells are not activated at the time of
grafting, unlike their cultured progeny that
express MyoD. Clonal assays suggest that the
lower regenerative capacity of cultured cells
reflects their more rapid differentiation. A simi-
lar situation is encountered with hematopoietic
stem cells, which begin to differentiate and to
lose their tissue reconstitution capacity when
Not only do purified muscle satellite cells
contribute to muscle repair when engrafted
into regenerating mdx muscles but some also
persist as progenitor cells, adopting a satellite
cell position and expressing Pax7. These re-
sults, therefore, point to muscle satellite cell
self-renewal. The fact that (Pax3)GFPþ cells
can be recovered from the muscles into which
they were originally transplanted and shown to
differentiate into muscle cells in culture also
argues in favor of self-renewal. We therefore
conclude that the satellite cell selection proce-
dure described here results in cells that can
both repair and contribute to the progenitor
cell population of damaged muscles. There
may be other stem cell types that can be
mobilized to contribute to this process (27),
but the muscle satellite cell population isolated
by the flow cytometry parameters that we
have defined is clearly a major contributor to
muscle regeneration and a potential therapeu-
References and Notes
1. S. B. Charge, M. A. Rudnicki, Physiol. Rev. 84, 209
2. Z. Qu-Petersen et al., J. Cell Biol. 157, 851 (2002).
3. G. M. Mueller, T. O’Day, J. F. Watchko, M. Ontell,
Hum. Gene Ther. 13, 1081 (2002).
4. D. Skuk, M. Goulet, B. Roy, J. P. Tremblay, Exp.
Neurol. 175, 112 (2002).
5. J. E. Morgan, C. N. Pagel, T. Sherratt, T. A. Partridge, J.
Neurol. Sci. 115, 191 (1993).
6. J. E. Morgan, R. M. Fletcher, T. A. Partridge, Muscle
Nerve 19, 132 (1996).
7. A. Irintchev, M. Zeschnigk, A. Starzinski-Powitz, A.
Wernig, Dev. Dyn. 199, 326 (1994).
8. A. Hollnagel, C. Grund, W. W. Franke, H. H. Arnold,
Mol. Cell. Biol. 22, 4760 (2002).
9. D. D. Cornelison, M. S. Filla, H. M. Stanley, A. C.
Rapraeger, B. B. Olwin, Dev. Biol. 239, 79 (2001).
10. J. R. Beauchamp et al., J. Cell Biol. 151, 1221 (2000).
11. P. Seale et al., Cell 102, 777 (2000).
12. M. Buckingham et al., J. Anat. 202, 59 (2003).
13. F. Relaix, D. Montarras, M. Buckingham, unpublished
14. F. Relaix, D. Rocancourt, A. Mansouri, M. Buckingham,
Nature 435, 948 (2005).
15. P. S. Zammit et al., J. Cell Biol. 166, 347 (2004).
16. Z. Yablonka-Reuveni, A. J. Rivera, Dev. Biol. 164, 588
17. D. D. Cornelison, B. J. Wold, Dev. Biol. 191, 270 (1997).
18. J. C. van Deutekom, G. J. van Ommen, Nat. Rev.
Genet. 4, 774 (2003).
19. E. P. Hoffman, J. E. Morgan, S. C. Watkins, T. A.
Partridge, J. Neurol. Sci. 99, 9 (1990).
20. R. J. Jankowski, B. M. Deasy, B. Cao, C. Gates, J.
Huard, J. Cell Sci. 115, 4361 (2002).
21. R. I. Sherwood et al., Cell 119, 543 (2004).
22. I. M. Conboy, T. A. Rando, Dev. Cell 3, 397 (2002).
23. D. Montarras et al., unpublished data.
24. J. X. DiMario, F. E. Stockdale, Exp. Cell Res. 216, 431
25. J. R. Beauchamp, J. E. Morgan, C. N. Pagel, T. A.
Partridge, J. Cell Biol. 144, 1113 (1999).
26. J. Antonchuk, G. Sauvageau, R. K. Humphries, Cell
109, 39 (2002).
27. T. Partridge, Cell 119, 447 (2004).
28. We thank D. Rocancourt and C. Cimper for technical
assistance. Supported by the Pasteur Institute and
the CNRS, with additional grants from the Associ-
ation Franc ¸aise contre les Myopathies, ‘‘the Cellules
Souches’’ Grand Programme Horizontal of the
Pasteur Institute, and the EuroStemCell Integrated
Project of the European Union 6th Framework
Programme. J.M., C.C., and T.P. received support
from the Medical Research Council of the UK, the
Muscular Dystrophy Campaign, and the Engineering
and Physical Sciences Research Council. T.P. occu-
pies a Blaise Pascal chair awarded by the Ecole
Normale Supe ´rieure. D.M. dedicates this paper to
R. B. Seaver.
Supporting Online Material
Materials and Methods
Figs. S1 to S4
References and Notes
12 May 2005; accepted 9 August 2005
Published online 1 September 2005;
Include this information when citing this paper.
R E P O R T S