Absence of CD34 on Murine Skeletal Muscle Satellite Cells Marks a Reversible State of Activation during Acute Injury

Background Skeletal muscle satellite cells are myogenic progenitors that reside on myofiber surface beneath the basal lamina. In recent years satellite cells have been identified and isolated based on their expression of CD34, a sialomucin surface receptor traditionally used as a marker of hematopoietic stem cells. Interestingly, a minority of satellite cells lacking CD34 has been described. Methodology/Principal Findings In order to elucidate the relationship between CD34+ and CD34- satellite cells we utilized fluorescence-activated cell sorting (FACS) to isolate each population for molecular analysis, culture and transplantation studies. Here we show that unless used in combination with α7 integrin, CD34 alone is inadequate for purifying satellite cells. Furthermore, the absence of CD34 marks a reversible state of activation dependent on muscle injury. Conclusions/Significance Following acute injury CD34- cells become the major myogenic population whereas the percentage of CD34+ cells remains constant. In turn activated CD34- cells can reverse their activation to maintain the pool of CD34+ reserve cells. Such activation switching and maintenance of reserve pool suggests the satellite cell compartment is tightly regulated during muscle regeneration.


Introduction
Since their discovery, satellite cells have been characterized as the resident stem cells of the skeletal muscle, responsible for postnatal myofiber growth and regeneration [1,2]. Classically, satellite cells have been defined by their position underneath the basal lamina of muscle fibers. More recently, satellite cells have been distinguished by nuclear Pax7 immunostaining and/or Myf5 nlacZ/+ reporter expression, and the presence of various surface receptors including a7 integrin (herein referred as a7), b1 integrin, CD34, NCAM, c-met, and CXCR4 [3,4,5,6,7,8,9,10,11,12]. Although satellite cells are unanimously recognized by anatomical location, there is no single surface marker specific or exclusive to the entire population. This issue is compounded by the heterogeneity of satellite cells between muscles, as reported with Pax3 expression, and within muscles, based on non-uniform expression levels of Myf5-driven reporters [13,14,15]. Such heterogeneity between and within various muscles exemplifies the complexity of the satellite cell pool that has yet an established canonical differentiation lineage. Beauchamp et al. (2000) reported that the majority of satellite cells could be identified by Myf5-knock-in reporter activity and CD34 expression. In this study approximately 20% of satellite cells monitored in freshly isolated myofibers from extensor digitorum longus (EDL) muscle, could not be detected by CD34 immunostaining. Although the CD34-satellite cell population was not characterized, it was suggested these cells may represent a more primitive population of muscle stem cells [5]. In separate studies using Pax3 and Pax7 reporter mice, it was observed that the majority of myogenic cells of skeletal muscle reside within the CD34+ fraction [16,17]. It has been further described that myogenic cells can be isolated by fluorescence-activated cell sorting (FACS) as CD452/Sca12/CD34+, and more recently as CD452/CD312/Sca12/CD11b2/a7+/CD34+ [18,19].
Although CD34 has been used to identify and isolate satellite cells, many cell types, including muscle endothelial cells, express CD34 [20]. Thus, in order to study satellite cell heterogeneity, we initially set out to develop a method to isolate pure populations of satellite cells using the CD34 antigen while excluding other cells that may also express CD34. Utilizing FACS based on previously published reports from this and other laboratories, we initially isolated satellite cells by removing CD45+ hematopoietic cells, CD31+ endothelial cells, other non-satellite Sca1+ cells, and then selecting the remaining CD34+ fraction [18,20,21]. Although the bulk of CD452/CD312/Sca12/CD34+ sorted cells were myogenic, we repeatedly observed non-myogenic cells in culture. Next, in accordance with the literature, we incorporated a7 for satellite cell FACS isolation [10,14,19,22]. Indeed, this approach eliminated non-myogenic cells in culture. Subsequently, we observed by FACS-analysis that the CD452/CD312/Sca12/ a7+ population was comprised of a CD34+ majority and CD342 minority. In the murine hematopoietic system it has been observed that single CD342/low stem cells can reconstitute the entire hematopoietic system and give rise to both CD34+/2 cells [23]. Therefore, in analogy with the hematopoietic system, we initially hypothesized that within the satellite cell compartment the minority of CD342 cells may represent more primitive cells upstream of the more abundant CD34+ population. However, our results described herein indicate the absence of CD34 on skeletal muscle satellite cells in vivo marks a state of activation dependent on muscle injury and not necessarily a hierarchy as initially predicted.

Results
Despite CD45/CD31/Sca1 Negative Selection, CD34 is Insufficient for Purifying Satellite Cell by Flow Cytometry Using our previously developed method to isolate endothelial cells from a collection of skeletal muscles (pooled hind limb, pectorals and triceps muscles), we initially FACS-sorted satellite cells as CD452/CD312/Sca12/CD34+ [20]. Reverse-transcription PCR (RT-PCR) of these freshly sorted cells revealed expression of Pax7 and myogenic factors Myf5 and MyoD (data not shown). In culture, the majority of CD452/CD312/Sca12/CD34+ sorted cells produced colonies comprised of small round cells that ultimately formed twitching myotubes. However, cultured CD452/CD312/Sca12/CD34+ cells also gave rise to larger non-myogenic cells that never formed myotubes (data not shown).
To improve satellite cell purification we incorporated a7 and investigated the myogenic potential within the CD452/CD312/ Sca12 population based on the presence or absence of a7 and CD34 antigens. By FACS-mediated single cell deposition only CD34+/a7+ (CD452/CD312/Sca12) sorted cells, which were small and round, gave rise to myogenic colonies (n = 3). In contrast, CD34+/a72 (CD452/CD312/Sca12) cells appeared larger and more spread-out and did not form myotubes (n = 3)( Figure 1A). Quantitative reverse-transcription PCR (q-RT-PCR) of freshly sorted cells (n = 2) revealed that between these two populations, only CD34+/a7+ cells express the definitive satellite cells genes Pax7 and Myf5 ( Figure 1B). Cryosections from uninjured tibialis anterior (TA) muscle, immunostainied for Pax7, a7 and CD34, revealed the presence of both populations. However, only CD34+/a7+ cells were Pax7+ and within the satellite cell position. CD34+/a72 cells were Pax72 and located in the interstitium ( Figure 1C). Altogether, our results indicate CD34+/a7+ cells are present in the satellite cell compartment while CD34+/a72 cells are not myogenic in culture nor satellite cells by anatomical definition.
By FACS-analysis we determined that a72 cells constitute 11.8% (61.2% SEM) of the CD452/CD312/Sca12/CD34+ population (n = 5); note with the exception of transplant studies and reporter mice, all experiments characterizing CD34+/2 cells were completed with 2 month old C57BL/6 male mice. We believe sorted CD34+/a72 cells, which correspond with CD34+/ a72/Pax72 cells observed by immunostaining, were the source of non-myogenic CD452/CD312/Sca12/CD34+ cells. Collectively, our results are consistent with previous reports that CD34 is present on the majority of skeletal muscle satellite cells. However, because CD34 cell surface expression is not exclusive to satellite cells, it is insufficient to FACS purify satellite cells unless used in conjunction with a second positive antigen such as a7.
CD342/a7+ (CD452/CD312/Sca12) Cells Represent a Minority of Limb Muscle Pax7+ Satellite Cells When purifying a7+ (CD452/CD312/Sca12) cells, we observed CD34+ majority and CD342 minority populations, both of which produced myogenic colonies in culture. These CD34+/2 populations sorted as CD452/CD312/Sca12/a7+ shall be abbreviated CD34+ and CD342 for the remainder of this manuscript. FACS-analysis indicates CD342 cells make up ,17.6% (62.2% SEM) of the a7+ (CD452/CD312/Sca12) population in pooled muscle preparations from individual mice (n = 5). Further antibody staining revealed the presence of CD342/Pax7+ cells in the satellite cell position underneath the basal lamina; confirming CD342 cells are an integral part of the satellite cell niche in vivo and not a mere product of cell isolation procedures ( Figure 2). Satellite cells constitute only 2-5% of the nuclei in adult muscles and their abundance varies between muscle type and age [24,25]. Therefore, direct quantification from cross sections would have required that many samples and cross sections be analyzed for accurate numbers of CD34+ and CD342, Pax7+ satellite cells. Thus in order to quantify the abundance of these two satellite cell populations, we performed flow cytometry analysis of pooled muscles as well as individual muscles.
Analysis of myogenic colonies generated by FACS-mediated single cell deposition of CD34+/2 satellite cells (1 cell per well, 96 wells per cell type from each mouse, n = 3) revealed that both populations contributed myogenic clones with similar growth pattern based on the number of nuclei counted from each colony ( Figure 3A, cell doubling). However, clonability was higher in the CD34+ versus the CD342 population ( Figure 3A, #colonies/per 96 wells). The difference observed between the number of wells with colonies, suggests the CD34+ population exhibits greater cloning efficiency. Alternatively, our culture conditions may not be optimal for seeding CD342 sorted cells.
As noted above, all Pax7+ cells were a7+ ( Figure 1C). However, not all a7+ cells were Pax7+. The microvascular bed, especially vascular smooth muscle cells stained positive for a7 (data not shown) [26,27]. However, in the event that any endothelial cells may be a7+, such putative endothelial cells are CD31+ and are excluded from the CD452/CD312/Sca12/a7+ population. Indeed, RT-PCR of freshly sorted cells (n = 3) for von Willebrand Factor and Tie1, genes expressed by skeletal muscle endothelium, confirms the absence of endothelial cells within both sorted cell populations ( Figure 3B) [20]. Additionally vascular smooth muscle cells within skeletal muscles immunostained negative for CD34 and thus were presumed to be excluded from the sorted CD34+ population (not shown) [28]. Although such vascular smooth muscle cells could potentially be present in the CD342 satellite cell population, our sort approach should eliminate the presence of such cells from the final population. As previously published, satellite cells have also been defined as small non-granulated cells based on FACS-analysis [13]. Likewise, in our FACS isolation of satellite cells, we gate on smaller events by forward scatter (FSC)-Area vs. FSC-Height to remove larger a7+ cells that may originate from vascular smooth muscle ( Figure S1).
To assess satellite cell purity within each population, we cytocentrifuged freshly sorted CD34+/2 cells (n = 3 mice) and immunostained with anti-Pax7. From both sorted populations we obtained a high purity of myogenic cells as indicated by the presence of nuclear Pax7 staining ( Figure 4A, left panels). On average, 89% (61.7% SEM) of CD34+ and 70% (62.6% SEM) of CD342 cells stained positive for Pax7. Although the majority of cytocentrifuged cells from both populations stained positive, the difference in the frequency of Pax7+ cell was significant (Student's t-test, P = 0.04) indicating there is some contribution of Pax7-cells within the CD342 population.
We also cytocentrifuged CD34+ and CD342 cells sorted from adult 3F-nlacZ-2F reporter mice to assess the possible presence of differentiated progeny of satellite cells in the CD34+/2 popula-tions. In this transgenic mouse, muscle fibers and cultured differentiated myoblasts and myotubes express nuclear localized LacZ [5,15,29]. Nuclei positive for b-galactosidase (b-gal) were identified by X-gal staining (data not shown). We observed that CD342 cells were completely negative, while approximately 0.01% of CD34+ cells stained positive for b-gal activity with X-gal. With the exception of the few CD34+/b2gal+ cells observed, lack of 3F-nlacZ-2F reporter activity indicated the vast majority of cells in the CD34+/2 sorted populations were not differentiated myoblasts (the phase at which this transgene is activated) [5,29]. In contrast, we did identify b-gal+ cells from cytocentrifuged 3F-nlacZ-2F unsorted cells stained with X-gal as a positive control (not shown).
To further compare the CD34+/2 populations, we surveyed the expression of genes known to be transcribed by satellite cells ( Figure 4B). Due to possible gene regulation changes in culture that may not necessarily represent in vivo expression, we chose to analyze only freshly sorted cells by q-RT-PCR. Initially we checked CD34 and a7 expression to validate our FACS-analysis ( Figure 4B). In accordance with antigen levels, CD342 cells express very low levels of CD34 mRNA, whereas CD34+ cells transcribe much higher levels (25 times greater than CD342 cells). The expression of a7 was similar in both CD34+/2 populations and not statistically significant (Student's t-test, P.0.05). Next, we compared the two populations for the expression of a panel of genes that are associated with different phases of myogenesis ( Figure 4B). Pax3, Pax7, and Myf5 are expressed by quiescent and proliferating satellite cells [4,12,30,31]. MyoD expression is initiated upon satellite cell activation and sustained in proliferating and differentiating progeny, while myogenin expression is associated specifically with myogenic differentiation [11,32]. Interestingly, expression levels of Pax3/7 genes and muscle regulatory factors (MRF's) Myf5, MyoD, and myogenin were significantly higher in CD34+ vs. CD342 cells (Student's t-test, P#0.05). The detection of myogenin suggests the presence of some differentiating myogenic cells, which may account for the presence of a minute number of cytocentrifuged CD34+ cells that stained positive with X-gal for the differentiation-linked 3F-nlacZ-2F reporter as detailed above. Our detection of myogenin and MyoD transcripts, in addition to Myf5, in preparation of freshly isolated satellite cells is in agreement with previous studies describing gene expression in freshly isolated satellite cells [33,34]. Expression of satellite cell associated genes CXCR4, c-met and NCAM was also significantly higher in CD34+ cells (Student's t-test, P#0.05). Collectively, this expression comparison suggests that CD34+ cells transcribe higher levels of satellite cell characteristic genes than CD342 cells. Alternatively, the CD342 population may contain a residual population of non-myogenic cells that contributes to the apparent reduced expression of these genes.

The CD34+ and CD342 Populations are Consistently Maintained Across Different Muscles
In the aforementioned experiments, we isolated CD34+/2 cells from pooled hindlimb, pectoralis and triceps muscles. In contrast, the study by Beauchamp et al. (2000) quantified the percentage of CD34+ satellite cells only on EDL single fiber isolations. Thus, to address the possibility that CD342 satellite cells reside only in specific skeletal muscles, we FACS-analyzed individual muscles (n = 5) using same selection strategy used for pooled muscle isolations ( Figure 5A). Because the number of hematopoietic cells can vary between mice depending on circulatory responses, such as mild inflammation, we restricted our analysis to the CD452 fraction. For each muscle, Figure 5B represents the average percentage of CD34+/2 cells compared to the total number of non-hematopoietic mononuclear cells. Figure 5C represents the average percentage within the a7+ (CD452/CD312/Sca12) population in order to determine if CD34+ cells consistently represent the majority.
Results from individual muscle analyses indicate that the average percentage of CD34+ cells compared to the CD452 fraction is consistent between hindlimb muscles, whereas pectoralis and triceps contain nearly twice the amount of CD34+ cells. Such variation in the number of CD34+ cells may be attributed to individual muscle demands for myogenic cells and/or myofiber types. Consequently, the percentage of CD34+ cells was statistically different (single factor ANOVA, p#0.05) between muscle groups, whereas the percentage of CD342 cells was not (P = 0.075), indicating this population is consistently present in all muscle groups analyzed. Despite this variation, CD34+ cells were consistently the majority within the a7+ (CD452/CD312/ Sca12) population. The ratio of CD34+ over CD342 cells was not statistically significant (single factor ANOVA, p.0.05), suggesting the distribution of these two populations is maintained between muscle groups.
Following Acute Injury, CD342 Cells Represent the Majority of Activated a7+ (CD452/CD312/Sca12) Myogenic Cells To elucidate the role of CD342 skeletal muscle cells of the skeletal muscle we examined their response to acute injury using the cardiotoxin (CTX) model. Such injury is characterized by a massive myogenic response leading to almost full muscle regeneration by week two following initial tissue destruction [35,36]. We injected CTX in left TA's and FACS-analyzed each damaged muscle individually at 3, 7, and 14 days post injury (n = 5 per time point)( Figure 6A&B). Just as the previous analysis of individual muscles, we compared here the percentage of CD34+/2 cells within the CD452 fraction ( Figure 6A) and the relative percentage within all a7+ (CD452/CD312/Sca12) cells ( Figure 6B). For this experiment we predicted that the more differentiated population will increase early after injury and continuously decrease from day 7 to14 as the muscle reaches full regeneration. In addition, if there is a canonical relationship between CD34+/2 cells, we have anticipated an inverse trend where one population decreases while the other increases during the course of regeneration. As expected, 3 days post injury there was a massive increase in the number of a7+ (CD452/CD312/Sca12) cells which by histology were not vascular smooth muscle cells as indicated by asmooth muscle actin (aSMA) staining ( Figure S2). Surprisingly after CTX injury, the CD34-cells became the majority and declined to near uninjured levels by day 14. In contrast, the average percentage of CD34+ cells remained between 3-5% within the CD452 fraction ( Figure 6A) despite becoming the minority within a7+ (CD452/CD312/Sca12) population ( Figure 6B). The fluctuation of CD342 cells between time points  within CD452 fraction and a7+ population was statistically significant (single factor ANOVA, p#0.05). However, the percentage of CD34+ cells within the CD452 fraction remained relatively constant (p = 0.076).
Since CD34+ cells became the minority within the a7+ (CD452/CD312/Sca12) population early in injury but the overall percentage of these cells did not increase, we postulated this population remained quiescent while CD342 cells were activated. Thus, we compared Myf5, MyoD, and myogenin expression by q-RT-PCR in freshly sorted cells from damaged limb muscles at 3 days post CTX injury ( Figure 6C). Cells were sorted from pooled CTX injected muscles (single TA, quadriceps and gastrocnemius) from individual mice (n = 3). Without injury both CD34+/2 cells expressed relative high levels of Myf5 but low levels of MyoD and myogenin mRNA, indicating both populations are committed to the myogenic pathway, but remain quiescent in uninjured muscle ( Figure 4B). In contrast, following injury the average relative expression of all three MRF's increased in both populations ( Figure 6C). Myf5 and MyoD expression was similar between both populations whereas myogenin was 3.9 times greater and statistically significant (Student's t-test, p#0.05) in CD342 cells. In accordance with MRF expression during myogenesis, upregulation of Myf5 and MyoD following injury indicates both CD34+/2 cells are activated while higher myogenin expression suggests that more differentiated cells reside within the CD342 population [11].
The significance of myogenin expression was confirmed by immunostaining as the majority of myogenin+ cells observed within damaged areas 3 days post CTX injury, were a7+/CD342 ( Figure 7A). Taking into account myogenin histological and FACS data we concluded CD342 cells were directly responding to injury. In contrast, significant upregulation of MyoD but not myogenin by CD34+ cells suggested this population was activated but not differentiating. However, the role of CD34+ cells following injury was unclear as their overall numbers did not increase while the majority of a7+ cells observed within damaged areas were CD342.
Although Both CD34+/2 Populations Proliferate in Response to Injury, CD34+ Cells Retain More BrdU Because the percentage of CD34+ cells did not increase with CTX damage, it was possible that there was a general increase in all cell types or that CD34+ cells did not proliferate in response to injury. To explore this alternative, we compared BrdU incorporation in order to quantify proliferation between CD34+/2 cells and asses if the absence of CD34 simply marks mitotically active myogenic cells. In order to test this hypothesis we analyzed BrdU incorporation without injury, 2.5-3 days post injury within the peak of proliferation, and after regeneration 14 days post injury (n = 3 for each timepoint analyzed)( Figure 7B) [35]. Bone marrow cells isolated from the same animals as a positive control had a high degree of BrdU incorporation following 3 days pulse. In contrast, BrdU was not detected in CD34+ or CD342 cells from uninjured limb. However, upon injury, BrdU was incorporated into the CD34+/2 populations. When isolated from mice pulsed for 12 hours between day 2.5-3 post CTX injury, both CD34+/2 populations had incorporated equally high levels of BrdU.
Interestingly, in cells isolated from mice pulsed for the first 3 days following injury and then chased to day 14, BrdU incorporation declined in CD342 cells but remained the same for CD34+ cells. Consequently, the amount BrdU retention by day 14 was roughly 10 fold greater for CD34+ vs. CD342 cells as represented by the log distance from the mean of each fluorescence peak. Given that the level of BrdU uptake at day 2.5-3 was similar for both populations, we estimate that the CD342 population divided .3 cell doublings during the day 4-14 chase.
Without injury, the lack of detectable BrdU incorporation correlates with q-RT-PCR data to indicate both populations are quiescent. Following injury both populations have equal levels of BrdU incorporation indicating a similar degree of cell division occurred within the peak of proliferation. Beyond 3 days after injury, CD34+ cells retained a similar level of BrdU indicating they ceased dividing, while CD342 continued to proliferate as their incorporation decreased during the chase. Interestingly, although CD34+ cells did incorporate BrdU following injury, their percentage did not increase but remained relatively constant. Therefore we postulated CD34+ cells give rise to more committed cells while tightly maintaining the parent population during the course of regeneration. All together, FACS, histological, and BrdU results indicate CD34+ cells divided early in response to injury and perhaps gave rise to CD342 cells, which further proliferated during the course of regeneration.

CD34+/2 Switching is Dependent on Muscle Injury
To test the possibility that activated CD34+ cells are becoming or giving rise to CD342 cells, we sorted both populations from injured chicken actin promoter driven EGFP transgenic mice and transplanted into injured and uninjured quadriceps of C57BL/6 mice. Recipient muscles were then analyzed by FACS to examine if GFP donor cells gained or lost CD34 when transplanted into each respective environment. Injured recipient quadriceps were injected with CTX one day prior to cell transplants. All recipient muscles were analyzed 3 days post transplant in order to detect donor cells prior to fusion (n = 3 for each condition). CD34 analysis of transplanted GFP cells indicated a distinct pattern of switching dependent on the state of muscle injury. As predicted, the majority of injury activated CD34+ donor cells gave rise to, or directly became CD34 negative. The percentage of CD342 cells generated from CD34+ donors was slightly more in injured recipient muscles, although this difference was not significant (Student's t-test, p.0.05)( Figure 8A, left graph). Conversely, the majority of CD342 cells transplanted into injured muscle remained CD342 ( Figure 8A, right graph). To our surprise a large portion of CD342 cells transplanted into uninjured muscle became and/or gave rise to CD34+ cells. In contrast to the CD34 expressing phenotype acquired from negative donors following transplant into uninjured muscles, the percentage of CD342 cells detected in transplanted injured muscles was very consistent and statistically significant (Student's t-test, P#0.05)( Figure 8A, right graph). These results indicated that the presence or absence of CD34 on satellite cells is conditionally dependent on muscle environmental demands and not necessarily hierarchal as initially hypothesized.
peaks represent unstained controls, green peaks are experimental stained sample. (B) The average percentage of each population per muscle group relative to all CD452, non-hematopoietic cells. By single factor ANOVA, differences between muscles was statistically significant for CD34+ (# P#0.05) but not CD342 (P = 0.076). (C) The average percentage of CD34+ and CD342 cells within the a7+ (CD452/CD312/Sca12) population. The ratio of CD34+ over CD342 cells was not statistically significant by single factor ANOVA. A-C are from n = 5, 2 month old C57BL/6 mice. Error bars represent 6SEM. doi:10.1371/journal.pone.0010920.g005 Because we only studied a short timeframe after injury, it was unclear if later during the course of regeneration, CD342 cells ever transitioned back to CD34+. To examine the possibility of conversion following the peak of proliferation, we sorted and transplanted CD342 cells isolated from muscles of Myf5 nlacZ/+ mice 3 days post CTX injury directly into uninjured (n = 3) and  Immunostaining of damaged TA, 3 days post injury, reveals that myogenin+ cells within the sight of injury are a7+/CD342. Insets show colocalization of DAPI with nuclear myogenin and membranous a7 but not CD34 staining. Insets were produced from the original sized merged injured (n = 3) quadriceps injected the previous day with CTX (C57BL/6 recipients). In order to detect the Myf5 nlacZ/+ transgene in donor cells that may have recently fused or formed immature myofibers, we harvested recipient muscles 7 days post transplant. Cyrosections were initially stained with X-gal to identify and validate the presence of Myf5 nlacZ/+ donor cells. Once b-gal+ cells were identified ( Figure S3), adjacent cryosections were stained with antibodies against b-gal, CD34, and aSMA as a marker of newly regenerated myofibers [37,38,39]. Results indicate that activated CD342 donor cells were myogenic as in both injured and uninjured recipient muscles they formed new muscle fibers identified as a-SMA+ with b-gal+ centrally located nuclei ( Figure 8B). In addition, we observed mononuclear b-gal+/ CD34+ cells indicating some CD342 cells revert back to a reserve state following injury repair ( Figure 8C). Altogether, the analysis of the FACS population dynamics during regeneration and BrdU pulse-chase studies, suggest that the CD342 population is the major myogenic population responsive of regeneration after injury. In addition, the transplant studies indicate that some CD342 negative cells can localize to the satellite cell position and express CD34. We propose a mechanism for this reversible switching of CD34 in response to injury to generate enough differentiating progeny while maintaining the pool of reserve satellite cells during regeneration (Figure 9).

Discussion
In recent years several surface receptors, including CD34, have been described to recognize satellite cells of the skeletal muscle [5]. However, no single surface receptor has proven exclusive to satellite cells, and CD34 is no exception. By utilizing FACS we set out to isolate pure populations of satellite cells using a host of antigens in order to select out contaminating cells that also express CD34. Our results reveal the population of skeletal muscle cells characterized as CD452/CD312/Sca12/CD34+/a72, lack expression of myogenic transcription factors Pax7 and Myf5, do not form myotubes in vitro, nor conform to the satellite cell anatomically defined position under the basal lamina. Therefore, we conclude CD34 alone is inadequate for selecting pure populations of satellite cells from the skeletal muscle.
Since the initial observation by Beauchamp et al. (2000), CD34 has been widely used to identify and isolate satellite cells [9,13,19,40,41,42]. Contrary to the multitude of studies that have examined CD34+ satellite cells, the CD342 minority remained uncharacterized. Initially, we predicted that similar to the hematopoietic system, a linear relationship existed within the satellite cell compartment of skeletal muscle, where more primitive CD342 cells give rise to more committed CD34+ cells. Contrary to our initial hypothesis, our results indicate that following acute injury CD342 cells represent the more myogenically active population. Furthermore, the absence or presence of CD34 is reversible depending on the state of injury.
Currently, an inducible CD34 Cre-Lox reporter model is not available to directly distinguish between downregulation of CD34 or asymmetric division as possible means for CD34 conversion. However, BrdU analysis, myogenin expression/staining and transplant data indicate CD342 cells directly respond and contribute to muscle regeneration following acute injury. Remarkably, despite similar BrdU incorporation at the peak of proliferation between day 2.5-3 post injury, the percentage of CD34+ cells did not change over the course regeneration. Because the absence of CD34 marks a state of activation which seems reversible, the steady percentage of CD34+ cells throughout injury suggests a proportion of myogenic cells are programmed and/or stimulated to maintain a reserve pool of satellite cells during the course of regeneration. This reversible switch from activation to dormancy has also been observed with hematopoietic stem cells as they maintain the stem cell pool following bone marrow injury [43].
Despite their response following acute injury, CD342 cells were quiescent in uninjured muscle. This leaves into question the role of this population during normal muscle homeostasis. Although both CD34+/2 populations were equally myogenic in culture, freshly sorted CD342 cells had lower expression of myogenic-related genes such as Pax7, Pax3, Myf5, MyoD, myogenin, CXCR4, and c-met. Furthermore, we and other groups have observed that within 24 hours of being seeded satellite cells lose CD34 in culture, suggesting CD34 is downregulated upon activation [5]. Therefore, CD342 cells appear to represent more ''primed'' satellite cells poised for injury response that can advance further in the myogenic pathway or upregulate CD34 and revert back to a reserve state depending on the muscle microenvironment. Although CD34's function has yet to be elucidated, reversibility has been reported with hematopoietic stem cells, suggesting a similar role or pathway of CD34 regulation may exists between tissue specific stem cells [44,45]. Interestingly, despite the major myogenic potential and contribution of CD34+/2 cells to muscle regeneration, a minority of Sca1+ (CD452/CD312/CD34+) cells sorted from injured muscle was myogenic in culture, suggesting another possible state of transition or stem cell reservoir exists within limb muscles (data not shown) [46].
Although the true function(s) of CD34 remains uncertain, it has been described that this highly glycosylated sialomucin receptor not only plays a role in signaling but is important in promoting both cellular adhesion and repulsion [44,47]. Interestingly in one study it has been reported that CD34 functions in mast cells to prevent adhesion [48]. Conversely a separate study has reported that CD34 functions to promote lymphocyte adhesion via Lselectin in high endothelial venule cells [49]. Thus, it is possible that the adhesion role of CD34 is dependant on the presence or absence of additional surface receptors as a means to modulate binding or repulsion by various cell types. As aforementioned satellite cells downregulate CD34 in culture, suggesting this receptor functions in maintaining an undifferentiated state. Perhaps such maintenance is linked to adhesion and therefore downregulation of CD34 is required for migration and subsequent engraftment within regions of damaged muscle.
By characterizing skeletal muscle satellite cells based on the presence or absence of CD34 we have identified a key myogenic population responsible for muscle regenerating following acute injury as CD342 (a7+/CD452/CD312/Sca12). In addition we show that the fraction of CD34+ cells remains steady during homeostasis and injury regeneration. Such evidence suggests, photographs. Scale bar = 50 mm. (B) FACS-analysis of BrdU incorporation in sorted CD34+/2 cells after 3 days pulse, indicates relative to bone marrow (BM) cells (top right histogram), without injury both populations are quiescent (top histogram). In mice pulsed for 12 hours between 2.5-3 days post CTX injection, both CD34+/2 cells show nearly equal levels of BrdU incorporation (middle histogram). Mice pulsed for the first 3 days following injury and then chased to day 14 reveal that CD342 cells retain less BrdU whereas CD34+ cells maintained a similar levels to the day 2.5-3 pulse (bottom histogram). Green peaks represent IgG Isotype controls, Violet BM mononuclear cells, Blue CD34+ cells, and Red CD342 cells. Schematics next to histograms portray timelines for each BrdU experiment. Each experimental pulse was conducted with n = 3, 2 month old C57BL/6 mice. doi:10.1371/journal.pone.0010920.g007

Ethics Statement
This study has been reviewed and approved by the University of Washington's Institution Animal Care and Use Committee.

Mice and Animal Care
All mice used for satellite cell characterization (histology, FACSanalysis of individual muscles and injury, single cell culture, cytocentrifugation, BrdU and expression analysis) were C57BL/6, predominantly 2 month old males. Reporter mice used for cytocentrifugation and/or culture were 1.5 month old Myf5 nLacZ/+ and 4 month old 3F-nlacZ-2F [5,50,51], previously described in our studies [12,29]. Transplant donor cells were derived from chicken b-actin promoter driven EGFP or Myf5 nLacZ/+ reporter mice ages ranging from 6-9 month old [52]. Recipients were C57BL/6, ages ranging from 2-12 months old. Unless specified otherwise, each experimental replicate (n = ) represents samples derived from individual animals that were independently processed and analyzed. Mice were housed and maintained in a modified barrier facility located at the University of Washington. All animal experiments were performed in accordance with guidelines approved by the University of Washington's Institution Animal Care and Use Committee.

FACS
Methods for muscle processing and FACS were followed as previously described [20]. Typically from pooled preparations including both hind limb muscles, pectorals and triceps of individual 2 month old C57BL/6 males, we recover 1.5610 6 mononuclear cells per gram of muscle processed. Samples were initially stained with biotinylated anti-CD34 in 100 ml PBS with 0.3% BSA per 10 6 cells for 45 minutes (min) on ice, then washed and resuspended in an antibody cocktail including conjugated streptavidin and directly conjugated anti-CD45/CD31/Sca-1/a7 for a second 45 min incubation. For single cell deposition DAPI (Sigma) was added for viability and left in solution prior to sorting respective donor cell type. The X-axis indicates the state of each recipient muscle prior to transplant (n = 3 per condition). Irrespective of the recipient state, the majority of CD34+ donor cell became or gave rise to CD342 after transplant. Although not statistically significant, this switch slightly increased in CTX injured recipients. In contrast a large proportion of CD342 cells became or gave rise to CD34+ when transplanted into uninjured muscle but significantly remained CD342 in injured muscle. Student's t-test calculated * P#0.05. Error bars represent 6SEM. (B) Sorted CD342 from CTX injured Myf5 nlacZ/+ mice were transplanted into uninjured (n = 3) and injured (n = 3) quadriceps muscles. In addition to a mock injection, a7-(CD45/CD312/Sca12) sorted cell were transplanted into uninjured quadriceps as negative controls (Figures S3 and S4). Recipient muscles were harvested 7 days after transplant and stained for a-smooth muscle actin (aSMA), b-galactosidase (b-gal) and CD34. Cells engrafted under both conditions as indicated by b-gal+ central nuclei in aSMA+ fibers (arrowheads). Insets show co-localization of nuclear b-gal with DAPI and cytoplasmic a-SMA staining. Insets were produced from 406 magnification photographs taken within the same field. (C) Engraftment of CD34-cells into uninjured muscle reveals conversion of CD342 cells to CD34+ cells. Interestingly CD34+ (line arrow) and CD342 (arrowhead), b-gal+ cells were observed near b-gal+ myonuclei (solid arrow) indicating donor CD342 cells can form new myofibers, remain CD342 or gain CD34 expression. Scale bars = 50 mm. doi:10.1371/journal.pone.0010920.g008 . Such trends indicate that CD34+ population represents the pool of less advanced satellite cells that can generate more differentiated CD342 cells while maintaining their numbers during injury repair. In turn, CD342 cells continue proliferating beyond day 3, indicating this population supplies the differentiated progeny required for complete muscle regeneration. (B) Although more differentiated cells reside within the CD342 population following injury, the absence of CD34 does not terminally mark differentiation as CD342 cells can revert back to the CD34+ pool of less advanced satellite cells. Thus activated myogenic cells can switch CD34 expression in response to injury, suggesting that CD34 may play a role in maintaining a reserve pool of satellite cell in order to prevent depletion during muscle regeneration. doi:10.1371/journal.pone.0010920.g009 in order to remove DAPI+ non-viable cells. FACS antibody data, dilutions, combinations and machine specifications are listed in Methods S1. Data was acquired at 10,000 events per sample and cells sorted with a BD Aria I and later Aria II, both using Diva software. Subsequent analysis and flow cytometry plots were generated using FlowJo v7.2.5 (TreeStar, Inc.). Average population values and standard deviations used for error bars were calculated from the analysis of individual mice. Graphs, Student's t-test and ANOVA statistical analyses were generated using Microsoft Excel 2003. For transplants, culture and cytocentrifugation sorted cells were collected in culture media. For reverse and quantitative PCR, sorted cells were collected in 350 ml RLT lysis buffer (Qaigen) supplemented with b-mercaptoethanol per manufacturers instructions and immediately placed on dry ice and stored at 280uC.

Cell Culture
Clonal cultures were prepared by FACS-Aria mediated single cell deposition into 96 well trays. Non-clonal cultures cells were directly sorted into 1.7 ml tubes containing 750 ml media and seeded manually onto 24 well dishes. All tissue culture dishes were coated with 0.67% (w/v) Type-A Gelatin (Sigma) in H 2 O and allowed to dry prior to use. Culture media consisted of Ham's F10 supplemented with 15% Horse Serum and final concentration of 2 mM CaCl 2 , 100 units/ml Penicillin with 100 mg/ml Streptomycin (all from HyClone), and 20 ng/ml bFGF (R&D). Cells were maintained at 37uC, 5%O 2 , and 5%CO 2 and media was changed after 4 days. At day 8 in culture, cells were fixed with 2% formaldehyde (diluted formalin in PBS) and stained with DAPI for counting clones or X-gal for presence of Myf5 nlacZ/+ or 3F-nlacZ-2F reporter activity. After DAPI staining, photographs of each single cell derived colony was taken and used to quantify proliferation. To avoid counting errors we utilized the ImageJ v1.40 g (Wayne Rasband, NIH) cell counter plugin (Kurt De Vos, University of Sheffield) which automatically tallies events labeled by users. Subsequently every DAPI labeled nuclei including myonuclei were accounted for in the final cell count. The average cell doubling for each CD34+/2 population was calculated based on the number of DAPI+ nuclei counted per colony using the following formula in Microsoft Excel; colony cell doubling = log (# nuclei counted/2).

Stainings
All muscle cryosections were cut 8 mm, fixed with 2% formaldehyde for 5 min, washed 36 with PBS, then stained with respective antibodies. For cytocentrifugation staining, freshly sorted cells were spun onto microscope slides at 800 g RCF, fixed with 4% formaldehyde for 5 min, washed, and then stained. Stainings utilizing biotinylated antibodies were first blocked with Vector labs Streptavidin-Avidin blocking kit. Stainings using mouse primary antibodies were completed using the M.O.M staining kit (Vector labs) with some modification to manufacturer's instruction. Cryosections were incubated overnight at 4uC with the M.O.M. mouse IgG blocking reagent and streptavidin conjugated fluorophores were used in place of avidin conjugates. All antibodies were diluted in PBS containing 1% BSA (fraction IV, from Fisher) except when prescribed by M.O.M. kit instructions. DAB staining for Pax7 (performed on freshly isolated, cytocentrifuged cells) was completed using the R.T.U. Vectastain Elite and ImmPact DAB substrate kits (both from Vector labs) following manufacturer's instructions. For X-gal staining, cells and tissue sections were fixed for 10 min at room temperature with 2% formaldehyde/0.2% glutaraldehyde and incubated overnight at 37uC with staining solution containing a final concentration of 1 mg/ml X-gal, 5 mM potassium ferricyonide, 5 mM potassium ferrocyonide, and 2 mM CaCl 2 (all from Fisher). All stainings included negative controls in which primary antibodies or X-gal was omitted (depicted in Figure S4). Brightfield photographs were taken with Fisher Micromaster digital inverted microscope with Infinity optics using Micron v1.05 software. Immunofluorescent stainings were acquired with a Zeiss Axiovert 200 microscope equipped with monochrome camera. Controls for immunofluorescent stainings were taken at the same or greater exposure time as primary stained samples. Monochromatic photographs were colored then merged with Adobe Photoshop v9.0.2. When necessary to reduce background, certain photos' brightness and contrast levels were adjusted within individual channels prior to coloring. A list of primary and secondary antibodies, dilutions and microscope specifications are described in the Methods S1.

Reverse-Transcription and Quantitative-RT-PCR
Methods and reagents for PCR reactions have been previously described by us [20]. Briefly RNA was purified from sorted cells using Qaigens RNeasy kit. All reactions were two-step beginning with cDNA synthesis followed by conventional or quantitative PCR for specific target genes. The following thermal cycling conditions were used: 95uC -79 initial activation followed by 94uC-300;57uC-300;72uC-450, for 35 cycles. Quantitative PCR reactions were run on an ABI 7900HT PCR system using sybr green PCR master mix. Results were analyzed using SDS 2.2 software and relative expressions calculated by comparative Ct method. cDNA from unsorted muscle mononuclear cells ( Figure 1B) or whole muscle ( Figures 4B and 6C) were used for calibration. The endogenous control was GAPDH. Ct values are listed in the Table  S1. Each sample was derived from specific cell populations sorted from individual mice. Reactions for each respective sample and gene were run in triplicate. The relative expression between samples was used to calculate the average expression values and SEM for error bars represented in each figure. Primer sequences for each target gene are listed in Methods S1.

BrdU Incorporation and Analysis
For each pulse 250 ml of 4 mg/ml BrdU (Sigma) in PBS was administered daily via intraperitoneal injection. CD34+/2 cells were sorted from limb muscles, fixed with 4% formaldehyde for 5 minutes, washed with PBS and left overnight at 4uC. Tibialis and femur bone marrows were collected as a positive control for BrdU incorporation by flushing the tibia and femur with a 28 gauge needle. Fixed cells were processed and stained with PE conjugated anti-BrdU antibody (BD) according to manufacturer's protocol. Briefly cells were further fixed with 70% EtOH, denatured with 2 M HCL, and stained in PBS containing 0.5% Tween20 and 0.5% BSA. Sorted cells were identified as DAPI positive and BrdU incorporation was defined by comparing positive bone marrow cells with IgG isotype stained controls.

Supporting Information
Methods S1 Includes antibodies and dilutions, fluorescent microscope specifications, antibody/fluorophore combinations used for each FACS experiment, BD flow cytometer specifications, and primers. Found at: doi:10.1371/journal.pone.0010920.s001 (0.12 MB DOC) Figure S1 FACS size selection strategy. In order to remove duplets and debris we first gate on the tight population of small, low granulated events located at the bottom left of the Forward Scatter Area (FSC-A, x-axis) vs. Side Scatter Area (SSC-A, y-axis) graph. Next we select events by Forward Side Scatter Area (FSC-A, x-axis) vs. Width (FSC-W, y-axis) to remove the smallest and largest events. Finally we remove events that are too granulated by Side Scatter Area (SSC-A, x-axis) vs. Side Scatter Width (SSC-W, y-axis) then move to antigen selection depicted in Figure 5A. For single cell deposition DAPI negative cells (live cells) were selected prior to antigen selection (not shown). Found at: doi:10.1371/journal.pone.0010920.s002 (1.18 MB TIF) Figure S2 Following injury the majority of a7 integrin+ cells are negative for CD34, a-smooth muscle actin, and CD31. Cryosection staining for a-smooth muscle actin (aSMA), CD31 and CD34 illustrates the majority of a7 integrin+ cells arising 3 days post CTX injury, are not endothelial or smooth muscle cells. (A) aSMA and CD31 staining shows the smooth muscle cells are confined to large vessels within injured muscle. (B) aSMA and a7 integrin staining confirms that the a7 integrin+ cells are not vascular smooth muscle cells within damaged areas of muscle (C) Staining for CD34 and a7 integrin shows the a7 integrin+ cells in the injured area are CD34-, while the majority of CD34+ cells represent vessels in the less injured region. Scale bars = 50 mm.