PU.1 (Sfpi1), a pleiotropic regulator expressed from the first embryonic stages with a crucial function in germinal progenitors

Male germinal differentiation in mammals comprises two distinct periods delimited by birth (reviewed by Hogan et al., 1994). During embryonic development, the primordial germ cells (PGCs) specified in the epiblast at embryonic day (E) 7.5 divide and colonize the gonadal ridge, where they become associated with somatic cells in the cord structures characteristic of the fetal testis. At the end of prenatal development, the germ cells (gonocytes) stop dividing. The spermatogonial stem cells (SSCs), which resume proliferation during the first days after birth(de Rooij and Grootegoed,1998), were functionally defined by their ability to reconstitute a functional seminiferous epithelium after transplantation into a sterile host(Brinster, 2002; Brinster and Zimmermann, 1994). The cyclic process of spermatogenesis initiated from SSCs will thereafter be maintained throughout the whole of adult life.

The two critical turning points in this process are the specification of PGCs in the early embryo and the appearance of dividing SSCs shortly after birth. Genes expressed at the earliest stages of specification of the germ line have been identified recently. They include fragilis (Ifitm3)(Saitou et al., 2002; Tanaka et al., 2005), stella(PGC7, Dppa3) (Bortvin et al.,2004; Nakamura et al.,2007; Payer et al.,2003; Saitou et al.,2002; Sato et al.,2002) and Blimp1 (Prdm1)(Ohinata et al., 2005; Saitou et al., 2005; Vincent et al., 2005). PGCs appear first in the proximal epiblast, in a region in which fragilis is upregulated at E6.5. This restricted PGC population (approximately 40 cells)expresses Blimp1 (E7.25) and thereafter (E7.5) stella. After birth,only a limited number of genes crucial for establishment and self-renewal of SSCs have been identified. Requirement for the Ret receptor was postulated based on the requirement for its ligand GDNF(Meng et al., 2000), and a partial requirement for Bcl6b was described recently(Oatley et al., 2006).

Our starting point was the establishment of a transcriptome profile of purified preparations of SSCs, on the assumption that crucial genes would appear among those expressed at elevated levels in the stem cells. Work on adult SSCs is made difficult by their small number. We previously reported an efficient procedure, in which a truncated version of the Stra8promoter is used to express in SSCs a neutral heterologous surface protein including two domains of the human CD4 antigen (huCD4)(Giuili et al., 2002). Once fractionated by immunomagnetic sorting, the huCD4-positive fraction is homogeneous for the expression of known SSC markers and highly efficient in the colonization of a sterile recipient. Hybridization on DNA arrays provided a shortlist of transcripts upregulated in the purified fraction. Interestingly, the list included PU.1 (Sfpi1), a member of the Ets family, which has been extensively studied for its important regulatory functions in the renewal of progenitor cells and in early differentiation, so far exclusively in the various lineages of hematopoietic differentiation (reviewed by Lloberas et al., 1999; Metcalf et al.,2006). The availability of a null mutation generated by insertion of the green fluorescent protein (GFP) reporter gene in the first exon(Back et al., 2004), allowed us to demonstrate here a strict requirement for PU.1 in the fetal differentiation of the male germ line and to show promoter activity starting at the earliest embryonic stages.

129Sv ES cells were grown at 37°C, 5% CO₂, 95% humidity in dishes coated with a feeder layer of mitotically inactivated primary mouse embryonic fibroblasts. Cells were kept in Dulbecco's modified Eagle's medium supplemented with 15% FCS, 100 IU/ml penicillin/streptomycin, 100 IU/ml non-essential amino acids, 0.1 mM 2-mercaptoethanol and 1000 IU/ml LIF.

A truncated version of the mouse Stra8 promoter was used as described previously (Giuili et al.,2002) to direct expression to the surface of SSCs of a fusion protein including two domains of the human CD4 protein (Stra8-CD4HAglo mice). Progeny were genotyped by PCR amplification of tail DNA using primers hCD4F and hCD4R (Table 1). PU.1^G/+ and PU.1^G/G mice(Back et al., 2004) were genotyped by PCR with primers VP102 (wild-type exon 1), VP104 (mutated exon 1)and WF189 (intron 1) (Table 1). Amplified fragments were 650 bp for the PU.1⁺ allele and 900 bp for PU.1^G. Families were maintained in the C57BL/6×DBA/2 F1 genetic background. Investigations were conducted in accordance with French and European rules for the care and use of laboratory animals.

The day of appearance of a vaginal plug was taken as E0.5. Plug-positive females were sacrificed by cervical dislocation for collecting blastocysts,E6.5, E12.5, E15.5 and E17.5 embryos. E6.5 and older embryos were genotyped by PCR with primers VP102, VP104, and WF189(Table 1). E12.5, E17.5 and E18.5 embryos were dissected for collection of the embryonic and fetal gonads.

Total germ cells were prepared as described previously(Vidal et al., 2001). Immunomagnetic isolation of huCD4-positive cells from total testicular cells of Stra8-CD4HAglo mice was performed using the CD4-positive Isolation Kit(Dynal, Oslo, Norway) according to the manufacturer's instructions. For each SSC preparation, 50 transgenic males were used for a final yield of 2×10⁶ cells corresponding to the expected number of 20,000 stem cells per testis (de Rooij and Grootegoed, 1998).

Total germ cells prepared from 20 mice were loaded in the Beckman (Palo Alto, CA, USA) elutriation rotor JE-5.0, at a flow rate of 7 ml/minute and a constant speed of 2000 rpm. Cells were collected in 11 fractions of 400 ml each, obtained by changing the flow rate from 7 to 50 ml/minute at constant speed. Fraction purity was checked by microscopic analysis after Hoechst 33253 staining.

Total RNA was extracted from cell fractions using Trizol (Invitrogen, Cergy Pontoise, France) according to the manufacturer's instructions. Ten to 20μg of stem cell RNA was obtained per SSC preparation. Quality of RNA was checked by agarose gel electrophoresis and using an Agilent bioanalyser.

Histology and immunohistochemistry analyses were performed according to published protocols (Wagner et al.,2005). Briefly, the samples were fixed overnight at 4°C in paraformaldehyde (3% in PBS) and embedded in paraffin. Routine histological examination was performed after Hematoxylin staining. Sections (3 μm) were incubated (16 hours, 4°C) with primary antibodies against PU.1 (T-21,sc-352, Santa Cruz Biotechnology, Santa Cruz, CA), PCNA (PC10, sc-56, Santa Cruz Biotechnology), each diluted 1:100 in PBS, 0.1% Triton X-100, 3% BSA. Antigen detection was performed with a biotinylated goat anti-rabbit antibody(Vector Laboratories, Burlingame, CA), followed by incubation with peroxidase-coupled streptavidin (Sigma, St Louis, MO). Visualization was achieved with DAB substrate (SK-4100, Vector Laboratories)(Wagner et al., 2006). For PCNA staining, the subsequent antigen detection was performed using the M.O.M. Kit (PK-2200, Vector laboratories). ES cells were stained with the PU.1 antibody as described for U2OS cells(Wagner et al., 2005). Omission of the first antibody served as a negative control; additionally,blocking peptide against PU.1 (sc-352P, Santa Cruz Biotechnology) was used to confirm the specificity of staining.

Apoptotic cells were detected by Tunel staining of paraffin sections using the In Situ Cell Death Detection Kit (Roche Molecular Biochemicals) as described previously (Wagner et al.,2005). Three sections from three different animals were analyzed.

Distribution of GFP fluorescence within the embryo was determined by confocal microscopy with a Leica CLSM microscope equipped with an argon-krypton ion laser emitting light at 488 and 514 nm and producing an excitation wavelength of 515 and 580 nm, and with a UV laser (excitation 352 nm, emission 405 nm). A series of optical sections of approximately 0.95 μm was recorded through the z-axis, using an objective of 43× and a zoom of 1.99.

DNase I-treated total RNA (1 μg) was reverse-transcribed using the M-MLV reverse transcriptase (Promega, Madison, WI, USA) according to the manufacturer's instructions. PCR amplification was performed with the Ready Mix Red Taq PCR Mix (Sigma) using the oligonucleotide primers shown in Table 1.

RNA preparations were analyzed by in vitro amplification and hybridization to DNA microarrays containing about 14,000 genes (Mouse Genome 430A 2.0,Affymetrix, Santa Clara, CA, USA). Absolute and comparison analyses of the mouse oligonucleotide arrays were conducted using the statistics-based Affymetrix software MAS-5.0 (GeneChip Software) with the default settings. We used the MAS5 algorithm (Affymetrix) to determine the `present' and `absent'calls for each probe set. Quantitative expression values were obtained with the robust microarray analysis (RMA) algorithm, which uses a method based on quantile normalization to calculate an expression value for each probe set. The complete microarray data are available in the NCBI GEO repository(http://www.ncbi.nlm.nih.gov/geo/).

cDNA samples from SSC and total fractions were analyzed by real-time PCR using the qPCR MasterMix Plus for SYBR Green (Eurogentec, Seraing, Belgium)and an ABI Prism 7000 sequence detector according to the manufacturer's protocol (Applied Biosystems, Foster City, CA). Amplification was performed by applying the comparative Ct method, as described in the manufacturer's protocol. Sequences of oligonucleotide primers are shown in Table 1.

Pregnant female mice received 50 mg/kg of 5-bromo-2′-deoxyuridine(BrdU, Roche Molecular Biochemicals) by intraperitoneal injection and were sacrificed 5 hours later. Paraffin sections were stained by immunolabeling with anti-BrdU antibody (Roche Molecular Biochemicals) as described above.

The donor testes were dissected at E17-19. For transplantation, the recipient testis (6-week-old ICR nu/nu mice, Charles River, Wilmington, MA)was exteriorized through a midline abdominal incision after anaesthetizing the recipient mouse. Using fine forceps, a small cut was made in the tunica albuginea, and the testis graft inserted into the testicular parenchyma. One week to 1 month after transplantation, the recipient testis with the transplant was embedded in paraffin and histological sections were cut at 8-μm intervals from each testis piece and stained with Hematoxylin and Eosin for microscopic examination.

SSCs were prepared by immunomagnetic sorting with anti-huCD4 antibodies from males of the Stra8-CD4 transgenic family. In these families(Giuili et al., 2002), a neutral surface marker is expressed under control of a truncated version (411 bp) of the promoter of the Stra8 gene. Unlike the wild-type promoter,the truncated version directs transgene expression to a minor population of premeiotic cells, positively identified as SSCs on the basis of expression of marker genes and of their high efficiency of establishment upon transplantation. Among other markers routinely used to ascertain the purity of the fractions (expression of cyclin A2, but not A1, of integrins β1 andα6), the two most characteristic complementary criteria are shown in Fig. 1A, namely expression of the endogenous Stra8 gene, characteristic of both SSCs and differentiated spermatogonia(Oulad-Abdelghani et al.,1996), and absence of Kit transcripts, which are expressed only in differentiated spermatogonia(Shinohara et al., 2000). Contamination by cells engaged in the germinal differentiation process was minimal, as shown by the negative results registered for Sycp1, which is characteristic of the pachytene spermatocytes(Meuwissen et al., 1992) and Prm1, a marker of spermatids(Peschon et al., 1987). Possible contamination by postmeiotic cells was in addition limited by starting the purification process from young males (3-4 weeks), in which the first wave of spermatogenesis would have just passed the second meiotic division (Bellve et al., 1977). Yields were ∼20,000 cells per testis, in agreement with previous estimates(de Rooij and Grootegoed,1998).

Three independent preparations of 20-40 μg SSC RNA were analyzed by hybridization on DNA microarrays (14,000 genes, Affymetrix Mouse Genome 430A 2.0). In order to identify genes selectively expressed in SSCs, we selected genes that were detected as `present' in all three huCD4<sup>+</sup> samples and `absent' in all three huCD4^- samples. We further imposed that the measured detection values exhibit at least a 2-fold difference between huCD4⁺ and huCD4^- samples, with a P-value(Student's t-test) of less than 0.05. These combined criteria led to a list of 84 genes (Table 2). In addition, to estimate the `background' of non-specific genes within that list, we performed similar analyses between permuted combinations of samples,in which two samples had been switched between the positive and negative groups. In each case, less then ten genes were found, suggesting that most of the selected genes are indeed differentially expressed between SSCs and non-SSCs. Differential expression of four genes randomly taken in the list(Cd14, Cd97, Nek7, Sfpi1) was also confirmed by quantitative RT-PCR(data not shown). We also noted that a number of genes, although excluded from this most stringent list because of low but significant expression in at least one of the huCD4^- samples, showed higher expression levels in the positive fraction. RT-PCR assays of five transcripts in this category(Cd81, Nptn, Fos, Cdc42, Lyn) again showed a differential expression in the two fractions (data not shown).

Several genes encoding transcription factors were present among the genes in the most stringent shortlist: Lyl1 encoding a bHLH factor(Capron et al., 2006), Elk3 encoding an Ets-domain factor(Chung et al., 2006) and Rbpsuh (Rbpj) encoding the transcriptional mediator of Notch signaling (Vasyutina et al.,2007).

Although all of these genes might play important regulatory roles in controlling SSC gene expression, we were especially intrigued by the presence of PU.1 (Sfpi1). This finding warranted further analysis because the published literature (reviewed by Lloberas et al., 1999; Metcalf et al., 2006)consistently refers to PU.1 as a transcription regulator that acts exclusively in the progenitors of blood cells. Both PU.1 RNA and the protein were unequivocally detected in huCD4⁺ cells by quantitative RT-PCR and by immunolabeling (Fig. 1B,C).

Immunodetection of PU.1 protein in adult testis sections revealed a complex pattern. As expected from analysis of the huCD4⁺ fraction, the protein was detected in a small number of spermatogonia in the most peripheral region of the tubule (Fig. 2A). Among these cells, the paired (A_paired) and aligned(A_aligned) cells characteristic of the early spermatogonial stages(de Rooij and Grootegoed,1998) were clearly identified(Fig. 2G-I). Whereas the bulk of differentiated spermatogonia were negative, expression of the protein was also evidenced in meiotic spermatocytes(Fig. 2A, asterisk). This finding was somewhat unexpected, a contamination by pachytene spermatocytes of the huCD4⁺ fraction being made unlikely by the absence of the characteristic meiotic transcripts Sycp1(Fig. 1A) and Ccna2(cyclin A2, not shown) (see Giuili et al.,2002). Meiotic expression was nevertheless confirmed by in situ analysis in the testis of 3-week-old males, in which the first synchronous wave of spermatogenesis leads to the accumulation of meiotic cells, which were uniformly and strongly labeled by anti-PU.1 immunostaining(Fig. 2C-E). This was further confirmed by PCR assays performed on elutriation-purified spermatocytes(Fig. 2F). The quantitatively minor representation of pachytene spermatocytes in the total huCD4^-fraction accounts for the initial detection of higher levels of PU.1transcripts in huCD4⁺ cells.

In the embryonic testis, PU.1 was detected in the germinal compartment, at least from E12.5 onward (Fig. 3). Further studies took advantage of a mouse mutant, PU.1^G, generated by insertion of a GFP cassette into the first exon of the gene to provide both a reporter for promoter activity and a null mutation (Back et al.,2004). PU.1^G/G homozygotes die shortly after birth and heterozygotes are healthy and fertile. Embryos of crosses between heterozygotes were individually genotyped and testis development was analyzed in the PU.1^G/G progeny, with their G/+ and +/+littermates used as controls. Histological analysis of homozygotes at E15.5 and E17.5 seemed at first to indicate that the testis was normally constituted, with a myoid cell layer surrounding internal cells in a structure similar to the testis cords of the wild type(Fig. 4A,B). The number of gonocytes per section was, however, smaller than in the controls. Among a series of genes characteristic of prenatal germinal differentiation, RT-PCR analysis detected transcripts of Oct4(Yoshimizu et al., 1999), Dazl (Lin and Page,2005), Taf4b(Falender et al., 2005) and vasa (Tanaka et al., 2000),but expression of Plzf (Zbtb16)(Buaas et al., 2004; Costoya et al., 2004) and stella (Saitou et al., 2002)was strongly reduced or absent in the null mutant(Fig. 4C and Fig. 5).

The defect in spermatogenesis of the PU.1^G/G mutant was related to a decrease in cell proliferation, as evidenced by decreased BrdU incorporation during development (Fig. 6). The number of germ cells incorporating BrdU was significantly decreased at E12.5 and E13.5. By contrast, BrdU incorporation in the somatic component of the testis cords (essentially Sertoli cells) was decreased to a lesser extent and only at later times, an observation that will be addressed in future studies. At postnatal stages, a functional defect of PU.1^G/G testes was evidenced by testis graft experiments,an approach made necessary by the early death of the mutant homozygotes. Testes dissected shortly before birth (E18.5) were grafted under the tunica albuginea of recipient nude mice. Spermatogenesis is known to proceed normally in such grafts up to the production of functional sperm(Honaramooz et al., 2002). Grafts of the PU.1^G/+ testes showed, after 1 week, a strong proliferative activity as evidenced by PCNA immunolabeling(Fig. 7Ae,g), leading, after 1 month, to the development of a structurally normal seminiferous epithelium(Fig. 7Ba,c). By contrast,grafts of their PU.1^G/G littermates showed, 1 week after transplantation, only a limited proliferative activity of Sertoli cells in the periphery of the tubules (Fig. 7f,h), which were essentially empty after 2 weeks (not shown) and 4 weeks (Fig. 7Bb,d). Thus, not only was the number of gonocytes reduced owing to a decreased proliferation rate, but the surviving germ cells were unable to enter a normal differentiation pathway.

Although not absolutely required for spermatogenesis(Bortvin et al., 2004), stella is the earliest marker of embryonic differentiation; it is already expressed in the blastocyst and later restricted to the germinal lineage(Saitou et al., 2002; Sato et al., 2002) (our unpublished results). The correlation between PU.1 and stella expression at the late fetal stage led us to check whether, as is stella itself, PU.1 is a marker of early embryonic development. Confocal microscopy detected GFP fluorescence in PU.1^G/+heterozygotes as early as the blastocyst stage (E3.5), at which the inner cell mass showed heterogeneous but strong labeling(Fig. 8A). ES cells were accordingly checked and found positive by immunolabeling with anti-PU.1 antibodies (Fig. 8B). In the E6.5 postimplantation embryo (Fig. 8C), GFP fluorescence indicated extensive expression in the epiblast and extraembryonic ectoderm, with a marked concentration in the general area of the node, which was maintained during the following days in a variety of structures. Although both genes are expressed at the same early stage, we could exclude the possibility that the PU.1 protein directly controls stella transcription, as transcripts of the latter were clearly detected in the early PU.1^G/G embryo (E10.5, data not shown). The pattern of PU.1 expression is in fact distinct and more extensive than that of stella, Blimp1 and fragilis, currently considered as the relevant markers of germ line specification and similarly expressed at very early stages (Saitou et al., 2002).

Previous work underlined the pleiotropic roles of the PU.1 protein, which exerts distinct functions in the progenitors of the various hematopoietic lineages (reviewed by Lloberas et al.,1999; Metcalf et al.,2006). The picture must now be expanded to include other classes of stem cells, at least the SSCs of the adult mammalian testis and embryonic PGCs. The list of progenitors under PU.1 control might not yet be complete,considering the widespread expression of the gene during embryogenesis. Considering the lineages for which sufficient data exist, namely the lymphopoietic and erythropoietic progenitors(Back et al., 2004; Iwasaki et al., 2005) and now the germinal lineage, a common picture seems to be emerging, in which the gene would act in the maintenance of cells in the undifferentiated state.

In the germinal differentiation process, our initial finding was a regulated expression of PU.1 RNA in the adult testis, first in SSCs and early spermatogonia, and later in meiotic spermatocytes. The availability of the null mutant PU.1^G was initially thought to provide the right tool to unravel its role in these cell types. It turned out,however, that expression of the locus starts at a much earlier stage of germinal differentiation - in fact, at the time of the first specification of germ cells and even before. Furthermore, although the structure of the fetal testis appeared at least grossly normal, it became clear that proliferation of the germ cells was strongly affected, resulting in a much reduced number of gonocytes at the end of fetal development. The remaining cells were in fact unable to enter the normal differentiation pathway, the lack of recognizable germinal progression in the grafted testes defining an absolute requirement for PU.1 expression in the fetal differentiation of germ cells.

A relationship between PU.1 and stella expression had been considered because of the concomitant expression of the genes at very early developmental stages and by the absence of stella RNA in the fetal testis of the PU.1-null mutants. A direct transcriptional regulation of stella by the PU.1 protein was, however, excluded by the observation that the gene is expressed in the early embryo (E7.5) of the PU.1-negative mutant. It must also be remembered that abrogation of stella expression does not affect germinal differentiation (Bortvin et al.,2004), and we also noted that the widespread expression of PU.1 in the postimplantation embryo does not correspond to that of stella, which is rapidly restricted to the germinal lineage.

The finding that PU.1 is required for the embryonic differentiation of germ cells leaves open the question of its role in the adult testis. An analysis would require a conditional mutation in which gene expression would be arrested exclusively after birth. An intriguing and complex pattern, with expression in SSCs and possibly early spermatogonia, extinguished in differentiated spermatogonia and starting again in meiotic cells, might actually suggest distinct activities in these different cell types. Altogether, a relatively extended domain of expression, starting at the earliest stages of development and required in at least two independent lineages, suggests a common function in the maintenance of embryonic and adult stem cells and in the control of their differentiation. Given the multiplicity of protein variants encoded by the PU.1 gene(Lloberas et al., 1999), it remains to be determined whether they exert distinct functions in different lineages.

We thank Yan Fantei-Caujolles and Sandra Kanani for expert handling of databases; Jean Claude Hammou for pathological expertise; Christelle Thibault for the generation of microarray data; Cedric Mathews for confocal microscopy analysis; and K. D. Wagner for critical reading of the manuscript. V.O. was supported by scholarships from Région PACA-Inserm and from Association pour la Recherche sur le Cancer, France. N.W. is a fellow of Fondation de France and was supported financially by A. Schedl. This work was made possible by a grant of Ligue Nationale Contre le Cancer as Equipe Labellisée and by grants from Ministère de la Recherche to M.R. and from Inserm (ACI FNS2001) to F.C.