Identification of Neuregulin as a Factor Required for Formation of Aligned Spermatogonia*

In the absence of somatic cells, medium conditioned by the SNL fibroblast line (SNL-CM) is able to stimulate primary cultures of rat type-A single spermatogonia to develop into chains of aligned spermatogonia at the 8-, 16-, and 32-cell stages. By comparison, medium conditioned by an MSC-1 Sertoli cell line is ineffective. Glial cell line-derived neurotrophic factor (GDNF)-like molecules were identified in SNL-CM and recombinant forms of GDNF, neurturin, and artemin were shown to stimulate formation of aligned spermatogonia, but principally to only the 4- and 8-cell stages. Because SNL-CM and GDNF-like molecules stimulated the formation of spermatogonial chain length differently, we purified components of SNL-CM to identify the additional contributing factor(s). A fraction was isolated that was dependent on GDNF, but required for effective formation of 16- and 32-cell chain lengths. Sequence analysis identified the factor as mouse neuregulin-1. At picomolar concentrations, recombinant neuregulin-1 in combination with GDNF effectively stimulated formation of aligned spermatogonia up to the 32-cell stage. Neuregulin in the absence of GDNF was relatively ineffective. Soluble receptors for neuregulins blocked the effects of GDNF and SNL-CM, suggesting that both neuregulin and GDNF are required for effective formation of long spermatogonial chains. Addition of neuregulin-1 to cultures on MSC-1 feeder layers resulted in spermatogonial behavior similar to that seen on feeder layers of SNL fibroblasts. In fact, SNL cells were found to express 100-fold higher levels of neuregulin-1 transcripts than MSC-1 cells. Thus, we identify neuregulin as a factor required for spermatogonial amplification and differentiation in culture.

Spermatozoa are produced from spermatogonial stem cells in the testes through the processes of spermatogenesis and spermiogenesis (1). Type A-single (A s ) spermatogonia have been proposed to represent the stem cell of the mammalian testis (2)(3)(4)(5)(6). Based on an ability to reconstitute spermatogenesis in recipient testes (7,8), enriched fractions of spermatogonial stem cells have been isolated by a variety of techniques including selection on laminin matrices (9 -13), selection with immunomagenetic particles (13)(14)(15)(16), flow cytometry (13,(17)(18)(19)(20), and ablation of differentiated germ cells from the testis (9,21). Methods to derive and propagate spermatogonial lines for long periods of time in culture also have been established (15,(22)(23)(24)(25)(26). So now, highly pure spermatogonial stem cells can be used to study spermatogenesis. Understanding the molecular basis of spermatogenesis could potentially lead to treatments and cures for male infertility, provide targets for contraceptive development, assist in the preservation of endangered species, and afford alternative methods to produce transgenic animals valuable in agriculture and medicine (27,28).
However, primary culture systems supporting spermatogenesis from purified spermatogonial stem cells have yet to be established (29 -31). Isolated populations of differentiating spermatogenic cells show limited viability once placed in culture (14,(32)(33)(34)(35). An exception is a spermatogonial line transformed by overexpression of telomerase, which appears to selfrenew for an extended period of time in culture; this line can be conditionally stimulated to differentiate into round spermatids by supplementing the cultures with stem cell factor (36). Differentiation to the spermatocyte stage in culture has yet to be reported for other established primary (15,(22)(23)(24)26) or transformed spermatogonial lines (37,38).
In rodents, spermatozoan progenitor cells progress through several pre-meiotic steps as clonal units of distinct spermatogonial types (i.e. types A s -A pr -A al -A 1 -A 2 -A 3 -A 4 -Int -B) (39)(40)(41)(42)(43). The cells comprising each clonal unit are connected to each other by intercellular bridges that are formed as remnants of incomplete cytokinesis (44,45). The number of interconnected progenitor cells within each spermatogenic unit amplify as they differentiate through up to 11 mitotic divisions, followed by two meiotic divisions, prior to advancing through the subsequent steps of spermiogenesis (39,40). It has been estimated that at least 4% of the mouse genome is dedicated specifically to supporting these processes (46); more recent studies suggest that even this percentage may be a low estimate (47).
The goal of this study was to identify factors that promote the early stages of spermatogenesis in culture. Two mouse feeder cell lines that differentially regulate rat spermatogonial stem cell fate in vitro were particularly useful in defining one factor (10). When freshly isolated spermatogonia that bind to laminin are cultured on an SNL fibroblast line in medium containing serum, the relative stem cell concentration is rapidly reduced, whereas total germ cell numbers increase (10). Based on a panel of molecular markers, the spermatogonia in these cultures appear to be differentiating through the early steps of spermatogenesis (10). In sharp contrast, when these cells are cultured on an MSC-1 Sertoli cell line, the spermatogonia maintain relative stem cell numbers (10). These studies suggest that SNL fibroblasts express molecules capable of stimulating spermatogonial development through the early stages of spermatogenesis in culture, whereas MSC-1 cells are deficient in the production of such factors. Here, we identify neuregulin-1 as one factor required for the early steps of spermatogenesis, namely the formation of type A-aligned (A al ) spermatogonia.

EXPERIMENTAL PROCEDURES
Materials and Chemicals-Dispase and rat tail collagen I-coated culture dishes were from Fisher, Inc. Media, trypsin solutions, and antibiotic-antimycotic solutions were from Invitrogen. Fetal bovine serum (FBS) 2 was from Atlanta Biologicals, Inc. Phosphate-buffered saline was from JRH Biosciences, Inc. Mitomycin-C, mouse laminin, and 2-mercaptoethanol were from Sigma. Alexa Fluor-594-conjugated, goat anti-rabbit IgG, and Hoechst 33342 were from Invitrogen. Glial cell line-derived neurotrophic factor (GDNF) neutralizing antibodies and recombinant forms of GDNF, neuregulins, and FcErbBs were from R&D Systems, Inc.
Testicular Cell Cultures-Seminiferous tubules were isolated from the testes of 22-23-day-old wild-type (Harlan, Inc.) or homozygous SD-Tg(ROSA-EGFP)2-4Reh Sprague-Dawley rats. Rats in the SD-Tg(ROSA-EGFP)2-4Reh line were produced by pronuclear injection, and exhibit germ cell-specific expression of GFP (48). This allows germ cells (GFP ϩ ) to be clearly identified in cultures of testis cells prepared from this transgenic rat line (10,26). Testes from rats of these ages contain late pachytene spermatocytes as the most advanced germ cell type (49). The tubules were enzymatically and mechanically dissociated into a cellular suspension to generate cultures of testis cells. The testis cell cultures were then used as a source for isolating enriched populations of laminin-binding germ cells, laminin non-binding germ cells, and tubular somatic cells by our previously established procedures (10,11). The isolated testis cell populations were used to initiate spermatogenesis colony forming assays, as described below. When isolated by these procedures, the laminin-binding germ cell fraction consists of Ͼ90% undifferentiated, type-A spermatogonia and Ͻ5% somatic cells (see "Results"). Such spermatogonia isolated from SD-Tg(ROSA-EGFP)2-4Reh rats are also referred to herein as GFP ϩ spermatogonia.
Spermatogenesis Colony Forming Assays-To analyze colony formation in vitro, the number of aligned spermatogonia formed from freshly isolated testis cells (day 0) was scored on culture day 9. Standard incubation conditions were: 0.5-1 ϫ 10 4 cells plated/cm 2 , DHF12 medium supplemented with 10% FBS, 30 M 2-mercaptoethanol, minus or plus added test reagents, at 32.5°C in 5.5% CO 2 . On day 9, the cultures were fixed for 10 min in ice-cold 4% paraformaldehyde, labeled with the anti-TEX14 IgGs using our reported protocol for immunocytochemistry (10), and then the number of GFP ϩ /TEX14 ϩ A al spermatogonia formed in each culture were scored by counting on a fluorescent microscope. For characterization and isolation of factors from SNL-CM, cultures of spermatogonia were initiated on S-laminin. Compared with laminin, S-laminin stimulated the formation of 4-fold more total A al spermatogonia (TEX14 ϩ , GFP ϩ )/well at the 8 -32-cell stages in response to GDNF (supplemental Fig. S1). At plating densities of less than 10 4 cells/cm 2 on S-laminin, individual colonies of GFP ϩ spermatogonia were composed of individual chains formed by intercellular bridges based on TEX14 expression. Therefore, under these conditions the number of independent chains/well is considered to be equal to the number of GFP ϩ colonies/well.
To analyze colony formation in vivo, WT Sprague-Dawley rats at 12 days of age were injected (intraperitoneal) with 12.5 mg/kg busulfan (4 mg/ml in 50% Me 2 SO) and then used as recipient males at 24 days of age. Donor rat germ cells (GFP ϩ ) isolated from culture were loaded into injection needles fashioned from 100 l of glass capillary tubes at ϳ3 ϫ 10 3 GFP ϩ cells/65 l of culture media (DHF12 medium supplemented with 10% FBS, 30 M 2-mercaptoethanol) containing 0.04% trypan blue. The entire 65-l volume was then transferred into the seminiferous tubules of anesthetized rats by retrograde injection through the rete of their right testes (50). The number of GFP ϩ colonies formed/testis were scored by using an Olympus IX70 fluorescence microscope (Olympus, Inc.) to visualize donor cell transgene expression in the seminiferous tubules at 32 days following transplantation.
Collection of Conditioned Medium-SNL fibroblasts were grown as monolayers in plastic culture dishes (10 cm) containing 10 ml of cDMEM. The cDMEM was removed from confluent cultures by aspiration, and then the monolayers in each dish were rinsed with 12 ml of serum-free cDMEM (sfDMEM)/dish. The sfDMEM used to rinse the monolayers was thoroughly removed from each culture and then replaced with 10 ml of fresh sfDMEM/dish. The cultures were then maintained for 2 days at 37°C in 5% CO 2 . After 2 days, the conditioned sfDMEM (SNL-CM) was removed from the cultures and transferred into sterile 50-ml tubes. The collected SNL-CM was centrifuged at ϳ3000 ϫ g for 10 min at 4°C, using a Beckman GPR tabletop centrifuge. The supernatant liquid was retained and stored frozen in sterile T75 flasks at Ϫ40°C. Concentration of SNL-CM-For pilot experiments, 400 ml of SNL-CM was processed under sterile conditions using an Amicon 8400 ultrafiltration apparatus fitted with 3000 M r cut-off filters (YM3, Millipore). This step yielded 26 ml of retentate fluid (15ϫ SNL-CM) and 374 ml of filtrate fluid. To identify factors that stimulate spermatogonial development by purification techniques, 10 liters of SNL-CM were collected and processed by the same method to yield ϳ125 ml of retentate fluid (80ϫ SNL-CM) and ϳ875 ml of filtrate fluid. Retentate and filtrate fluid were stored frozen at Ϫ40°C until further use.
Purification of Spermatogenic Factors from Conditioned Medium-The preparation of 80ϫ SNL-CM was fractionated using high performance liquid chromatography (HPLC). The HPLC runs were supported using a Hewlett-Packard 1100 Series HPLC work station equipped with online monitoring, and off-line analysis software (HP Chem-Station). The 80ϫ SNL-CM was adjusted to 0.5% trifluoroacetic acid and injected onto a preparative C4, HPLC column (22 ϫ 250 mm, TP214 Vydac, Inc.) equilibrated at 1% acetonitrile, 0.1% trifluoroacetic acid. Bound components of the SNL-CM were then eluted from the column using a two-step linear gradient of increasing acetonitrile concentration over 120 min (1% acetonitrile, 0.1% trifluoroacetic acid from 0 to 15 min, then 1 to 80% acetonitrile, 0.1% trifluoroacetic acid from 15 to 120 min) at a flow rate of 4 ml/min. Column fractions (8 ml/fraction) containing peaks of activity from duplicate columns runs (60 ml, 80ϫ SNL-CM/run) were pooled, concentrated on a SpeedVac (Savant Instruments, Inc.), suspended to 2 ml in sterile column buffer (100 mM ammonium acetate, pH 4.9), and then further fractionated over a 25/60, Sephacryl-200 column (GE Healthcare) pre-equilibrated in column buffer using a flow rate of 1.5 ml/min. Column fractions (5 ml/fraction) containing activity were pooled, concentrated on a SpeedVac, suspended to 1 ml in 20% acetonitrile, 0.1% trifluoroacetic acid, and then further fractionated over a semi-preparative, C4, HPLC column (10 ϫ 250 mm, TP214, Vydac, Inc.) using a two-step linear gradient (20% acetonitrile, 0.1% trifluoroacetic acid from 0 to 5 min, then 20 -50% acetonitrile, 0.1% trifluoroacetic acid from 5 to 90 min) at a flow rate of 3 ml/min. Fractions containing activity were pooled, concentrated on a SpeedVac, suspended to 1 ml in 20% acetonitrile, 0.1% trifluoroacetic acid, and then further fractionated over an analytical C4, HPLC column (4.6 ϫ 250 mm, TP214, Vydac, Inc.) using a two-step linear gradient (20% acetonitrile, 0.1% trifluoroacetic acid at 0 -5 min, then 20 -40% acetonitrile, 0.1% trifluoroacetic acid from 5 to 90 min) at a flow rate of 0.75 ml/ml. A portion of the fraction (10% of total fraction volume) containing the peak of biological activity was then fractionated by electrophoresis over an 8% acrylamide gel. Based on pilot experiments, seven slices of acrylamide (1 mm, height) were dissected from the lane loaded with the active fraction spanning the 36-and 64-kDa prestained molecular weight markers (Sea Blue, Invitrogen, Inc.). The dissected gel slice fractions were place in 1.5-ml microcentrifuge tubes containing 0.5 ml of DHF12-FBS, minced with a scalpel, and agitated for 16 h at 1000 rpm at 22-24°C. The fractions were centrifuged at 10,000 ϫ g for 15 min in a table top microcentrifuge. The supernatant fluid was maintained and tested for biological activity.

Production of Sertoli Cell-modified Laminin (S-Laminin)-
The rat Sertoli cell line, ASC-17D, was obtained from Barry R. Zirkin (Johns Hopkins Medical School) and was maintained at 32°C in 5.5% CO 2 in DHF12 medium supplemented with 8.5% FBS and 1% antibiotic/antimycotic solution (ASC medium). Culture plates were pre-coated with 2.2 g/cm 2 mouse laminin for 4 h at 22-24°C. The ASC-17D cells were suspended in ASC medium then plated into the laminin-coated wells (1.5 ϫ 10 5 cells/cm 2 ) and maintained at 32°C in 5.5% CO 2 for 2 days. To remove the ASC-17D cells from the S-laminin, cultures were treated for 4 h with fresh ASC medium containing 10 g/ml of mitomycin-C. The mitomycin-C-containing medium was removed and the cells were rinsed 2 times with 0.5 ml/well of ASC medium prior to the subsequent addition of fresh ASC medium. The cultures were then maintained for another 6 days. Medium was replaced with fresh ASC medium on the third day. Six days after mitomycin-C treatment the medium was again removed and replaced with 0.5 ml of sterile water/well. The water was removed from the wells after 10 min. The plates of S-laminin were allowed to air dry in the cell culture hood for 3-4 h and then were stored at 4°C until use.
TEX14 Antibody Production and Purification-Synthetic peptides, TEX14-1 (CNMTLLDPTKGSTREKKTKDQ) and TEX14-2 (CTKGSTREKKTKDQDMVEQKR) were produced based on an amino acid sequence that was deduced from PCRamplified cDNA. The cDNA was first synthesized from total RNA from rat LB germ cells. The partial sequence encoded the carboxyl terminus of the rat TEX14 protein (NCBI accession U46694). Amino-terminal cysteine residues were added for conjugation to maleimide-activated keyhole limpet hemocyanin (Pierce, Inc.). Purified conjugates were used to raise poly- clonal antisera in rabbits. Preimmune and immune sera were processed over protein-A affinity columns (Hi-Trap, Amersham Biosciences, Inc.). Purified IgG fractions from immune sera were further purified by affinity chromatography using the antigenic TEX14-1 and TEX14-2 peptides (Sulfolink Kit, Pierce, Inc.). All preimmune (PI-1 and PI-2) and immune (anti-TEX14-1 and anti-TEX14-2) IgG preparations were adjusted to ϳ2 M by protein assay (microBCA, Pierce Inc.) and Coomassie staining (GelCode Blue, Pierce Inc.) on reducing SDS-polyacrylamide gels. Commercial sources of purified, rabbit IgG (Zymed Laboratories Inc., Inc. and Research Diagnostics, Inc.) served as standard to determine IgG concentrations.

RESULTS
Rat germ cells that bind to laminin are highly enriched in type A spermatogonia (Ͼ90%) and spermatogonial stem cells (10 -12). Freshly isolated spermatogonia that bind to laminin consist of single (87.9 Ϯ 2% S.D., n ϭ 4 cultures) or paired (12.1 Ϯ 1.8% S.D., n ϭ 4 cultures) DAZL ϩ , Plzf ϩ cells. The laminin selection results in greater than 24-fold more DAZL ϩ spermatogenic colonies formed/10,000 cells plated than unfractionated testis cells after 9 days in culture on SNL fibroblasts (supplemental Fig. S2A). Through the use of antibodies that specifically label the cytoplasmic bridges of developing germ cells   (1 nM) on the development of A al spermatogonia in the presence and absence of serum-free SNL-CM (60 l/ml, 15x-3,000MW retentate). Data are the average from duplicate wells of a representative experiment. C, effects of GDNF-neutralizing IgG (GNAb: ϩ) and isotype control IgG (GNAb: Ϫ) on the ability of SNL-CM and/or rat GDNF to stimulate formation of A al spermatogonia. Medium with or without GDNF (100 pM) and/or serum-free SNL-CM (60 l/ml, 15x-3,000MW retentate) was supplemented with the respective IgG components at 3 g/ml. The mean Ϯ S.E. from triplicate wells are presented. D, migration of GDNF-like factors isolated from the reverse phase-HPLC step shown in A on a non-reducing, polyacrylamide SDS gel. Gel fractions (1 mm height) containing molecules that migrated with a M r between 5,000 and 150,000 were dissected from the lane loaded with the active faction, extracted, and then tested for their ability to stimulate formation of A al spermatogonia (gray bars). The peak of activity was extracted from gel fraction 7, which migrated with a M r of ϳ31,000 based on molecular weight standards. a Aal 8 -32, number of A-aligned spermatogonia at the 8-, 16-, and 32-cell stages, which can be stimulated to develop from day 0 GFP ϩ spermatogonia/fraction after 9 days in culture on S-laminin in the presence of 100 pM recombinant rat GDNF. b Serum-containing SNL-CM (SC-CM). c 3000 molecular weight cut-off filter was used to concentrate 10 liters of SF-CM to 125 ml (3000 MW Retentate). d Serum-free SNL CM (SF-CM).
Based on these properties, factors that promote the formation of chains of A al spermatogonia were identified.
Because purified type A spermatogonia do not survive efficiently under routine culture conditions in the absence of somatic cells (33,53,54), we hypothesized that SNL cells produce factors that maintain the initial steps of spermatogenesis in vitro (10,31). To test this, freshly isolated spermatogonia selected on laminin were cultured for 9 days on a matrix of S-laminin (see "Experimental Procedures"), with or without medium conditioned by SNL cells (SNL-CM). In cultures supplemented with SNL-CM, greater than 50-fold more of the 8-, 16-, and 32-cell stage chains were present by day 9 (Fig. 1). By comparison, medium conditioned by a MSC-1 mouse Sertoli cell line was minimally effective at promoting chain formation (Fig. 1). Active components in serum-free SNL-CM that stimulated formation of A al spermatogonia could be concentrated by ultrafiltration using a filter that retains molecules greater than about 3000 kDa (Table 1). These results suggested that SNL cells secreted factors required for formation of A al spermatogonia.
Because GDNF seems required for the maintenance of spermatogenesis in vivo, and has been shown to stimulate DNA synthesis and self-renewal of spermatogonia in vitro (55)(56)(57), GDNF family ligands were tested as SNL-CM candidate factors. In cultures of germ cells on S-laminin, recombinant forms of GDNF, neurturin, and artemin each stimulated an increase in the number of A al spermatogonia per well at the 4-and 8-cell stages (supplemental Fig. S3 and supplemental Table 1). Few 16-cell chains and no 32-cell chains were observed in these experiments. These peptides also caused a decrease in the number of A s and A pr spermatogonia/well (supplemental Fig. S4 and supplemental Table S1). The maximal effective concentrations for stimulating a decrease in the number of A s and A pr cells per well were achieved at ϳ1 nM GDNF, ϳ1 nM neurturin, and Ͼ1 nM artemin (supplemental Fig. S4). The maximal effective concentrations for stimulating an increase in the total number of A al cells/well at the 4 -16-cell stages were achieved at ϳ1 pM GDNF, ϳ10 pM neurturin, and ϳ1 nM artemin (supplemental Fig. S3). Persephin was also tested in these studies, but had no apparent effects. In summary, as with the SNL-CM, these peptides were able to stimulate the formation of A al spermatogonia on S-laminin, but were not as effective as SNL-CM for stimulating development A al spermatogonia, especially the longer chains containing 16 -32 cells.
The SNL-CM (400-ml) was then fractionated by HPLC and fractions were identified that were more effective than recombinant GDNF at stimulating formation of A al spermatogonia to the 8-, 16-, and 32-cell stages ( Fig. 2A). Additionally, the effects of SNL-CM were synergistic in combination with maximal concentrations of GDNF, neurturin, or artemin for stimulating formation of 16 -32-cell chains (Fig. 2B). However, neutralizing antiserum raised to GDNF dramatically reduced the ability of both GDNF and SNL-CM to stimulate formation of A al cells beyond the 4-cell stage (Fig. 2C). Like GDNF, the purified molecules were inactivated by pretreatment with trypsin (data not shown), and migrated with a M r of ϳ31,000 on non-reducing SDS gels (Fig.  2D). Also like recombinant GDNF, the gel-purified factor was estimated to stimulate spermatogonia to develop to the 4 -16-cell stages at low picomolar concentrations (Fig. 2D). Together, the results in Fig. 2 showed that SNL-CM contained GDNF-like factors plus distinct factors that could increase the effectiveness of GDNF to stimulate formation of A al spermatogonia. The question then arose as to the nature of the factor(s) in SNL-CM responsible for these effects. Fig. 3A illustrates a major peak of activity eluting from HPLC at fraction 43 (black arrow), which is "GDNF-independent activity" given no exogenous GDNF was added to the incubation mixture (ϪGDNF). When a maximal concentration of GDNF was added (ϩGDNF), the fractions bracketed as GDNF-dependent activity dramatically stimulated A al 16 -32cell chain formation (Fig. 3B). A concentration response of the pooled fractions containing this GDNF-dependent activity, with or without GDNF, demonstrates a strong dependence on GDNF (Fig. 3C). We proceeded to purify and identify the factor(s) in the GDNF-dependent activity peak.
The various cell chain lengths that develop in response to this fraction (estimated based on TEX14 ϩ -labeling) are shown in Fig. 4A. To estimate the apparent M r of the factor, part of the pooled GDNF-dependent activity was applied to non-reducing SDS gels. The estimated M r was 45,000 (Fig. 4B). That the factor was likely a protein is shown in Fig. 4C where the GDNF-dependent fraction was treated with trypsin. Following such treatment the activity was lost. Furthermore, following organic solvent extraction (n-butanol) the activity in the isolated pool partitioned with the aqueous phase (data not shown). The activity in the fraction was processed over a series of chromatography steps (Table 1 and Fig. 4D), and identified based on mass spectrophotometry as mouse neuregulin-1 (NCBI accession number AAT68241) (Fig. 5A).
Like the factor purified from the active pool (Figs. 3 and 4), all three forms of recombinant human neuregulin functioned at picomolar concentrations to augment the effectiveness of GDNF in stimulating formation of the longer A al chains of spermatogonia (Fig. 5, B and C). For example, 300 pM N1␤1 176 -246 increased the number of spermatogonial chains at the 16 -32-cell stages by greater than 10-fold when compared with GDNF alone (Fig. 5B, right). At nanomolar concentrations, all three forms of recombinant neuregulin were able to stimulate formation of some A al cells in the absence of GDNF (Fig.  5B, left), but were still 7-20-fold less effective than in the presence of GDNF (Fig. 5B, right). In the presence of GDNF, N1␤1 20 -246 was the most potent (EC 50 ϳ 1.5 pM), and was ϳ65-fold and 4000-fold more potent than N1␤1 176 -246 (EC 50 ϳ 100 pM) or N1␣1 176 -241 (EC 50 ϳ 6 nM), respectively (Fig. 5B). Mouse EGF did not appear to effect the development of spermatogonia in either the absence or the presence of GDNF (Fig. 5B).
Using FACS-purified A s spermatogonia (Ͼ99.8% GFP ϩ spermatogonia), which are depleted of somatic testis cells (26), we next showed that soluble extracellular domains of neuregulin receptors (Fc-ErbB3/4) could neutralize the effects of SNL-CM and/or GDNF on stimulating formation of A al spermatogonia beyond the 4-cell stage (Fig. 6, supplemental Fig.  S5). Addition of recombinant neuregulin-1 to the cultures could block the observed inhibitory effects of the soluble receptor domains, as development of A al spermatogonia to the 8-cell stage was recovered by 185 h (day 8) in culture (Fig. 6, supplemental Fig. S5). These experiments provided evidence that germ cells express receptors for neuregulin.
Expression of neuregulin receptors on germ cells was supported by reverse transcriptase-PCR experiments (Fig. 7, supplemental Table 2). We detected an enrichment of transcripts encoding ErbB2 and ErbB3 in fractions of germ cells purified from the testes of 9-and 25-day-old rats by FACS (Fig. 7, A and IgG (i.e. arrows in Apr) and nuclei with Hoechst 33342 dye. B, the relative mobility of the GDNF-dependent activity is shown after fractionation over a non-reducing, polyacrylamide SDS gel. Gel-slice fractions (1 mm height) containing molecules that migrated with a M r between 5,000 and 150,000 were dissected from the lane loaded with activity, extracted, and then tested for the ability to stimulate formation of A al spermatogonia under standard culture conditions in the absence (blue bars) or presence of 100 pM GDNF (red bars). C, inactivation of GDNF and the GDNF-dependent activity by trypsin. Note, GDNF was added to the GDNF-dependent activity samples following their treatment with trypsin and inhibition of the trypsin with serum. D, dose responsiveness of the GDNF-dependent activity after further purification over a final analytical reverse phase-HPLC column step (Table 1). Shown are responses to 100 pM recombinant rat GDNF plus different volumes of the fraction containing the peak of activity that eluted from the HPLC column. The effects of 100 pM GDNF alone are designated by the dashed line (arrow). B). Also, by a distinct cell isolation method, transcripts encoding ErbB2 and ErbB3 were again identified to be enriched in germ cell fractions purified from the testes of 23-dayold rats (Fig. 7B). In contrast, transcripts encoding ErbB1 were enriched in interstitial testis cells, and transcripts encoding ErbB4 were enriched in fractions of tubular somatic cells (Fig. 7B). Transcripts for ErbB1 and ErbB4 were also detected in some germ cell fractions; the band intensities of these PCR products were weak relative to signals obtained with somatic cell fractions (Fig. 7B).
Consistent with the identification of neuregulin in the SNL-CM, and the apparent relative lack of neuregulin-like activity in MSC-1 conditioned medium, when RNA concentrations were estimated by quantitative PCR, neuregulin-1 transcripts were ϳ100-fold more abundant in SNL cells than in MSC-1 cells (Fig. 8A). When neuregulin was added to cultures of freshly isolated spermatogonia on MSC-1 feeders, it stimulated an increase in the total number of germ cells/culture at day 9 (Fig. 8B). Moreover, in these same cultures, neuregulin stimulated a decrease in the concentration of stem cells within the germ cell population (Fig.  8C). These results are based on the number of GFP ϩ cells isolated from each culture (Fig. 8B) and the ability of the germ cells in each culture to colonize the seminiferous tubules of recipient rat testes (Fig. 8C). Thus, here, the behavior of spermatogonia in culture on MCS-1 cells, in the presence of added neuregulin, reflects the reported behavior of spermatogonia that expand in number as they differentiate in culture on SNL cells without added neuregulin (10).

DISCUSSION
We have identified neuregulin-1 as a factor capable of promoting the formation of aligned spermatogonia in a GDNF-dependent manner. There also appeared to be a factor in fibroblast conditioned  , and aligned (right) spermatogonia/2-mm 2 culture area that developed from FACS-purified, GFP ϩ spermatogonia (Ͼ99.8% GFP ϩ ). Spermatogonia were cultured under standard conditions in medium containing 30% SNL-CM and 100 pM GDNF supplemented with 6 g/ml human Fc-IgG protein (Fc-Control), a mixture of 3 g/ml Fc-ErbB3 plus 3 g/ml Fc-ErbB4 fusion proteins (Fc-ErbB3/4), or the Fc-ErbB3/4 mixture plus 100 nM human neuregulin-1 ␤1 (Fc-ErbB ϩ NRG). The relative numbers of GFP ϩ single, paired, and aligned cell stages were scored for each culture at the indicated time points (h) using an IX-70 fluorescent microscope (Olympus, Inc.) and the imaging program Simple-PCI (C-Imaging Systems, Compix, Cranberry Township, PA). To score cultures, images of 2-mm 2 microscopic fields were captured at three separate pre-determined coordinates in each culture, and then the average numbers of each cellular stage/field were calculated for each culture. Mean Ϯ S.E., n ϭ 3 cultures/time point. medium that promoted some spermatogonial differentiation in the absence of added GDNF; the identity of this factor remains unknown; it could be GDNF itself.
Neuregulin-1 is a member of the EGF superfamily and is a ligand for two transmembrane receptors in the ErbB family, ErbB3 and ErbB4. There are four different erbB (erythroblastoma virus B) genes that encode ErbB1, ErbB2, ErbB3, and ErbB4. Through binding to ErbB3 or ErbB4, recombinant forms of neuregulin-1 have been shown to signal proliferation, survival, differentiation, chemotaxis, and/or apoptosis dependent on the cell type (59 -61).
As for neuregulins acting as important signaling molecules in the testis, transcripts encoding the ␣-1 and ␤-1 isoforms of neuregulin-1 have been detected in peritubular cells from the rat testis, and neuregulin-like activity was identified in medium conditioned by rat peritubular cells (62). In these studies the N1␤1 176 -246 and N1␣1 176 -241 forms of neuregulin-1 stimulated an increase in the abundance of transcripts for transferrin and androgen-binding protein in cultures of rat Sertoli cells; this was associated with an increase in the concentration of both proteins in the culture medium (62). Based on reverse transcriptase-PCR analysis from other laboratories, transcripts encoding ErbB1, ErbB2, ErbB3, and ErbB4 are selectively expressed by different populations of isolated somatic testis cells (Leydig, Sertoli, and peritubular), and in seminiferous tubules dissected from all stages of the rat epithelial cycle (62,63). Thus, neuregulins and their receptors are expressed by somatic testis cells.
However, the expression of neuregulins or neuregulin receptors in germ cells has not been reported until now. The ability of recombinant neuregulin to stimulate development of FACS-purified GFP ϩ spermatogonia, and the ability of soluble neuregulin receptors to block development in response to GDNF demonstrate that receptors FIGURE 7. Transcripts encoding ErbB2 and ErbB3 are highly enriched in purified spermatogenic cells. A, freshly purified germ cell (GC) and somatic cell (SC) fractions were collected from the testes of 9-(left) and 25-(right) day-old GCS-EGFP rats by FACS. The GC fractions collected were ϳ91 and ϳ99% EGFP-positive for the day 9 and day 25 GC fractions, respectively. The SC fractions collected were ϳ97 and ϳ93% EGFP-negative for the day 9 and day 25 GC fractions, respectively. Based on the age of the rats used, the GC fractions contained mixtures of spermatogonia and spermatocytes, with the day 9 germ cell sample containing less numbers of spermatocytes. The SC fractions contained mixtures tubular and interstial somatic cell types. B, detection of ErbB family transcripts by reverse transcriptase-PCR in purified fractions of rat germ and somatic testis cells that were obtained by two different methods: Testis Cell Cultures enriched in laminin-binding germ cells (LB), laminin non-binding germ cells (LNB), tubular somatic cells (SC), and interstitial somatic cells were isolated from 23-day-old rats as previously described (10,11). Freshly Sorted Testis Cells from 9-and 25-day-old rats were isolated as described above in A, which consisted of highly purified germ cell and somatic cell fractions. Transcripts encoding Dazl, Kcne1, and Insl3 (see supplemental Table 2) were used as markers to test each fraction for enrichment with germ cells, tubular somatic cells, or interstitial cells, respectively. Transcripts were also analyzed in an unfractionated testis from a 23-day-old rat (Testis Day 23). for neuregulins are present on the germ cells. In support of this hypothesis, we show for the first time that transcripts encoding the neuregulin receptor ErbB3, and its primary signal transducer, ErbB2, are enriched in fractions of purified spermatogonia. In contrast, transcripts for the other known neuregulin receptor, ErbB4, are enriched in fractions of tubular somatic cells. Thus, it is concluded that neuregulins do have the potential to stimulate spermatogenesis directly in the germline by binding to ErbB3 and activating ErbB2/ErbB3 signaling complexes expressed on spermatogonia. Furthermore, neuregulins could also regulate spermatogenesis by binding to and activating ErbB4 signaling complexes expressed on the tubular somatic cells that associate directly with germ cells in the seminiferous epithelium.
It has already been shown that addition of the ErbB1 ligand, EGF, to medium is not essential for maintenance of germ line stem cells in culture (15,16,24), and we also did not observe an effect of EGF on spermatogonial development (Fig. 5B). However, in organ cultures, EGF has been shown to stimulate DNA synthesis in differentiating type-A spermatogonia and to stimulate an increase in stem cell factor expression, effects that are thought to be mediated by EGF signaling to somatic testis cells (63). This is consistent with expression of ErbB1 transcripts in testicular somatic cells (Fig. 7). Similarly, in Drosophila, the ability of spermatogonial stem cells to differentiate is blocked by mutations that disrupt signaling through the EGF receptor, which is also known to be expressed by supporting somatic cells (64,65). Thus, in both insects and mammals, EGF probably regulates spermatogonial differentiation through its effects on somatic cells.
Because the phenotypes of neuregulin-1, ErbB2, ErbB3, and ErbB4 knock-out mice are embryonic lethal (66 -70), and because culture systems supporting spermatogenesis in vitro have not been fully developed, the role of neuregulins in spermatogenesis remains unclear. Nevertheless, the expression of transcripts encoding ErbB receptors and neuregulins in the testes, and the ability of neuregulins to stimulate formation of A al spermatogonia, particularly in the presence of GDNF, suggest neuregulin signaling is important for spermatogenesis in vivo.