Mouse primordial germ cells lacking β1 integrins enter the germline but fail to migrate normally to the gonads

Primordial germ cells (PGCs) are the founder cells of the gametes. In mammals, they are set aside from other cell lineages of the embryo during gastrulation, appearing at embryonic day 7.5 (E7.5) in the mouse, as a small population of cells near the base of the allantois (Ginsburg et al., 1990). They become incorporated into the developing hind-gut endoderm (E8), from which they migrate to the genital ridges as the mesentery of the hindgut forms (E9.5-10.5). By E12.5, PGCs are no longer motile and aggregate with each other and somatic cells of the urogenital ridges. These cell aggregates are the primitive sex cords, forerunners of ovarian follicles in the female, and seminiferous tubules in the male. In Drosophila, genetic analysis has revealed that PGCs migration occurs in four discreet phases: migration through gut endoderm, migration out of gut endoderm to lateral mesoderm, association with gonadal mesoderm, and gonad coalescence (Moore et al., 1998). The genes required for these processes in mammals are largely unknown.

It has been shown, using quantitative short-term adhesion assays, that mouse PGCs adhere to the extracellular matrix (ECM) glycoproteins fibronectin, laminin, and collagen IV, and that their adhesiveness to fibronectin and laminin alters during and after migration (Garcia-Castro et al., 1997). The molecular basis of these adhesive events is unknown. PGCs are a small population of cells (approximately 1,000 at E10.5), buried inside the embryo, and systematic biochemical analysis of surface receptors has not previously been possible. In situ hybridization and immunocytochemical analysis has shown that at post- migratory stages, the integrin subunit α6 is expressed in male and female gonads (Frojdman and Pelliniemi, 1995, 1994; Thorsteinsdottir et al., 1995), and a monoclonal antibody against this subunit interferes with reassembly of male E14.5 sex cords in vitro (Pesce et al., 1994). To date, no genetic evidence has demonstrated a role for cell-ECM or cell-cell molecules in the early development of the germline in mammals.

The integrins are a family of heterodimeric receptors that mediate diverse functions, including cell-ECM attachment, outside-in signaling, and cell migration (Hynes, 1992). A variety of cell migrations in embryonic development are thought to be integrin-dependent, although confirmation by genetic analysis is still required for many of them (Brakebusch et al., 1997). In addition, integrins are known to be required for the proper formation of tissues, including skin, myocardium and kidney (Baudoin et al., 1998; DiPersio et al., 1997; Dowling et al., 1996; Fässler et al., 1996; Georges-Labouesse et al., 1996; Kreidberg et al., 1996; Muller et al., 1997; van der Neut et al., 1996). We sought to determine if integrins play a role in development of the early mammalian germline.

Using cell sorting, RT-PCR and immunocytochemistry, we detected expression of integrin subunits α3, α5, α6, αV, β1 and β3 by PGCs. To facilitate this study, we generated a transgenic mouse expressing the Green Fluorescent Protein (GFP) specifically in the germline of post-gastrulation mice. Using a genetic approach, we investigated the functional requirements of the integrin subunits expressed by PGCs. To determine if β1 integrins are required for germline development, we analyzed chimeric β1^−/−→wild-type embryos. In addition, we analyzed the early germline of embryos with targeted deletion of integrin subunit α3, α6 or αV. The results presented here establish a role for β1 integrins in the differentiation of mouse germ cells.

Primary antibodies used in this study are listed in Table 1. Cy-3-, FITC-, and PE-conjugates were purchased from Jackson ImmunoResearch.

Oct4ΔPE:GFP in pBluescript (Stratagene), a derivative of Oct4ΔPE:β-galactosidase, was a kind gift from Hans Schöler (Yeom et al., 1996). In this construct, GFP is derived from pEGFP (Clontech). A 19 kb NotI fragment of Oct4ΔPE:GFP was gel purified, extracted several times in phenol, and dialyzed in 10 mM Tris-HCl 0.25 mM ETDA pH 7.5. Transgenic mice were generated by pronuclear injection of the construct at 2 μg/ml, and were established on a FVB background (Jackson Laboratory). Mice with targeted deletions of the integrin subunit α6, α3 or αV have been described elsewhere (Bader et al., 1998; Georges-Labouesse et al., 1996; Kreidberg et al., 1996). Generation of chimeric embryos containing integrin β1^−/− cells has been described previously (Fässler and Meyer, 1995). To analyze for germline transmission, chimeras were mated with C57B6 animals. The chimeras used for germline transmission analysis had an agouti coat color contribution of at least 15%. A minimum of 4 litters were documented per chimeric animal, although routinely more than 10 litters were analyzed. Wild-type embryos were obtained from CD1 mice (Charles River) with the morning of vaginal plug considered day 0.5. Sexing of embryos was performed by PCR (Hogan et al., 1994).

For paraffin sections, embryos were fixed in either 2% trichloracetic acid, 4% paraformaldehyde, or 4% formaldehyde for 30 minutes, dehydrated in a graded ethanol series, cleared in xylene, and embedded in Paraplast (Sigma). 5-10 μm-thick sections were mounted on slides, dewaxed in xylene, rehydrated in a graded ethanol series, washed briefly in phosphate-buffered saline (PBS), and blocked with 10% goat serum +0.02% sodium azide in PBS (PSA). For frozen sections, tissue was embedded in OCT (Tissue-Tek), snap-frozen in isopentane submerged in liquid N₂, and cut into 10 μm-thick sections. After brief fixation in acetone, slides were washed in PBS and blocked with PSA. Subsequent steps in the staining of paraffin and frozen sections were identical. Primary antibodies were diluted 1:100 in PSA, with the exception of TG-1 (diluted 1:5). After thorough washing in PBS, secondary antibodies were added after dilution in PSA (FITC 1:100, Cy-3 1:500). After several washes in PBS, slides were mounted in 90% glycerol, 10% H₂O with 100 μg/ml DABCO (Sigma). β-galactosidase staining was performed as reported previously (Hogan et al., 1994).

Flow cytometry was performed essentially as described previously (McCarrey et al., 1987), except that a two-color method was employed using anti-Forssman and anti-E cadherin antibodies. Alternatively, PGCs from Oct4ΔPE:GFP transgenic embryos were sorted by GFP expression only. From 8,000 to 50,000 PGCs were collected per experiment, depending on the stage of the embryo. Approximately 50 embryos were used for each sort and PGC-rich tissues were dissected as follows: hind-gut, mesentery and urogenital ridges at E10.5; gonads and mesophroi at E11.5 and E12.5. Flow cytometry was performed with a FacsVantage and analyzed with Cellquest (Becton Dickinson). Primary and secondary antibodies were diluted 1:100. After sorting, an aliquot of cells was stained for alkaline phosphatase activity to confirm the specificity of the sort (Cooke et al., 1993).

Poly(A)⁺ RNA was isolated from FACS-purified cells using the Micro Fast Track mRNA isolation kit (Invitrogen). cDNA was generated using the cDNA Cycle kit (Invitrogen). To rule-out genomic DNA contamination, half of each mRNA sample was processed without the addition of reverse transcriptase. PCR consisted of an initial denaturing step of 5 minutes at 95°C; 5 cycles of 30 seconds at 95°C, 1 minute at 65°C, 30 seconds at 72°C; followed by 35 cycles of 95°C, 1 minute at 60°C, 30 seconds at 72°C; followed by 10 minutes at 72°C (Almeida et al., 1995). [³²P]dATP was added to PCR reaction mixtures at E10.5 to increase sensitivity. Primers were based on published mouse cDNA sequences or published PCR primers (see Table 1).

Embryonic stem (ES) cell maintenance and embryonic germ (EG) cell derivation have been reported previously (Fässler et al., 1995; Resnick et al., 1992). To generate EG cells, the hind gut, mesentery, mesonephroi and gonads were dissected from E11.5 β1^−/−→wild-type chimeras. This tissue was triturated and plated on an irradiated feeder layer in ES cell medium with 800 μg/ml G418 (GIBCO) to select for integrin β1^−/− cells (Hirsch et al., 1996). EG cell formation was determined by colony morphology and alkaline phosphatase expression.

During migration and early gonad assembly, PGCs are greatly outnumbered by neighboring somatic cells. Therefore, it is not possible to specifically detect PGC gene expression by simple analysis of PGC- rich tissues. In this study, three methods were used to identify the integrin subunits expressed by PGCs. (i) PGCs were isolated by double antibody- labeled flow cytometry and integrin expression analyzed by RT-PCR. (ii) Immunoctyochemistry with integrin-specific antibodies. (iii) A transgenic mouse was generated with germline- specific expression of the Green Fluorescent Protein (GFP). Using this line, the integrin expression profile of living PGCs were analyzed using flow cytometry.

Previously, it has been shown that PGCs can be sorted through the use of monoclonal antibodies or by transgene expression (Abe et al., 1996; McCarrey et al., 1987). In order to discriminate PGC gene expression from that of other cells, PGCs were sorted by two-color flow cytometry using monoclonal antibodies against E cadherin and the Forssman antigen (see Fig. 1) (Fenderson et al., 1990; Wu et al., 1993; M. Gomperts and J. H., unpublished observation). Using this technique, we isolated PGCs to greater than 99% purity, as evidenced by alkaline phosphatase staining (Fig. 1). Using the sorted PGCs, we isolated mRNA and performed RT-PCR with specific primers to assess gene expression (see Fig. 1). Using primers specific for integrin subunits α6, α3, and β1, we were able detect gene expression of these molecules at E10.5, E11.5 and E12.5 (α6, β1) or E10.5 only (α3) (Fig. 1, data not shown). We were unable to detect PGC expression of integrin subunit α7 by this method (data not shown). As positive controls, we amplified PGC cDNA with primers specific for the sequences of the transcription factor Oct-4 and the receptor tyrosine kinase Kit, both previously shown to be expressed in the germline (Williams et al., 1992; Yeom et al., 1996). As a negative control, we utilized primers for the Kit ligand, Steel, which is expressed by somatic cells of the urogenital ridge, but not by PGCs. We detected PGC expression of Oct-4 and Kit, but not Steel.

We performed immunocytochemistry with integrin-specific antibodies at E10.5 and E12.5. At E10.5, PGCs are usually in the dorsal mesentery or the urogenital ridges and have the morphology of motile cells. At E12.5, PGCs are largely non-motile and assemble into primitive sex cords; in the male, they are large and uniform in appearance, in the female, these are small and irregularly shaped (Fig. 2). The β1 integrin subunit was expressed by both PGCs and somatic cells at E10.5 and E12.5 (Fig. 2B,D; data not shown). In the sex cords, staining was concentrated in areas adjacent to the basement membrane (Fig. 2B,D). In contrast, gonadal α6 integrin subunit expression was confined to the primary sex cords (Fig. 2F). Staining for the α3 integrin subunit revealed limited, diffuse expression mainly in the mesenchyme of E12.5 male and female gonads (Fig. 2E, data not shown). Neither α6 nor α3 integrin subunits were detected on migratory E10.5 PGCs by immunocytochemistry (data not shown).

We generated a transgenic mouse expressing green fluorescent protein (GFP) in germline cells. In this animal, GFP is driven by the promoter of Oct-4 with a deletion of the proximal enhancer, which has previously been shown to be germline- specific after gastrulation (Yeom et al., 1996). To confirm that GFP⁺ cells in this transgenic line were PGCs, we stained sections of E10.5-E13.5 Oct4ΔPE:GFP⁺ embryos with anti- SSEA-1, which has previously been shown to label PGCs (Donovan et al., 1987). As shown in Fig. 3, GFP⁺ cells were also SSEA-1⁺. The expression of GFP was found to be highly germline-specific, both by whole-mount confocal analysis and by flow cytometry (Figs 3C, 4A). In the adult, luminal germ cells of the testis and maturing oocytes expressed GFP (Fig. 3E-F).

Using transgenic Oct4ΔPE:GFP⁺ embryos, we analyzed integrin expression of live PGCs using two-color flow cytometry with integrin subunit-specific antibodies. In this experiment, PGCs were identified on the basis of GFP expression (green channel, x-axis), and integrin expression was identified based on PE-labeled secondary antibodies (red channel, y-axis). PGCs were found to express high levels of integrin subunits αV and β3 (Fig. 4B; Table 2). Expression of integrin subunits β1, α6 and low levels of α5 were also detected (Table 2). PGC surface expression of several other integrin subunits was not detected (Table 2).

Integrin β1^+/− mice are viable and fertile, whereas integrin β1^−/− embryos die very early in embryogenesis, before establishment of the germline (Fässler and Meyer, 1995; Stephens et al., 1995). However, integrin β1^−/− ES cells will contribute to specific tissues in chimeric animals (for review see Brakebusch et al., 1997). To determine if integrin β1^−/− cells can form functional gametes, we generated a total of 16 integrin β1^−/−→wild-type and 3 integrin β1^+/−→wild-type chimeric adult males from ES cells derived in two independent experiments (Fässler and Meyer, 1995; Fässler et al., 1995). To assay for germline transmission, each of the integrin β1^−/−→wild-type chimeric males was allowed to father 4-10 litters, totalling over 500 offspring. In addition, we generated 3 female integrin β1^−/−→wild-type chimeras, which each produced 4 litters. No germline transmission was ever found using integrin β1^−/− ES cells, whereas control integrin β1^+/− ES cells were transmitted to the germline in all cases.

To determine if β1 integrins are required for embryonic cells to enter the germline, integrin β1^−/−:β-gal⁺→wild-type and β1^+/−:β-gal⁺→wild-type chimeric embryos were analyzed by combined immunofluorescent and histochemical staining. Embryos with a very low contribution of β1^−/− cells were not analyzed. We were able to detect β1^−/− PGCs in 17 out of 21 chimeric embryos, as late as E13.5. In culture, PGCs will form ES cell-like cells, called EG cells (Resnick et al., 1992). We were able to derive an integrin β1^−/− EG cell colony from a chimeric E11.5 embryo (data not shown). These results demonstrate that cells lacking β1 integrins can enter the germline.

Since PGCs form in the absence of β1 integrins, we next sought to determine if these cells behave normally in vivo. The locations of integrin β1^−/− PGCs in sections of three chimeric E10.5, three chimeric E11.5, and seven E13.5 chimeric embryos were scored (Table 3; Fig. 5). Since highly chimeric (β1^−/−:β-gal⁺→wild-type) embryos develop abnormally (Fässler and Meyer, 1995), the embryos chosen for quantitative analysis were of normal crown-rump length for gestastional age. As an internal control for defective PGC migration due to β1^−/− somatic cells, host wild-type PGCs were analyzed in the same embryos. As an additional control, we scored the location of β1^+/− PGCs in 2 β1^−/−:β-gal⁺→wild-type embryos at E11.5. At E10.5 and E11.5, horizontal sections of embryonic tissue caudal to the heart were analyzed. PGCs were scored as being located in the hind-gut endoderm, the mesentery (including the mesoderm of the hind-gut) or the urogenital ridges (or any structure dorsal to the hind-gut mesentery). At E13.5, sagittal sections of embryos were analyzed throughout the coelomic cavity. PGC location was scored as being normal (within the gonad) or ectopic (outside of the gonad). Since some PGCs begin losing expression of SSEA-1 at E12.5 (Cooke et al., 1993), we were not able to identify all PGCs at E13.5.

At E10.5, about 80% of wild-type PGCs were found at the urogenital ridges (Figs 5A, 6A-B). In contrast, only 20% of integrin β1^−/− PGCs were found at the urogenital ridges, with the majority remaining in the hind-gut mesentery (Fig. 5A). At E11.5, the majority of integrin β1^−/− PGCs continued to be found in the hind-gut mesentery (Fig. 5B). In β1^+/−→wild-type chimeras, integrin β1^+/− PGCs were found in a similar distribution to that of wild-type PGCs (Fig. 5B). At E13.5, over 50% of integrin β1^−/− PGCs were still located outside the gonads, compared to 1% of wild-type PGCs (Fig. 5C). At this stage, ectopic integrin β1^−/− PGCs were always found in the mesonephric kidney or immediately adjacent to the basement membrane of the coelomic epithelium, medial to the gonads (Fig. 6E-F). The distribution of integrin β1^−/− PGCs along the anterior-posterior axis at all stages was normal. Integrin β1^−/− somatic cells contributed to the tissues of the urogenital ridges, including the coelomic epithelium, the mesenchyme, and the mesonephric duct (Fig. 6C).

Integrins α3β1 and α6β1 are expressed in the earliest stages of gonad development (Figs 1, 2; Table 2), and mice with targeted deletion of either subunit usually survive until birth (Georges- Labouesse et al., 1996; Kreidberg et al., 1996). Integrin subunit αV is highly expressed by migratory and postmigratory PGCs (Fig. 4; Table 2). Targeted deletion of integrin αV results in embryonic lethality, although about 20% of null embryos survive until birth (Bader et al., 1998). To determine if these integrin subunits have essential roles in PGC migration, we examined embryos with targeted deletion of integrin subunit α3, α6 or αV. By gross examination of at least 3 embryos null for these integrin subunits at E10.5 and E11.5, we were unable to detect any obvious defects in PGC migration. To confirm this, we performed quantitative analysis of the location of PGCs in mutant embryos at E11.5 and E12.5. As shown in (Fig. 4D), we saw no major defects in PGC migration in integrin α3^−/−, α6^−/−, or αV^−/− embryos. Sex cord assembly in embryos lacking α3 integrins, α6 integrins, or both α3 and α6 integrins was overtly normal (Fig. 7). From this analysis we conclude that α3, α6 and αV integrins are not required for PGC migration, and that α3 and α6 integrins are not required for sex cord assembly.

In this work, we have started to characterize ECM receptors expressed by PGCs during migration and gonad assembly, and to test their roles in these processes. To make this characterization possible, we have utilized new tools to study PGC gene expression. We have developed a method to highly purify wild-type PGCs using two-color antibody-based flow cytometry. Subsequently, RT-PCR was used to analyze gene expression of these purified PGCs. Second, we have generated a transgenic mouse expressing high levels of GFP in the germline, which allows PGCs to be directly live-sorted without the use of fluorescent antibodies or enzyme substrates.

Using a combination of flow cytometry, immunocytochemistry and RT-PCR, we detected expression of integrin subunits α3, α5, α6, αV, β1 and β3 by migratory and/or post-migratory PGCs. Expression of several other subunits, including αIIb, β2 and β4, was not detected. Due to lack of suitable reagents at the time of this study, we did not explore the possibility that integrin subunits α8, α9 and α10 are expressed by PGCs. Integrin α8 is reported to be expressed in the gonadal mesesenchyme only (Muller et al., 1997). Integrin α9 expression has not been reported in the gonad, and initiation of expression is late in embryogenesis (Wang et al., 1995). Previous reports have shown that fertility is retained in the absence of integrin subunits α1, α7, α8, β3, β2, β5 and β6, and are thus not essential for early germline development (reviewed by Hynes, 1996). Caution must be taken not to over- interpret fertility data, since a reduced population of founding germ cells may be sufficient for testicular or ovarian function.

We were unable to detect any germline transmission in the absence of β1 integrins. Since embryos with a contribution of >25% integrin β1^−/− cells do not develop normally (Fässler and Meyer, 1995), it is possible that germline transmission was not detected because of globally insufficient chimerism. Indeed, previous GPI analysis of testicular tissue from chimeric integrin β1^−/−→wild-type adult mice demonstrated a consistently low contribution of integrin β1^−/− cells (Fässler and Meyer, 1995). However this deficiency of integrin β1^−/− cells was tissue-specific, since many tissues in these adult mice (e.g. skeletal muscle, gut) supported chimerism well in excess of 10%. The scope of breeding in this study should have had sufficient power to detect a very low level of germline chimerism (<1%), since we examined over 500 offspring fathered by integrin β1^−/−→wild-type chimeras. Therefore, we feel that globally weak chimerism is not a likely explanation for the absence of germline transmission reported here.

PGCs are derived from cells of the proximal epiblast, an area that also gives rise to extraembryonic mesodermal tissues, including the proximal allantois and the hematopoietic cells of the yolk sac (Lawson and Hage, 1994). Integrin β1^−/− hematopoietic stem cells arise and are capable of differentiation but are unable to migrate into the fetal liver (Hirsch et al., 1996). We show here that this developmental defect is paralleled in the early germline, as integrin β1^−/− PGCs arise but are impaired in their colonization of the gonads. In contrast to the hematopoietic system, this phenotype is not absolute, as some integrin β1^−/− PGCs are able to colonize their target. Similarly, although PGC migration was found to be severely defective in W^e/W^e embryos, a small number of PGCs were able to colonize the urogenital ridges (Buehr et al., 1993). In Drosophila, mutations that affect PGC migration often do not prevent all cells from reaching the gonads (Broihier et al., 1998; Moore et al., 1998). We cannot exclude the possibility that, in chimeric embryos, some integrin β1^−/− PGCs are carried or pulled to the urogenital ridges by wild-type PGCs, since PGCs are known to interact with each other during migration (Gomperts et al., 1994). Tissue-specific gene targeting and in vitro analysis of integrin β1^−/− PGCs will be needed to more thoroughly understand the role of β1 integrins in PGC colonization of the gonads.

From the analysis presented here, we cannot distinguish between a decrease in PGC motility or homing in the absence of β1 integrins, since interference with either process would presumably result in an abundance of ectopic PGCs. Given that the majority of ectopic integrin β1^−/− PGCs were found in the hind-gut mesentery at E10.5 and E11.5, we were surprised to find none in this structure at E13.5. It has been shown that ectopic PGCs undergo apoptosis (Pesce et al., 1993). It is likely that the microenvironment of the mesentery at this late stage no longer supports PGC survival, whereas ectopic areas nearer to the gonad are more permissive.

Integrins α3β1, α6β1 and α6β4 function as laminin receptors (Hall et al., 1990; Hemler et al., 1989; Tomaselli et al., 1990). The disruption of integrin subunit α3 or α6 by targeted deletion revealed no essential function of these molecules during migration or sex cord assembly. Given the very high level of expression of the integrin subunit α6 in the fetal ovary and testes, this molecule may play a role later in development of the germline or it may be required for somatic cell function. In male integrin α6^−/− gonads, we were able to detect putative pre-Sertoli cells (data not shown), although we do not know if these cells would function properly in adult animals, as α6^−/− pups die at birth (Georges-Labouesse et al., 1996). Laminin α1 is required for normal PGC migration in Drosophila (Jaglarz and Howard, 1995). Laminin may play a conserved role in PGC migration in mammals, although the results of this study would argue that this is not accomplished through interaction with α3- or α6-containing integrins.

We show here that PGCs express αV- and β3-containing integrins, which are reported to bind fibronectin and a diverse array of other ligands. It has previously been proposed that mouse PGCs use fibronectin as a substrate for migration (Alvarez-Buylla and Merchant-Larios, 1986). Although we cannot exclude this as a possibility, the data presented here make this scenario less likely. We show here that PGC migration is normal in integrin αV^−/− embryos, despite the high expression of this molecule by wild-type PGCs. Recent reports also argue against an essential role for fibronectin receptors in the germline: αV- and α5-containing integrins are not required for germline transmission in chimeric mice (Bader et al., 1998; Taverna et al., 1998), and mice and humans lacking β3 integrins are fertile (Coller et al., 1994; Hodivala-Dilke et al., 1999). In addition, mice lacking integrin α8β1 (a fibronectin receptor) are fertile (Muller et al., 1997). We are currently investigating the possibility that, in the absence of αV integrins, PGCs upregulate the expression of integrin α5β1 or other fibronectin receptor(s).

In this work we have systematically studied the integrin expression profile of murine PGCs. In doing so, we have generated a transgenic mouse expressing high levels of GFP specifically in the embryonic, fetal and adult germline. In addition, we have shown that β1-containing integrins are not required for embryonic cells to enter the germline, but are required for normal germline transmission and PGC targeting to the gonads. This is the first genetic evidence for an involvement of β1 integrins in the development of the early germline.

We are very grateful to Hans Schöler for the Oct4ΔPE:GFP construct. We would like to thank Cord Brakebusch for help with stem cell culture and many valuable discussions; Peter Beverley, Peter Ekblom, Hubert Eng and Donna Mendrick for gifts of antibodies; Janet Peller and the University of Minnesota Cancer Center Flow Cytometry Core for assistance with cell sorting; and Sandra Horn and Robert Ehlenfeldt for technical assistance. Financial support for this work came from the National Life and Health Insurance Medical Research Fund, the Harrison Fund, the Institute of Human Genetics, the National Institutes of Health (HD33440-01), and the Association pour la Recherche contre le Cancer.