Genetic Factors on Mouse Chromosome 18 Affecting Susceptibility to Testicular Germ Cell Tumors and Permissiveness to Embryonic Stem Cell Derivation

Testicular germ cell tumors (TGCT) are the most common solid tumors in men 15 to 34 years of age, accounting for >1% of all male cancers (1, 2). More than 90% of testicular cancers are TGCTs (3). Family history is a significant risk factor, with a frequency in affected males that is 10 to 13 times higher among brothers, 4 times higher among sons of affected fathers (4–7), and a 75-fold increased risk for monozygotic twins (8). Unfortunately, linkage and association studies to identify genes that control susceptibility have been difficult because of a shortage of multigenerational pedigrees with a sufficient number of affected individuals, the sterility that often results from treatment, and the genetic complexity of the disease. Using a candidate gene approach, Nathanson and colleagues identified a 1.6-Mb deletion (gr/gr) on Yq11 that confers susceptibility to TGCTs (9). Recently, a genome-wide association study implicated Kitl, a component of the c-kit signal transduction pathway (10, 11). Mutations in the Kitl gene that enhance TGCT susceptibility have also been described in mice (12, 13). Despite these advances, the genetic factors that contribute to the disease remain poorly understood. The difficulty of finding TGCT genes in humans makes a mouse model relevant to characterizing the genetic control of susceptibility.

Strain 129 males develop spontaneous TGCTs at a rate of 1% to 8%, depending on the substrain (14, 15). These TGCTs are an established model of spontaneous testicular teratomas and teratocarcinomas that occur in human infants (14, 15). In both infants and mice, TGCTs arise from primordial germ cells (PGCs; ref. 12) and are thought to lack both carcinoma in situ and characteristic karyotypic abnormalities such as isochromosome 12p that are found in adult TGCTs in humans (16). As in humans, genetic control of TGCT tumorigenesis in mice is complex, and linkage studies have met with only modest success. The first linkage study failed to uncover any statistically significant linkages, although the best involved distal chromosome 19 (17). A separate cross implicated a ∼23 cM quantitative trait locus (QTL) on chromosome 13 (18). This QTL contains >400 genes, which represents almost half of all the genes on chromosome 13.⁴

Chromosome substitution strains (CSS) are a powerful alternative to segregating crosses for discovering genes involved in TGCT susceptibility. CSSs, which are also known as consomic strains, are inbred strains in which an entire chromosome is replaced by its homologue from a different inbred strain. CSSs have several logistical and statistical advantages (19, 20). Fine mapping is accomplished either with crosses or preferably with panels of congenic strains in which the substituted chromosome is partitioned into segments of various lengths on the inbred host background (21–24).

Based on the weak linkage to chromosome 19 (17), Matin and colleagues made the first autosomal CSS in mice to test whether chromosome 19 had QTLs involved in TGCTs (25). MOLF is an inbred strain that is derived from Mus musculus molossinus and is genetically distant from 129/Sv (26–28). Among 129-Chr 19^MOLF males, 82% developed at least one TGCT, showing that MOLF-19 had at least one TGCT-promoting QTL (25). A panel of 13 congenic strains was then made to map the QTLs (21). Analysis of TGCT prevalence in the congenic strains showed that at least five TGCT QTLs were located on this substituted chromosome (21). Moreover, closely linked QTLs are readily resolved in a panel of congenic strains, which is difficult to accomplish in segregating crosses because too few closely linked crossovers occur to provide strong evidence for independence. In addition, the boundaries of these candidate intervals are genetically, rather than statistically, defined (cf. refs. 21–24).

To identify other 129-derived TGCT factors that must exist to account for the complex patterns of inheritance, we began by testing for the action of QTLs with dominant effects on susceptibility in a panel of incipient CSSs and found statistically significant evidence for QTLs suppressing TGCTs in two CSSs, 129-Chr 2^MOLF and 129-Chr 18^MOLF. An incipient CSS is one that has not yet been backcrossed 10 generations (19, 29). Then, with a panel of congenic strains that together span the length of chromosome 18, we identified four TGCT QTLs on this substituted chromosome. Finally, we showed that early embryonic cells from 129-Chr 18^MOLF are less likely to yield pluripotent cell lines in vitro than those from 129/Sv.

MOLF/Ei (MOLF) and 129S1/SvImJ (129/Sv) mice were purchased from The Jackson Laboratory. MOLF mice were maintained on LabDiets 7960 rodent chow (PMI Nutrition International). 129/Sv and backcrossed mice were maintained on 5010 rodent chow (PMI Nutrition International). Mice were housed in the Case Western Reserve University Animal Resource Center and maintained on a 12:12-h light/dark cycle. Mice were provided water and diet ad libitum. The mouse work was approved by the Case Western Reserve University Institutional Animal Care and Use Committee.

To make the 129-Chr^MOLF panel, MOLF mice were mated with 129/Sv and the progenies were backcrossed to 129/Sv. N2 mice were genotyped and mice with nonrecombinant chromosomes of interest were backcrossed to the host strains. This step was repeated until at least the N10 generation. Nonrecombinant N10+ mice were brother-sister mated to make the substituted chromosome homozygous.

Mice were genotyped with either SNaPshot single base extension (Applied Biosystems) or microsatellite markers. For single base extension, SNaPshot was performed as described by the manufacturer's protocol with the following changes: 1 unit of shrimp alkaline phosphatase (USB Corporation) and 0.2 units of Exonuclease I (USB Corporation) each were added to 15 μL of PCR product to remove unused deoxynucleotide triphosphates and PCR primers. The custom single base extension oligonucleotide, which anneals immediately upstream of the SNP, was used at a working concentration of 0.2 to 1.0 μmol/L, depending on the efficiency of each primer. Reactions were multiplexed at least four and at most six times. After primer extension, the reactions were subjected to capillary electrophoresis on ABI DNA Analyzers (Applied Biosystems) in the Case Western Reserve University Genomics Core. For SNP genotyping, at least four and at most six SNPs were typed per chromosome. Two SNPs were as close to each of the ends of each chromosome as possible, and two to four SNPs were evenly spaced in between. There were no gaps >23 cM. The identities of the SNPs and the primers used for genotyping are listed in Supplementary Table S1.

For microsatellite marker genotyping, polymorphic MIT markers were identified (Center for Inherited Disease Research). DNA for PCR genotyping was isolated from mouse tail as previously described (18) and resuspended in 10 mmol/L of Tris. PCR amplification was performed in a 96-well PTC-225 tetrad thermal cycler (MJ Research). PCR conditions were as follows: 94°C for 2 min followed by 39 cycles of 94°C for 60 s, 55°C for 35 s, and 72°C for 45 s. A final extension of 72°C for 5 min was performed. PCR products were resolved in 4% agarose gels (Invitrogen 15510-027) in 1× TAE and visualized with ethidium bromide. Genotypes were inferred by length polymorphism. Four MIT markers were typed per chromosome. Two markers were as close to the ends of each chromosome as possible, and two were evenly spaced in between. There were no gaps >34 cM. The identities of the microsatellite markers are listed in Supplementary Table S2.

Three- to 5-wk-old males (a minority were >5 wk) were euthanized with CO₂ inhalation or cervical dislocation. Incisions were made in the abdomen to expose the testes. TGCTs are identifiable with macroscopic inspection and scored for laterality (14). TGCT prevalence was analyzed using the χ² goodness-of-fit test or Fisher's exact test and compared with the host TGCT prevalence. The significance threshold for all calculations was set at P < 0.05 after Bonferroni correction for multiple testing.

We used the following method to identify donor chromosomes that harbor QTLs that confer dominant effects on TGCT prevalence:

where “total % affected” is the fraction of all males that had at least one TGCT, m is the fraction of backcrossed mice that inherited the MOLF allele (their genotype is designated M1), “% affected with QTL” is the percentage of males that are heterozygous for the MOLF allele and affected with at least one TGCT, 1 − m is the fraction of males that inherited two copies of the 129-derived allele (their genotype is designated 11), and “% affected without a QTL” is the percentage of all males that are homozygous for the 129 allele and affected with at least one TGCT.

The m and (1 − m) terms, which represent the actual segregation ratios for the alternative alleles (cf. Table 1), were included to avoid an assumption of normal (1:1) Mendelian segregation, i.e., m = 0.5. If heterozygosity for MOLF-derived alleles does not affect susceptibility, then the overall fraction of affected males is equivalent to the baseline rate in 129/Sv males. This fraction will be higher (or lower) than the baseline rate if heterozygous males have enhancer (or suppressor) effects.

The total percentage of affected males (term 1) was estimated by surveying backcrossed males for TGCTs. The percentage of affected males without the putative MOLF-derived QTL (term 3) was 5% (Table 1), which is the baseline TGCT prevalence in 129/Sv males in this survey and is consistent with published estimates (14, 15, 21, 30–32). Because the value for two of the three terms was known, we were able to solve for the unknown term, i.e., the percentage of affected males with a putative QTL for each CSS. Because of the modes of inheritance of the Y chromosome and the mitochondria, this method was not needed for CSSs 129-Chr Y^MOLF and 129-Chr Mito^MOLF.

N12 mice that were heterosomic for MOLF-derived chromosome 18 were backcrossed to 129/Sv and N13 progeny were typed for six microsatellite markers. Recombinant N13 mice were backcrossed to 129/Sv and N14 heterozygous progeny were intercrossed to produce homozygous congenic mice. Each congenic strain was then maintained with brother-sister mating. The identities of the microsatellite markers used for genotyping are listed in Supplementary Table S3.

We sequenced portions of chromosome 18 to discover additional SNPs that could be used to map the recombination breakpoints in selected congenic strains. The identities of the PCR primers used for breakpoint mapping are listed in Supplementary Table S4.

Testes were fixed in 10% phosphate-buffered formalin (Fisher Scientific SF100-4) at 4°C for at least 48 h, rinsed once in 1× PBS at room temperature, and equilibrated in 30% sucrose in 1× PBS at 4°C for at least 2 d before embedding and freezing in optimal cutting temperature compound (Sakura Finetek USA 4583). Embedded gonads were sectioned at 5 to 10 μm with a Leica CM3050 cryostat and slides were dried in the dark at room temperature for 1 h before staining with H&E.

Mouse embryonic stem (ES) cell lines were derived as described previously (33). In brief, blastocysts were flushed from the uterus at embryonic day 3.5 (E3.5) into HEPES-buffered mKSOM (34). Individual blastocysts were explanted onto γ-irradiated mouse embryo fibroblasts in mouse ES cell medium that consisted of knockout DMEM (Invitrogen) supplemented with 15% knockout serum replacement, 5% ES cell–qualified fetal bovine serum (Invitrogen), 10³ units/mL of recombinant mouse leukemia inhibitory factor (ESGRO, Millipore), 2 mmol/L of l-glutamine, 1× nonessential amino acids (both from Invitrogen), and 0.1 mmol/L of 2-mercaptoethanol (Sigma). After 5 to 6 d, the inner cell masses had reached a suitable size and were picked from the plate with a pulled glass pipette and incubated for 5 min at room temperature in 0.25% trypsin-EDTA. Trypsin was neutralized with serum-containing medium and the inner cell masses were partially dissociated individually with the pulled glass pipette. The partial dissociates were plated individually onto a feeder layer of irradiated mouse embryo fibroblasts in a single well of a 24-well plate (1.9 cm² surface area) and cultured with mouse ES cell medium. After 5 d, individual wells were trypsinized and passaged into a single well of a six-well plate (9.6 cm² surface area). ES cell colonies became readily evident between days 2 and 7. Colonies were individually picked and dissociated for further expansion. Dissociated colonies that gave rise to ES cell lines were expanded and cryopreserved. To minimize variability, comparison of ES cell line derivation efficiencies between strain 129/Sv and 129-Chr 18^MOLF was performed during an overlapping time period and all reagents used were identical. ES cell derivation efficiency was analyzed using the χ² goodness-of-fit test and compared with the standard value for strain 129 (30%; see refs. 33, 35–37). The significance threshold for all calculations was set at P < 0.05 after Yates' correction for continuity.

Mouse embryo fibroblasts were maintained with DMEM supplemented with 10% fetal bovine serum, 2 mmol/L of l-glutamine, and 0.1 mmol/L of 2-mercaptoethanol. Mouse embryo fibroblasts were irradiated with 30 to 60 Gy and seeded as feeders at a density of 5.2 × 10⁴ cells/cm² for mouse ES cell derivation and maintenance cultures. All cells were grown on Nunclon Δ–treated dishes or multiwell plates (Fisher Scientific) and coated for 2 h at 37°C with 0.1% (w/v) gelatin (Sigma). Antibiotics at concentrations of 50 units/mL and 50 μg/mL for penicillin and streptomycin, respectively, were used only for primary explants of blastocysts.

We selected nine MOLF chromosomes to substitute onto the 129 inbred background, i.e., chromosomes 2, 3, 7, 11, 16, 18, X, Y, and the mitochondria. These chromosomes were selected based on prior association with TGCTs (7, 18, X, Y; see refs. 9, 17, 38) or for diversity in chromosome length and architecture (2, 3, 11, 16, and mitochondria). At each backcross generation, males were surveyed for TGCTs (Table 1). For the autosomal CSSs and 129-Chr X^MOLF, we tested whether the frequency of affected males with the putative, dominantly acting QTL differed significantly from the control 129/Sv rate. Two CSSs entail donor chromosomes that segregate but do not undergo genetic recombination and are not subject to dominance effects, 129-Chr Y^MOLF and 129-Chr Mito^MOLF. For those two CSSs, we simply tested whether the total percentage of affected males differed significantly from the percentage in 129/Sv controls.

We examined 140 control 129/Sv males and a total of 1,916 post-N2 129-Chr^MOLF backcross males for TGCTs among the nine incipient CSSs (Table 1). Only two incipient CSSs, 129-Chr 2^MOLF and 129-Chr 18^MOLF, provided evidence for QTLs with dominant effects on the baseline rate of susceptibility in 129/Sv males; both showed a significantly reduced percentage of affected males (Table 1). These results are consistent with substitution of at least one 129-derived susceptibility gene or at least one MOLF-derived TGCT suppressor on each chromosome.

We prioritized the development of these CSSs and surveyed 365 homosomic 129-Chr 18^MOLF males for TGCTs. Remarkably, none of the 365 adult 129-Chr 18^MOLF males examined were affected with a TGCT (P < 10⁻⁶; Fig. 1). These results validate the evidence from the incipient CSSs for at least one TGCT QTL on chromosome 18. 129-Chr 2^MOLF was only recently made homozygous, the prevalence of affected males is being assessed, and results will be reported elsewhere.

To determine the number and location of TGCT QTLs on chromosome 18, we constructed a panel of eight congenic strains from 129-Chr 18^MOLF (Fig. 1). In total, we surveyed 1,979 congenic males for TGCTs. All but four of the TGCTs were histologically verified. All TGCT cases were unilateral.

We analyzed TGCT prevalence in the congenic strains using a method involving sequential comparisons between pairs of congenic strains (22–24).⁵

This systematic process tested whether the unique MOLF-derived segment in each pairwise congenic comparison harbors a QTL that affects that rate of affected males in the comparison strain. The TGCT prevalences in C1 and C6 were both significantly reduced relative to the 129/Sv control (Fig. 1), suggesting the presence of TGCT suppressor QTLs in both of these MOLF-derived congenic intervals. The reduced frequency of affected C1 males (0.87%) defines a small (2 cM) QTL near the centromere, an interval labeled region 1 (Fig. 1). Similarly, C6 males were affected at a significantly reduced rate (1.20%), and although C6 lacks region 1 near the centromere, it has a ∼6 cM MOLF-derived chromosome segment at the telomere (region 2; Fig. 1). The simplest explanation for these results is that two independently acting TGCT QTLs, which respectively confer a 4-fold to 6-fold reduction in susceptibility, are located at opposite ends of chromosome 18.

The TGCT prevalence in C7 was enhanced compared with C6 (3.85% versus 1.20%, P = 0.0008; Fig. 1), which is evidence for a TGCT enhancer in the portion of C7 that does not overlap the congenic segment in the C6 strain. This ∼19 cM interval defines the location of a third TGCT QTL (region 3; Fig. 1).

Finally, with a single exception in the C4 congenic strain, four of the eight congenic strains (C3, C4, C5, and C9) lacked TGCTs as adults, recapitulating the absence of TGCTs in the 129-Chr 18^MOLF parent strain. The four strains share a ∼4 cM region of chromosome 18 between markers M21 (21 cM) and M25 (25 cM), defining the location of a fourth TGCT QTL (region 4; Fig. 1).

Youngren and colleagues provided evidence that they interpreted as epigenetic control of the 129-Chr 19^MOLF CSS (21). In a panel of congenic strains made from 129-Chr 19^MOLF, the length of the congenic segments, rather than their locations on the chromosome, predicted TGCT prevalence (21). The only exceptions were two congenic strains containing the MOLF-derived telomeric segment, suggesting that the proposed epigenetic effect involved the centromeric and central portions but not the distal portion of chromosome 19 (21). Using the panel of eight congenic strains made from 129-Chr 18^MOLF, we tested whether length rather than location predicted TGCT prevalence in a manner similar to the 129-Chr 19^MOLF congenic strains. The correlation (r = −0.53) between the length of the congenic segments and TGCT prevalence was not statistically significant (Supplementary Fig. S1), arguing against a chromosome-wide epigenetic defect and supporting the existence of QTLs located at discrete sites on the chromosome.

129 is the most permissive mouse strain for ES cell derivation (39), which interestingly is the only strain to show an appreciable frequency of spontaneous TGCTs (14, 15, 39). The genetic elements of strain 129 that result in high-frequency reprogramming of PGCs to a tumorigenic, pluripotent state could also facilitate the unique ability to efficiently derive pluripotent cell lines from strain 129 early embryos. We therefore compared the frequency of ES cell derivation in 129-Chr 18^MOLF and control 129/Sv mice. We obtained four 129/Sv ES cell lines from 18 blastocysts (22.2%), which is not significantly different from the accepted and reproducible average for this strain (30%; P = 0.391; Table 2). By contrast, the ES cell derivation efficiency of the 129-Chr 18^MOLF strain was significantly reduced (4.8%; P = 0.008) as we only obtained one ES cell line from a total of 21 blastocysts (Table 2). These results suggest that the genetic elements contributing to the formation of TGCTs in vivo may also influence the ability of early embryo cells to form pluripotent ES cell lines in vitro.

TGCTs are the most common type of testicular cancer as well as the most common tumor in men ages 15 to 34 years (1–3). Despite their prevalence, the genetic control of susceptibility is poorly understood. Family history suggests a strong genetic component, but studies have found a small number of genetic variants that each account for a relatively small fraction of the overall inherited risk (9–11). The 129/Sv inbred strain is an established model in which to investigate the genetic components of TGCT susceptibility (14, 15, 30–32). A previous study showed that QTLs affecting TGCT prevalence could be mapped using chromosome substitution strains (25). We therefore constructed and screened a panel of incipient 129-Chr^MOLF CSSs to test whether other MOLF chromosomes had QTLs affecting TGCT prevalence. We identified two CSSs (129-Chr 2^MOLF and 129-Chr 18^MOLF) with dominant QTLs. To map these QTLs on chromosome 18, we constructed and surveyed a panel of eight congenic strains derived from 129-Chr 18^MOLF. The results of our study show that four regions of chromosome 18 contain QTLs affecting TGCT prevalence.

The key to predicting how the four regions on chromosome 18 affect TGCT prevalence is knowing which genes are in each QTL. Region 1 is a ∼2 cM segment containing ∼27 genes. Region 1 belongs to a conserved synteny involving 10p in humans, a region that has not been previously associated with TGCTs (40). One of the genes in region 1, Map3k8, is a mitogen-activated protein kinase and a known proto-oncogene; overexpression of Map3k8 in various cell lines promotes proliferation and transformation (41, 42). Interestingly, Spry4 is a negative regulator of mitogen-activated protein kinase activity (43), and SNPs in Spry4 were recently associated with human TGCT in a genome-wide association study (10). Spry4 is located on mouse chromosome 18, but outside the candidate regions, and coding nonsynonymous nucleotide differences are not known between the 129/Sv and MOLF alleles of Map3k8 or Spry4.⁶

Perhaps the most interesting candidate gene in region 2 is Fbxo15, one of the few known targets of stem cell pluripotency factor Oct3/4 (44). Oct3/4 is aberrantly expressed in human TGCTs (45), which may link Fbxo15 to TGCT susceptibility. The 129/Sv and MOLF alleles of Fbxo15 differ in a single G→A substitution in exon 5, which leads to a valine to isoleucine (V→I) shift at amino acid position 188.⁶ Interestingly, region 2 belongs to a conserved synteny with 18q22 in humans, which has been weakly associated with human TGCT susceptibility in two linkage studies (maximum heterogeneity LOD score = 1.81 and 1.44, respectively; ref. 40).

Region 3 is a large ∼19 cM segment containing ∼148 genes. Unlike regions 1 or 2, region 3 contains at least one factor that increases TGCT prevalence. Region 3 belongs to a conserved synteny with 5q, 18p, and 18q, with 18q having been weakly associated with human TGCTs (40), and 5q being significantly associated with human TGCT susceptibility (11). Because the locations of the recombination breakpoints of C5 and C6 are only roughly mapped, the boundaries of region 3 are uncertain. Additional mapping with subcongenic strains should reduce the number of candidate genes and facilitate identification of the TGCT enhancer at this QTL.

Region 4 is a ∼4 cM segment containing ∼48 genes. Region 4 belongs to a conserved synteny with 5q in humans, which has been associated with significant TGCT risk (11). Region 4 contains Eif1a, a translation factor that is functionally related to Eif2s2, whose misexpression is linked to TGCT suppression in 129/Sv males heterozygous for the agouti yellow mutation (A^y) and those that have reduced expression of Eif2s2 (46). There are no known nonsynonymous differences in coding sequences between the 129/Sv and MOLF alleles of Eif1a.⁶

Interestingly, none of the four regions contain the dead-end (Dnd1) gene, an allele of which, Ter, is a known modifier of TGCT susceptibility (47, 48). Similarly, none of the regions contain sprouty (Spry4), the mouse orthologue of a human gene associated with TGCTs in genome-wide association studies (10). Therefore, the genes and factors in the four regions on chromosome 18 have not been previously associated with TGCT susceptibility. Nonetheless, both Dnd1 and Spry4 map close to the proximal boundary of region 4. If region 4 contained a long-range transcriptional regulator, then either gene could account for the modifier effect of region 4. Unfortunately, the transcriptional control of both genes is poorly described.

Although ES cells are a widely used research tool, the mechanisms that permit their derivation from early embryos remain unclear. The genetic contribution to the ability to derive ES cell lines is often overlooked and poorly understood as the overwhelming majority of ES cell lines in use today are derived from strain 129. Using standard techniques, the majority of other strains have proven refractory to ES cell derivation. In the limited number of other strains in which ES cells can be derived, the efficiencies are consistently much lower than that of strain 129. For example, despite the attempts of multiple groups, the nonobese diabetic strain has not yielded ES cell lines. Interestingly, NOD129F1 blastocysts yield ES cell lines at a frequency nearly equivalent to that of strain 129 itself (49). Furthermore, genetic manipulation of the nonobese diabetic strain by constitutive expression of Myc or Klf4 results in the ability to derive ES cells from this classically nonpermissive strain (50). These results show that heritable genetic or epigenetic factors influence the ability to derive ES cell lines from early embryos.

Strikingly, strain 129 is also the only mouse strain to show an appreciable frequency of TGCTs. We therefore investigated ES cell derivation efficiency in 129-Chr 18^MOLF, which is resistant to TGCTs despite having all but one chromosome derived from 129. We found that the ES cell derivation efficiency in 129-Chr 18^MOLF was significantly reduced as we could only derive one ES cell line from 21 blastocysts, suggesting that the genetic elements contributing to the formation of TGCTs from primordial germ cells might also contribute to the derivation of ES cell lines from early embryos. Further analysis of the ES cell derivation efficiency and TGCT prevalence in 129-Chr 18^MOLF subcongenic strains should allow for the precise mapping, and ultimately, the identification of the factors responsible for the acquisition and maintenance of the pluripotent state whether in vitro for ES cells or in vivo for TGCTs. An understanding of these factors may have a direct effect on the clinical treatment of TGCTs.

No potential conflicts of interest were disclosed.

Grant support: National Cancer Institute grant CA75056, National Center for Research Resources grant RR12305, and Case Comprehensive Cancer Center and the Center for Stem Cell and Regenerative Medicine.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.