Cyclic AMP and c-KIT Signaling in Familial Testicular Germ Cell Tumor Predisposition

Abstract

Background:

Familial testicular germ cell tumors (FTGCTs) are hypothesized to result from the combined interaction of multiple low-penetrance genes. We reported inactivating germline mutations of the cAMP-binding phosphodiesterase 11A (PDE11A) as modifiers of FTGCT risk. Recent genome-wide association studies have identified single-nucleotide polymorphisms in the KITLG gene, the ligand for the cKIT tyrosine kinase receptor, as strong modifiers of susceptibility to both familial and sporadic testicular germ cell tumors.

Design:

We studied 94 patients with FTGCTs and 50 at-risk male relatives from 63 unrelated kindreds, in whom the PDE11A gene had been sequenced by investigating the association between KITLG genome-wide association study single-nucleotide polymorphisms rs3782179 and rs4474514 and FTGCT risk in these patients and in 692 controls. We also examined cAMP and c-KIT signaling in testicular tissues and cell lines and extended the studies to 2 sporadic cases, one with a PDE11A defect and one without, as a comparison.

Results:

We found a higher frequency of the KITLG risk alleles in FTGCT patients who also had a PDE11A sequence variant, compared with those with a wild-type PDE11A sequence. In NTERA-2 and Tcam-2 cells transfected with the mutated forms of PDE11A (R52T, F258Y, Y727C, R804H, V820M, R867G, and M878V), cAMP levels were significantly higher, and the relative phosphodiesterase activity was lower than in the wild-type cells. KITLG expression was consistently increased in the presence of PDE11A-inactivating defects, both at the RNA and protein levels, in familial testicular germ cell tumors. The 2 sporadic cases that were studied, one with a PDE11A defect and another without, agreed with the data in FTGTCT and in the cell lines.

Conclusions:

Patients with FTGCT and PDE11A defects also carry KITLG risk alleles more frequently. There may be an interaction between cAMP and c-KIT signaling in predisposition to testicular germ cell tumors.

Testicular germ cell tumors (TGCTs) account for 95% of testicular cancer, the most common solid malignancy affecting young men (1). Epidemiological, family, and twin studies provide substantial evidence for a genetic basis for this disease, but attempts to identify a rare, highly-penetrant gene as the basis for familial TGCTs (FTGCTs) have failed (2–4). This has led to the hypothesis that the genetic susceptibility to FTGCTs may result from the combined interaction of multiple low-penetrance genes. Somatic activating mutations in the protooncogene KIT (c-KIT) have been described in sporadic TGCT tumor tissue samples, in approximately 8% of cases (5, 6), suggesting the involvement of abnormalities in the c-KIT signaling pathway in the pathogenesis of TGCTs. Common constitutional variants in the KITLG gene have also been associated with these tumors and implicated as strong modifiers of testicular cancer risk by genome-wide association studies (GWASs) (7, 8). In the latter, both sporadic and familial patient subsets demonstrated similarly strong associations with the same KITLG variants, and we have confirmed this association relative to familial TGCT in our own cohort of multiple-case families (9).

Recently we reported PDE11A mutations and functional polymorphisms as candidate modifying alleles for familial TGCTs (10). Phosphodiesterase (PDE) 11A (PDE11A) is a dual-specificity PDE, binding with cAMP and GMP; PDE11A is expressed in several tissues, but its highest expression has been reported in the testis, prostate, and adrenal glands (11). Therefore, although this interesting new association still needs validation, PDE11A may be an important regulator of the cAMP levels and signaling in these tissues.

In the present study, we investigated whether these independent (located at different genomic loci) KITLG and PDE11A variants coexist in individual patients with familial TGCT and, if so, whether there is functional interaction between the 2 pathways of c-AMP and c-KIT signaling in testicular tissue. We also studied 2 sporadic TGCT patients, one with a PDE11A variant and one without, and investigated the 2 signaling pathways in 2 cell lines. The data point to an interesting and previously unrecognized potential link between these 2 signaling molecules in the predisposition to FTGCT.

Materials and Methods

Genetic investigation for KITLG variants

Genetic analysis was performed in 94 patients with TGCTs from 64 families enrolled in the National Cancer Institute Clinical Genetics Branch Familial Testicular Germ Cell Study (Protocol 02-C-0178) in 50 unaffected male blood relatives, the same patients reported in our earlier publication (10), and in a control group that included 692 unselected healthy individuals enrolled in the New York Cancer Project (12). Two sporadic cases of TGCTs were also included, one with an inactivating PDE11A sequence change (F258Y) and the other with wild-type (WT) PDE11A. Written informed consent was obtained from all participants, and the study (NCT-00039598) was approved by the Institutional Review Board of the National Cancer Institute, National Institutes of Health (Bethesda, Maryland). Patients and control individuals were all tested for PDE11A mutations, as we have previously described (10).

There are 2 TGCT GWAS reports that comprise the basis for implicating KITLG as a testis cancer genetic modifier (7, 8). Because both the University of Pennsylvania study (7) and our own familial cohort consisted entirely of US white men, the GWAS KITLG single-nucleotide polymorphisms (SNPs) for the current analysis were selected from that GWAS. The presence of the KITLG variants that were most strongly associated with TGCT risk in the recently published GWAS (7), rs3782179 and rs4474514, was investigated in patients with FTGCT and their family members using restriction enzymes BsaJI and BglII (New England Biolabs Inc, Ipswich, Massachusetts), respectively; when positive, their presence was confirmed by sequencing. The 692 New York Cancer Project samples were then genotyped using the Mass-Array matrix-assisted laser desorption ionization time-of-flight mass spectrometry-based SNP genotyping system, according to the manufacturer's protocol (Sequenom, San Diego, California), as previously published (13). Dideoxyterminator mixes and primers are available upon request.

Allele frequencies for KITLG variants were compared in cases and in controls, and data were stratified by PDE11A mutation status.

Immunohistochemical analysis

PDE11A, c-KIT, and KITLG [stem cell factor (SCF)/kit-ligand] protein expressions were assessed in the tumor tissues by immunohistochemistry (IHC), using specific rabbit polyclonal antibodies (GTX14624 for PDE11A; GeneTex, Inc, Irvine, California; sc17806 for c-KIT; Santa Cruz Biotechnology, Santa Cruz, California; and sc13126 for KITLG; Santa Cruz Biotechnology), in collaboration with Histoserv, Inc (Germantown, Maryland). Serially cut, paraffin-embedded tissue slides were stained in all cases for hematoxylin and eosin, PDE11A, c-KIT, and KITLG antibodies following the manufacturer's protocols. Appropriate positive and negative controls were obtained. For PDE11A antibody, positive staining is cytoplasmic; for c-KIT and KITLG antibodies, positive staining is membranous/cytoplasmic.

In vitro studies

Two different cell lines were used to perform the experiments: TCam-2 and NTERA-2. Cells were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum and 1% antibiotic-antimycotic at 37°C in 5% CO₂ humidified atmosphere. All cell culture reagent and media were obtained from Invitrogen, Inc (Carlsbad, California). TCam-2 cells [originally described by Mizuno et al (14) and a kind gift of Dr Leendert Looijenga] and testicular embryonic cancer NTERA-2 cells (obtained from American Type Culture Collection, Manassas, Virginia) were transiently transfected with plasmid DNA expressing either WT or mutated forms of PDE11A described in patients with familial TGCTs (R52T, F258Y, Y727C, R804H, V820M, R867G, and M878V), using Lipofectamine 2000 (Invitrogen), as we described elsewhere (11). After harvesting, the relative activity of PDE in the cells was measured using the BIOMOL GREEN reagent (QuantiZyme assay system; BIOMOL International LP, Plymouth Meeting, Pennsylvania). cAMP levels were determined using the cAMP [3H] Biotrak assay system (GE Healthcare Life Sciences, Pittsburgh, Pennsylvania). Functional studies were done in triplicate, and each experiment was repeated at least twice.

Protein levels were determined by Western blot, using the antibodies for c-KIT and for KITLG mentioned previously, following standard procedures (15). mRNA expression for c-KIT and KITLG was analyzed by real-time quantitative PCR in the ABI Prism 7700 sequence detector (Applied Biosystems, Inc, Foster City, California), using specific primers (available upon request). Relative quantification was performed using the 2^-ΔΔCT method (16).

Statistical analysis

All statistical analyses were performed with the SPSS 16.0 (SPSS Inc, Chicago, Illinois). Continuous genetic data were expressed as mean ± SD. All in vitro experiments were performed in triplicate. A 2-sample t test was used for paired samples. A P < .05 was considered statistically significant.

Results

Genotyping for PDE11A and KITLG

We studied the frequency of PDE11A alterations and the KITLG SNPs, rs3782179 and rs4474514, in blood samples from 94 patients with familial TGCT and 692 control individuals. Among the FTGCT group, 19 patients from 14 unrelated families harbored the PDE11A variations (c.155G>T/p.R52T, c.773T>A/p.F258Y, c.2180A >G/p.Y727C, c.2411G>A/p.R804H, c.2458G>A/p.V820M, c.2599C>G/p.R867G, and c.2632A>G/p.M878V), as previously reported (10). PDE11A sequence variants were also present in samples from 63 control individuals. The variants c.2411G>A/p.R804H and c.2599C>G/p.R867G that were observed in FTGCT patients, and also the previously reported c.919C>T/p.R307X and c.1655_1657del/inFS15X variants, were also found in this group of controls (10, 11, 17). Allele frequencies for the KITLG SNP rs3782179 in TGCT and in control samples are shown in the Table 1 (more details are shown in Supplemental Tables 1 and 2, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org). Due to strong linkage disequilibrium, the results for SNP rs4474514 were essentially the same as those for rs3782179 and are therefore not presented.

Table 1.

Genotypes of the KITLG SNP rs3782179 in Patients With FTGCT and Control Individuals

KITLG rs3782179 Allele Frequency
	FTGCT Patients T*/TC/C	Controls T*/TC/C
PDE11A variants +	15/4/0	41/18/4
PDE11A variants −	66/9/0	431/174/24
Total	81/13/0	472/192/28

Asterisk indicates the risk allele, T.

The frequency of the KITLG rs3782179 SNP major T allele (which is the TGCT risk conferring allele) was slightly higher in patients with PDE11A variants than those without but was not significantly different from that in controls. When we analyzed the allele frequencies of the KITLG SNPs and the PDE11A variants, comparing all TGCT patients with control samples, the differences were strongly significant (P = .0007, Supplemental Table 3). Indeed, the risk of FTGCT was 2.6-fold higher in individuals harboring both a PDE11A sequence variant and the KITLG rs3782179 risk allele, compared with carriers of the KITLG risk allele only (odds ratio 2.60; 95% confidence interval 1.47–4.60).

A total of 86.2% of the TGCT patients were homozygous carriers of the KITLG rs3782179 risk allele vs 68.2% in the control group (P = .0003), confirming its role in TGCT susceptibility (7). Among the PDE11A variant-positive individuals, the homozygous carrier rate of the rs3782179 risk allele was also more frequent in the TGCTs (79%) compared with the controls (65.1%), a difference that was, however, not statistically significant (Supplemental Table 4).

We also studied 50 unaffected blood relatives of FTGCT patients, in whom the rs3782179 risk allele frequency was similar between carriers and noncarriers of PDE11A sequence variations (Supplemental Table 5). Allele frequencies and comparative analysis of the KITLG variant rs4474514 (data not shown) were identical with the results observed for KITLG variant rs3782179 (see tables), again because these 2 loci are in such strong linkage disequilibrium.

PDE11A, c-KIT, and KITLG protein expression

Formalin-fixed, paraffin-embedded tumor tissue along with control tissue from the same subject was studied in 30 FTGCT patients. IHC showed that carriers of inactivating PDE11A variations had lower PDE11A protein expression than the control tissue but higher than the control staining for c-KIT and KITLG (Figure 1). All 7 patients were homozygous for the KITLG variant rs3782179 risk allele (Supplemental Table 6). In contrast, tumor tissue from 7 (that were available) of the 10 FTGCT patients who had PDE11A WT sequence, showed very low or no KITLG protein expression in tumor cells despite their carrying 2 copies of the KITLG rs3782179 risk allele (Figure 1). These IHC studies suggested that there might be an association between the presence of a mutant PDE11A enzyme and higher c-KIT signaling, which provided the background for the in vitro studies below.

Figure 1.

Expression of PDE11A, c-KIT, and KITLG in testicular tumor tissue of a patient positive for the R804H PDE11A missense mutation (A, B, and C, respectively) and in testicular tumor cells of a patient with wt PDE11A genotype (D, E, and F, respectively). PDE11A expression was negative and c-KIT and KITLG expressions were positive and intense in the tumor of the patient harboring the PDE11A mutation, whereas low positive expression of PDE11A and c-KIT, with no expression of KITLG was seen in the patient WT for PDE11A.

In vitro and studies in sporadic TGCTs

Experiments in the human testicular embryonic carcinoma (NTERA-2) cell line transfected with expression vectors harboring the above-mentioned PDE11A substitutions were then performed to investigate in vitro a possible connection between the cAMP and c-KIT signaling pathways. As previously reported (10), cAMP levels were significantly higher in cells transfected with each of the PDE11A variations, relative to WT sequence (Figure 2A). Accordingly, PDE activity was lower in PDE11A variant-transfected NTERA-2 cells vs WT cells (Figure 2B). KITLG expression was consistently increased in the presence of the PDE11A sequence variants, at both the protein and RNA levels (Figure 3, A and B, respectively). KITLG mRNA expression was significantly higher in the PDE11A constructs (R52T, F258Y, A349T, D609N, R687G, and M878V) than its expression in PDE11A wild-type cells.

Figure 2.

cAMP levels in NTERA-2 cells transfected with PDE11A mutants (A) and relative activity of PDE in NTERA-2 cells (B).

Figure 3.

Protein expression of c-KIT and KITLG in N-TERA cells transfected with PDE11A mutant constructs (A) and gene expression of c-Kit and SCF (KITLG) in N-TERA cells transfected with PDE11A mutant constructs (B).

We repeated the in vitro studies using TCam-2 cells (a seminoma cell line) that are not commercially available. Similar to NTERA-2 cell results, cAMP levels were significantly higher and the PDE activity was significantly lower in variant-transfected TCam-2 cells vs WT cells (Figure 4, A and B, respectively). At the protein level, KITLG expression was consistently increased in the presence of PDE11A sequence variants (Figure 4C).

Figure 4.

cAMP levels (A) and activity of PDE (B) in TCam-2 cells transfected with PDE11A wild type and mutants. C: Western blot analysis of protein lysate from TCam-2 cells transfected with PDE11A wild type and mutants. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

We then looked for cKIT and KITLG expression in tissue from 2 sporadic TGCT cases obtained at the time of surgery, one with an inactivating PDE11A sequence variant (F258Y) and the other without (WT PDE11A sequence). KITLG mRNA levels were more than 2 times higher in the PDE11A mutation-positive tumor than the PDE11A WT tumor (Figure 5A). In the tumor lysates, a higher expression of KITLG was also detected in the PDE11A mutation-positive tumor vs the PDE11A WT tumor in which the KITL expression was almost undetectable (Figure 5B).

Figure 5.

A, c-Kit (cKIT) and SCF (KITLG) mRNA expression in the tumor from a patient with sporadic TGCT with a PDE11A variation and in a control as determined using real-time PCR. B, Western blot analysis of protein lysate from sporadic testicular tumors. The patient harbors a F258Y PDE11A mutation. A sporadic testicular cancer without PDE11A mutations was used as a control. A higher expression of KITLG was detected in patient's tumor compared with the control in which KITL expression was almost undetectable. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Discussion

TGCTs are the most common cancers in young white men (18, 19), and although the molecular causes of these tumors remain elusive, evidence for a substantial genetic contribution to TGCT susceptibility has been developed in recent years, the following has been found: first, an estimated 1.4% of men with newly diagnosed TGCT report a positive family history for this cancer (20); second, siblings of men with TGCT have an 8- to 10-fold increase in TGCT risk (21); and third, a 37- to 76.5-fold higher risk of developing TGCT has been reported in dizygotic/monozygotic twin brothers of men with the disease (22). However, linkage analysis and candidate gene approaches have failed to demonstrate a single high-penetrance susceptibility gene, suggesting that FTGCT susceptibility may result from the interaction of multiple common, low-penetrance genetic variants (3).

We recently reported that PDE11A-inactivating sequence variants may modify the risk of familial and bilateral TGCT (10). PDE11A is a regulator of cAMP signaling in the adrenal cortex and other steroidogenic tissues; it is also highly expressed in normal human testis. In our previous study, the prevalence of PDE11A-inactivating variants (19%) was significantly higher among patients with FTGCT than in unaffected family members and controls; functional studies showed that all the described genetic variants reduced PDE activity when transfected into human H295R and rat Leydig cell lines (10). In addition, PDE11A protein expression was decreased or absent in testicular tumor samples from familial TGCT subjects who were carriers of the sequence variants (10).

Recently 3 GWASs have provided strong evidence for 6 new TGCT susceptibility loci in the regions of KITLG, SPRY4, BAK1, DMRT1, TERT/CLPTM1L, and ATF7IP (7, 8, 23). The KITLG gene is located on chromosome 12 and encodes the ligand for the receptor tyrosine kinase c-KIT. Its statistical association with TGCT is the strongest cancer-related GWAS finding to date (odds ratios in risk allele heterozygotes range from 1.3 to 2.6, compared with risk allele homozygotes, which range from 4.6 to 6.7). The c-KIT signaling pathway has previously been implicated in both the normal biology of germ cells and the pathogenesis of sporadic TGCT. Somatic mutations (or overexpression) of the protooncogene KIT are detected at various frequencies in several neoplasms, including testicular cancer (5, 6). The murine homologue of the KITLG gene, Kitl, controls primordial germ cell survival, proliferation, and migration (24, 25); loss of the transmembrane Kitl isoform increases susceptibility to TGCT in mice and Kitl homozygous knockout animals are infertile due to progenitor germ cell development failure. The discovery that 4 of the 6 TGCT-related GWAS hits (ie, all except TERT/CLPTM1L and ATF7IP) (26) are important in normal fetal testis development and fertility has provided a unifying etiological theme that has altered the focus of research into the molecular pathogenesis of TGCT.

In the current study, we investigated whether FTGCT patients with PDE11A variants more frequently carried risk-conferring KITLG alleles; we also investigated cAMP and c-KIT signaling in vitro. We demonstrated that the copresence of PDE11A-inactivating sequence variants and the KITLG rs3782179 risk allele T was more common than the presence of the latter alone in patients with TGCT; this increased frequency was also higher than that observed in controls. Using logistic regression analysis, we treated the KITLG genotype as a continuous variable (ie, 0 vs 1 vs 2 risk alleles) to determine whether there was statistical evidence of a gene-gene interaction in our data, but we found no evidence to support this hypothesis (data not shown). Thus, although variants in each gene were independently associated with TGCT risk, there was no statistical evidence of interaction between the 2 genes. However, we had very limited statistical power to detect such a phenomenon. This question would be worth reconsideration in a larger data set when one becomes available.

IHC showed strong KITLG expression in TGCT cells from men harboring both the PDE11A genetic variants and KITLG risk-conferring genotype. Although IHC is a nonquantitative technique, the comparison between normal cells and structures within the same surgical specimen suggested that indeed PDE11A inhibition was copresent with c-KIT signaling activation.

In addition, in vitro, NTERA-2 cells transfected with the specific PDE11A mutations found in our patients showed not only significantly higher cAMP levels and lower PDE activity but also consistently increased KITLG expression, both at the RNA and protein levels. Similar results were obtained (Figure 4) when these experiments were repeated using the TCam-2 cell line that was derived from a primary testicular tumor sample of pure classical seminoma from a 35-year-old patient and currently is the only cell line to be generally accepted as representative of the seminoma subtype of germ cell tumors (14, 27–30).

Finally, the data were replicated when we compared fresh tissue from 2 sporadic TGCT cases (Figure 5) obtained at surgery, one with a PDE11A variant (F258Y) and one without. The tumor with the PDE11A variant showed evidence for higher KIT signaling, suggesting, like a number of other studies, that familial and sporadic TGCTs differ little, if at all, in their molecular pathogenesis. Of course, larger studies of exclusively sporadic TGCT cases must be performed to confirm this observation.

In conclusion, these data support a possible functional link between the cAMP and c-KIT signaling pathways in increasing the risk of TGCT caused by certain KITLG alleles. This is the first demonstration of such a novel association in any tissue; the possibility that gene variants in both pathways increase the risk of familial testicular cancer warrants further exploration in the testis and in other tissues.