Gain-of-function amino acid substitutions drive positive selection of FGFR2 mutations in human spermatogonia

Abstract

Despite the importance of mutation in genetics, there are virtually no experimental data on the occurrence of specific nucleotide substitutions in human gametes. C>G transversions at position 755 of FGF receptor 2 (FGFR2) cause Apert syndrome; this mutation, encoding the gain-of-function substitution Ser252Trp, occurs with a birth rate elevated 200- to 800-fold above background and originates exclusively from the unaffected father. We previously demonstrated high levels of both 755C>G and 755C>T FGFR2 mutations in human sperm and proposed that these particular mutations are enriched because the encoded proteins confer a selective advantage to spermatogonial cells. Here, we examine three corollaries of this hypothesis. First, we show that mutation levels at the adjacent FGFR2 nucleotides 752-754 are low, excluding any general increase in local mutation rate. Second, we present three instances of double-nucleotide changes involving 755C, expected to be extremely rare as chance events. Two of these double-nucleotide substitutions are shown, either by assessment of the pedigree or by direct analysis of sperm, to have arisen in sequential steps; the third (encoding Ser252Tyr) was predicted from structural considerations. Finally, we demonstrate that both major alternative spliceforms of FGFR2 (Fgfr2b and Fgfr2c) are expressed in rat spermatogonial stem cell lines. Taken together, these observations show that specific FGFR2 mutations attain high levels in sperm because they encode proteins with gain-of-function properties, favoring clonal expansion of mutant spermatogonial cells. Among FGFR2 mutations, those causing Apert syndrome may be especially prevalent because they enhance signaling by FGF ligands specific for each of the major expressed isoforms.

Both direct and indirect methods concur that the majority of human germ-line mutations are of paternal origin (1-4). Notable examples of a nearly exclusive paternal origin are provided by dominant pathological mutations in three genes, FGFR2 (causing Apert, Crouzon, and Pfeiffer syndromes), FGFR3 (causing achondroplasia and related bone dysplasias), and RET (causing multiple endocrine neoplasia type 2), all of which encode members of the receptor tyrosine kinase family (5-11). In addition, the average age of the fathers originating these mutations is elevated by 2-5 years (12, 13). This paternal age effect, which provided the earliest clue to the occurrence of mutation in humans (14, 15), has generally been attributed to the requirement for repeated cycles of replication of spermatogonial stem cells, occurring once every ≈16 days, so that mutations accumulate with increasing age (“copy-error hypothesis”). This effect might be accentuated by sex differences in methylation, age-related increases in DNA damage, or deterioration in chromatin structure, replication fidelity, or proofreading mechanisms (1, 4, 16).

The localized nature and high apparent rates of these receptor tyrosine kinase mutations make them potentially amenable to analysis of their biological origins. We developed a sensitive assay (17) to quantify low levels (< 10^-5) in sperm of the FGFR2 755C>G transversion (encoding Ser252Trp), which is the most frequent FGFR2 mutation arising de novo in human livebirths (18). The heterozygous state of this mutation causes Apert syndrome, a characteristic combination of craniosynostosis and syndactyly (19).

The local nucleotide and encoded protein contexts of FGFR2 755C (part of a CpG dinucleotide) are shown in Fig. 1 a and b. In the protein, Ser-252 resides at the start of the extracellular Ig-like IgIIIa region, and hence, is present in both major isoforms of FGFR2 (IgIIIa/IIIb = FGFR2b and IgIIIa/IIIc = FGFR2c); these isoforms are generated by mutually exclusive alternative splicing (Fig. 1a) and bind different repertoires of ligands (20, 21). Genetic, biochemical, and structural arguments suggest that the predominant pathophysiological effects of the Ser252Trp substitution are mediated through the FGFR2c isoform by a combination of enhanced affinity for natural ligands (for example, FGF2) and illegitimate binding of ligands (for example, FGF10) that are normally specific for FGFR2b (21-24). Crystallographic analysis of Ser252Trp-FGFR2c bound to FGF2 shows that this enhanced affinity is principally caused by formation of a hydrophobic patch in the receptor that stabilizes contacts with the flexible N-terminal region of the ligand (25).

Fig. 1.

Structural features of human FGFR2 and sequence context around nucleotide 755G. (a) Cartoon of partial gene structure showing exons IIIa, IIIb, IIIc (encoding the extracellular IgIII domain), and TM (encoding the transmembrane segment). The positions of the 749-112G/A SNP, nucleotides 755C in exon IIIa, and 943G in exon IIIc are indicated. Brackets enclose positions of restriction sites AvaI and EcoRV, diagnostic for exons IIIb and IIIc, respectively, in the rat Fgfr2 ortholog. Alternative splicing to generate the spliceforms FGFR2b and FGFR2c is shown above. (b) DNA sequence around the start of exon IIIa (intron sequence is in lowercase and exon sequence is in capitals). Nucleotides are clustered into codons with the corresponding encoded amino acids. The box encloses the MboI restriction site (nucleotides 752-755) used to select for mutant sequences. Below, all 12 possible single-nucleotide substitutions (bold) at the MboI site are shown with the encoded amino acid in brackets. Only the 755C>G (Apert) and 755C > T (normal/Crouzon) mutations have been observed in the heterozygous state in humans.

Our measurement of the level of FGFR2 755C>G in sperm exploits the fact that the mutation abolishes a recognition site for the restriction enzyme MboI (Fig. 1b) so that predigestion of genomic DNA with MboI enables enrichment of mutant molecules. After PCR amplification and a further round of MboI digestion/PCR, nucleotides at FGFR2 position 755 are quantified, in relation to a spiked mutant FGFR2 reference sequence, by using pyrosequencing technology and statistical analysis (17).

Key conclusions from previous work on sperm were as follows (i) The level of 755C>G mutations in sperm varies widely between healthy men (from < 10^-6 to 1 in 6,200), but is relatively stable over months in individual men. (ii) Average mutation levels are positively correlated with donor age, and mirror epidemiological observations of paternal age for Apert syndrome births. (iii) Surprisingly, levels of the 755C>T transition (encoding Ser252Leu), which is usually clinically silent in the heterozygous state but occasionally causes Crouzon syndrome (26, 27), are also elevated (but overall, are ≈1.7-fold lower than 755C>G) and also positively correlate with donor age. Levels of the third possible substitution, 755C>A (encoding a stop codon) are not increased. (iv) In individuals heterozygous for a nearby SNP, 749-112G/A, most donors producing high levels of 755C>G show a marked predominance of mutations on one or other allele of the SNP, an effect that is less marked for 755C>T. (v) In sperm from a healthy 37-year-old donor, we found a double-mutation 755_756CG>TC (encoding Ser252Phe) present at ≈1 in 175,000. A similar heterozygous double-mutation 755_756CG>TT, also encoding Ser252Phe (Table 1), was previously reported in two children with Apert syndrome (26, 28).

Table 1. Tabulation of multiple nucleotide mutations that include position 755, which were reported previously or are described here.

Nucleotides	Amino acids	Phenotype	Ref(s).
755_756CGG>TT	Ser252Phe	Apert syndrome	26, 28
755_756CG>TC	Ser252Phe	Apert syndrome	17, this work
755_757CGC>TCT	Ser252Phe; Pro253Ser	Pfeiffer syndrome	26
755C>T; 943G>T	Ser252Leu; Ala315Ser	Syndactyly	29, this work
755_756CG>AC	Ser252Tyr	Not reported	This work

To explain these observations, we proposed that specific FGFR2 mutations, arising at low frequency in diploid (mitotic) spermatogonial stem cells, confer a proliferative advantage to the mutated cell relative to its WT neighbors and lead to clonal expansion within the testis over time. We term this process “protein-driven selection” of mutations, in contrast to copy error, which describes a neutral process of mutation accumulation. Modeling of this process estimated that the 755C>T mutations arise ≈2.2-fold more frequently than 755C>G, which is compatible with the expected predominance of C>T transitions at CpG dinucleotides (2, 30). On the other hand, the selective advantage of clones harboring 755C>G was estimated as ≈3.8-fold greater than 755C>T, which is compatible with biochemical data that FGFR2c-Ser252Trp enhances FGF2 binding affinity substantially more than FGFR2c-Ser252Leu (22). Although the double mutations could have originated in a one-step process, alternatively they may have developed sequentially, whereby the relatively weak proliferative advantage of a cell-harboring 755C>T was enhanced by a second-hit FGFR2 mutation arising on the same allele. If the resulting double-nucleotide mutant encoded an FGFR2 protein with increased FGF binding affinity (such as Ser252Phe), this would increase the cell's proliferative advantage, and hence, enrich for these very rare events (which have an estimated germline rate of 10^-11; ref. 30). Additional examples of multiple nucleotide mutations that include 755C>T are shown in Table 1.

Three issues that arose from the previous work (17) are investigated here. First, in protein-driven selection, the background mutation rate need not be altered; we have exploited the MboI digestion strategy to compare mutation levels at FGFR2 positions 752-754 with those previously determined for position 755. Second, the mechanisms underlying the remarkable clustering of double mutations at position 755 remain uncertain. We present three instances of double mutation that expand the spectrum of mutations and illuminate their origins. Third, we have directly tested the prediction that FGFR2 must be expressed in spermatogonia by using recently isolated cell lines. These findings reinforce the key role of protein-driven selection and suggest a molecular explanation of the particularly high birth rate of mutations causing Apert syndrome.

Materials and Methods

Human Subjects. All protocols were approved by the Oxfordshire Research Ethics Committee. Subject 204F2 presented at 5 months of age with craniosynostosis and syndactyly, features diagnostic of Apert syndrome. Physical examination of her father, 32.3 years old at the time of the birth, was normal. An asymptomatic 49-year-old male (individual II-2 in ref. 31), heterozygous for both the 943G>T (Ala315Ser) mutation and 749-112G/A SNP in FGFR2, provided two semen samples at 1-week intervals. His carrier daughter was homozygous (GG) at the SNP, establishing that the 943T allele was in cis with the G allele of the SNP.

Analysis of Genomic DNA. Genomic DNA was extracted by using a Nucleon BACC2 kit (Tepnel Life Sciences, Manchester, U.K.). Primer pairs and amplification conditions for FGFR2 exon IIIa were as described (18). The DNA was sequenced on a 3100 DNA Sequencer (Applied Biosystems).

Reanalysis of Previous Samples. We previously collected blood and semen samples from healthy donors and quantified mutations at FGFR2 755C in triplicate 10-μg DNA aliquots, two of which were mixed with different amounts of a triple FGFR2 mutation (Table 1) (“spike DNA”) to provide an internal reference. The procedure involved gel purification of MboI-resistant DNA, PCR amplification, a repeat MboI digest and PCR, followed by Pyrosequencing and statistical analysis (17).

To quantify mutations at positions 752G, 753A, and 754T (Fig. 1b) in relation to those at 755C, we used Pyrosequencing to assay the remainder of the final unspiked PCR products that were used for the original 755C analysis (17). Sufficient material was available from all 11 of the original blood controls and 95 of 99 of the original sperm controls. Pyrosequencing was performed on a PSQ-HS96A system (Biotage, Uppsala). The sequencing primers used, and corresponding dispensation orders, were as follows. 752: 5′-CTCTCTCCACCAGAGC-3′, dispensation order ATATCATCGATAGCTGCTCA; 753: 5′-CTCTCCACCAGAGCG-3′, dispensation order TCTCGTCGCAGCGCTAGCTGC; and 754, 5′-TCTCCACCAGAGCGA-3′, dispensation order CGACGCTAGCTGACTCACG. The molecular species detected by each dispensation are shown in Fig. 6 a-c, which is published as supporting information on the PNAS web site.

Thirteen molecular species were assumed to be present in the mixture (WT and the 12 single-nucleotide substitutions shown in Fig. 1b). Relative amounts of each species were quantified from the complete Pyrosequencing data set by using the general approach described (17). Estimation of WT and 755A, 755G, and 755T mutants was validated on cloned DNA (Fig. 6d). Absolute amounts were obtained by equalizing the sum of mutations (A, G, and T) at position 755 to the value obtained previously (17). Correlation coefficients between the published and new estimates of 755A, 755G, and 755T were 0.37 (reflecting the large uncertainty for these low estimates), > 0.99, and 0.96, respectively. In Fig. 2a, the previously published values are illustrated.

Fig. 2.

Quantification of mutations at the 752-755 MboI site of FGFR2. (a) Estimated levels of all 12 possible single-nucleotide mutations (see Fig. 1b) in 95 sperm and 11 blood samples (shown separately), arranged in order of decreasing level of total mutations at position 755. (b) Pie chart showing overall proportion of total mutations at each nucleotide in sperm (Left) and blood (Right). The areas of the two circles are proportional to the overall mutation levels from the two sources.

To search for 755_756CG>AC or AT double mutations, we reexamined the original Apert Pyrosequencing data (17) for discrepancies in peak height between dispensations 5(A) and 6(G). One individual (ID number 37, the 60-year-old father of a child heterozygous for the 755C>G Apert mutation) consistently showed a deficiency of peak 6 in relation to peak 5 in pyrograms from the triplicate samples. The unspiked PCR product was reamplified with EcoR1-AG12 and EcoR1-AG63 (17), digested with BglI plus SfiI (to select for double mutations), and EcoRI, and ligated into pUC18-EcoRI (Amersham Pharmacia). Five BglI-resistant clones were sequenced with M13F and M13R (17) with identical results.

Analysis of Sperm Samples from a 943G>T Heterozygote. DNA from sperm was extracted and analyzed for mutations at FGFR2 755C as described (17). To normalize the mutation level with previous estimates, we processed, in parallel, eight additional sperm DNA samples that we had analyzed previously and used two independent serial dilutions of the spike DNA. Multiple regression (least-squares method) forcing a 0 intercept was used to estimate the best predictor of mutation level.

Rat Fgfr2 RNA Preparation and Expression. Two rat spermatogonial cell lines, GC-5spg and GC-6spg, were cultured under standard conditions (32). RNA was isolated by using RNA-Bee (Campro Scientific, Veenendaal, The Netherlands) according to the manufacturer's instructions. cDNA was produced from ≈7 μg of RNA by random hexamers RT-PCR with use of 400 units of Moloney murine leukemia virus reverse transcriptase (Promega) in the presence of 1× first-strand buffer/2 mM dNTP/20 mM DTT/40 units of RNase inhibitor in a 20-μl volume (negative controls without reverse transcriptase were performed in parallel). One-eighth of the resulting cDNA was PCR-amplified with 0.1 μM primers ExIIIa-F (5′-CCCATATCCAGTGGATCAAACATG-3′) and TM-R (5′-GTGATCTCCTTCTCTCTCACAGGTGC-3′) in the presence of 2.5 units of TAQ/Pwo polymerase (Roche), 1× buffer 2 and 200 μM dNTP in a 40-μl volume (35 cycles at 94°C for 30 s, 55°C for 30 s, and 72°C for 45 s, followed by a 10-min extension at 72°C). The resulting 253/256-bp PCR fragments were digested for 2 h with 20 units of AvaI (specific for the Fgfr2b spliceform), EcoRV (specific for the Fgfr2c spliceform), or left undigested, and separated on a 2% TBE (45 mM Tris/45 mM boric acid/1 mM EDTA, pH 8.3)-agarose gel.

Results

Analysis of Mutations at FGFR2 Nucleotides 752-755. We extended our analysis of FGFR2 nucleotide 755C to include the three other nucleotides (752G, 753A, and 754T) represented in the MboI recognition site. Analysis of each position is instructive in a different way (Fig. 1b). Nucleotide 752G, like nucleotide 755C, is part of a CpG dinucleotide; if the high levels of mutation at 755C were related to cytosine methylation alone, then we would anticipate a comparable magnitude of mutation at 752G. All substitutions at 753A are synonymous and are therefore likely to be selectively neutral; mutations at 754T result in substitutions at Ser-252 of amino acids that are structurally different from Trp and Leu, and hence, are not expected to enhance FGF binding affinity (21, 25).

We estimated levels of 10 potential species present at positions 752-754 in relation to total substitutions at position 755 in sperm (n = 95) and blood (n = 11) samples, and compared these measurements with those obtained previously for the same samples at position 755. Fig. 2a shows the results for the individual samples and Fig. 2b summarizes these data. In sperm, the average level of mutation at 755C>G and 755C>T was 19-fold higher than the average for the other 10 substitutions. 752G>A, which is expected to be the predominant mutation at a CpG dinucleotide (2, 30), was the next most common substitution (1 in 485,000); however, accurate quantification of mutations at this and lower levels is likely to be unreliable, owing to increasing contamination with background PCR errors and DNA damage. In blood, for which there is no evidence for selection of the 755C>G or 755C>T mutations (17), levels of mutation at the two CpG residues (752G and 755C) were comparable (Fig. 2b). These data show that, within the target MboI site, only the 755C>G and 755C>T mutations have disproportionately elevated levels in sperm, excluding the possibility that this region of FGFR2 is generally highly mutable.

Double Mutations in FGFR2: Sequential Origin and Shared Properties of Encoded Proteins. We identified a child with the classical features of Apert syndrome, in whom a heterozygous double mutation in FGFR2, 755_756CG>TC encoding Ser252Phe, was present (Fig. 3a). FGFR2 analysis of her parents showed a normal sequence in the mother, but the heterozygous mutation 755C>T was present in the clinically unaffected father; this mutation had been transmitted from his clinically unaffected mother (Fig. 3a). The findings in this family illustrate two points. First, it represents the fourth known case in a liveborn individual of mutation of two or three consecutive nucleotides involving FGFR2 codon Ser-252 (Table 1); fewer than one such mutation is expected to occur in the entire human population (30). The same 755_756CG>TC change was previously identified as a de novo double mutation in the sperm of a healthy donor (17). Second, the sequential origin of the two mutations is demonstrated: a 756G>C transversion occurred on a paternal FGFR2 allele that already carried a 755C>T mutation. The father did not wish to provide a semen sample for further analysis.

Fig. 3.

Identification of double-mutation events in FGFR2.(a) A sporadic case of Apert syndrome (subject 204F2). (Top) DNA sequencing of FGFR2 showing heterozygosity for the 755_756CG>TC mutation in the proband (Right). Only the 755C>T mutation is present in her father (Left). (Middle) The proband manifests the typical facial appearance and syndactyly of Apert syndrome (Right), whereas her father has clinically normal hands (Left). (Bottom) Family pedigree, showing patterns of digestion of FGFR2 exon IIIa with DpnII (an isoschizomer of MboI) and HinfI. The specific HinfI digestion product in the proband (black circle) indicates that the two substitutions identified on DNA sequencing are present on the same allele. WT, normal control. (b) Identification of a Ser252Tyr mutation in the sperm of a 60-year-old man. (Upper) Pyrogram from the donor's unspiked sperm sample. E and S denote addition of enzyme and substrate, respectively. The fifth (A) and sixth (G) nucleotide dispensations (green arrows) normally yield peaks of equal height; the absence of a peak at the sixth dispensation indicates that in this sample, 755C>A mutations are not followed by 756G. (Lower) Sequencing of a mutant clone from this PCR product reveals the double-mutation 755_756CG>AC (arrowheads). The pyrogram shows that this mutation is present at comparable levels to 755G (black arrow) and 755T (red arrows). (c) Measurement of levels of the mutations 755C>G (Left) and 755C>T (Right) in two sperm samples from a 49-year-old man heterozygous for the 943G>T (Ala315Ser) mutation. Red triangles show levels for this individual (mean ± 95% equal tail probability interval) normalized with respect to previous measurements on 99 normal donors (white circles) (17).

The double mutations at the Ser-252 codon described up to this point all encode phenylalanine (Table 1). As with the canonical Apert mutation Ser252Trp, Ser252Phe introduces a large aromatic side chain, which (given the similarity of the consequent phenotype) is likely to participate in analogous hydrophobic interactions with bound FGF (25). We considered whether this property might be mimicked by any other substitutions of Ser-252: tyrosine, the third amino acid with an aromatic side chain, is the obvious candidate, but (like phenylalanine) this would require a double substitution at codon 252 (755_756CG>AT or AC). Retrospectively, we sought evidence for this mutation in our previous Pyrosequencing data. We noted an atypical pyrogram from one sperm sample that showed absence of the second peak diagnostic for 755A, suggesting that an additional substitution at position 756 had occurred (Fig. 3b). Cloning and sequencing revealed the double-mutation 755_756CG>AC encoding Ser252Tyr, estimated to be present at a level of 1 in 22,100 in the sample. This finding confirms our prediction that Ser252Tyr mutations (expected to be even rarer than Ser252Phe because none of the contributing single-nucleotide substitutions is positively selected) can also be enriched in sperm. The heterozygous phenotype resulting from this mutation may resemble Apert syndrome but has not yet been reported in humans.

To further explore the occurrence of double mutations in sperm, we considered the previous clinical observation (29) of a family in which a 755C>T (Ser252Leu) mutation had arisen de novo on a paternal allele already bearing a constitutional 943G>T (Ala315Ser) substitution in the IIIc exon of FGFR2 (see Fig. 1a). These two changes are pathologically synergistic from both the phenotypic and functional viewpoints. The heterozygous state of either single substitution is clinically mild (26, 31); but, when both substitutions exist on the same FGFR2 allele, syndactyly of severity comparable to Apert syndrome is evident (29). The functional corollary is that illegitimate binding by FGF10, a ligand that is normally selective for the FGFR2b isoform, occurs to mutant FGFR2c when both substitutions are present (21).

To assess whether the 755C>T and 943G>T mutations synergize in the testis, we analyzed the levels of FGFR2 755C substitutions in two sperm samples from a 49-year-old man heterozygous both for 943G>T (31) and the 749-112G/A SNP. Whereas the levels of 755C>G were within the normal range, 755C>T was present at the highest level yet recorded (an average of 1 in 5,147) (Fig. 3c). Allele-specific PCR showed that 91% (95% equal tail probability interval of 87-94%) of the 755C>T mutations were present on the same FGFR2 allele as the 943G>T (Ala315Ser) substitution. Ignoring other bias or stochastic effects on the frequency of originating 755C>T mutations between the two FGFR2 alleles, the 10-fold predominance on the Ala315Ser allele indicates the extra proliferative advantage conferred by the double mutation to the mutant spermatogonial cell, compared with 755C>T arising on the WT allele. This process is likely to be mediated through illegitimate signaling by an FGF10-like ligand (21).

Both the Fgfr2b and Fgfr2c Spliceforms Are Expressed in Spermatogonial Cells. An important prerequisite for protein-driven selection is that FGFR2 must be expressed in the cell population where selection is taking place. Previous studies on the whole testes had not demonstrated expression of FGFR2 in spermatogonial cells (33, 34). We analyzed RNA prepared from two rat spermatogonial stem cell lines, GC-5spg and GC-6spg, with characteristics similar to type A spermatogonia, based on morphology, expression, and stem cell properties; however, the growth rate of GC-6spg is approximately double that of GC-5spg (32). By using RT-PCR and diagnostic restriction digests, we demonstrated exclusive Fgfr2c expression in cell line GC-5spg; however, cell line GC-6spg showed a different pattern, with predominant Fgfr2b expression and only weak Fgfr2c expression (Fig. 4). These results show that FGFR2 is expressed in spermatogonial stem cells. The different pattern of expression of Fgfr2 spliceforms in the two cell lines suggests that they are in differing cellular states and that expression of the distinct isoforms occurs at different spermatogonial stages.

Fig. 4.

Expression of Fgfr2b and Fgfr2c spliceforms in two rat spermatogonial stem cell lines. Digestion of RT-PCR products with AvaI (specific for Fgfr2b) and EcoRV (specific for Fgfr2c) compared with uncut product and a DNA-negative control (-), for cDNA from cell lines GC-5spg (Left) and GC-6spg (Right). The product from GC-5spg digests completely with EcoRV, indicating that only the Fgfr2c spliceform is present. By contrast, the majority of product from GC-6spg digests with AvaI, indicating predominance of Fgfr2b; the faint EcoRV digestion product shows that a minority of Fgfr2c is also present.

Discussion

Our previous study (17) documented high mutation levels and allelic skewing of FGFR2 mutations at 755C>G and (to a lesser extent) 755C>T in the sperm of many healthy men. Here, we have extended these observations by showing that similarly high mutation levels are not present at three adjacent nucleotides (752-754), that double mutations involving 755C can arise from sequential events, and that both Fgfr2b and Fgfr2c spliceforms are expressed in rat spermatogonial stem cell lines. These observations all support selection of FGFR2 mutations at the protein level in the testis.

Three key concepts are required to understand the seemingly complex patterns of mutation. First, where the mutation has been observed in the heterozygous state in humans, the severity of the clinical phenotype seems to provide a good indication of the strength of gain of function conferred by the mutant protein (21, 22, 26, 29). Second, distinct gain-of-function mechanisms operate in the skull and limb: craniofacial phenotypes result from enhanced signaling through physiological FGFR2c pathways, whereas limb malformations are mediated through illegitimate FGFR2c activation by FGF10-like ligands (21-24, 29). Third, we propose that FGF signaling regulates spermatogonial cell fate such that the increased affinity of FGF interaction in cells expressing mutant FGFR2 favors their net proliferation. Despite previous difficulties in identifying FGFR2 in spermatogonial cells, our demonstration of Fgfr2 expression (Fig. 4) is not surprising, given the requirement of FGF for successful spermatogonial cell culture (32, 35, 36) and the role of FGF receptors in determining the fate of other stem cell types (37).

Applying these concepts, Fig. 5 diagrammatically illustrates the scenarios by which various FGFR2 mutations could attain detectable levels (> 10^-6) in sperm. Importantly, these scenarios can explain the high levels of specific single and double mutations without requiring any elevation in the underlying germ-line mutation frequency. The scenario in Fig. 5d provides particularly compelling evidence for protein-driven selection because the structural basis for the synergy in cis observed in sperm (Fig. 3c) between the physically remote FGFR2 mutations 943G>T and 755C>T (separated by 2.7 kb in genomic DNA) is revealed within the conformation of the encoded mutant protein (21). We previously argued (17) that the major factor explaining the paternal age effect in Apert syndrome is the time-dependent clonal accumulation of cells carrying rare replication errors, rather than an intrinsically high and/or age-dependent frequency of the initiating mutations. This proposal is in accordance with evidence that specific mechanisms exist to maintain very low mutation frequencies in stem cells (38, 39).

Fig. 5.

Diagrammatic representation of mutation and selection processes in spermatogonial cells, leading to differing prevalence of various FGFR2 mutations. Pairs of parallel lines denote the two FGFR2 alleles with the identity of nucleotides at positions 755 and 943 (d only) specifically indicated. Radiating clusters of arrowed lines represent independent mutational events with the number of arrows reflecting relative mutation frequency. Gray sectors indicate expansion over time of mutant clones bearing the indicated genotype. (a) 755C>G arising in a WT individual: the C>G transversion is relatively infrequent, but the Ser252Trp substitution confers a strong selective advantage. (b) 755C>T arising in a WT individual: C>T transitions occur frequently at this CpG dinucleotide, but the Ser252Leu substitution confers a weak selective advantage compared with Ser252Trp, leading to slower clonal expansion. (c) 756G>C arising in the spermatogonial cell of an individual constitutionally heterozygous for 755C>T. Mutations arising on 755T alleles encode Ser252Phe; selective advantage is comparable to Ser252Trp. Mutations arising on 755C alleles are synonymous and selectively neutral. Note that the processes illustrated in b and c, if they occurred sequentially in a single spermatogonial cell, would lead to double mutations in the sperm of WT individuals. (d) 755C>T arising in an individual constitutionally heterozygous for 943G>T. Mutations on 943T alleles encode the synergistic double-substitution Ser252Leu; Ala315Ser in cis, leading to strong selective advantage. Mutations arising on the WT (943G) allele are more weakly selected as in b.

The observation that many constitutively activating, paternally derived mutations arise in the IIIc exon of FGFR2, and hence, only affect FGFR2c signaling (6, 18, 40), would appear to indicate that this is the specific isoform that is physiologically active in spermatogonial cells. Hovever, both phenotypic (29) and functional (21) evidence suggests that the double-mutation Ser252Leu; Ala315Ser, which exhibits selection in the testis (Fig. 3c), acts by enabling illegitimate FGF10-FGFR2c signaling. This evidence for an FGF10-like ligand in the spermatogonial environment suggests that physiological FGFR2b signaling may also be taking place, which is supported by the expression data (Fig. 4) and by the identification of two rare germ-line Alu element insertions into FGFR2, of paternal origin, that activate ectopic FGFR2b expression and manifest an Apert syndrome phenotype (41). We propose that the especially high rate of de novo Apert syndrome FGFR2 mutations in humans (18) occurs because these mutations uniquely activate both legitimate and illegitimate pathways of FGFR2c signaling (21, 23, 24).

The concept of protein-driven positive selection is not new, being familiar in the context of somatic mutations occurring during neoplasia (42), in the development of immunity (43), and in microorganisms (44). However, a role in germ-line mutation in vertebrates has not been widely recognized. Several features of FGFR2 mutation in Apert syndrome, including the very high apparent rates of specific single and double mutations, exclusive paternal origin, and allelic skewing, indicate that protein-driven selection, rather than a neutral copy-error mechanism, is the overriding factor shaping their occurrence. We suggest that this concept extends to additional disorders due to gain-of-function mutations in FGFR2 (45), FGFR3 (46), and RET, which show similar biological features and have been popular exemplars for thinking about male-driven mutation. The extended reproductive life of the human male might render our species particularly vulnerable to these selfish germ-line mutations (47).