Cellular evidence for selfish spermatogonial selection in aged human testes



Owing to a recent trend for delayed paternity, the genomic integrity of spermatozoa of older men has become a focus of increased interest. Older fathers are at higher risk for their children to be born with several monogenic conditions collectively termed paternal age effect (PAE) disorders, which include achondroplasia, Apert syndrome and Costello syndrome. These disorders are caused by specific mutations originating almost exclusively from the male germline, in genes encoding components of the tyrosine kinase receptor/RAS/MAPK signalling pathway. These particular mutations, occurring randomly during mitotic divisions of spermatogonial stem cells (SSCs), are predicted to confer a selective/growth advantage on the mutant SSC. This selective advantage leads to a clonal expansion of the mutant cells over time, which generates mutant spermatozoa at levels significantly above the background mutation rate. This phenomenon, termed selfish spermatogonial selection, is likely to occur in all men. In rare cases, probably because of additional mutational events, selfish spermatogonial selection may lead to spermatocytic seminoma. The studies that initially predicted the clonal nature of selfish spermatogonial selection were based on DNA analysis, rather than the visualization of mutant clones in intact testes. In a recent study that aimed to identify these clones directly, we stained serial sections of fixed testes for expression of melanoma antigen family A4 (MAGEA4), a marker of spermatogonia. A subset of seminiferous tubules with an appearance and distribution compatible with the predicted mutant clones were identified. In these tubules, termed ‘immunopositive tubules’, there is an increased density of spermatogonia positive for markers related to selfish selection (FGFR3) and SSC self-renewal (phosphorylated AKT). Here we detail the properties of the immunopositive tubules and how they relate to the predicted mutant clones, as well as discussing the utility of identifying the potential cellular source of PAE mutations.


Advanced age is associated with many structural, physiological and molecular changes in the testis (Paul & Robaire, 2013). Fertility is reduced in older men, consistent with the documented reduction in sperm motility (Auger et al., 1995; Levitas et al., 2007) and impaired sperm morphology (Auger et al., 1995; Eskenazi et al., 2003; Ng et al., 2004). Sperm DNA parameters also change with age, including increases in global DNA methylation (Jenkins et al., 2013), DNA fragmentation (Spano et al., 1998; Wyrobek et al., 2006), telomere length (Allsopp et al., 1992; Baird et al., 2006; Kimura et al., 2008; Aston et al., 2012) and number of de novo mutations (Kong et al., 2012; Michaelson et al., 2012; Jiang et al., 2013). The last parameter, most commonly involving single nucleotide substitutions, is attributed to copy errors that occur during the numerous divisions that spermatogonial stem cells (SSCs) undertake in the lifetime of the individual (~23 divisions per year in adults). Hence, de novo mutations in critical coding or regulatory regions of the genome may cause disease in the next generation. In this review, we describe how specific gain-of-function mutations arising spontaneously in SSCs may lead to a clonal expansion of the mutant cells, thereby significantly increasing the number of spermatozoa carrying pathogenic mutations. This process, termed selfish spermatogonial selection, likely occurs in all men. The most common clinical outcome of this selfish selection is the birth of a child with a paternal age effect (PAE) disorder; very rarely, it may also lead to a testicular tumour (Goriely & Wilkie, 2012). Our search for evidence of selfish spermatogonial selection directly in human testes identified a subset of seminiferous tubules with increased density of spermatogonia that fulfil the criteria expected for selfish mutant clones (Lim et al., 2012). We will describe the morphology and prevalence of these cellular events, and discuss the biological processes that are anticipated to drive their clonal expansion in the ageing testis.

Paternal age effect disorders and mutations

PAE and sperm studies

Older fathers are at increased risk for their offspring to have congenital disorders such as achondroplasia, Apert, Noonan and Costello syndromes and multiple endocrine neoplasia types 2A and 2B (Goriely & Wilkie, 2012). These spontaneously arising dominant genetic disorders, termed PAE disorders, are caused by specific gain-of-function mutations in fibroblast growth factor receptors (FGFRs), other tyrosine kinase receptors, and components of the RAS/mitogen-activated protein kinase (MAPK) signalling pathway (Table 1). The causative de novo mutations that affect a few well-characterized locations in the genome arise almost exclusively from the unaffected fathers, who are typically 2–7 years older than the population average. Considering the relatively high spontaneous birth rate of these disorders (e.g. 1/30 000 births for achondroplasia) it is estimated that some of these specific point mutations arise from the male germline up to ~1000-fold more frequently than the expected background mutation rate (reviewed by Goriely & Wilkie, 2012). Direct analysis of sperm DNA from healthy men sampled from the general population has confirmed that PAE mutations are indeed present at levels substantially above the background mutation rate in most men (Goriely et al., 2003, 2009; Yoon et al., 2009; Giannoulatou et al., 2013). Fathers of affected children produce spermatozoa with the corresponding mutation at levels that tend to be above average, but are within the normal range for the particular mutation (Goriely et al., 2003). The levels of individual mutations vary widely between different men, probably because the originating mutations are rare and arise stochastically. Overall, the average mutation levels correlate with paternal age and account for the birth prevalence of the particular disorder (Goriely et al., 2003, 2009; Yoon et al., 2009; Giannoulatou et al., 2013). These observations indicate that the production of PAE mutations is not restricted to fathers of affected children, but is a universal process, with levels of mutant spermatozoa increasing over time.

Table 1. Paternal age effect mutations investigated in DNA from spermatozoa and whole testes
    Investigated in sperm/testis DNAReported in tumourse
Germline disorderBirth prevalenceGeneMutationsSpermTestisSpSd
  1. SpS, spermatocytic seminoma; NA, not applicable (no corresponding germline disorder has been identified).

  2. a

    Few other mutations in same gene also cause the same disorder.

  3. b

    Many other mutations in same and/or other genes cause the same disorder.

  4. c

    Other reports were omitted owing to technical limitations of the studies (see Wilkie, 2005; Goriely & Wilkie, 2012).

  5. d

    Combined data from Goriely et al. (2009) and Giannoulatou et al. (2013).

  6. e

    Data from COSMIC database September 2013 (v66), also see Goriely & Wilkie (2012).

Apert syndrome~1/65 000 FGFR2 p.S252WGoriely et al. (2003), Yoon et al. (2009)cQin et al. (2007), Choi et al. (2008, 2012)0/25Yes
p.P253RYoon et al. (2009)cChoi et al. (2008)0/24Yes
Crouzon/Pfeiffer syndrome~1/50 000–1/100 000p.S252LbGoriely et al. (2003, 2005)0/25No
Achondroplasia~1/30 000 FGFR3 p.G380Ra c Shinde et al. (2013)c0/44Yes
Thanatophoric dysplasia type II~1/370 000p.K650EGoriely et al. (2009)2/55 p.K650EYes
Costello syndrome~1/300 000 HRAS p.G12SaGiannoulatou et al. (2013)0/59Yes
p.G13Ca1/53 p.G13RYes
NANAp.Q612/59 p.Q61K 4/59 p.Q61RYes
Noonan syndrome~1/10 000 PTPN11 p.N308DbYoon et al. (2013)0/21Yes
Multiple endocrine neoplasia type 2B~1/150 000 RET p.M918TChoi et al. (2012)0/22Yes

Selfish selection

The apparent high mutation rates at the specific genomic sites that give rise to PAE conditions, and the age-related effect, are best explained by a phenomenon termed selfish spermatogonial selection. Mutations arising by chance at key positions in PAE genes (Table 1) during SSC divisions are associated with gain-of-function properties, conferring a proliferative or survival advantage on the mutant SSC. This leads to expansion of the mutant clone over time in a manner similar to oncogenesis (identical mutations occurring in somatic tissues are associated with cancer – see Table 1), resulting in increased levels of mutant spermatozoa. This clonal expansion mechanism was initially proposed based on the allelic ratio distribution of PAE mutations in sperm DNA: in samples with high levels of the Apert syndrome mutation in FGFR2, the mutations were typically skewed towards one or the other allele, implying rare originating mutational events followed by expansion of the mutant cells (Goriely et al., 2003). (By contrast, if the mutations had been occurring at such elevated levels because of large numbers of primary mutational events – for example, caused by many independent copy errors at mutational hotspots – the mutations would likely occur equally frequently on both alleles.) Furthermore, it could be anticipated that if the mutational events were recurrent and there was no selection occurring, then mutations would be detected at low levels throughout the testis. However, by dissecting whole human testes into approximately 200 pieces and testing them for specific mutations, it was demonstrated that PAE mutations have a restricted spatial distribution, with small regions containing high levels of mutation surrounded by larger mutation-free regions, consistent with a localized clonal expansion mechanism (Qin et al., 2007; Choi et al., 2008, 2012; Shinde et al., 2013; Yoon et al., 2013).

Spermatogonial divisions and spermatogonial selection

Although there are multiple lines of evidence to support selfish spermatogonial selection, the precise mechanism(s) of clonal expansion have yet to be determined, not least because understanding of the physiology of cell turnover in human spermatogenesis is still far from complete. The consensus model of human spermatogenesis dates back over four decades to Clermont's original descriptions of human and primate seminiferous tubules (reviewed by Clermont, 1972). In this model, Adark spermatogonia represent the reserve stem cell population that divide rarely to form Apale spermatogonia. Apale spermatogonia are more regularly proliferating stem cells that divide to form Apale spermatogonia or the more mature B spermatogonia, which undergo meiosis to form spermatocytes and eventually spermatozoa. In humans, the entire process of spermatogenesis takes about 72 days. The duration of the seminiferous epithelium cycle is approximately 16 days (~23 cycles per year) and can be classified into different stages based on the presence and morphology of different cell types (Muciaccia et al., 2013). The spermatogonial divisions in humans are believed to be symmetrical (Clermont, 1972; Ehmcke & Schlatt, 2006), however, in other tissues homeostasis can be maintained by symmetrical or asymmetrical divisions of stem cells (Fuchs & Chen, 2013).

Mathematical modelling of selfish spermatogonial selection based on a simple self-renewal mechanism assumed that wild-type SSCs divide asymmetrically and suggested that mutant SSCs undertaking occasional symmetrical divisions (approximately one for every 100 regular asymmetrical divisions) could explain the localized clonal distribution of PAE mutations in testes (Crow, 2006; Qin et al., 2007; Choi et al., 2008, 2012; Giannoulatou et al., 2013). Such a small selective growth advantage (<1%) has also been predicted for driver mutations in cancer (Bozic et al., 2010). A recent selfish selection model (Yoon et al., 2013) showed that selection is also compatible with the predicted symmetrical divisions of human spermatogonia. In this selection model mutant SSCs are slightly more likely to form two SSCs (self-renew), rather than to form two more differentiated cells, leading to an expansion of the mutant SSCs over time (Yoon et al., 2013). Thus, although the mechanisms controlling normal human spermatogenesis at the cellular level are still unknown, a selective advantage resulting in a modest increase in self-renewing symmetrical divisions would lead to the expansion of selfish mutant spermatogonial clones, irrespective of the precise details of the lineage dynamics.

In contrast to the limited information on human SSCs, spermatogenesis in the mouse, which differs somewhat in terms of its cellular organization (for review see Dym et al., 2009), is better documented. Murine SSC clones are highly dynamic: clonal expansion or contraction (loss) is a stochastic process that can be explained by neutral drift (Nakagawa et al., 2007; Klein et al., 2010b). Over time, mouse seminiferous tubules become less diverse in terms of SSC clones; some clones are retained and expand, while others are lost to maintain homeostasis (Fig. 1Aii,Bii,Cii). Although expansion or loss of clones is likely to be a chance decision between neighbouring wild-type clones, it has been postulated that a clone with a selective advantage may preferentially be retained and grow in size (Nakagawa et al., 2007). Thus, clones with PAE mutations would be more likely to expand than to be lost (Fig. 1B,C). In skin, p53 mutant cells have a growth advantage over their wild-type counterparts only when a selective pressure (UVB radiation) is present (Klein et al., 2010a), indicating that the preferential expansion of mutant cells may be the result of a selective advantage.

Figure 1.

Model of the predicted expansion mechanism for selfish mutations in spermatogonia represented at the testis (i) seminiferous tubule (ii) and cellular (iii) level. (A) (i) Within the testis, groups of one to four highly convoluted seminiferous tubules (blue) are encapsulated within lobules (black borders). (ii) On the basis of the neutral drift model of mouse spermatogonial stem cell (SSC) dynamics (Klein et al., 2010b), a seminiferous tubule initially contains a large number of clones derived from single SSCs, each encompassing a small region of the tubule (blue and grey regions represent individual clones). (iii) Cross section of a seminiferous tubule, showing the location of the spermatogonia (ovals outlined in grey) at the periphery, with differentiation of the SSCs proceeding towards the lumen (grey arrows). A selfish mutation (red) arises by chance in a SSC. (B) and (C) (i) Over time (black arrows), clonal expansion occurs. (ii) In normal development, some clones are anticipated to be lost whilst others expand to maintain tissue homeostasis. This is a stochastic process governed by neutral drift between neighbouring wild-type clones. The increased probability of self-renewing symmetrical divisions of the mutant spermatogonia lead to their preferential expansion along (ii) and around (iii) the seminiferous tubule. This model predicts that a small selective advantage may be sufficient to drive clonal expansion. The heterozygous mutant clones produce wild-type and mutant spermatozoa in equal proportions (iii). (D) and (E) (i) In rare cases, probably because of additional mutagenic events (red dotted arrows), the mutant clones may develop into spermatocytic seminoma (E), possibly via an intratubular form (D). Spermatocytic seminoma and intratubular spermatocytic seminoma are composed of three cell types that vary in size and morphology (Diii). Such cellular variation is not seen in immunopositive tubules (Biii, Ciii). The possible link between immunopositive tubules and intratubular spermatocytic seminoma is further discussed in section 'Immunopositive tubules and intratubular spermatocytic seminoma'.

Role of PAE proteins in spermatogonia

So far, all identified PAE genes encode components of the tyrosine kinase receptor-RAS/MAPK signalling pathway and are expressed in spermatogonia (Table 1). Two of the major ligands in SSC renewal are glial-derived neurotrophic factor (Gdnf) (Kubota et al., 2004), which binds the growth factor receptor Ret, and fibroblast growth factor-2 (Fgf2) (Ishii et al., 2012), which binds Fgfr2 and Fgfr3 (Ornitz et al., 1996). Fgfr2, mutations in the human orthologue of which cause Apert, Crouzon and Pfeiffer syndromes, is expressed in rat SSC cell lines (Goriely et al., 2005), and its paralogue FGFR3 (mutations of which cause achondroplasia and thanatophoric dysplasia) is expressed in human spermatogonia (von Kopylow et al., 2012a,b; Ewen et al., 2013).

Further functional studies implicate other PAE genes in regulating SSC self-renewal and differentiation. Reduced expression of Gdnf in mouse results in progressive depletion of the germ cell pool (Meng et al., 2000). Conversely, overexpression of Gdnf inhibits SSC differentiation resulting in the formation of spermatogonial clusters that develop into seminomatous tumours (Meng et al., 2001). Gdnf binding to Ret (activating mutations in RET cause multiple endocrine neoplasia types 2A and 2B) and its co-receptor Gfrα1 leads to Mek/Erk and Pi3k/Akt activation and SSC proliferation in vitro (Oatley et al., 2007; He et al., 2008; Hasegawa et al., 2013). Chemical inhibition of Mek1, an activator of Erk1/2, nullifies the proliferative effects of Gdnf addition (He et al., 2008) and causes expression of differentiation markers, indicating a role for Erk signalling in maintaining the undifferentiated state (Hasegawa et al., 2013). Gain-of-function mutations in PTPN11, which encodes the tyrosine phosphatase SHP2 that is recruited to stimulated growth factor receptors, cause Noonan syndrome. Loss of Shp2 in spermatogonia halts their proliferation, resulting in the progressive loss of germ cells – an effect that may be mediated, at least in part, by impaired Erk signalling (Puri et al., 2013).

Activating mutations in HRAS cause Costello syndrome while mutations in its paralogues KRAS and NRAS cause Noonan syndrome. In mouse, all three orthologues are expressed in SSCs and Ras is activated in SSCs cultured in the presence of Egf, FGF2, or Gdnf (Lee et al., 2009). Inhibiting Ras blocks SSC proliferation in vitro, while transfecting SSCs with activated Hras (encoding p.G12V) bypasses the requirement for cytokine stimulation (Lee et al., 2009). Transplanting germ cells with activated Hras to sterile recipient testes results in germ cell colonies, some of which were associated with normal spermatogenesis, whereas other regions resembled seminomatous tumours similar to those caused by Gdnf overexpression (Lee et al., 2009).

Therefore, as all the molecules known to be associated with PAE disorders cluster within a network controlling homeostasis in spermatogonia, it is predicted that activation of this pathway by gain-of-function mutations may lead to enhanced proliferation of the mutant cells.

Selfish spermatogonial selection and testicular tumours

Based on the apparent selection and clonal expansion of spermatogonia with PAE mutations, coupled with the fact that identical mutations in somatic tissues are associated with cancer (Table 1), it was postulated that selfish spermatogonial selection may lead to testicular tumours (Hansen et al., 2005). Spermatocytic seminoma (SpS) is a rare, slow-growing tumour that rarely metastasises. Unlike the more common classical seminoma that develops from precursor carcinoma in situ cells derived from gonocytes (Skakkebaek et al., 1987; Sonne et al., 2009), SpS develops from adult germ cells, thought to be spermatogonia (Rajpert-De Meyts et al., 2003; Lim et al., 2011). SpS is also distinct from classical seminoma in terms of age of onset, with SpS typically found in older men (mean age of onset ~54 years). Owing to the spermatogonial origin and age-related profile of this tumour we hypothesized that SpS may be an extreme outcome of the selfish selection process.

To identify if there is a connection between the two phenomena, we screened a panel of SpS cases for mutations in PAE-associated and other candidate genes (Goriely et al., 2009; Giannoulatou et al., 2013). A mutation in FGFR3 (encoding p.K650E), that causes thanatophoric dysplasia type II as a germline mutation, was identified in two cases (Table 1). In addition, mutations in two of three oncogenic hotspot codons of HRAS, encoding p.G13 and p.Q61, were identified in seven cases in total (Table 1). The HRAS mutations identified in SpS are distinct from those that cause Costello syndrome and have never been reported as germline mutations, but identical somatic mutations are associated with various cancers (COSMIC v66; www.sanger.ac.uk/cosmic) (Goriely et al., 2009; Giannoulatou et al., 2013). Interestingly, the mutation-positive samples were from significantly older patients than the mutation-negative samples, suggesting that two genetically and epidemiologically distinct groups of SpS may exist (Goriely et al., 2009; Giannoulatou et al., 2013). The majority of SpS cases were also shown to be positive for expression of either FGFR3 or HRAS, often in a mutually exclusive pattern. Positivity for FGFR3 or HRAS did not always correlate with the mutational profile of the tumours (Goriely et al., 2009). The mutual exclusivity of mutations and the inverse correlation of immunohistochemical positivity, suggest that mutations at any point of the receptor tyrosine kinase/RAS/MAPK cascade may lead to upregulation of the signalling pathway.

The identification of mutations in PAE genes, and the overexpression of PAE proteins in SpS, both suggest a role for selfish spermatogonial selection in the pathogenesis of late onset testicular tumours (Fig. 1). In some testes with SpS, the lumen of a subset of the seminiferous tubules in the adjacent normal tissue may be filled with abnormal cells of various sizes. This rare presentation has been proposed to be a precursor to SpS, known as intratubular SpS (ISS) (Muller et al., 1987; Lim et al., 2011). Therefore, on the basis of the model of the pathogenesis of mutant clones and SpS, ISS may represent an intermediate stage (Fig. 1), although the spread of an established tumour is an alternative interpretation.

In conclusion, selfish mutations in spermatogonia are unique in that they are associated with tumourigenesis in the host and congenital disease in potential offspring. Conceptually, this unifies the processes of somatic and germline mutation, usually viewed as distinct genetic events occurring in different cells and at different times during an organism's life, to a single initiating event during spermatogenesis. Although PAE mutations probably occur in the testes of all men as they age, SpS is rare, suggesting that additional mutagenic events are required for it to develop (Goriely et al., 2009; Goriely & Wilkie, 2012).

Cellular/morphological evidence of selfish spermatogonial selection

Expectations from previous data

The mutation analysis performed on sperm DNA (Goriely et al., 2003, 2005, 2009; Yoon et al., 2009; Giannoulatou et al., 2013) suggests that selfish spermatogonial selection is a universal process. Further, the evidence from whole testes describes large isolated foci of apparent clones of mutant cells (Qin et al., 2007; Choi et al., 2008, 2012; Shinde et al., 2013; Yoon et al., 2013). Owing to the inherent technical difficulty of detecting mutations at low levels, these experiments have been limited to a small number of genomic locations associated with well-characterized disorders (Table 1), but numerous other mutations are expected to be selected in the paternal germline (Goriely & Wilkie, 2012; Goriely et al., 2013). Therefore, mutant clones are anticipated to be common throughout the testis of ageing men. Considering that the additive birth prevalence of the eight disorders listed in Table 1 is in the order of 1/5500, it follows that approximately 1/5500 spermatozoa (corresponding to 1/2750 diploid spermatogonia) of men of the average paternal age (estimated to be 32.8 years in 2011 for England and Wales) are anticipated to carry selfish mutations i.e. these are rather common events. Mutant spermatogonial clones are predicted to be more prevalent and larger in older men because of the age-related nature of selfish spermatogonial selection.

Based on the anatomy of the seminiferous tubules, it is expected that growth of selfish clones would, at least initially, be restricted by the basement membrane and therefore follow a tubular course (Fig. 1B,C). Although whole testis studies showed a clonal distribution of PAE mutations (Qin et al., 2007; Choi et al., 2008, 2012; Shinde et al., 2013; Yoon et al., 2013) because these were based on DNA analysis, the normal tissue architecture was destroyed and therefore the morphology and cellular appearance of these mutant clones is unknown. As the mutant spermatogonia are predicted to expand in a manner similar to oncogenesis, we hypothesized that they might have a distinct morphological appearance that would be apparent in cross sections of seminiferous tubules (Fig. 1B,C). Alternatively, as the clones must, at least initially, be able to maintain spermatogenesis, their morphological appearance may resemble normal tubules but they might differ in their immunohistochemical characteristics based on the dysregulation of the RAS/MAPK pathway.

Immunopositive tubules in aged testes

In the first and only study to seek evidence of selfish selection directly in human testes (Lim et al., 2012), we screened one or two small regions (corresponding to ~20 consecutive 5 μm sections) of histologically normal formalin-fixed paraffin embedded testes from six men aged 69 years and over. Serial sections were stained with an antibody against melanoma antigen family A4 (MAGEA4) protein, a robust marker of all spermatogonial cell types (He et al., 2010) that is also expressed in SpS (Rajpert-De Meyts et al., 2003; Lim et al., 2011). Sections were independently analysed for the presence of clusters of spermatogonia or other morphological appearances that may reflect a mutational origin. In three of the six testes, this approach identified a subset of seminiferous tubules with a distinct histopathological appearance; in these tubules, termed ‘immunopositive tubules’, dense clustering of spermatogonia, sometimes forming multiple layers at the basal lamina or clusters in the lumen of the tubule, was apparent (Lim et al., 2012) (Fig. 2).

Figure 2.

Immunopositive tubules at low and high magnification in 5 μm sections of testis from a 75 year old man. (A) At low magnification, three groups (demarcated by yellow dotted lines) of cross-sectioned seminiferous tubules with darker staining of FGFR3, MAGEA4 and pAKT are apparent. Owing to the convoluted nature of seminiferous tubules, it is likely that the immunopositive cross sections within the yellow lines are part of the same tubule. Also, given their close proximity, we postulate that the three groups are different regions of the same tubule. (B) Higher magnification of regions in blue boxes in A. In normal tubules (outside yellow lines), MAGEA4 is expressed in a single layer of spermatogonia and FGFR3 and pAKT are found in a subset of spermatogonia adjacent to the basal lamina. In the immunopositive tubules, dense clusters and multiple layers of cells positive for MAGEA4, FGFR3 and pAKT are present adjacent to basal lamina. Very small clusters (asterisks) are present in the lumen of some immunopositive tubules. Sections were labelled with antibodies and counterstained with haematoxylin, as previously described (Lim et al., 2012). Other immunopositive tubules from the same testis region (sample 2–1) were previously reported in Lim et al. (2012). Scale bar = 100 μm.

Immunopositive tubules are positive for FGFR3 and pAKT

To confirm the spermatogonial nature of the clustered cells in the immunopositive tubules [weak MAGEA4 staining of spermatocytes had previously been reported (Aubry et al., 2001)], and to further investigate the putative association with SpS, we stained adjacent sections of the three testes containing immunopositive tubules for other spermatogonial markers expressed in SpS (Goriely et al., 2009; Lim et al., 2011). This revealed that the immunopositive tubules in each of the three testes display enhanced FGFR3 staining (Lim et al., 2012) (Fig. 2). In normal tubules, FGFR3 is expressed only in a subset of spermatogonia, believed to be Adark type (von Kopylow et al., 2012a; Ewen et al., 2013), but the majority of the spermatogonia in immunopositive tubules are FGFR3-positive. Subsequent staining for signalling proteins downstream of FGFR3 revealed that the spermatogonia in the immunopositive tubules were also positive for phosphorylated AKT (pAKT), in a pattern similar to that of the FGFR3 staining (Lim et al., 2012).

Akt signalling has a direct role in SSC self-renewal/proliferation in response to Gdnf; antagonising this pathway in cultured SSCs using chemical inhibitors reduces cell growth (Lee et al., 2007, 2009; Oatley et al., 2007). Phosphorylation of Akt occurs after the addition of Gdnf (Lee et al., 2007; Hasegawa et al., 2013). SSCs transfected with activated Akt proliferate in the absence of Gdnf, and can successfully colonize and proliferate when transplanted into sterile testes (Lee et al., 2007). Akt also has roles independent of Gdnf, as Akt-transfected cells require supplementation with FGF2 (Lee et al., 2007) and in cultured wild-type spermatogonia Akt signalling controls expression of both Gdnf- and non-Gdnf-regulated SSC transcription factors (Oatley et al., 2007).

In conclusion, the Lim et al. (2012) study has shown that the clustered spermatogonia in immunopositive tubules express a protein with an established role in selfish selection and SpS (FGFR3), in addition to a key signalling molecule in spermatogonial self-renewal (pAKT) (Fig. 2). We have therefore proposed that the appearance of dense clustering and multiple layering of spermatogonia positive for FGFR3 and pAKT is compatible with the clonal expansion of selfish mutant spermatogonia (Fig. 1). It remains to be established whether FGFR3 and pAKT are co-expressed in the same cell, and if immunopositive tubules represent an enrichment of cells that normally express FGFR3 and pAKT, or if their expression is the downstream effect of a PAE mutation, as is predicted for FGFR3 and HRAS in SpS.

Potentially analogous findings from previous studies

Patterns of multiple layering of spermatogonia and clusters of lumenal spermatogonia similar to those we have described for immunopositive tubules (Lim et al., 2012) have previously been reported in sections of normal human testes from elderly men (aged 68–72 years) viewed under an electron microscope (Holstein et al., 1984). On the basis of their morphology, the cells were defined as Apale spermatogonia, and it was hypothesized that as these tubules only contained spermatogonia and Sertoli cells, the dense clustering of Apale spermatogonia was attributable to the failure of the cells to differentiate into B spermatogonia (Holstein et al., 1984). In this study (Lim et al., 2012), apparently normal differentiation was present in the immunopositive tubules in two of the three positive testes, contrasting Holstein's findings. This distinction is important as the mutant clones must be capable of differentiating into functional spermatozoa if the mutations are to be transmitted to the next generation. In the third positive testis no spermatocytes were seen, similar to Holstein's report, but B spermatogonia were present (indicated by SAGE1 positivity) (Lim et al., 2012). Another report of two- or multi-layered spermatogonia described in a subset of tubules of infertile men (aged 27–44 years) revealed that the spermatogonia were within a lumenal extension of the blood–testis barrier (Bergmann et al., 1989), indicating that the spermatogonia were within their normal niche. The tubules with multi-layered spermatogonia were independent of other events associated with the infertility (Bergmann et al., 1989). We postulate that these earlier reports reflect the same phenomenon as we have described, but may represent more severe, macroscopically distinct manifestations. As discussed later, we suggest that cases with severe clustering of spermatogonia, with little or no differentiation, may represent mutant clones with highly activating mutations that inhibit differentiation and/or lead to oncogene-induced senescence of spermatogonia.

In some of the immunopositive tubules we observed spermatogonia detached from the basement membrane and located in the lumen, either as individual cells or as clusters of tens to hundreds of cells that we termed ‘microclones’ (Lim et al., 2012). Such lumenal clusters of spermatogonia were also reported by Holstein (Holstein et al., 1984). Lumenal microclones were also present in the three testes in which immunopositive tubules were not detected (Lim et al., 2012). Previous reports of lumenal spermatogonia in human testes involved cases where Sertoli cells lacked tight junctions. This phenomenon has been reported in normal tissue (from men aged 38, 66, 73, 77 years) at the terminal segment of the junction between seminiferous tubules and the rete testis (the transitional zone) (Lindner, 1982), in infertile men (aged 27–44 years) (Bergmann et al., 1989) and in oestrogen-treated transsexuals (aged 22–48 years) (Schulze, 1988). The status of the tight junctions was not investigated in our study. However, although microclones are relatively commonly detected (Lim et al., 2012), because of their small size and abnormal positioning within the tubules they are unlikely to account for the large regions of PAE mutations observed in whole testes (Qin et al., 2007; Choi et al., 2008, 2012; Shinde et al., 2013; Yoon et al., 2013).

Cellular proliferation in immunopositive tubules

This study suggested that the spermatogonia in the immunopositive tubules are unlikely to be rapidly dividing, as few cells were positive for proliferation marker Ki67 (Lim et al., 2012). This was consistent with the FGFR3 positivity of the immunopositive tubules, as expression of FGFR3 and Ki67 is mutually exclusive in spermatogonia in normal tubules (von Kopylow et al., 2012a; Ewen et al., 2013). Although there was no apparent difference in proliferation of spermatogonia in normal and immunopositive tubules (Lim et al., 2012), a clear distinction is not to be expected as the models of selfish spermatogonial selection only predict an increase in the type of division, and not the number of divisions, in selfish clones (Qin et al., 2007; Choi et al., 2008, 2012; Giannoulatou et al., 2013; Yoon et al., 2013). It is anticipated that highly proliferating spermatogonia would exhibit limited differentiation, in effect reducing the number of mutant spermatozoa being produced. Therefore, the level of proliferation in immunopositive tubules is consistent with the outcome of mathematical predictions concerning selfish selection (section 'Spermatogonial divisions and spermatogonial selection').

Immunopositive tubules and intratubular spermatocytic seminoma

Owing to the rarity of SpS, there have been very few reported cases of ISS, with only a small fraction of the presumably asymptomatic ISS reported in testes without SpS (Eble, 1994). However, if ISS is the precursor to SpS (Muller et al., 1987; Lim et al., 2011), cases of ISS should be more apparent in testes without SpS. The small number of reported cases of ISS without overt SpS may be attributable to the limited screening of non-cancerous, non-infertile testes from elderly men. Although ISS was not detected in any of the six testes analysed in our study (Lim et al., 2012), the intratubular clusters of spermatogonia in the immunopositive tubules are reminiscent of mild cases of ISS. Similar to SpS, ISS is formed of three distinct cell types that vary in their size: small cells (6–8 μm) with condensed chromatin, medium cells (15–20 μm) with round nuclei (the most prominent cell type) and large (80–100 μm), occasionally multinucleated cells (Eble, 1994) (Fig. 1Diii). However, such variation in cell types was not seen in the immunopositive tubules (Lim et al., 2012). The predicted transition from immunopositive tubules to ISS/SpS may involve additional mutations. However, to delineate the possible connection between these two abnormal intratubular entities will require the identification of further cases of ISS. Screening these for expression of FGFR3 and pAKT, as well as establishing pAKT expression in SpS, may clarify the relationship between these phenomena. In addition, identification of PAE mutations in cases of ISS without SpS would establish whether ISS represents an intermediate stage between immunopositive tubules and SpS (Fig. 1).

Size and distribution of immunopositive regions of tubules is consistent with selfish spermatogonial selection


Pulse-labelling studies have shown that clones derived from a single SSC can occupy regions of seminiferous tubules up to 8 mm in length in old mice (Nakagawa et al., 2007; Klein et al., 2010b). However, the contribution of single stem cells to clones in human testes has not yet been determined. The immunopositive regions of seminiferous tubules extend at least into the millimetre range, demonstrated by staining of longitudinal sections (Fig. 3) and the three-dimensional reconstruction of neighbouring contiguous tubular cross sections (Lim et al., 2012). Typically, the identified immunopositive tubular cross sections cluster together, with small groups of immunopositive regions close to each other (Fig. 2A). In humans, seminiferous tubules are partitioned by septa into incomplete lobules each typically containing 1–4 tubules (van Lommel, 2003; Glass, 2005). Therefore, when immunopositive tubules are physically close and located within the same lobule, they are likely to belong to the same tubule, suggesting that these represent single clones that could extend into the centimetre range. Clones of 5 mm–5 cm in length would represent approximately 1/150 000th–1/15 000th of the total length of seminiferous tubules in humans [2 testes × 500 tubules × 750 mm each (Glass, 2005)]. In theory, these selfish clones would produce mutant spermatozoa in the range of 1/30 000–1/300 000 (considering the heterozygous nature of PAE mutations), indicating that the immunopositive tubules would be capable of producing mutant spermatozoa at levels that are compatible with those detected in human spermatozoa DNA studies (Goriely et al., 2003, 2009; Yoon et al., 2009; Giannoulatou et al., 2013).

Figure 3.

Cross section of a seminiferous tubule displaying normal and immunopositive regions. (A) In sections 09 and 10 (5 μm each) of a testis from a 79 year old man (sample 3–1 in Lim et al., 2012), cross sections reveal a region where an immunopositive tubule is surrounded by tubules with normal staining. Sections 14, 22 and 23 demonstrate that the immunopositive regions and normal regions are part of the same tubule. There is little apparent difference in proliferation between these two regions as indicated by Ki67 staining. (B) Higher magnification view of the boxed regions in section 14. (i) Dense clustering of MAGEA4-positive spermatogonia in the immunopositive region. Haematoxylin staining reveals that there are no differentiated cells in the tubule, but Sertoli cells are present (yellow arrows). (ii) In the normal region MAGEA4 positive cells form a single layer at the basal lamina and the lumen contains spermatocytes as revealed by haematoxylin staining. Scale bar = 100 μm.


For each of the three testes in which immunopositive tubules were detected, an additional tissue block from the same testis was also screened. Further immunopositive tubules were identified in two of the three cases, although at lower frequencies (Lim et al., 2012). This suggests there are regional variations in the frequency of immunopositive tubules and that the lack of detection in three of the original testes may be attributable to a sampling effect (in each case, the volume screened represented less than 0.2% of the total testis). Such regional variation is predicted by the clonal nature of selfish spermatogonial selection and was also reported in the whole testis studies (Qin et al., 2007; Choi et al., 2008, 2012; Shinde et al., 2013; Yoon et al., 2013). Thus, expansion of single mutant clones along seminiferous tubules, that are restricted geographically to lobules, can explain the localized distribution of PAE mutations in testes.

The occasional longitudinal section of seminiferous tubules revealed that the immunopositive regions may be contiguous with regions of the tubule with a normal histological appearance (Lim et al., 2012) (Fig. 3). This may represent the progression front of normal and mutant clones and may provide an interesting template to study the growth patterns of the clones. If the mutant clone is expanding, is there increased proliferation at the progressing front of the clone? Is there a visible effect on the neighbouring wild-type clones – are the wild-type cells lost through differentiation or do they undergo apoptosis?

Other signalling pathways in selfish spermatogonial selection

To date, all documented examples of mutations that accumulate through selfish spermatogonial selection involve genes encoding members of the tyrosine kinase receptor/RAS/MAPK pathway, but in theory any functional mutation in a gene expressed in spermatogonia may also confer a selective advantage. As detailed above, PI3K/AKT signalling plays a well-established role in SSC self-renewal (Lee et al., 2007; Oatley et al., 2007; Hasegawa et al., 2013) and immunopositive spermatogonia are positive for pAKT (Lim et al., 2012). De novo germline (Riviere et al., 2012) and post-zygotic mutations in the PI3K/AKT pathway cause overgrowth disorders (Kurek et al., 2012; Lee et al., 2012; Orloff et al., 2013; Rios et al., 2013) and somatic mutations are associated with cancer (Carpten et al., 2007; Miron et al., 2010; Flatley et al., 2013). Therefore, one may predict that mutations activating this pathway in spermatogonia could lead to selfish spermatogonial selection, potentially associated with embryonic lethality. It is also predicted that mutations that are less activating than those associated with PAE disorders, may lead to a more modest selective advantage. Such mutations could contribute to complex disease or cancer predisposition (Goriely & Wilkie, 2012; Goriely et al., 2013).

Conclusions and future directions

Our study (Lim et al., 2012) suggests that clusters of immunopositive spermatogonia are relatively common phenomena in aged testes (constituting up to 5% of seminiferous tubules in thin sections). With today's understanding of selfish spermatogonial selection, we propose that these spermatogonial clusters represent the clonal expansion of mutant selfish cells, based on their morphology, expression of relevant markers, and their size and distribution. As younger men also produce spermatozoa with PAE mutations (Goriely et al., 2003, 2009; Yoon et al., 2009; Giannoulatou et al., 2013) and have foci of mutations in their testes (Qin et al., 2007; Choi et al., 2008, 2012; Yoon et al., 2013), immunopositive tubules should also be apparent in younger men. The lack of reporting of such clusters of spermatogonia may be attributed to a number of reasons: (i) the limited studies of ‘normal’ (i.e. fertile, non-cancerous) testes; (ii) selfish clones are present in isolated regions of testes and are expected to be less common and smaller in younger men – screening few sections is less likely to hit upon a clone; (iii) if detected, multiple layering of spermatogonia may have been disregarded as artefactual staining [in some instances MAGEA4 is weakly expressed in spermatocytes (Aubry et al., 2001)]. Of note, the latter explanation was disproven by staining with other markers of spermatogonia (Lim et al., 2012).

We have highlighted that although selfish clones are predicted to be common, there may be marked regional differences in their distribution. However, by screening a reasonable number of testicular sections, and with use of robust spermatogonial markers, immunopositive clones should be readily apparent in human testicular samples. As our initial study was limited to six testes (in three of which immunopositive tubules were detected), documentation of further immunopositive tubules will be important to put these findings into a broader context. Are all immunopositive tubules positive for FGFR3 and pAKT? Is the lack of differentiation, seen in one of our samples (Fig. 3B) and also reported by Holstein, a common phenomenon and is it caused by a shift in the proliferation/differentiation balance or oncogene-induced senescence (Ota et al., 2009; Krejci et al., 2010)?

To confirm conclusively that immunopositive tubules are the cellular manifestation of selfish spermatogonial selection and the source of PAE mutations observed in sperm DNA, direct evidence of the molecular nature of these altered appearances will be required. This may be achieved using laser capture micro-dissection followed by DNA amplification and sequencing. If confirmed, this approach could facilitate the testing of many hypotheses. Are tubules with more subtle appearances associated with milder mutations that cause less severe disorders or contribute to complex disease? Do tubules that lack differentiating cells harbour stronger activating mutations and does this explain why some mutations associated with SpS (e.g. HRAS p.G13R and p.Q61R/K) have not been reported as germline mutations (Goriely et al., 2009; Giannoulatou et al., 2013)? Or, in the case of highly activating mutations, are spermatozoa indeed produced, but the mutations are incompatible with foetal development? For example, FGFR3 p.K650E causes the neonatal lethal disorder thanatophoric dysplasia type II – could even stronger activating mutations cause embryonic lethal disorders? The isolated tubules could also provide a template to test molecularly whether mutations of genes in other pathways play a role in selfish spermatogonial selection.

Detailing of the antigenic profile and mutation status of the selfish clones will not only reveal insights into the biology of selfish spermatogonial selection but may also add to the limited knowledge of germ cell homeostasis in humans. We have estimated that the individual immunopositive clones may extend over long distances – whole-mount staining of seminiferous tubules may reveal the true size and distribution of these clones as well as the progression front of normal and mutant clones.

In developed countries, there has been a trend for parents to have children later in life; in the UK today, more than one-third of children are born to fathers aged 35 years or over (England and Wales Birth Census data series FM1). In addition to PAE disorders, there have been numerous other associations between older paternal age and increased risk of offspring being affected by disorders such as autism (Grether et al., 2009), schizophrenia (Frans et al., 2011) and cancer (Yip et al., 2006) (reviewed in Goriely et al., 2013). However, the sperm DNA and whole testis mutation studies have been limited to a few specific genomic sites located in PAE genes, owing to the major technical challenges of detecting low-frequency mutations. Isolation of the cellular source of these mutations, which we propose are immunopositive tubules, will provide a DNA template more suitable for the detection of milder mutations present at lower frequencies, as well as the possibility of testing candidate genes for novel mutations not currently associated with selfish spermatogonial selection. This will broadly expand current knowledge of the impact that advancing paternal age will have on the genomes of our offspring.


We thank the Wellcome Trust for funding this study (Programme grant 091182 to A.G. and A.O.M.W.).