Main

Mutations of the X-chromosomal AR gene cause a wide spectrum of defective masculinization of genetic male (46,XY) individuals because of androgen insensitivity (1–3). Depending on the severity of the AR defect, AIS extends from a predominantly male appearance, through incomplete masculinization with ambiguous genitalia, to complete AIS with normal female external genitalia (4, 5). Difficulties in the clinical management are often based on considerable genotype-phenotype variabilities in this disease (6–9).

Recently, we found that somatic mosaicism caused by postzygotic mutational events can be an important genetic determinant of in vivo androgen action in AIS caused by expression of the wild-type AR (10). Moreover, de novo mutations and especially somatic de novo mutations of the AR gene occur at a particularly high rate (11). Previously, somatic mosaicism has also been detected in a mother of two affected children (12). Nevertheless, only very few data exist concerning the clinical relevance of this obviously underestimated genetic condition in AIS.

We herein present a detailed study on the clinical and molecular spectrum of somatic mosaicism in this disease. To enable a relevant overview, we considered all five mosaic AIS patients whom we have identified since 1993. Mutation detection in four of these patients has previously been reported (10, 11). Molecular identification of a fifth mosaic subject has been included in the current study. We particularly focused on the following comparative analyses of different aspects of AR function:1) the actual AIS phenotypes, 2) the expected phenotypes in case of inherited germline mutations as described in the literature, 3) the serum SHBG response to stanozolol as a biochemical measure of in vivo androgen action (13), and 4) the transactivation activities of the mutant ARs in cotransfection experiments reflecting the function of the mutant alleles. The current study demonstrates the particular clinical relevance of somatic mosaicism in AIS as a potent genetic determinant of the phenotype of affected individuals.

METHODS

The study was approved by the local ethical committee.

Patients

Patient 1.

Family history was inconspicuous. Physical examination revealed a 23-y-old woman, with predominantly female external genitalia, clitoral hypertrophy (20 mm), and pubic hair Tanner stage 4. The patient had bilateral axillary hair, breast development Tanner stage 5, and a normal female voice. Her karyotype is 46,XY [Fig. 1, see also (10)].

Figure 1
figure 1

Phenotypes of patients 1–5, phenotype classification (13), and somatic mutations including their localization within the AR gene.

Full size image

Patient 2.

Family history was inconspicuous. Physical examination revealed a 6-mo-old infant, with ambiguous external genitalia with bifid scrotum, very thin and short phallus (15 mm), large foreskin, severe perineoscrotal hypospadias, and bilateral inguinal masses. Gender assignment was male. His karyotype is 46,XY (Fig. 1).

Patient 3.

Family history was inconspicuous. Physical examination revealed a newborn with ambiguous external genitalia, small phallus and pronounced labia majora, and urogenital sinus with blind, short vagina. Gender assignment was female. Her karyotype is 46,XY [Fig. 1, see also (14)].

Patient 4.

Family history was inconspicuous. Physical examination revealed an 11-y-old girl with completely female external genitalia. Her karyotype is 46,XY (Fig. 1).

Patient 5.

Family history was inconspicuous. Physical examination revealed a newborn with ambiguous external genitalia characterized by clitoris-like micropenis, penoscrotal hypospadias, and bifid scrotum with palpable gonads. Gender assignment was female. Her karyotype is 46,XY [Fig. 1, see also (15)].

Serum concentrations of testosterone and dihydrotestosterone as well as LH were normal for age in patients 2, 3, and 5. Elevated serum levels for LH and testosterone had been detected in patient 1 (10, 14, 15). In patient 4, serum LH was 5.3 U/L, and serum testosterone was 7.5 nM at the age of almost 12 y just before gonadectomy.

Phenotype classification and SHBG androgen sensitivity test

The AIS phenotypes of patients 1–5 were classified according to a previously developed scheme (13). Only the major groups 1 to 5, which reflect increasing virilization deficit because of increasing in vivo AR malfunction (1 = male, 2 = predominantly male, 3 = ambiguous, 4 = predominantly female, 5 = female external genitalia) have been considered for correlation with the SHBG androgen sensitivity test.

The standardized SHBG androgen sensitivity test has been performed in patients 1, 2, and 5 as previously described (13). In brief, the anabolic steroid stanozolol (17β-hydroxy-17α-methyl-5α-androstano-[3, 2-c]pyrazol) was administered orally for 3 consecutive days (0.2 mg/kg per day in a single evening dose). Serum SHBG concentration was determined once before application of stanozolol (d 0) and compared with the lowest level at d 5, 6, 7, or 8. Results were expressed as a percentage of the initial level. The SHBG responses were compared with the expected response as published previously, for a total of 23 AIS patients having germline mutations (13). Patient 3 was not available for functional tests. The parents of patient 4 did not consent to the SHBG androgen sensitivity test.

DNA analyses

Mutation detection in patients 1, 3, 4, and 5 has previously been reported (10, 11). Molecular diagnoses of patient 2, his sister, mother, and a male control have been included in this study. In the latter three subjects, DNA analyses were based on genomic DNA extracted from peripheral blood leukocytes derived from a single blood sample. Studies on patient 2 were performed on genomic DNA from peripheral blood leukocytes derived from two independent blood samples as well as from cultured genital skin fibroblasts. Verification of DNA sequence alteration in exon 3 of this patient has been performed by Hha I restriction site analysis on PCR-amplified DNA fragments containing the locus of the mutation.

Culturing conditions of cells

Genital skin fibroblasts of patients 2 and 4 were obtained during reconstructive surgery of the external genitalia (patient 2) or during gonadectomy (patient 4). Male control fibroblasts were derived from foreskin specimens. They were cultured at 37°C with 5% CO2 in MEM (Life Technologies, Grand Island, NY), supplemented with 10% FCS, 1% (vol/vol) MEM non-essential amino acids (Life Technologies), and penicillin (200 IE/mL)/streptomycin (0.2 mg/mL). CHO cells and COS1 cells were maintained at 37°C and 5% CO2 as well. They were cultured in DMEM with the nutrient mix F12 (GIBCO, Karlsruhe, Germany), 10% (vol/vol) FCS, and antibiotics as described above. For transactivation studies 10% dextran/charcoal-treated FCS was used.

Androgen binding studies

Androgen binding properties of cultured genital skin fibroblasts of patients 1 and 5 have previously been reported (10, 15). Additional binding analyses on genital skin fibroblasts of a male control subject and patients 2 and 4 were performed in this study as previously described (15). In brief, cultured genital skin fibroblasts were incubated in duplicate with increasing concentrations of the synthetic androgen R1881 (DuPont-New England Nuclear, Boston, MA) (patient 2, 0.02–3.0 nM; control subject and patient 4, 0.02–15 nM) in either the presence or the absence of a 200-fold molar excess of the unlabeled steroid. For Scatchard calculations and statistics, Microsoft Excel software (Microsoft Corp., Richmond, WA) was used. Binding parameters were either calculated (normal control subject and patient 2) or graphically deduced from the Scatchard plot (patient 4) as previously described (15) based on publications by Feldman (16) and Clark et al. (17). Binding data of the control subject and patient 4 were verified in a second independent experiment.

AR expression plasmids and transfection experiments

Only the mutant AR alleles of patients 3 and 4 were functionally compared by cotransfection analyses in this study. The His733 (patient 3) and the Cys774 (patient 4) point-mutated AR were created using the QuikChange site-directed mutagenesis kit (Stratagene Cloning Systems, La Jolla, CA) according to the manufacturer, on the basis of the wild-type AR expression plasmid pSVAR0 (15, 18). Mutagenic primers were His733: sense, 5′-CAC-GTG-GAC-GAC-CAT-ATG-GCT-GTC-ATT-CAG-3′; antisense, 5′-CTG-AAT-GAC-AGC-CAT-ATG-GTC-GTC-CAC-GTG-3′; Cys774: sense, 5′-G-GTT-TTC-AAT-GAG-TAC-TGC-ATG-CAC-AAG-TCC-CGG-3′; antisense, 5′-CCG-GGA-CTT-GTG-CAT-GCA-GTA-CTC-ATT-GAA-AAC-C-3′; Plasmid constructs were verified by plasmid sequencing as previously described (10).

CHO cells were cotransfected using the Ca2+-phosphate precipitation method (19) with minor changes (10). A mixture of 50 ng/mL precipitate of either mutant (His733, Cys774) or wild-type AR expression plasmids, 2 μg/mL of androgen-responsive MMTV luciferase reporter plasmid (Organon, Oss, the Netherlands) (10), 10 ng/mL of constitutively expressed pRL-SV40 Renilla luciferase plasmid (Promega Corp., Madison, WI), adjusted to a final amount of 20 μg/mL plasmid DNA with pTZ19 carrier plasmid, was used. Cells were incubated either without hormone or with 0.01, 0.1, 1.0, 5.0, and 10.0 nM of R1881. All transfections were performed in triplicate. Three independent experiments were performed. All three constructs were assayed at the same time to enable comparability of the results. Firefly luciferase counts were corrected for transfection efficiency by Renilla luciferase activity, determined in both by the dual luciferase reporter gene assay (Promega). In addition, all plasmid constructs were transfected into COS1 cells verifying AR protein expression by Western immunoblot analysis (data not presented).

RESULTS

Classification of AIS phenotypes and SHBG androgen sensitivitytests.

AIS phenotypes were attributed to the five major groups of a previously developed classification scheme (13) (Fig. 1). In patients 1, 2, and 5, SHBG androgen sensitivity tests have been performed. All results were in marked discrepancy to the corresponding phenotypes (Fig. 2). In patient 1, basal SHBG concentration was 27.8 nM (normal for age). In response to stanozolol, SHBG increased slightly to 31.3 nM (112.7% of initial value). This finding is usually associated with complete AIS (13). In patient 2, basal SHBG was 70.9 nM (normal for age); the poststanozolol SHBG concentration decreased to 32.1 nM (45.3% of initial value). Basal SHBG concentration of patient 5 was 101.7 nM (normal for age). In response to stanozolol, SHBG decreased to 56.6 nM (55.7% of initial value). The SHBG decline found in patients 2 and 5 would normally suggest uninhibited androgen sensitivity or at best just slightly defective androgen action in vivo (13).

Figure 2
figure 2

SHBG decrease in response to stanozolol expressed as percentage of initial SHBG concentration. Phenotypes were classified according to Sinnecker et al. (13). In brief: 2 = predominantly male, 3 = ambiguous, 4 = predominantly female, 5 = female. Controls had no androgen insensitivity. Circles represent single AIS patients with germline mutations as previously published (13), lines represent the median, and gray boxes indicate the range within each AIS phenotype group. Squares represent data from AIS patients 1, 2, and 5 of this study having somatic mutations of the AR gene.

Full size image

Mutation detection analyses.

Mutation detection in patients 1, 3, 4, and 5 has previously been described (10, 11). In patient 2, a 596 GCC(Ala)/ACC(Thr) somatic mutation of the AR gene was found. This mutation predicted for abolishment of a Hha I restriction site. Partial Hha I digestion of PCR fragments from this patient containing the mutation, based on genomic DNA from peripheral blood leukocytes from two independent blood samples as well as from cultured genital skin fibroblasts, demonstrated the coexistence of mutant and wild-type AR DNA sequences, indicating postzygotic origin of the mutation leading to somatic mosaicism (Fig. 3). Complete digestion proved the presence of only the wild-type AR DNA sequence in the mother, the sister, and the male control (Fig. 3). Fig. 1 summarizes the somatic mutations, their localization within the AR gene, and the predicted amino acid changes resulting from the mutant alleles of all five patients.

Figure 3
figure 3

Hha I restriction recognition site analysis of a 177-bp genomic AR DNA PCR product of exon 3 leading to a 117-bp and a 60-bp fragment. Lane 1, marker;lane 2, patient, first genomic DNA sample derived from peripheral blood leukocytes, not digested;lane 3, patient, the same sample as in lane 2, Hha I digest;lane 4, patient, second independent genomic DNA sample derived from peripheral blood leukocytes, not digested;lane 5, patient, the same sample as in lane 4, Hha I digest;lane 6, patient, genomic DNA derived from cultured genital skin fibroblasts, not digested;lane 7, patient, the same sample as in lane 6, Hha I digest;lane 8, sister, genomic DNA derived from peripheral blood leukocytes, not digested;lane 9, sister, the same sample as in lane 8, Hha I digest;lane 10, mother, genomic DNA derived from peripheral blood leukocytes, not digested;lane 11, mother, the same sample as in lane 10, Hha I digest;lane 12, male control, genomic DNA derived from peripheral blood leukocytes, not digested;lane 13, male control, the same sample as in lane 12, Hha I digest. Partial digestion of the PCR products derived from genomic DNA of the patient indicates the coexistence of mutant and wild-type AR DNA sequences (lanes 3, 5, and 7). Open arrows indicate the weak formation of the expected 117 bp fragment. The absence of visible 60-bp fragments in lanes 3, 5, and 7 compared with lanes 9, 11, and 13 is most probably because of the low amount of wild-type AR DNA in the genomic DNA of the patient.

Full size image

Androgen binding studies.

Scatchard analysis on R1881-binding on genital skin fibroblasts of patient 2 revealed normal androgen binding properties (kd: 0.08 nM, Bmax: 42.84 fmol/mg protein).

Because an extended range of androgen concentration in the Scatchard analysis (0.02 to 15 nM R1881) may resolve a two-component binding system caused by a somatic mutation of the AR within the ligand binding domain (15), we investigated the genital skin fibroblasts of patient 4 in this way as well (Fig. 4). The occurrence of a hyperbola indicated the expression of two functionally different ARs (16, 17). Two independent experiments (A and B) revealed the following results: A:kd1: 0.03 nM, Bmax1: 2.1 fmol/mg protein;kd2: 8.5 nM, Bmax2: 29.3 fmol/mg protein; B:kd1: 0.02 nM, Bmax1: 1.0 fmol/mg protein;kd2: 5.6 nM, Bmax2: 24.0 fmol/mg protein. Binding data of the first binding component demonstrate very low expression of the wild-type AR, supporting the somatic mosaicism. Binding properties of the second binding component are well in accordance with the markedly elevated dissociation constant of the Cys774AR as characterized previously by detailed in vitro studies by Marcelli et al. (20). The cause for the apparent discrepancy between detectable (our results) and not detectable (20, 21) specific androgen binding within genital skin fibroblasts in association with the 774 CGC(Arg)/TGC (Cys) mutation has not been investigated in this study.

Figure 4
figure 4

Scatchard analyses on R1881 binding in cultured genital skin fibroblasts of a male control subject (left) and patient 4 (right).

Full size image

The following binding parameters were calculated from two independent experiments (C and D) of the male control subject: C:kd: 0.15 nM, Bmax: 31.9 fmol/mg protein; D:kd: 0.14 nM, Bmax: 36.0 fmol/mg protein. Elevation of R1881 concentration greater than 3.0 nM did not show additional specific androgen binding because of saturation of wild-type AR binding sites (Fig. 4).

Transient transfection experiments.

No significant differences of expression levels caused by the different ARs could be observed (data not shown). In reporter gene assays, the His733AR (patient 3) and the Cys774AR (patient 4) revealed almost completely abolished induction of the MMTV luciferase reporter gene over the whole range of hormone concentrations (Fig. 5).

Figure 5
figure 5

Relative induction of the androgen-responsive MMTV luciferase reporter gene in the presence of 0.01–10 nM R1881 related to basal activity in the absence of hormone. Comparisons of the wild-type AR (pSVAR0), the His733AR (patient 3), and the Cys774AR (patient 4) are shown.

Full size image

DISCUSSION

Postzygotic de novo mutations of the AR gene occur at the expected high rate of an X-linked recessive mutation in AIS (11). Although the resulting somatic mosaicism of mutant and wild-type AR alleles can be an important molecular determinant of in vivo androgen action (10), only very little is known about its clinical relevance. Therefore, we investigated in detail the genotype-phenotype relationship in mosaic AIS individuals, considering different aspects of AR function in vivo and in vitro.

Patients 1, 3, and 5 displayed characteristic genotype-phenotype discrepancies. Patient 1 showed obvious symptoms of external virilization (AIS type 4) despite the presence of a premature stop codon of the AR gene (10). This type of mutation has usually been associated with complete loss of in vivo androgen action (5, 22). As expected, the Stop172 AR construct was inactive in cotransfections (10).

A similar constellation has been observed in patient 5. This patient also showed unexpectedly strong virilization (AIS type 3) taking into account that the corresponding inherited mutation has usually been associated with complete AIS (23, 24) or at best incomplete testicular feminization (25). Correspondingly, cotransfection studies demonstrated a marked functional defect of the Met866AR (15, 23, 24).

The mutation 733 CAG(Gln)→CAT(His) of patient 3 has not been published previously as a germline mutation by other groups. For this reason, the virilization of this patient (AIS type 3) could not be compared with the virilization expected in case of a germline mutation. Cotransfections demonstrated an almost complete loss of function of the His733AR, also in the presence of 10 nM R1881, which is comparable to a supraphysiologic dihydrotestosterone concentration. Therefore, we assume that severe if not complete androgen resistance would have occurred in this case.

In patients 2 and 4, no obvious genotype-phenotype discrepancies have occurred. Patient 4 has presented with a complete female phenotype (AIS type 5). Her phenotype corresponded well with the underlying cotransfection analyses, which displayed an almost complete absence of transactivation activity because of the Cys774AR (Fig. 5). Accordingly, complete AIS has been exclusively reported in patients having the 774 CGC(Arg)→TGC(Cys) AR gene mutation (20, 21, 23). Obviously, the very low expression level of the wild-type AR as detected by Scatchard analyses on genital skin fibroblasts of patient 4 (Fig. 4) was insufficient in inducing virilization of the external genitalia during fetal life. However, interpretation of the binding data should be restrained, because the ratio of fibroblasts expressing either the mutant or the wild-type AR in cell culture does not necessarily reflect the conditions in vivo. Nevertheless, these data do not rule out with adequate certainty that increasing androgens during puberty might still lead to undesired virilization caused by activation of the wild-type AR. This could result in a phenotype similar to patient 1, characterized by marked clitoromegaly, or in other symptoms of unwanted androgen action. For this reason, gonadectomy has been performed in patient 4 recently at the age of 12 y. We propose that all AIS patients with somatic mosaicism of an AR gene mutation who are reared as females should be gonadectomized before the onset of puberty to prevent undesired virilization.

Patient 2 was the second patient whose (prepubertal) phenotype (AIS type 2) displayed no obvious discrepancy compared with the function of the mutant AR allele (26). In the literature, the germline mutation 596 GCC(Ala)→ACC(Thr) has been described in association with marked in vivo AR function (Reifenstein syndrome), normal androgen binding, also supported by our Scatchard analysis on genital skin fibroblasts of patient 5, and with considerable transactivation activity in vitro (26). We believe that these data, and the presence of a somatic mosaicism, allow us to assume that there is sufficient in vivo androgen action to respond favorably to possible future androgen treatment. Thus, in contrast to patients 3 and 5, patient 2 is reared as a boy.

Summarizing the presented genotype-phenotype correlations, three of five mosaic patients showed markedly stronger virilization than expected from the function of the mutant AR alleles in cotransfection studies, or from data concerning the same or functionally equivalent germline mutations. In two of our five patients, final assessment of the phenotypes that would have occurred naturally if no surgical intervention had been performed was uncertain. Neither the appearance nor the extent of pubertal virilization caused by the somatic mosaicism could be predicted. Our findings, therefore, indicate that mosaic AIS patients should be treated differently from patients with germline AR mutations. All clinical decisions should consider the possibility of relevant androgen action caused by activation of the wild-type AR within a proportion of the somatic cells (10, 15). Our data underline the particular importance of accurate determination of the mode of inheritance in all AIS patients who have de novo mutations of the AR gene. Somatic mosaicism should be specifically excluded. Besides, these data provide essential information for conclusive genetic counseling as previously discussed in more detail (11, 12).

A surprising result of our study was the failure of the SHBG androgen sensitivity test in all three mosaic patients who were tested. In patient 1, poststanozolol SHBG concentration was unexpectedly high, usually suggesting complete AIS. In patients 2 and 5, SHBG response was in the range commonly expected for completely normal or at most just slightly impaired in vivo AR function (13). The molecular background for these discrepancies could be a different ratio of mutant to wild-type AR alleles in the liver compared with the urogenital mesenchyme based on the mosaic composition of the tissues. Hence, these data indicate that the SHBG androgen sensitivity test is unsuitable for the assessment of mosaic AIS subjects. However, apparent discrepancies in the SHBG response compared with the actual phenotype can provide helpful information, already suggesting somatic mosaicism before completion of the final molecular diagnosis.

The phenotypic modification caused by somatic mosaicism in AIS as demonstrated in this study is well in accordance with observations on different other X-chromosomal genes [(27) and references therein]. It suggests that expression of wild-type gene products based on somatic mosaicism could represent a conceivable mechanism explaining in part the phenotypic variability observed in other X-linked diseases like Duchenne muscular dystrophy (28) or hemophilia A (29) as well.