A Unique Exonic Splicing Mutation in the Human Androgen Receptor Gene Indicates a Physiologic Relevance of Regular Androgen Receptor Transcript Variants¹

In a patient with partial androgen insensitivity syndrome (AIS), we identified a single inherited presumably silent nucleotide variation (AGC -> AGT) in exon 8 (codon 888) of the AR gene. However, in the patient’s genital skin fibroblasts, a considerably shortened transcript of 5.5 kb (normal: 10.5 kb) was detected, which misses a part of exon 8 and a prominent portion of the 3′-untranslated region. The translation product includes eight missense amino acids from codon 886 onward followed by a premature stop codon. As shown by in vitro expression analysis, the mutant protein lacks any residual function. However, reverse transcribed PCRs and sequence data indicate the existence of two additional splicing variants of 6.4 kb and 7.8-kb length both in patient and normal control genital skin fibroblasts. These splicing variants comprise the complete coding region but a shortened 3′-untranslated region. Thus, a distinct alternative pre-messegner RNA-processing event leading to two additional transcripts occurs generally in genital skin fibroblasts. In addition, this process partially prevents aberrant splicing in the patient and produces a small fraction of normal, functionally intact AR-protein that could explain the partial masculinization in this patient.

This first report of an exonic splicing mutation in the AR-gene indicates a physiologic relevance of the regular AR-messenger RNA variants with shortened 3′-untranslated regions and their functional translation products in human genital development.

NORMAL SEXUAL DIFFERENTIATION in males is induced by androgen action. Signaling of androgenic steroids in their target cells is enabled by the androgen receptor (AR), a phosphoprotein that complexes these ligands, homodimerizes, and finally acts as a transcription factor on genes, which induce virilization (1–3). The AR is encoded by an X-chromosomal gene divided into eight exons (4). The amino acid sequence encoded by exon 1 is partly involved in transactivation of androgen target genes, the region encoded by exon 2 and 3 enables DNA-binding, and the region encoded by exon 4 to 8 is involved in ligand binding (5).

Androgen insensitivity syndrome (AIS) is a common cause for virilization disorders in patients with 46, XY karyotype. The clinical spectrum of this disorder reaches from infertility in males to a completely female phenotype (6). AIS is based on the inability of androgen-dependent target tissues to react to androgens. Most often, this is due to mutations within the AR-gene (7–11). In most cases one defined point mutation causing a single amino acid exchange or a premature termination induces defective DNA- or ligand-binding of the AR (11).

Previously, in LNCaP cells derived from prostatic tissue and in prostatic tissue itself, the existence of a shortened AR-transcript variant in addition to the full-length AR- messenger RNA (mRNA) has been reported (4, 12, 13). This seems to be the result of a variable splicing event within the 3′-untranslated region (3′-UTR). To this date, it is unknown whether transcript variants exist in other androgen target tissues and are translated to functional AR-proteins in physiologically relevant amounts to influence the male phenotype.

We present a patient with ambiguous genitalia bearing a novel germline point mutation in exon 8 of the AR-gene. In this context, we demonstrate the existence of two AR-transcript variants found in normal control and patient genital skin fibroblasts (GSF) and discuss their physiological activity and possible influence on the phenotype of our patient.

Materials and Methods

Patient

The propositus with a 46,XY karyotype and partial AIS [type 2b according to the classification of Sinnecker et al. (6), corresponding to type 3 according to Quigley et al. (3)] was studied at the age of 4 months. The individual suffers from perineal hypospadias (grade IV), scrotum bipartitum, undescended testes, and micropenis. Phallic enlargement and rugation of the labia maiora provide clear clinical signs for residual androgen action. Mullerian structures were not seen on ultrasound and genitography. The family history was uninformative. At this time, baseline values for luteinizing hormone (0.7 U/L) and testosterone (1.0 nmol/L) were prepubertally low. After stimulation with hCG (1000 IU daily for 5 days, im), serum-testosterone concentration rose markedly to a value of 36.4 nmol/L (normal > 10 nmol/L), excluding a testosterone biosynthesis defect and suggesting androgen insensitivity (14). A genital skin biopsy was performed; at this time genital skin tissue was preserved for cell culture. Informed consent for detailed studies was given by the mother.

Mutation analysis of the AR-gene

Genomic DNA from the patient and the mother was isolated from peripheral blood leukocytes by standard procedures. The whole coding region of the AR gene including all exon/intron boundaries was amplified by PCR in 14 segments using primers derived from published sequences (15); then each fragment was tested for sequence variations by single-strand conformation analysis (SSCA) as previously described (7, 16). Briefly, PCR products were heat denatured, electrophoresed on nondenaturating polyacrylamide (PAA) gels, and silver stained. Single strands from PCR products that showed aberrant migration, compared with normal controls, were cycle sequenced using the thermosequenase fluorescent labeled primer cycle sequencing kit with 7-deaza-dGTP (Amersham Pharmacia Biotech, Freiburg, Germany) and analyzed by an ALF Express automated sequencer (Amersham Pharmacia Biotech) according to the manufacturer’s instructions.

Cell culture

Biopsy specimens from the patient’s genital skin were dissected mechanically and incubated in medium (DMEM-F12, 5% CO₂, 10% charcoal-stripped FCS, antibiotics) at 37 C to grow fibroblasts. GSF obtained from the foreskin of a normal prepubertal boy served as a control. Only cells of the second and the third passage were used for AR-expression studies.

RNA analysis, RT-PCRs, and Northern blots

Whole RNA from cultured GSF from the patient and the normal control was isolated using Rneasy columns as indicated by the manufacturer (QIAGEN, Hilden, Germany). RNA was first quantified photometrically (absorbency at 260 nm measured in a DNA/RNA calculator from Amersham Pharmacia Biotech). Then 5-μL samples were electrophoresed and stained with ethidiumbromide on formaldehyde-denaturing 1% agarose gels to determine quality and integrity of RNA and to test quantification results. If gels were planned to be blotted and used for hybridization experiments (Northern blots; see below), 4-μg RNA per lane were loaded.

Extra long (XL) RT-PCRs for amplicons of 2 kb to 7 kb length and RT-PCRs for amplicons smaller than 1 kb were performed on 1-μg whole-RNA samples. In both cases, RT was made by specific antisense priming (primers see Table 1) with Superscript II reverse transcriptase (Life Technologies, Inc., Eggenstein, Germany) following the manufacturer’s protocols.

TABLE 1.

Primers employed for RT PCR amplification

Primer	Sequence and 5′ → 3′ annealing position (Ref.)	Used for1
HARE1s	5′-TGG ATG GAT AGC TAC TCC GG-3′	Competitive RT PCRs
	nucleotides 1654–1673 (15 )	RT PCRs
HARE6s	5′-TAC CGC ATG CAC AAG TCC CGG-3′	XL RT PCRs
	nucleotides 3348–3368 (15 )
HARE7s	5′-CTC ACC AAG CTC CTG GAC TC-3′	XL RT PCRs
	nucleotides 3917–3937 (15 )
HARE4a	5′-ACT ACA CCT GGC TCA ATG GC-3′	Competitive RT PCRs
	nucleotides 2845–2826 (15 )	Specific RT-priming
hARE8a	5′-GAG GAG TAG TGC AGA GTT ATA A-3′	RT PCRs
	nucleotides 4287–4266 (15 )	Specific RT-priming
hAR3′UTRa	5′-CAG AAC ACT AGC GCT TGG AG-3′	XL RT PCRs
	nucleotide 10333–10313 (13 )	Specific XL RT-priming

Information in this column can also be found in text and in Fig. 2.

¹

Corresponding nucleotide positions are indicated as in the published AR-gene- or -transcript sequences. Note that the primer hAR3′UTRa binds 43 bases upstream of the first polyadenylation site of the AR transcription unit.

Fig. 2.

Effects of the patient’s mutation on the AR-transcript. A, Northern blots on each 4-μg whole RNA from patient and normal control GSF hybridized to an AR-RNA probe. Bold-typed kilobase pair values indicate AR-transcript lengths. Note that the patient’s AR-transcript is shorter than wild-type AR-mRNA. Signals at 4.9 kb most likely denote AR-mRNA degradation products. B, Relative transcript amounts in patient and normal control GSF determined by competitive RT-PCRs. A(t/s) values per microgram whole RNA from patient (black curve) and control (gray curve) are depicted logarithmically as a function of the applied competitor amount (see also text). Boxed, italic values under the graph represent interpolated competitor concentrations needed to achieve balanced target-to-standard optical densities (A[t/s] = 1); they give information about relative transcript amounts. C, RT-PCR products on RNA from patient and control GSF obtained with primers hARE1s and hARE8a (spanning all exon boundaries). The patient and the normal control amplificates are of equal length. Top bands of the molecular weight marker (φ 174; HaeIII -digest) are 1353 and 1078 bp long. D, XL RT-PCR products on RNA from patient and control GSF obtained with primers hARE7s and 3′UTRa (comprising the major part of the 3′UTR). Boldface kb values represent specific products. Note that the patient’s material lacks the full-length amplificate but shares two shortened amplificates with the control and displays one additional band at 1.7 kb. The band sizes of the molecular weight marker (1-kb ladder) are: 1.64, 2.04, 3.05, 4.07, 5.09, 6.11, 7.13[… ] kb.

For characterization of the AR coding region, fragments smaller than 1 kb were amplified by RT-PCRs with Ampli-TAq DNA-polymerase (Perkin-Elmer Corp., Weiterstadt, Germany) using the primers indicated in Fig. 1 and Table 1. Cycling conditions and PCR-solution compositions were as described (10, 17). RT-PCR products were electrophoresed on 2% agarose gels, stained with ethidiumbromide, and evaluated on a standard UV plate. Semiquantification of AR transcription was achieved by competitive RT-PCRs as previously published (10, 17). Briefly, whole RNA from patient and control GSF was first standardized for ubiquitous ribosomal protein L7 transcription (semiquantitative RT-PCR with modified primers) (18). Then RNA samples were mixed with serially diluted concentrations of artificial RNA-standard (shortened target) and submitted to RT-PCRs using the primers hARE1s and hARE4a (Table 1). After electrophoresis on nondenaturating PAA gels and silver staining, optical densities of PCR product signals (target: 479 bp; standard: 324 bp) were evaluated by computerized densitometry (ImageMaster, Pharmacia). Signal optical densities of target (t) and standard (s) in each sample were compared, resulting in dimensionless numbers used as relative concentration equivalents of the transcript: A(t/s).

Fig. 1.

Complete AR transcription unit (10.5 kb) with enlarged open reading frame (amino acid [aa]-coding region), primer annealing areas, and expected (XL) RT-PCR product lengths. Primer sequences are given in Table 1.

To investigate long complementary DNA (cDNA) fragments enclosing the exon 8/3′UTR link, we employed XL RT-PCRs. Amplicons were between 2 kb and 7 kb long; for primer sequences and locations, see Fig. 1 and Table 1. For this purpose, the GeneAmp XL PCR kit (Perkin-Elmer Corp.) was used according to the producer’s recommendations. Initial examination of XL RT-PCR products was made after electrophoresis on 1% agarose gels and ethidiumbromide staining on a UV plate. Before sequencing, products were isolated from the gel and finally purified by QuiaQuick columns (QIAGEN) twice.

Northern blots and probe generation were performed as previously published (10, 17). In brief, digoxigenin-labeled AR-RNA probes were generated by RT-PCR with an antisense primer tailed by a T7-RNA-polymerase recognition sequence. After purification, the product was in vitro transcribed by T7-RNA-polymerase (Promega Corp., Heidelberg, Germany) using digoxigenin-labeled rNTPs (Roche Molecular Biochemicals, Mannheim, Germany) according to the manufacturer’s instructions. Finally DEPC-water was added to a final volume of 100 μL. Northern blot filters were prehybridized for 1–2 h and then hybridized with 10 mL Easy-hyb solution (Roche Molecular Biochemicals) and the probe (3μ L) at 68 C overnight. Stringent washings were performed at 68 C with 0.1 × SSC/0.1% SDS for 2 × 15 min. Chemoluminescence signals of hybridized probes were developed following the Digoxigenin users guide for Northern blots (Roche Molecular Biochemicals).

Western immunoblots

Western blots for AR proteins expressed in the patient and the normal control GSF strains were performed as described previously (19, 20). Specific anti-AR antibodies were a generous gift of Dr. A. O. Brinkmann (Erasmus University, Rotterdam, The Netherlands). In brief, GSF were first lysed to isolate AR proteins from whole-cell lysates by immunoprecipitation using the monoclonal anti-AR antibody F39. Subsequently, the samples were size fractionated by SDS-PAGE and blotted on cellulose-nitrate membranes. These were first incubated with the primary antibody SP061 directed against the amino acids 301–320 of the AR (19). Then a second incubation with an antirabbit-peroxidase conjugate was performed. Detection of AR proteins was accomplished by chemiluminescence.

Androgen-binding studies

Androgen-binding assays were performed according to previous protocols (20, 21). In brief, confluent GSF cultures from either the patient or the control subject were incubated with media containing various concentrations of (³H)R1881 (17β-hydroxy-17α-(³H)-methyl-4,6,11-oestrotrien-3-one (NEN Life Science Products, Boston, MA) in either the presence or absence of a 200-fold molar excess of unlabeled ligand. After 1 hour incubation at 37 C, aliquots of 50 μL culture medium were taken for measurements of total counts. Cells were washed, lysed, and submitted to liquid scintillation counting; 100-μL samples were reserved for protein determination. Results were evaluated by computerized Scatchard analysis.

AR expression plasmids

The sequence of the shortened AR-transcript specific for the patient’s GSF was introduced into the human AR expression vector pSVAR0 (21). First, RT-PCRs were applied on whole RNA from the patient GSF with the sense-primer hARE 1s (Table 1 and Fig. 1) and the antisense-primer hAR-BFa (5′-CCT CTT AAG GGA TCC AGG TCA CAA GTA CAT GGC ATC -3′). This primer is unique for a part of the aberrant sequence in the patient-specific AR-mRNA (in italics) and encloses the premature stop codon (underlined); the primer is tailed at the 5′-end by a sequence containing a BamHI recognition site (bold typed) for later cleavage. The PCR mixture contained 1 μL cDNA (first strand prepared for XL RT-PCRs), 200 μm dNTPs, 1× PCR-buffer (pH 9.0), 1.0 mm MgCl₂, 20pmole each sense and antisense primer and 1U Ampli-TAq DNA-polymerase (Perkin-Elmer Corp.). Cycling conditions were: 75-sec denaturation at 94 C, 90-sec primer annealing at 52 C, and 120-sec primer extension at 72 C for 35 cycles. After testing the success of the experiment on a 2% agarose gel, the RT-PCR product was immediately cloned into the PCR2.1 vector (TA cloning kit, Invitrogen, Leek, The Netherlands). After verifying the correct sequence by plasmid sequencing (Seqlab, Göttingen, Germany), PCR2.1 (comprising the patient-specific RT-PCR product as insert) and pSVAR0 were digested with BamHI and Asp1 (Roche Molecular Biochemicals) and subsequently purified from agarose gels. Ligation of the BamHI-Asp1 insert into the pSVAR0 vector was performed using T4-DNA-ligase according to the manufacturer’s instructions (Invitrogen). The ligation product was transformed in DH5α-Escherichia coli bacteria and single clones were purified by the QIAprep spin miniprep kit (QIAGEN). The correct plasmid sequence (further called pSVAR3′Del) was verified again.

Transient transfections

Either pSVAR0 or pSVAR3′Del were transiently transfected in eukaryotic chinese hamster ovary cells using the calcium phosphate precipitation method (23) with only minor changes as described previously (20). Activation of cotransfected androgen inducible MMTV-luciferase reporter gene was determined either in the absence of hormone or in the presence of 0.1-, 1.0-, and 10.0 nm dihydrotestosterone, respectively. All transfections were performed in triplicate, and three independent experiments were carried out. Transfection efficiency was determined by cotransfection of the constitutively expressed pRLSV40 Renilla luciferase plasmid (Promega Corp.). Firefly and Renilla luciferase activity were measured using the dual luciferase reporter gene assay (Promega Corp.).

Results

Analysis of the AR-gene

SSCA of PCR products amplifying exon 8 displayed an aberrant migration pattern on the PAA gel, compared with normal controls. Sequencing revealed a transition of cytosine to thymidine (AGC → AGT) in the third position of codon 888. The mutation has no influence on the predicted amino acid coding sequence itself as both triplets code for serine. The mother of the patient was found to be heterozygous carrier of this variation. By SSCA of the whole coding region and additional sequencing of exons 4 to 8 of the AR-gene, we could not detect any additional mutation. The polymorphic trinucleotid repeats within the first exon were determined to carry 21 CAG-repeats and 24 GGN-repeats.

Studies of the AR-transcript

By the Northern blot shown in Fig. 2A, we could prove the existence of an AR-transcript of approximately 5.5 kb length in the patient’s whole RNA. This is considerably shorter than the wild-type AR-transcript found in whole RNA from control GSF (10.5 kb). An additional band at 4.9 kb in patient and control RNA represents a commonly observed AR-mRNA degradation product (13). UV luminescence and ratio of ethidium-bromide stained 28S- and 18S-rRNA bands on the gel and L7 housekeeping gene transcription in both patient and control material were similar indicating equivalent RNA quantity and integrity. Then competitive RT-PCRs of an exon 1–4 amplicon of the AR-transcript (primers hARE1s and hARE4a; see Fig. 1 and Table 1) with 0.02-, 0.1-, and 0.5-attomole competitor/μg RNA were applied. As shown in Fig. 2B, calculated competitor amounts needed to obtain an A(t/s) value of 1.0 (indicating similar amounts of target and standard amplificates) were 0.06 and 0.18 attomole per μg RNA from the patient and the normal control, respectively. Comparison of these values revealed that the AR-transcript level in patient GSF amounts to approximately 30% of that in GSF from the equally aged normal control. RT-PCRs of an exon 1–8 amplicon of the AR-mRNA (primers hARE1s and hARE8a) including all exon boundaries and the mutation region (Fig. 1) yielded products of equal length in samples with normal and patient whole RNA as shown in Fig. 2C. By these RT-PCRs and experiments with other mutation-flanking primer pairs binding sequences within the coding region, no additional products could be demonstrated in patient material (not shown). XL RT-PCRs with primers flanking the mutation site and the major part of the 3′-UTR (primers hARE6s and hAR3′UTRa; Fig. 1) lead to the generation of three products (3.75 kb, 2.55 kb, and 1.95 kb) that are shorter than the wild-type amplicon (6.42 kb). To determine whether these products were specific, XL PCRs with primers hARE7s (hybridizing 264 bp downstream of hARE6s; Fig. 1) and hAR3′UTRa were performed on 10,000-fold diluted samples from gel-purified initial XL RT-PCR products. As demonstrated in Fig. 2D, the three products from the first XL RT-PCR must be derived from the AR-transcript because they are approximately 260 bp shorter (3.5 kb, 2.3 kb, and 1.7 kb). In the patient whole RNA, a full-length wild-type XL RT-PCR product (6.42/6.16 kb) could not be demonstrated. In contrast, on normal control RNA, a faint band specific for the full-length amplicon could be detected after a seminested XL RT-PCR with primers hARE7s and hAR3′UTRa. However, the specific 3.5-kb and 2.3-kb seminested XL PCR products found in patient material were also generated on normal control whole RNA (Fig. 2D). In contrast, the 1.7-kb amplificate could not be reproduced on normal material. Additional smears and faint bands (visible on the gel in Fig. 2D) were not mirrored by corresponding signals on electrophorated samples of the first XL RT-PCR, indicating those signals to be unspecific.

Finally, exon 8 sequences of the XL RT-PCR product derived from the patient-specific AR-transcript (1.96 kb/1.7 kb) and one shortened amplificate found in patient and normal material (2.55 kb/2.3 kb) were sequenced. As illustrated in Fig. 3, the product derived from the 1.7-kb patient-specific AR-transcript displays a deletion beginning 5 bp upstream of the mutant T in codon 888 shown below in the 2.3-kb short regular AR-transcript (Fig. 3). The last sense codon 886 (Met) is merged to a downstream sequence derived from the 3′UTR coding for 8 missense amino acids and followed by a premature stop-codon. Upstream of codon 886, sequencing up to exon 4 on the subcloned patient-specific amplificate revealed correct amino acid coding. In contrast, the 2.55 kb/2.3 kb amplificate exhibits the complete and correct coding sequence of exon 8 of the AR-gene.

Fig. 3.

Nucleotide sequences (from codon 884 in exon 8 going downstream) and resulting amino acid coding from the patient’s 1.7- and 2.3-kb XL RT-PCR products from Fig. 2D. The shared wild-type sequence appears bold typed; missense nucleotides and amino acids are in italics. Stop-codons are underlined. Note that the 2.3-kb XL RT-PCR product displays correct wild-type-AR amino acid coding, and in the 1.7-kb product a missense amino acid sequence is encoded and a premature termination codon is introduced at codon position 895.

To examine whether RNA from normal GSF contains small amounts of patient-specific AR-transcript, we designed a composite-primer (hAR-BFa) specific for a fraction of the 3′UTR sequence merged to codon 886 in the patient’s aberrant AR-mRNA shown in Fig. 3. Using this primer and the primer hARE1s, we could reproducibly amplify a specific (1.13 kb) RT-PCR product on patient material but not on normal control RNA. This demonstrates the absence of patient-specific AR-transcripts in normal GSF.

Studies of AR translation

In the patient’s GSF, we found an AR-protein that was slightly smaller than the normal AR 110/112 kDa doublet (Fig. 4). Correspondingly, a lighter AR-protein (87 kDa) that most probably is the result of a downstream translation-initiation at Met¹⁸⁹ (4) found in normal material also appeared to be smaller in the patient sample.

Fig. 4.

Western blot from patient and normal control GSF detected by labeled anti-AR antibodies. Kilodalton values indicate sizes of normal AR-protein isoforms (for further information see text). Patient-specific AR-proteins appear to be slightly shorter.

Scatchard analyses of androgen-binding data on the patient’s GSF revealed a normal K_d of 0.093 nm and a B_max of 4.8 fmole/mg protein, which indicate a very low level of otherwise normal androgen binding. Expectedly, the control cell line showed normal K_d- and B_max-values (0.073 nm and 46.11 fmole/mg protein, respectively).

In transient transfection assays, the pSVAR0 plasmid representing the wild-type androgen receptor showed a concentration dependent activation of the androgen inducible MMTV-luciferase reporter gene (Fig. 5). However, the plasmid construct pSVAR3′Del reflecting the deleted patient-specific translation product failed to induce the reporter gene, even in the presence of supraphysiological concentrations of dihydrotestosterone. This demonstrates the absence of any transactivation activity of the patient-specific AR-translation product.

Fig. 5.

Androgen-induced transactivation capacity of the patient-specific AR-protein (cDNA cloned as “pSVAR3′Del”) in comparison with a normal control (wild-type-AR cDNA clone“ pSVAR0”). The induction of luciferase activity calculated in relation to basal activity found in the absence of ligand of three independent experiments is depicted as a function of dihydrotestosterone incubation.

Discussion

So far, only few mutations that lead to extensive structural aberrations of the AR-transcript or -protein have been described (11). In these cases, exonic mutations normally lead to the introduction of a premature stop-codon, and intronic alterations within or nearby splice sites cause aberrations of the splicing process. First, exons adjacent to the mutated region can be completely absent in mature mRNA (10, 24, 25). Second, mutations can lead to the generation of various transcripts with either unaltered normal sequence, inserted nucleotides or absent exons (26, 27). This phenomenon can be explained by the activation of cryptic splice sites.

The mutation presented here leads to some extent to the generation of a shortened transcript in the patient (Fig. 2). This patient-specific AR-transcript of approximately 5.5 kb length indicates alternative splicing, caused by the activation of a cryptic splice-donor site within exon 8 in which the wild-type-sequence “… GGUGAGC… ” is changed by the mutation to “… GGUGAGU… ”. The mutated sequence is similar to the donor splice-site consensus sequence “… exon GGUAAGU intron… ” In aberrant AR-pre-mRNA splicing, the binucleotide “GU” now serves as splicing-signal leading to false intron-demarcation of the last part of exon 8 and a prominent fraction of the 3′UTR. In connection with an assumed acceptor splice-site located within the 3′UTR, the splicing process merges a downstream 3′UTR-fragment to codon 886, removing the internalized sequence. By this event, a sequence coding eight missense amino acids and a premature stop codon follows codon 886 (Fig. 3).

Alternative AR-mRNA splicing has been reported to occur in the prostatic cancer cell line LNCaP and in cells from prostatic tissue (4, 12, 13). In both cases, only 3′UTR sequences seem to be altered. Faber et al. (13) described a sequence surrounding position 8685 in the 3′UTR: “… intron CTTTAAC[N (27)]CAGATCA exon… .” It conforms to the acceptor splice-site consensus sequence and most probably serves as an alternative AR pre-mRNA splice site in LNCaP cells. In our patient, the aberrant splicing event leading to the formation of the patient-specific AR-transcript employs the new donor splice site behind codon 886 (located 3759 bp downstream of the transcript’s 5′-end) and presumably the above-mentioned sequence at position 8685 as acceptor splice site. As depicted in Fig. 6, this will lead to an AR-mRNA of approximately 5.57 kb length (deleted of 4.93 kb), matching with the 5.5-kb patient-specific AR-transcript found in the Northern blot. This consideration is further supported by the finding that the first four nucleotides of the 3′UTR sequence spliced to codon 886 (ATG) in the patient-specific AR-mRNA are identical to the first “exonic” bases following the putative acceptor splice site given above.

Fig. 6.

Illustration of hypothesized normal and aberrant AR-transcripts and the derived XL RT-PCR products. Also, putative splicing sites and effects are indicated. Transcripts are represented by full-size, XL RT-PCR products by thinner rows. Splice sites within the 3′UTR of the full-length AR-mRNA are depicted as small gray marks. The site of the mutation in codon 888 of the patient’s AR-gene is marked by an “x” within the full-length AR-mRNA. At the left side, the origin of GSF, in which the respective transcripts can be found, is designated. ORF, Open reading frame; Δ, deletion due to the respective splicing effect.

In patient material, we could not find indications for the existence of a full-length AR-transcript by RT-PCRs or Northern blotting (Fig. 2, A and D). Interestingly, seminested XL RT-PCR products of 3.5 kb and 2.3 kb length were generated on whole RNA from patient and normal control GSF (Fig. 2D). The 2.3-kb amplificate includes the complete 3′-end of the AR-mRNA coding sequence (Fig. 3) with a deletion exclusively restricted to the 3′UTR. Subtraction of this 3.86-kb deletion from the full-length transcript (10.5 kb) unveils a transcript size of approximately 6.4 kb. Correspondingly, the second shortened AR-transcript missing approx. 2.66 kb is about 7.8 kb long and most likely also comprises the complete coding region. As illustrated in Fig. 6, we believe that both shortened transcripts are the result of regular but rare splicing events applying the acceptor splice-site at position 8685 and 2 donor splice sites lying upstream but still within the 3′UTR. The 7.8-kb transcript seems to be identical to the similarly long 3′UTR-shortened AR-mRNA reported by Faber et al. (13) for LNCaP cells or the alternative AR-transcript described by Lubahn et al. (4) for prostatic tissue. In our experiments, both shortened wild-type AR-transcripts are not visible by Northern blotting indicating very low concentrations in control GSF, compared with the full-length transcript. This may explain the lack of other reports for alternative AR splice products in GSF or other androgen target tissues. Possibly, the 7.8-kb and/or 6.4-kb shortened wild-type AR-transcripts may exist in many or even all androgen-sensitive tissues.

The existence of shortened wild-type AR-transcripts in patient material may be explained by the following consideration (see also Fig. 6): During the splicing process, patient-specific AR-mRNA formation may be favored because of the activated strong splice-donor site in exon 8. However, a small portion of AR-pre-mRNA is spliced applying both donor-splice sites and the wild-type acceptor splice site described above within the 3′UTR leading to the generation of shortened wild-type AR-transcripts. Thus, as the acceptor splice site is eliminated, subsequent aberrant splicing cannot occur.

This hypothesis matches the results on the translation level. The Western blot (Fig. 4) demonstrates that the patient-specific AR-transcript is translated, resulting in a slightly shortened protein, compared with the wild-type AR (110/112 kDa or 87 kDa) found in control GSF. However, normal AR-protein was not discriminated in patient material by this approach; this could originate in only small quantities of normal AR displaying a hardly or undetectable band, which additionally may be covered by the nearby laying band of aberrant AR-protein. In contrast, androgen-binding analysis in patient GSF indicated the existence of a very small fraction of protein with normal androgen-binding properties. Therefore, we assume that the normal AR-protein fraction in patient GSF is too small to be detected by our Western blot, and it is above the detection level in the androgen-binding assay. Function of the patient-specific mutant AR-protein was excluded by androgen-induced transactivation in cotransfection assays (Fig. 5). This was expected because in other studies completely absent androgen-binding capacity is observed for AR-proteins bearing a premature termination in the hormone-binding region (3).

Therefore, we conclude that the clinically and biochemically apparent residual androgen action in the presented patient is most likely mediated by the translation products of small quantities of functional shortened wild-type AR-transcripts. It remains to be investigated whether the normal, alternative splicing effects described here may individually affect the variable genotype-phenotype correlation observed in AIS.

Acknowledgements

We are grateful to Dr. Hartmut Merz for his friendly permission to use parts of the laboratory equipment from the Department of Pathology of the Medical University of Lübeck and especially to Anke Müller for her excellent technical counseling. We thank Timo Gaiser for his advice allowing successful cloning of the patient-specific AR-cDNA. We further thank Nicole Getschmann for her excellent technical assistance.

¹

This work was supported by the Deutsche Forschungsgemeinschaft (Hi 497/3–2,3 and Hi 497/4–2 to O.H.)