Symposium: Genetic aspects of male (in)fertility

Transcription

1 RBMOnline - Vol 10. No Reproductive BioMedicine Online; on web 23 November 2004 Symposium: Genetic aspects of male (in)fertility Molecular pathology of the CFTR locus in male infertility Professor Mireille Claustres Professor Mireille Claustres was born in Avignon (Southern France). After a Master s degree in Psychology in Montpellier, France in 1969, she moved to Medicine University of Montpellier and received her MD degree in 1977, followed by degrees in medical biology and biochemistry at the University of Sciences of Montpellier. She was Assistant and Associate Professor in Medical Biochemistry, received her PhD degree in 1985, and was named Professor of Medicine in Since 1988, she has created and developed a Molecular Genetics Department devoted to orphan genetic diseases. Now, as head of the Medical Molecular Genetics Department, she is responsible for genetic testing (including prenatal and preimplantation genetic diagnosis) and academic teaching in genetics, and leads a research group in genetics including cystic fibrosis, dystonia and Duchenne muscular dystrophy. Mireille Claustres Laboratoire de Génétique Moléculaire et Chromosomique, CHU de Montpellier, Institut Universitaire de Recherche Clinique (IURC), 641 Avenue du Doyen Gaston Giraud, 34093, Montpellier Cedex 5, France Correspondence: Tel: ; Fax: ; Mireille.Claustres@igh.cnrs.fr Abstract Congenital bilateral absence of the vas deferens (CBAVD) is a form of infertility with an autosomal recessive genetic background in otherwise healthy males. CBAVD is caused by cystic fibrosis transmembrane conductance regulator (CFTR) gene mutations on both alleles in approximately 80% of cases. Striking CFTR genotypic differences are observed in cystic fibrosis (CF) and in CBAVD. The 5T allele is a CBAVD mutation with incomplete penetrance. Recent evidence confirmed that a second polymorphic locus exists and is a major CFTR modifier. The development of minigene models have led to results suggesting that CFTR exon 9 is skipped in humans because of unusual suboptimal 5 splice sites. An extremely rare T3 allele has been reported and it has recently been confirmed that the T3 allele dramatically increases exon 9 skipping and should be considered as a CF mutation. Routine testing for the most prevalent mutations in the CF Caucasian population will miss most CFTR gene alterations, which can be detected only through exhaustive scanning of CFTR sequences. Finally, a higher than expected frequency of CFTR mutations and/or polymorphisms is now found in a growing number of monosymptomatic disorders, which creates a dilemma for setting nosologic boundaries between CF and diseases related to CFTR. 14 Keywords: alternative splicing, congenital bilateral absence of vas deferens, cystic fibrosis transmembrane conductance regulator, polyvariant mutant genes, variable penetrance Introduction Congenital bilateral absence of the vas deferens (CBAVD) (MIM ) accounts for approximately 2% of cases of male infertility (Dubin and Amelar, 1971) and is present in 25% of patients with excretory azoospermia. Although the clinical diagnosis of CBAVD is relatively easy based on physical examination of the scrotum (Figure 1), it can take years until the vasal malformation is recognized, with an average time until correct diagnosis of about 4.3 years (Weiske et al., 2000). Aetiology of CBAVD has for a long time been unknown. Several families in which CBAVD was transmitted as an autosomal recessive trait have been reported in the literature (reviewed in OMIM, 2004), leading to the hypothesis that a genetic factor might be responsible for CBAVD. Infertility in males with cystic fibrosis (CF) (MIM ) was first suspected in the 1960s (Le Lannou et al., 1957) when it was realized that CF men were childless after several years or marriage, and their infertility was attributed to the absence of vas deferens (Kaplan et al., 1968). Because of the striking similarity between the anatomical characteristics and semen parameters for CF and CBAVD, these two so far unrelated disorders were postulated to have a common genetic origin as early as 1968 (Kaplan et al., 1968; Holsclaw et al., 1971). The link between CBAVD and CF was proven after the identification of the gene responsible for CF, when a French group tested the presence of the most common CF mutation (F508del) in 17 men with isolated CBAVD: 42% were found to be heterozygous for the F508 deletion, a frequency much higher than the expected frequency of CF carriers in the general population (one in 25 30, 4 5%) (Dumur et al., 1990). This finding led to the speculation that these patients might have a different, as yet unidentified, cystic fibrosis transmembrane conductance regulator (CFTR) gene mutation on the other allele; in this case CBAVD would be a specific mild form of CF. Anguiano et al. (1992) identified the first three compound heterozygotes, leading to the conclusion that CBAVD might be an incomplete (genital) form of CF, which was confirmed by several reports (Patrizio et al., 1993; Culard et al., 1994; Oates and Amos, 1994). However, the initial work

2 Figure 1. Seminal ph, volume, and fructose content in fathers not tested for CFTR mutations (A), in men with azoospermia without CFTR mutation (B), and in men with azoospermia and CFTR mutation (C) (from Von Eckardstein et al., 2000). Diagnosis of CBAVD is based on impalpable vas deferens on scrotal examination and/or a missing segment of the vas on transrectal ultrasonography (Oates and Amos, 1994). The leading signs of CBAVD are total absence of spermatozoa associated with an acidic ejaculate composed only of prostatic secretions, with low ph (average <6.8), low volume (<1.5 ml) and low concentrations of fructose of the ejaculate, as consequence of dysplasia or absence of the seminal vesicles (which contribute up to 70% of the normal ejaculate volume). Seminal plasma markers provide an effective method to predict CFTR mutations in men with obstructive azoospermia: ph <7.4 and fructose levels <2 µmol/ejaculate had the highest sensitivity and specificity for the prediction of the presence of CFTR mutation in CBAVD (de la Taille et al., 1998; Von Eckardstein et al., 2000). On surgical exploration the intrascrotal parts of the vasa deferentia are not detectable or are reduced to fibrotic cords. CBAVD is regularly accompanied by aplasia of the epididymal tail and body and various anatomical or functional abnormalities of the seminal vesicles including hypoplasia and aplasia (Goldstein and Schlossberg, 1998). Usually CBAVD patients have a normal or nearly normal testicular volume, with a tendency towards a lower volume (Ramos et al., 2004). Testicular biopsy demonstrates normal or hypospermatogenic activity. The simplified and less invasive PESA (percutaneous epididymal sperm aspiration) procedure has been recently proven to be an efficient diagnostic tool to confirm an obstructive azoospermia (Ramos et al., 2004). Renal ultrasonography is recommended to all patients to assess renal morphology and organ presence. In about 10 20% of the patients, a renal anomaly is also present. In addition to CFTR genetic testing, sweat testing and clinical evaluation for CF-related signs should be included in the evaluation of CBAVD. also revealed that most CBAVD patients of whom the entire coding sequence had been analysed did not seem to carry mutations on both copies of the gene, which was unexpected for an autosomal recessive inheritance. A crucial step in the assessment of a common genetic etiology for CF and CBAVD was the report from Chillon et al. (1995) that the so-called normal allele of many heterozygotes was carrying the same variation in a non-coding DNA sequence, called the 5T allele. This variant had been previously demonstrated to be an intronic splicing variant that reduces the amount of functional CFTR protein at the cell membrane (Chu et al., 1991, 1992). The frequency of the 5T allele in CBAVD patients was up to six times (30%) higher than in the general population (5%), and 34% of men with CBAVD had inherited a CFTR mutation on one gene and a 5T allele on the other one, a combination that would not result in CF but does result in enough of a reduction in CFTR protein to produce isolated CBAVD (Chillon et al., 1995). For the past 12 years, increasing amounts of data have documented the link between most cases of CBAVD and alterations within the CFTR gene (reviewed in de Braekeeler and Ferec, 1966; Lissens et al., 1996; Patrizio and Zielenski, 1996; Lissens and Liebaers, 1997; Meschede et al., 1998; Phillipson, 1998; Patrizio and Leonard, 2000; Quinzii and Castellani, 2000; Stuhrmann and Dork, 2000; Weiske et al., 2000). The present report is focused on the current knowledge of the molecular basis of CBAVD. Cystic fibrosis and mutations of the CFTR gene CF is the most common lethal, autosomal recessive genetic disorder among Caucasians, with an estimated incidence ranging from 1 in 1700 live births in Northern Ireland to 1 in 10,000 in Mediterranean populations of Europe. The disease was described in the 1950s when children were dying of pancreatic insufficiency, failure to thrive and respiratory infections, and it was called cystic fibrosis of the pancreas. The median age of diagnosis is 7 months, with 66% of the affected individuals diagnosed by 1 year of age. Without treatment, CF is typically a fatal disorder in infancy and childhood, secondary to fat malabsorption due to pancreatic enzymes deficiency, failure to thrive and progressive respiratory infections. The concentration of salt in the sweat of CF patients is greatly elevated (>60 mmol/l, 3 5 times normal), which is used as a diagnostic test for the disease. Today, improved management of symptoms often preserves life into adulthood, and the median length of survival has increased from roughly 1 year in 1950 to more than 30 years. However, as there is still no curative treatment of CF, patients continue to have a considerable burden of illness and die prematurely, largely because of chronic pulmonary damage generated by excessive inflammation and chronic infection. A better understanding of CFTR mutations will lead to novel therapeutic strategies. More than 96% of men with CF have 15

3 16 anomalies in the Wolffian structures causing obstructive azoospermia and infertility (Kaplan et al., 1968). Spermatogenesis persists, but the body and tail of the epididymis, vas deferens and seminal vesicles are atrophic or absent. More than 1200 mutations can affect CFTR function Cystic fibrosis is caused by mutations in the CFTR gene on the long arm of chromosome 7, which encodes a glycosylated transmembrane protein that functions as a chloride channel, regulating salt and water transport in plasma membranes of epithelial cells (Figure 2). CFTR is also involved in bicarbonate transport (Choi et al., 2001) and the regulation of other channels (reviewed by Schwiebert et al., 1999). Absent or critically low levels of CFTR lead to viscous secretions that obstruct the lumen of the pancreatic ducts, biliary tract, airways, sinuses, gastrointestinal tract and reproductive tissues. The most common CF mutation worldwide is a deletion of three bases in exon 10 encoding a phenylalanine residue at position 508 within the first NBD, named F508del (or F508), which impairs the ability of CFTR to fold in the endoplasmic reticulum, thereby enhancing the degradation of the protein. The F508del mutation accounts for approximately 67% of the CFTR genes in CF patients from European descent and is present in about 89% of all patients with classic CF. The frequency of F508del varies considerably between populations and shows a decreasing northwest-to-southeast gradient in Europe, being as high as 87% in Danish patients and as low as 40 60% in Mediterranean countries. It is postulated that the mutation diffused with the movement of Neolithic farmers from the Middle East. Based on assumptions of mutation rate of short tandem repeats and of the demographic history of the European population, the age of the F508del mutation has been estimated to be generations old (10,000 50,000 years) (Kaplan et al., 1994; Morral et al., 1994; Wiuf, 2001). Its high frequency in Caucasians is explained by a selective advantage of carriers who might have better resistance to pathogens (as an example, it is postulated that F508del heterozygotes had an advantage in surviving cholera; Gabriel et al., 1994), although it is still a matter of controversy. Since the discovery of the CFTR gene in 1989, reports from many laboratories documented a growing number of less common mutations (mostly missense or nonsense mutations). Today more than 1300 alterations in the CFTR gene are compiled in an electronic database maintained by Toronto Sick Children s Hospital (Cystic Fibrosis Mutation Database, 2004). The vast majority of CFTR mutations are point mutations, changing a few bases or only one base in the genomic CFTR sequence. The majority of the mutations are rare (less than a dozen have a frequency above 1%), or have been found in only a few number of cases across the world, or are unique to a particular individual or family. There are relatively few genomic rearrangements such as large deletions or insertions within the CFTR gene, however they may be underestimated as they cannot be detected by standard polymerase chain reaction (PCR)-based techniques used for diagnosis (Audrezet et al., 2004). Two categories of CF mutations (severe and mild) contribute to clinical variation in CF CFTR mutations express their effects trough a variety of molecular mechanisms that include either the failure of production of CFTR protein or the elaboration of a mutant protein with little or no functional CFTR activity on the apical membrane (Figure 3). The extent to which various CFTR alleles contribute to clinical variation in CF has been extensively evaluated by genotype phenotype studies (reviewed in Estivill, 1996; Kerem and Kerem, 1996a,b; Tsui and Durie, 1997; Zielenski, 2000). The only symptom that correlates well with the mutational genotype is the pancreatic function, which was found to be concordant among affected siblings. In CF patients from the same family, the pancreatic function, whether sufficient or insufficient, is almost the same, indicating that the pancreatic status is determined primarily by the genotype at the CFTR locus. Other symptoms, such as lung disease,may vary considerably among these sibships and are poorly correlated to CFTR genotype (Zielenski, 2000). In approximately 85% of CF patients, the exocrine pancreas is profoundly affected (PI, pancreatic insufficiency). Without sufficient fluid and bicarbonate due to complete loss of CFTR function, digestive proenzymes are retained in pancreatic ducts and prematurely activated, ultimately leading to tissue destruction and fibrosis, in addition to malabsorption (Durie, 1998; Choi et al., 2001; Reddy and Quinton, 2003). In the remaining 15% of CF patients, the exocrine pancreas is abnormal but functionally adequate and patients do not need pancreatic enzyme supplementation (PS, pancreatic sufficiency). Analysis of CF mutations in patients with CF-PI or CF-PS revealed two categories of alleles, severe and mild (Kristidis et al., 1992) (Figure 3). A severe mutation on a CFTR gene inherited from one parent confers a PI phenotype only if combined with another severe mutation on the gene inherited from the other parent, whereas a mild mutation confers a PS phenotype in a dominant fashion even if the other mutation is severe. Mutations that belong to classes 1, 2, and 3 result in the complete functional loss of CFTR protein from the epithelial cell surface and fall in the category of severe alleles. Mutations included in classes 4 and 5 are referred to as mild and provide enough residual CFTR function to compensate for lack of function corresponding to a severe allele, so that compound heterozygotes of a severe and a mild mutation are, usually, pancreatic sufficient. Distinct spectrum of CFTR gene mutations and genotypes in CF and in CBAVD From the compilation of six large studies, totalling 1150 CBAVD patients who have been extensively investigated for CFTR mutations, the three most common CFTR mutations in Caucasian men with CBAVD are F508del (21.5% of alleles, versus 67% in the CF population), the 5T variant [19% of alleles, which is an average 4-fold higher than in the general population (5%)], and

4 Figure 2. The CFTR gene and protein: the gene responsible for cystic fibrosis was discovered in 1989 by positional cloning (Kerem et al., 1989; Riordan et al., 1989; Rommens et al., 1989) and the cdna sequence was determined in 1991 along with the locations of intron/exon boundaries (Zielenski et al., 1991). The gene spans ~190 kb of genomic DNA on region q31.2 of chromosome 7 (Ellsworth, 2000) and consists of 27 short exonic sequences that code for proteins (exons) interspersed with very long non-coding intronic sequences. The exons are numbered 1 24 with subdivisions A and B for exons 6, 14 and 17, because these exons were recognized as separate units after the initial publication of the gene (Zielenski et al., 1991). The CFTR promoter harbours the characteristic of a housekeeping -type gene, and the gene is transcribed in freshly isolated normal bronchial epithelial cells at a relatively low rate (Yoshimura et al., 1991), producing about two CFTR mrna transcripts per cell (Trapnell, 1993). Following transcription, pre-mrna involves the joining together of all exons (and discarding the introns) in a continuous sequence to form a mature transcript (CFTR mrna) of 6.5 kb that will be translated into a 1480 amino acid protein of approximately 168 kda which forms a transmembrane ionic channel called CFTR, cystic fibrosis transmembrane conductance regulator (Riordan et al., 1989; Rommens et al., 1989). The CFTR gene is also named ABCC7 because it encodes a protein which is a member of the ATP-binding cassette (ABC) superfamily of membrane transporters that includes the multidrug resistance-associated protein (MDR). CFTR is an integral membrane protein that functions principally as a camp-regulated chloride channel that controls the ion and water content in glandular tissues. It consists of two membrane-spanning regions (TM), each comprising six subunits, two nucleotide binding domains (NBD) and a cytoplasmic regulatory domain (R), unique within the ABC family, which contains many potential sites for phosphorylation by protein kinases (reviewed by Akabas, 2000). The TM define the CFTR chloride channel, while the NBD and the R domain mediate channel gating. Surprisingly, wild-type CFTR is inefficient at its own biosynthesis. CFTR is subjected to a complex processing including glycosylation in the Golgi apparatus and ATP-dependent conformational maturation in the membranes of endoplasmic reticulum assisted by chaperones. As a result, only 30% of the immature wild-type CFTR molecules transit to the late secretory pathway and reach the plasma membrane. 70% are trapped in the endoplasmic reticulum-associated degradation pathway: they are ubiquitinated and retrotranslocated to the cytosol where they undergo proteasomal degradation (Gelman and Kopito, 2003). CFTR is expressed in the apical plasma membrane of secretory and reabsorptive epithelia and allows the transepithelial movement of water and solute by mediating chloride translocation across the plasma membrane. H2N refers to the N-terminal end and COOH to the C-terminal end of the CFTR protein. 17

5 18 Symposium - Molecular pathology of the CFTR locus - M Claustres

6 Figure 3. The molecular consequences of CFTR mutations (adapted from Welsh and Smith, 1993; Zielenski and Tsui, 1995; Estivill, 1996). Although insights into the mechanism of dysfunction by functional testing of mutants expressed in cell lines has been obtained for only a small proportion of the more than 1300 reported mutations in the CFTR gene, they have been classified into six main different groups according to the mechanism by which they alter CFTR chloride channel function. Class 1: mutations interfere with CFTR production. These result in the introduction of a premature signal for termination of translation (stop codon) in the mrna. The truncated proteins are unstable and are recognized by chaperone proteins in the endoplasmic reticulum and are rapidly degraded. The net effect is no CFTR protein at the apical membrane. Examples include nonsense (G542X), frameshift (3659delC) or severe splicing (1717 1G A) mutations. Class 2: mutations affect protein maturation. These lead to the production of a protein that cannot be trafficked to its site of function on the apical membrane. Mutant proteins are mostly retained in the endoplasmic reticulum, failing to mature into fully glycosylated forms and immediately targeted for proteolytic degradation in the proteasome. The end result is absence of CFTR protein at the apical membrane of the cell. The most common mutation responsible for CF is a deletion of a single amino acid, F508del. Another example is the missense mutation N1303K, which substitutes an asparagine residue for a lysine at position Class 3: mutations affect channel regulation. The mutated protein is properly trafficked and localized to the plasma membrane but cannot be activated or function as a chloride channel. All mutations so far attributed to this group are located within the NBD, such as missense mutations G551D in NBD1 or G1349D in NBD2. Class 4: mutations affect chloride conductance. The CFTR protein is correctly trafficked to the cell membrane and is capable of being activated but generates reduced Cl current by altering the rate of ion flow or the opening time of the channel. Most of class 4 mutations are located within the membrane-spanning domains, such as missense mutations R117H, R334W or R347P. They confer a milder phenotype, even in combination with a severe allele. Class 5: mutations reduce the levels of a normally functioning CFTR. Various mutations or sequence variations may be associated with reduced amount of functional CFTR at the apical membrane, due to partially aberrant splicing ( kbC T, G A, or the 5T allele) or inefficient trafficking (A455E). Mutants in this group are predicted to result in milder or monosymptomatic phenotypes. Class 6: mutations decrease stability of CFTR present or affect the regulation of other channels. Recently, two new classes of mutants have been added. Class 6A mutants lack the last residues of CFTR (Q1412X, S1455X, 4279insA, 4326delTC), which dramatically reduces the apical stability of the protein (Haardt, 1999). Class 6B mutants are unable to interact properly with other ion channels, such as missense mutation G551D and the ORCC (outwardly rectifying chloride channel) (Fulmer et al., 1995). Although they have been investigated in a few patients, most class 6 mutants should be considered as severe. Grouping mutations into different classes is useful for understanding the mechanisms of dysfunction. However a single mutation can cause more than one type of abnormality, and hence fall into multiple classes (for example, missense P574H is both a class 2 and class 4 mutation, missense G551D is both a class 3 and 6B mutation). With reference to the pancreatic status, mutations that produce no functional CFTR at the apical membrane (1 3 and 6) contribute to pancreatic insufficiency (PI) and are classified as severe, whereas mutations with residual function (4 and 5) tend to sustain pancreatic function (PS) and are classified as mild. However, certain missense mutations (such as G85E) may confer a variable pancreatic phenotype. 19

7 20 R117H (6.4 versus 0.5%) (Table 1). An extensive collection of mutations identified by 19 laboratories in France illustrates the extreme heterogeneity of CFTR alleles and the different distribution of mutations and genotypes in CF and CBAVD (Claustres et al., 2000) (Figure 4). Almost 98% of a total of 3710 patients with classic CF had at least one CF mutant allele identified at the time of the study, including 3312 patients (89%) with two mutations and 318 patients (9%) with only one mutation, and 310 different CFTR mutations were recorded. In the group of 800 men with CBAVD, a total of 137 different mutations scattered over the whole gene were identified, including at least 11 complex alleles (more than one sequence change on a single gene). Seventy-three (53%) of these mutations were also present in the CF cohort, while 64 others (47%) were detected in CBAVD only. In 3303 patients with CF and 381 men with CBAVD in whom the two mutations had been identified, we reckoned up to 481 and 131 different mutation genotypes respectively. Distribution and frequencies of genotypes were markedly different between CBAVD and CF (Figure 4). Overall, 88% of patients with CF had inherited two severe mutations and 12% had inherited a severe combined to a mild or variable mutation. None of the CBAVD patients was homozygote for F508del or compound heterozygote for two severe mutations; they had either a severe and a mild (88%) or two mild (12%) mutations. The combination of the 5T allele in one copy of the CFTR gene with a CF mutation in the other copy was the most common cause of CBAVD. CBAVD in non-caucasian populations Cystic fibrosis is presumed to be rare in Asian or Oriental populations, with incidence estimated as 1 in 100, ,000 live births in Japan (Yamashiro et al., 1997) and less than one in 90,000 live births among Orientals. By contrast, CBAVD/CFTR does not seem uncommon in these populations, as illustrated by recent series of CFTR screening (Table 1). Interestingly, only a few number of F508 carriers were identified, but an exceptional number of males were found with the 5T allele, which represented 43.7% of CBAVD alleles in Egyptian (Lissens et al., 1999), 44.4% in Taiwanese (Wu et al., 2004), 30% in Japanese (Anzai et al., 2003), and 20% in Turkish (Dayangac et al., 2004) series. The frequency of the 5T allele in these populations, as calculated from the study of the spouses of CBAVD patients, does not seem different from that of Caucasian populations. A notable proportion of CBAVD males have homozygosity for the 5T allele, which has also been reported in a Tamil male from Sri Lanka (Fokstuen et al., 2000). These results show that the 5T variant is involved in many cases of CBAVD even in populations where CF is rare, as initially observed by Dork et al. (1997). By contrast, the frequency and distribution of CFTR mutations are very different from that found in European populations, with Q1352H and D1152H as the most common mutations respectively in Japanese (13%) or Turkish (14.7%) CBAVD alleles (Table 1). The 5T variant as a CBAVD mutation The CFTR exon 9 is skipped in humans Analysis of CFTR mrna transcripts in epithelial cells from normal individuals demonstrated that all had a proportion (ranging from 9 to 66%) of CFTR mrna transcripts without exon 9 (Chu et al., 1991). To explore the possible cause of variable deletion of exon 9 in CFTR mrna transcripts, the region encompassing exon 9 was sequenced, and it was found that the degree of CFTR exon 9 skipping was inversely correlated with the length of a polymorphic polythymidine tract (Tn) upstream of the exon (Figure 5). Individuals carrying common 7T or 9T alleles at this locus produced 9 25% of transcripts without exon 9, whereas individuals carrying a 5T allele produced 62 66% of incomplete mrnas. Extensive quantification of CFTR transcripts in large series of individuals with different combinations of Tn alleles confirmed that the shorter the Tn tract the greater the relative amount of CFTR mrna transcripts without exon 9 present in respiratory epithelium, with rare individuals homozygous for the 5T allele predicted to produce up to 90% of aberrant mrnas (Chu et al., 1992). It was concluded that, in the presence of a 5T allele, the polypyrimidine tract is too short and is under-utilized as a splice acceptor site, resulting in reduced efficiency of exon 9 splicing. The CFTR mrna deleted for exon 9 maintains an open reading frame and would encode a CFTR isoform that is missing 60 amino acids (from 404 through 464) of the first NBD. Given the important role for NBD1, it was surprising to observe that normal individuals can have up to 66% of bronchial CFTR mrna transcripts that are missing exon 9, a region representing 21% of the sequence coding for the critical NBD1. As expected, it was demonstrated that the protein produced by CFTR mrna without exon 9 is not properly processed and is not capable of generating Cl conductance in response to camp (Delaney et al., 1993; Strong et al., 1993). It was postulated that, since the length of the polypyrimidine locus Tn in intron 8 directly influences the efficiency of exon 9 splicing in CFTR and determines the percentage of functional CFTR transcripts produced, this could cause an effect similar to a mild CFTR mutation, with low levels of expression of normal CFTR. These results also suggested that only a very small fraction of functional full-length CFTR is needed to maintain a clinically normal phenotype in bronchial epithelial cells in regard to CF, as individuals homozygous for the 5T allele initially described had no clinical signs of CF. This threshold level varies as 10 25% according to the strategy used for the quantification of CFTR transcripts (Chu et al., 1992, 1993; Mak et al., 1997; Rave-Harel et al., 1997). The 5T allele as a genetic modifier of a CF mutation A possible role for the 5T variant in disease was first suspected by the observation of variable phenotypes when it is associated on a single CFTR gene (in cis) with a missense mutation (R117H) which gives rise to a partially functional CFTR protein (Sheppard et al., 1993). A high frequency of R117H mutations was observed in CBAVD patients (Gervais et al., 1993). Analysis of Tn alleles co-segregating with R117H in groups of patients with CF and CBAVD revealed a tendency of the R117H allele to be associated with the 5T variant in CF patients, whereas it was associated with the 7T variant in CBAVD patients (Kiesewetter et al., 1993). This suggested that, when paired in trans with another CF mutation (carried on the other gene), the R117H-5T allele would cause a classic CF-PS phenotype, R117H-7T would cause CBAVD, and R117H-9T would not cause any disease (Kiesewetter et al., 1993). Indeed, the association R117H-9T was reported in a single case so far, in a 10-month-old African American male baby who sweated heavily, presented with a low serum sodium level, had a borderline sweat test value and no CF clinical symptom at the time of the report (Friedman

8 Table 1. Series review of F508del, R117H and IVS8-Tn frequencies in patients with CBAVD. Series F508 5T R117H Others a Proportion Reference (%) (%) (%) n (%) of identified alleles (%) Caucasian German (106) (30) 80.3 Dork et al. (1997) Spanish (110) (38) 83 Casals et al. (2000) French (800) (34) 79.4 Claustres et al. (2000) Canadian (134) (15) 62 Zielenski et al. (1995); Mak et al. (1999) Non-Caucasian Egyptian (20) na na Lissens et al. (1999) Turkish (51) b (50) 72.5 Dayangac et al. (2004) Taiwanese (27) c (50) na Wu et al. (2004) Japanese (19) (18) 47 Anzai et al. (2003) a Number of other different mutations identified and percentage of total alleles. b Mutation D1152H was the most prevalent, accounting for 14.7% of alleles. c Mutation Q1352H was the most prevalent, accounting for 13% of alleles. na = not available. Figure 4. Spectrum of CFTR genotypes in CF or CBAVD patients with two mutations identified. In a subset of 3303 patients with CF and 381 patients with CBAVD in which the two CFTR mutations were identified in France, a total of 481 different mutation genotypes have been generated in CF and 131 in CBAVD (Claustres et al., 2000). In the group of CF patients, 88% carried two severe mutations [including homozygosity for F508del (48%) and compound heterozygosity for F508del and another severe mutation (39%)], whereas 11.8% had a severe mutant in trans of a mild or variable mutation and only 0.2% inherited two mild mutations (including three patients with the combination F508del/5T). By contrast, CBAVD resulted either from a combination of a severe mutation (classes 1 3) with a mild mutation (classes 4 or 5) (88%) or from a combination of two mild mutations (12%). A total of 22 mutation genotypes were shared by the two groups of patients. The two most common compound heterozygous genotypes found in men with CBAVD were F508del/5T (28%) and F508del/R117H (6%). 21

9 Figure 5. The 5T allele: a splicing variant that promotes CFTR exon 9 skipping in humans. A polymorphic site in CFTR intron 8 (IVS8-Tn) influences the splicing of exon 9 and the level of CFTR protein production. In the general population, three alleles are found at the Tn locus, 5T (5% of alleles), 7T (84%) and 9T (11%). Transcripts derived from genes that carry five thymidines (5T) at this locus have high levels of exon 9 skipping, whereas those with seven or nine thymidines (7T and 9T respectively) have successively lower levels of skipping. Thus, humans produce two types of CFTR mrna, with or without exon 9, the proportion of incomplete transcripts being inversely correlated with the length of the Tn alleles. Individuals homozygous for the 5T allele may have 75 90% of mrna missing exon 9, whereas individuals homozygous for 7T or 9T have less than 25 or 15% respectively of mrna without exon 9. Other combinations of alleles produce intermediate levels (Chu et al., 1993; Rave-Harel et al., 1997; Teng et al., 1997; Larriba et al., 1998). Two rare alleles at the Tn locus have been recently reported. A 3T allele was described in a German male with CF-PS (Buratti et al., 2001) and in a French patient with CBAVD (Disset et al., 2004). A 6T allele has been found in a healthy German female carrier (Dork et al., 2001) and in two patients with CBAVD, one in France (Viel et al., 2004) and the other in Turkey (Dayangac et al., 2004). IVS8 = intervening sequence (intron) et al., 1997). Thus, the R117H mutation on the 7T background with a severe mutation in trans affects mostly the male reproductive tract and produces a sufficient level of partially functional CFTR in the lung to prevent a classical form of disease (although later onset mild lung disease may be observed in some cases), whereas on a 5T background, the 5T variant enhances the deleterious effect of mutation R117H, producing a CF-PS form because lower levels of partially functional CFTR are produced both in the lung and the reproductive tract. The work of Kiesewetter and colleagues suggested for the first time that the specific IVS8-Tn background on which a CFTR mutation resides could modulate disease severity in a tissue-specific manner. Several other mutations have been detected on a 5T background (Dork et al., 1997; Claustres et al., 2000), but they have not been extensively studied for genotype/phenotype relationship or functional expression. The F508del in Europeans seems to have arisen only once on the 9T haplotype and is thus well spliced, but produces a non-functional protein The 5T allele is a CBAVD mutation with incomplete penetrance Several independent studies totalling around 300 CBAVD patients showed that the frequency of the 5T allele in individuals with CBAVD (19 24%) was 4-fold significantly higher than that of the general population (4 5%) (Chillon et al., 1995; Costes et al., 1995; Jarvi et al., 1995; Zielenski et al., 1995; Dumur et al., 1996). Thus, the 5T allele in combination with a CFTR mutation on the other chromosome appeared to be one of the major causes of CBAVD. The study of Chillon et al. (1995) about the frequency of 5T allele in parents of CF children revealed that 9.33% of the mothers and 4.14% of the fathers were heterozygous for the 5T variant, suggesting that a man could carry both a CF-chromosome and a 5T-chromosome and be normally fertile. Furthermore, if this variant could cause CBAVD in all cases, then its frequency (0.05) combined with that of CF mutations in Europeans (0.04) would predict that approximately 1 in 222 [(1/500 carriers of a CF mutation and a 5T allele) + (1/400 5T homozygotes)] Caucasian men could suffer from CBAVD (Zielenski et al., 1995), which is 4.5-fold higher than the estimated incidence of CBAVD in the Caucasian population (1/1000) (Jequier et al., 1985). Shin et al. (1997) retrospectively reviewed the fertility status of 131 brothers of 105 men with CBAVD who attempted conception and observed that only seven (5%) were found to have CBAVD themselves. Although it is possible that they underestimated the true incidence of brothers with CBAVD by not defining the fertility status of married brothers who had not attempted to have children, their data suggested that this prevalence is five times lower than the 25% expected for an

10 autosomal recessive inheritance pattern. A discrepancy between the ratio of observed versus expected number of men with a particular phenotype is termed incomplete penetrance, which reflects here that inheritance of two abnormal alleles at the CFTR locus may not be sufficient to produce CBAVD in every case. It was apparent that the 5T variant could not be a fully penetrant disease causing allele. By comparing the frequency of the 5T allele in fathers (2.07) and in mothers (4.67) of CF patients, the degree of penetrance for the 5T variant in combination with a CF mutation with respect to CBAVD was estimated as 0.56 (1 2.07/4.67) (Chillon et al., 1995), a figure similar to 0.60 calculated from Canadian families (Zielenski et al., 1995). Tissues are differentially sensitive to exon 9 mis-splicing Major insights into potential mechanisms underlying the phenotypic divergence between CF and CBAVD were gained from the analysis of CFTR transcripts in the nasal epithelium and genital tissues in normal individuals with different Tn genotypes (Mak et al., 1997; Rave-Harel et al., 1997; Teng et al., 1997). In two independent studies, a significant higher proportion of transcripts without exon 9 were observed in vasal cells compared with nasal cells from the same individual (Mak et al., 1997) or from different individuals (Teng et al., 1997) (Figure 6). Therefore, nasal epithelial cells produced more functional CFTR than the genital tract. There is variability in the efficiency of the splicing mechanism, not only among different individuals but also between different organs of the same individual. These observations were further confirmed by Rave-Harel et al. (1997), who demonstrated that, in CBAVD males, the level of normal transcripts (exon 9+) was lower in the epididymal epithelium than in the nasal epithelium. Furthermore, they also showed that the level of normal CFTR transcripts in nasal epithelial cells correlated with the severity of lung disease. The discovery of differential splicing efficiency between tissues that express CFTR, provided new insights into the relationship between levels of normal CFTR and phenotype variation. First, the amount of CFTR required by each organ involved in CF or related diseases to maintain a normal phenotype may be variable. In some organs a small reduction in the level of normal transcripts might lead to dysfunction while others would not be Figure 6. Relationship between full-length CFTR mrna and phenotype. The efficiency of exon 9 splicing is lower in Wolffian tissues (adapted from Mak et al., 1997; Rave-Harel et al., 1997; Teng et al., 1997). (a) Relationship between CFTR IVS8-Tn genotype and proportion of transcripts without exon 9 in different tissues. Correct splicing of exon 9 is less efficient in the vas deferens (Mak et al., 1997; Teng et al., 1997) or in epididymal cells (Rave-Harel et al., 1997) than in nasal tissue. (b) A typical CBAVD patient has a severe mutation (F508del) in one CFTR allele and the variant 5T in the other. Mutation F508del (associated with a 9T allele in cis) will result in fulllength transcripts but absent CFTR protein at the membrane. From the 5T allele, this individual may produce a sufficient proportion of full-length (exon 9+) CFTR mrna in the lung (32% of total CFTR transcripts), which is adequate to maintain a normal pulmonary phenotype, but an insufficient level (26%) in the reproductive tract to confer a normal genital duct phenotype, thus presenting an isolated CBAVD without CF pulmonary manifestations. In contrast, a normal CF carrier carrying the F508del mutation in one copy of the CFTR gene but the 7T variant in the other allele will produce enough fulllength CFTR mrna in the reproductive tract (38%) from the 7T allele to sustain a normal phenotype. The threshold level for normal/abnormal phenotype depicted in the figure is arbitrary; it may vary between organs and between individuals and also varies between the published studies, depending on the different methods used in the quantitative analysis of exon 9+ transcripts (Mak et al., 1997). 23

11 24 affected. Second, the 5T allele may then become a diseasecausing mutation for the vas deferens due to the overall decrease in full-length CFTR mrna in this tissue. Third, CBAVD patients have no pulmonary disease probably because the amounts of functional mrna may exceed the necessary threshold transcript levels for a non-cf phenotype in the lung (Figure 6). Fourth, the high frequency of mild CFTR mutations in patients with CBAVD suggested that the vas deferens was one of the tissues most susceptible to the effects of changes in CFTR activity. Another polymorphism, the TG tract, explains the partial penetrance of the 5T allele as a CBAVD disease mutation Cuppens et al. (1998) postulated that another polymorphic locus based on TG repeats (with alleles ranging from 9 to 13 repeats) placed immediately upstream of the Tn site, might affect the efficiency of exon 9 inclusion in the CFTR mrna, in combination or independently from the Tn locus. The TG and T tracts are present in the pre-mrna (as UG and U) within the 3 splice site of exon 9 and hence could have a direct involvement in exon recognition by the splicing machinery. They quantified the CFTR transcripts from nasal epithelial cells and found that, on a 7T background, the TG11 allele gave a 2.8-fold increase in the proportion of CFTR transcripts that lacked exon 9, and TG12 gave a 6-fold increase, compared with the TG10 allele. Moreover, 5T CFTR genes derived from CBAVD patients were found to carry a high number (12 13) of TG repeats, while 5T CFTR genes derived from healthy CF fathers harboured a low number (11) of TG repeats. It was suggested that a longer TG repeat would place the branchpoint A nucleotide of the lariat in a less favourable position for splicing (Cuppens et al., 1998) (Figure 7a). Thus, both the alleles at the TGm and Tn loci determine the proportion of fulllength or mis-spliced CFTR transcripts and thereby affect net chloride transport activity of CFTR-expressing cells. Cuppens et al. (1998) proposed to name such mutant CFTR genes that harbour a particular combination of alleles and variants at polymorphic sites polyvariant mutant CFTR genes. This concept was recently confirmed by the results of an international collaborative study on the disease penetrance of the 5T allele. By comparing the TG tract of men with CBAVD, non-classic CF or fertile men (fathers of patients with CF), all carrying a severe mutation in trans of a 5T allele, it was found that longer TG repeats (12 or 13) are more often associated with diseases phenotypes than 5T alleles adjacent to short TG tracts (Groman et al., 2004) (Figure 7b). The combination TG11 5T was found in 78% of unaffected individuals, versus 9% of affected individuals, suggesting that TG11 5T is generally benign. Conversely, 91% of affected individuals had 12 or 13 repeats, versus 22% of unaffected individuals. The TG13 5T combination was found exclusively in affected individuals. When 5T is found in trans with a severe CF mutation, the odds of pathogenicity are 28 and 34 times greater for TG12 5T and TG13 5T, respectively, than for TG11 5T. Thus, the determination of TG repeat number is predictive of pathogenic 5T alleles. The pathogenic role of these polymorphic repeats is due to their effects on CFTR exon 9 at the level of pre-mrna splicing: longer UG repeats increase exon 9 skipping, which raises the proportion of non-functional CFTR protein. This is in line with the growing number of human diseases resulting from mutations or polymorphisms in cis-acting elements that disrupt use of alternative splice sites (Faustino and Cooper, 2003). Polyvariant mutants and coding SNP (single nucleotide polymorphisms) It has been postulated that common polymorphisms so far considered as neutral might have functional consequences on the CFTR mrna or the protein and play a role in the penetrance of the 5T allele. In contrast with the polymorphic loci TGm and Tn, the M470V locus is polymorphic at the amino acid level. The two alleles, M and V at position 470, were transiently expressed in COS cells, and their kinetics of maturation and degradation were studied by Cuppens et al. (1998). They have described potential functional differences for the two CFTR variants M and V470 (A or G at nucleotide position 1540 respectively), with the V470 variant resulting in decreased functional CFTR and the M470 maturing more slowly. We have reported a strong association of the 5T allele with the valine at position 470 (de Meeus et al., 1998a,b), which has been further confirmed in a Spanish sample (Casals et al., 2000), suggesting that the M470V locus could contribute to the variable expression of the 5T allele. In the study of Groman et al. (2004), TG11 5T and the highly penetrant TG13 5T allele were found on a M470 background in all of the cases in which phase could be established, whereas TG12 5T was found to be mostly associated with V470. Polymorphism M470V thus does not appear to be a major modifier of the penetrance of the 5T allele in CBAVD, but its role cannot be excluded, since combinations TG13 T5 M470 and TG13 T5 V470 have been observed in CBAVD and CF patients respectively (Cuppens et al., 1998). It is possible that some variations considered as neutral polymorphisms in cystic fibrosis, such as F508C, 1716G>A, R75Q or the double mutant [G576A+R668C] may be mild mutants involved in CBAVD cases when no other mutation is found on the allele despite careful whole coding sequences scanning. In normal individuals, variable levels of CFTR transcripts without exon 12 account for 5 30% of total CFTR mrna (Bremer et al., 1992; Slomski et al., 1992; Hull et al., 1994). Skipping of exon 12 removes an important part of the first NBF, rendering the CFTR protein non functional. Pagani et al. (2003a) analysed the splicing pattern resulting from several variants in CFTR exon 12, including G576A (G>C at nucleotide 1859 in exon 12), which is associated in cis with R668C (C>T at nucleotide 2134 in exon 13), and this complex allele produced only 22% and 7% of mrna with exon 12 in transcripts derived from nasal epithelial cells or from hybrid minigene transfection assays respectively. They also evaluated the effect of several splicing factors on the severity of splicing defect and suggested that variations in the concentration in splicing factors may be responsible for tissue phenotypic expression. Steiner et al. (2004) studied the CFTR mrna from epithelial nasal cells of patients affected with CFTRopathies and demonstrated that several common coding SNP within the CFTR gene are associated with a significant increase of exon 9 and 12 skipping. Partial penetrance could be a feature not only of the 5T allele, but also of other CFTR variants, and further clinical and molecular studies are required to address this issue.

12 Molecular basis for the skipping of exon 9 in humans Minigenes to replicate CFTR exon 9 alternative splicing Much attention has been devoted to understanding the mechanisms that influence the alternative splicing of exon 9. To this end, researchers have developed an in-vivo model system that replicates exon 9 alternative splicing consisting of a reporter minigene by means of which the effect of the different CFTR alleles can be experimentally analysed. In particular, minigenes carrying either exon 9 or exons 8, 9, 10 with flanking introns (Figure 8a) have been used to isolate the cis and trans factors, which control the splicing of exon 9 and modulate its efficiency (Nicksic et al., 1999; Nissim- Rafinia et al., 2000; Pagani et al., 2000; Buratti and Baralle, 2001; Buratti et al., 2001; Hefferon et al., 2002). By using transient transfections of minigenes containing different combinations of TGm and Tn repeats and studying their splicing pattern, the group of Baralle was the first to confirm that the longer the (TG)m tract the higher the proportion of transcripts without exon 9, but only when activated by the 5T allele (in their minigene system), with TG13T5>TG12T5>TG11T5 (Nicksic et al., 1999). A complex network of cis- and transacting factors modulate CFTR exon 9 alternative splicing The observation that identical pre-mrna transcripts are processed into alternatively spliced forms in a tissuespecific manner strongly suggested tissue-specific differences in their splicing environment, that could be due either to the presence of specific alternative splicing factors or to variations in the activities or levels of constitutive splicing factors (Mak et al., 1997; Rave-Harel et al., 1997; Teng et al., 1997). Experiments using minigenes showed that increasing the intracellular concentrations of a variety of cellular splicing factors including SR (serine argininerich) proteins, in particular the splicing factor 2 (SF2/ASF), and hnrnps (heterogeneous nuclear ribonucleoproteins) promoted the skipping of exon 9 from pre-mrnas (Nissim- Raffinia et al., 2000; Pagani et al., 2000). The group of Baralle demonstrated that this inhibitory effect of transacting factors on exon 9 inclusion was strictly dependent on the composition of the TGm and Tn alleles, and was further modulated by exonic splicing elements located in exon 9 [a 6-bp exonic splicing enhancer (ESE) and a 10-bp exonic splicing silencer (ESS)] and, more importantly, by an intronic splicing silencer (ISS) located in intron 9 (Pagani et al., 2000) (Figure 8). The nuclear protein TDP-43 (HIV-1 TAR DNA-binding protein) was identified as the factor interacting specifically with the (UG)m tract in the CFTR pre-mrna (Buratti et al., 2001). Overexpression of TDP-43 resulted in an increase of exon 9 skipping, whereas antisense inhibition of endogenous TDP-43 expression resulted in an increased inclusion of exon 9. TDP-43 binding to the UG polymorphic repeat reduces the proper recognition of the nearby 3 splicing site and, in association with the ISS element in intron 9, mediates exon skipping (Buratti et al., 2001, 2004). These data were recently confirmed by Wang et al. (2004), who showed that the mouse homologue of human TDP-43 inhibits human CFTR exon 9 splicing in a minigene system. Recent studies showed that promoter architecture also can interfere in the modulation of CFTR exon 9 skipping in minigene experiments, which suggests an interesting kinetic model in the coupling of transcription and alternative splicing (Pagani et al., 2003b). Zuccato et al. (2004) reported the identification of two polypyrimidinebinding proteins, TIA-1 (inducing exon 9 inclusion) and PTB (polypyrimidine tract-binding protein, inducing exon 9 skipping) as novel players in the regulation of CFTR exon 9 splicing. TIA-1 is the first trans-acting factor with a positive effect on exon 9 splicing and binds specifically to a polypyrimidine-rich controlling element (PCE) containing intronic splicing enhancers (ISE) located between the weak 5 splice site and the ISS in intron 9. Suboptimal 5 splice sites make exon 9 vulnerable to skipping Because many exons, both in CFTR and in other genes, have short polypyrimidine tracts in their 3 splice sites, yet are not skipped, other mechanisms of exon 9 skipping have been investigated by mutating the 5 splice sites both up- and downstream of exon 9 (Hefferon et al., 2002). Conversion of the upstream 5 splice site to consensus by replacing a pyrimidine at position +3 (Shapiro-Senapathy score of 84.3%) with a purine (score of 94%) resulted in increased exon skipping (50% exon 9 versus 30%). In contrast, conversion of the downstream 5 splice site to consensus by insertion of an adenine at position +4 (score 100% versus 75%) resulted in a reduction in exon 9 skipping, regardless of whether the upstream 5 splice site was consensus or not (Hefferon et al., 2002). These results suggested that human CFTR exon 9 is skipped because of suboptimal splice sites that make it inherently vulnerable to skipping. The potential role of the downstream 5 splice site in exon 9 skipping are further supported by data from sheep and mouse genomes. Sheep have upstream and downstream 5 splice site sequences very similar to those of humans but have an extensive polypyrimidine tract in the 3 splice site in the intron 8 (Y14) not accompanied by an adjacent TG variation. Although this structure would have predicted retention of exon 9, sheep skips exon 9 in all tissues analysed (Hefferon et al., 2002; Broackes-Carter et al., 2003). CFTR exon 9 skipping is absent in the mouse, which has two strong splice sites and a very short and non-polymorphic polypyrimidine tract (Y5) (Delaney et al., 1993; Nicsic et al., 1999; Ellsworth et al., 2000). Altogether, these observations led to the conclusion that exon 9 skipping in humans is due to the unusual composition of the flanking 5 splice sites (Hefferon et al., 2002). Rozmahel et al. (1997) have found that a fragment of CFTR gene spanning exon 9 and its flanking introns and polymorphic loci is present in multiple copies in the human genome, whereas it is present in only a single copy in different Old World monkeys who bifurcated from the Homo lineage more than 10 million years ago. They have suggested that a retrotransposition event followed by an amplification of the integration site occurred in the human genome. Thus, it is conceivable that insertion and rearrangement events occurring in intervening sequences surrounding human CFTR exon 9 placed fortuitous new cisacting elements in the proximity of its splice junction, which lowered exon definition and provided the basis for exon 9 25

13 a b Figure 7. Polyvariant mutants: the alleles at TGm modulate the partial penetrance of the 5T allele. (a) Sequences at the 5 and 3 splice sites in intron 8 (or IVS8, intervening sequence 8). Pre-mRNA splicing is a complex mechanism that relies on the correct identification on the short protein coding sequences (exons) among the large non-coding sequences (introns). Sequences in CFTR intron 8 share most of the canonical features of splicing sites, including the 5 -GT splice donor, 3 -AG splice acceptor at the intron exon 9 junction, branch-point A (in green) and pyrimidine-rich region at the splice acceptor site. However, the polypyrimidine tract is composed exclusively by thymidines and is polymorphic for its length. Moreover, it is placed immediately downstream of another polymorphic locus made of a variable number of TG repeats. This peculiar architecture of the intron 8 3 splice site modulates the alternative splicing of human CFTR exon 9 (Hefferon et al., 2002). Consensus sequences for 5 and 3 sites are given; y indicates a pyrimidine, and n indicates any nucleotide. Nine haplotypes TGmTn have been found in Caucasian populations, with TG11T7 the most common combination. Longer alleles at TGm associated with shorter alleles at Tn exacerbate exon 9 skipping (yellow, decreasing amounts of functional CFTR from top to bottom). In nasal epithelial cells from normal individuals, TG10T7 gives about 5% mis-splicing, TG11T7 14%, TG12T7 30%, and TG12T % (Cuppens et al., 1998). (b) Frequency of TGm-5T combinations found in fertile or infertile men carrying a CF mutation on a gene and a 5T allele on the second gene (adapted from Groman et al., 2004). 26

14 a b c Figure 8. The minigene approach to analyse CFTR exon 9 splicing. (a) A minigene is a genomic fragment that includes the exon of interest and the surrounding introns as well as the flanking constitutively spliced exons and is cloned in a eukaryotic expression vector then transfected in appropriate cells. Expression of transfected minigenes thus replicates constitutive and alternative splicing patterns, which are analyzed after RT-PCR with primers that can detect the inclusion and exclusion of the alternative spliced exon. These constructs are specifically designed to introduce by site-directed mutagenesis modifications in either the length or the sequence of the sites of interest. Minigenes can be cotransfected with putative splicing factors to test trans-acting factors or they can be transfected into different cell types to analyse them for their splicing ability. This part of the figure gives a schematic representation of the cis-acting intronic and exonic elements that affect the alternative splicing of exon 9. The elements identified so far include the weak upstream and dowstream 5 splice sites (Hefferon et al., 2002), the TGm and Tn polymorphic sites at the 3 -end of CFTR IVS8, the exonic splicing enhancer (ESE) and silencer (ESS) sequences in exon 9, the polypyrimidinerich controlling element (PCE) and the intronic splicing silencer (ISS) in IVS9 (Pagani et al., 2000, 2003; Zuccato et al., 2004). (b) Analysis of the size and the relative proportions of minigene RT-PCR products containing E9 (415 bp) and those lacking E9 (233 bp) by using fluorescent electrophoresis and Genescan softwares. A strong positive correlation is found between the length of the T tract and the proportion of mrna with exon 9, with T7>T5>T3 observed in all cell lines tested (Disset et al., 2004). (c) A decrease in the number of Ts in a TG12 background determines a cell-type-dependent reduction in transcripts exon 9+, with the lowest levels observed in epididymis-derived cell lines (Disset et al., 2004). 27

15 28 skipping in CFTR mrna (Hefferon et al., 2002). Recently, Hefferon et al. (2004) showed that the TG tract affects splicing through the formation of transient RNA secondary structures. Replacement of the TG tract in the minigene with random sequence abolished splicing of exon 9. Replacements with sequences that can self-base-pair resulted in a substantial increase in exon 9 inclusion, suggesting a role for RNA secondary structure in the influence of these sequences on splicing. Exon 9 alternative splicing is extremely sensitive to sequence variations Single base changes in exons may cause pathological splicing events by inducing exon skipping or inclusion from the mrna through disturbance of exonic elements that modulate splicing. Skipping of constitutive exons can occur for missense (substitution of amino acid) and silent (change at third position of the codon usage, that does not change the amino acid code) mutations through disruption of ESE or creation of ESS (Liu et al., 2001). Pagani et al. (2003c) demonstrated that single nucleotide substitutions in exon 9 might have a profound effect on the splicing efficiency of this exon, inducing exon inclusion (Q452P) or exclusion (A455E). Overall, 35 substitutions of 47 caused changes in the splicing pattern, this extreme sensitivity being probably related to the weak definition of the CFTR exon 9. They also identified at the 3 portion of exon 9 a composite regulatory element with juxtaposed enhancer and silencer properties. Another site for composite regulatory elements of splicing (CERES) had been previously described by the same group in CFTR exon 12 (Pagani et al., 2003a). These studies illustrate that some missense mutations may exert part of their phenotypic expression and disease variability through their effect on splicing of mrna and not only through CFTR channel dysfunction. This is the case of missense A455E, which was found in this study to increase exon 9 skipping (50% in a TG11T9 context). Overall, these results suggest that, in addition to adjacent splice sites, the entire sequence of exon 9 is important for exon recognition and processing. The rare 3T allele is a CF diseasecausing mutation T3 is a very rare allele which, so far, has been reported only twice worldwide. The first case was a German patient presenting with pulmonary symptoms with recurrent infections, elevated sweat chloride concentrations and pancreatic sufficiency (CF-PS) (Buratti et al., 2001). He carried a TG13T3 sequence (no other mutation) in trans with TG10T9-F508del on the second allele. The proportion of exon 9 skipping was evaluated by semi-quantitative analysis of RT- PCR products derived from blood lymphocytes, indicating that only 6% of correctly spliced transcripts (with exon 9) were generated from the TG13T3 allele. In minigene assays, the T3 allele led to a highest degree of exon 9 skipping, ranging from 55% (TG13T5) to 86% (TG13T3), which was concordant with the CF phenotype of the patient (Buratti et al., 2001). We reported a novel TG12T3 allele in a CBAVD patient who carries a TG11T7 and the benign variation F508C on the other chromosome (Disset et al., 2004). By using a minigene assay designed to include all the cis-acting elements reported so far in the vicinity of exon 9 (Figure 8a), we have analysed the splicing patterns derived from minigenes transiently expressed in six different CFTR-expressing epithelial cell lines, providing new insights into the mechanisms by which the proportion of skipped transcripts may be increased in male reproductive tissues. We demonstrated that longer TG tracts promote exon 9 skipping even in a T7 background, in contrast with previous minigene studies where this effect was observed only on a 5T context (Nicksic et al., 1999), but in accordance with initial observations from the study of endogenous CFTR transcripts (Cuppens et al., 1998). A decrease in the number of Ts in a TG12 background determined a cell-type-dependent reduction in exon 9+ transcripts, with the lowest levels observed in male genital tissues derived cells (Figure 8b). Previous studies had postulated that CBAVD could be related to less efficient splicing mechanisms in the male genital tract compared with the respiratory tract. However, by comparing the rate of decrease in exon 9 inclusion expressed in the different cell lines, we showed that elevated basic levels of full-length CFTR transcripts (for example, in the testis derived cell line) does not prevent a drastic reduction induced by TG12T5 or TG12T3 alleles. Our results showed that tissuespecific trans-acting splicing factors do not contribute to the different patterns of exon 9 splicing found between the cell lines. The results rather support the concept that the ratio of general splicing factors plays a role in the tissue variability of exon 9 alternative splicing (Disset et al., 2004). An alternative hypothesis is that deeper intronic trans-acting factors might be involved that are not present in the minigene constructs. We also confirmed that the 3T allele dramatically increases exon 9 skipping and should be considered as a CF mutation predicted to be fully penetrant in disease, in contrast with the 5T allele. It is thus conceivable that an individual carrying the extreme allele TG13T3 in combination with a severe mutation (F508del) will have a CF-PS phenotype, whereas an individual carrying a TG12T3 in combination with a mild variant (F508C) will only express a CBAVD phenotype. The complete skipping of exon 9 in all CFTR transcripts is predictive of severe CF, as observed in patients with mutations that destroy the consensus (AG) dinucleotide of the acceptor splice site in intron 8 (Cystic Fibrosis Mutation Database, 2004), which is expected to block exon 9 pre-mrna splicing and result in deleted transcripts. CFTR mutations and other infertile phenotypes Besides CBAVD, a higher than expected frequency of CF mutations or polyvariant haplotypes has been observed in other forms of infertility. Infertile patients with congenital unilateral absence of vas deferens (CUAVD) CUAVD is a rare condition, occurring in less than 1/1000 males. Martin et al. (1992), reporting on two brothers, one with CUAVD and the other with CBAVD, were the first to point out an association with the CFTR gene. Men with CUAVD and an obstructed contralateral vas deferens may be classified as having CBAVD (Schlegel et al., 1996). Mickle et al. (1995) defined CUAVD as palpable absence of one scrotal vas deferens and were able to subdivide them into clinically and genetically distinct groups. Patients with CUAVD and a contralateral obstruction of the seminal tract (at either the

16 inguinal or pelvic level) had a high frequency of CFTR mutations (88%), no renal agenesis, and this group is a CFTRassociated phenotype. Men with CUAVD and a patent contralateral seminal tract (i.e. they were not azoospermic) had no CFTR mutation, and this group displays a high frequency (40 80%) of ipsilateral renal agenesis (Mickle et al., 1995; Mak et al., 1999; Weiske et al., 2000; McCallum et al., 2001; Kolettis and Sandlow, 2002). CUAVD is a heterogeneous condition, and the two different forms (with and without renal anomalies) are probably the result of distinct pathophysiological processes. Obstructive azoospermia with intact vas deferens Several studies documented the assumption that non-cbavd azoospermia may in some cases be related to the CFTR gene, including idiopathic forms of epididymal obstruction (Jarvi et al., 1995; Mak et al., 1999) and bilateral ejaculatory duct obstruction (BEDO), where a high frequency of CFTR mutations has been found and several compound heterozygotes reported (Meschede et al., 1997; Mak et al., 1999). In a study involving 16 men with isolated anomalies of the seminal vesicles (IASV), only one was found to be heterozygous for a missense mutation and one for the 5T allele, which gives a frequency not different from the general population, so that IASV is not a CFTR-related entity (Meschede et al., 1997). The association of chronic bronchopulmonary disease with azoospermia due to a complete bilateral obstruction of the epididymes characterize Young s syndrome, but, in contrast to CBAVD or CF, there is no anatomical malformation of the seminal ducts. No CF mutations have been convincingly demonstrated in Young s syndrome. Despite some clinical common aspects, CF and Young s syndrome are two distinct entities. Non-obstructive causes of male infertility and subfertility A higher than expected frequency of CFTR mutations or variants has been reported in other more common forms of infertility. A CFTR mutation frequency of 17.5% (versus 5% in the general population) was detected in 80 infertile men with reduced sperm quality (Van der Ven et al., 1996). However, these observations were not confirmed in larger samples from southern France (Pallares-Ruiz et al., 1999), The Netherlands (Tuerlings et al., 1998), Germany (Stuhrmann and Dork, 2000) or Slovenia (Ravnik-Glavac et al., 2001). Although it seems that CFTR mutations do not play a major role in spermatogenetic failure, a minor effect on spermatogenesis or sperm maturation is still possible. Since it seems that the male reproductive tract is the most sensitive system of all CFTRaffected tissues to even minor CFTR dysfunction, it has been suggested that, in absence of well-defined CFTR mutations, polyvariant mutants producing less functional CFTR, such as the common single nucleotide polymorphism (SNP) M470V in exon 10, may predispose to oligospermia and testicular failure (Cuppens et al., 1998). The V470 homozygous genotype was found to be significantly more frequent among non-cbavd infertile patients (Boucher et al., 1999; Pallares-Ruiz et al., 1999; Larriba et al., 2001). It is difficult to compare SNP frequencies between published studies, partially because of the poor comparability of the populations being studied in terms of sample size, anatomic findings, and extent of genetic screening. Infertility may be caused by more than one genetic defect Cases where compound genetic defects coexist have been recently reported. A Chinese patient presenting with CBAVD and testicular atrophy with late maturation arrest of spermatogenesis was found to be heterozygous for the 5T allele and also to carry a pericentric inversion of chromosome 6 (Black et al., 2000). In another male with CBAVD, a compound heterozygosity for a CF mutation and the 5T allele was coexisting with a Robertsonian translocation (Drouineaud et al., 2003). With extended genetic analysis, a growing number of men may be found with more than one genetic defect, such as Klinefelter patients (47,XXY) also carrying a CFTR mutation, or men having both an AZF deletion of the Y chromosome and a CFTR mutation (Dohle et al., 2002). Uterine CFTR-mediated bicarbonate ion and sperm capacitation: the female counterpart? No female equivalent of CBAVD exists due to the different embryological origin of the definitive female reproductive tract. It was thought that the main genetic factor contributing to the reduced fertility observed in about 50% of women with cystic fibrosis is the thick dehydrated cervical mucus acting as a barrier for sperm passage. Intrauterine insemination (IUI) has been used to overcome this, and the first case of successful pregnancy and birth after IVF in a woman with CF homozygous for F508del was reported in 2000, after eight failed attempts at IUI (Rodgers et al., 2000). Recent findings suggest that the infertility of women with CF may not be caused exclusively by the failure of spermatozoa to penetrate cervical mucus, but also from an inability of spermatozoa to capacitate within the uterus and oviduct because of defective CFTR-mediated bicarbonate secretion (Wang et al., 2003). Bicarbonate ion is one of the essential components for mammalian spermatozoa to be capacitated in vitro. Acting in concert with calcium in spermatozoa, HCO 3 elevates intracellular camp levels (Evans and Florman, 2002), which in turn activate multiple pathways that drive the process of sperm capacitation, including hyperactivation of flagellar motility. In vivo, it is known that CFTR gene expression occurs in endocervical cells throughout the menstrual cycle (Hayslip et al., 1997) and that the female reproductive tract can secrete high concentrations of bicarbonate ion. The study by Wang et al. (2003) provided for the first time a direct link between invitro results and in-vivo observations, on the basis of several findings: (i) endometrial epithelial cells possess a CFTRmediated bicarbonate transport mechanism, (ii) spermatozoa are not capacitated efficiently in medium conditioned by cells in which CFTR expression has been suppressed by antisense RNA or by cells expressing the F508del mutation, (iii) the failure of F508del cells to capacitate spermatozoa is rescued by transfection with wild-type CFTR or by addition of HCO 3 to conditioned medium. In cystic fibrosis, attention has been focused essentially on chloride-channel properties of CFTR. However, the demonstration that CFTR-dependent bicarbonate secretion can drive sperm capacitation in vitro again stresses the importance of the bicarbonate, the 29

17 30 neglected ion (Quinton, 2001). Indeed, CFTR can transport bicarbonate and chloride by distinct mechanisms (reviewed by Hug et al., 2003) and the severity of CF can be predicted more accurately by bicarbonate conductance than by chloride conductance (Choi et al., 2001; Wine, 2001). Genetic heterogeneity in CBAVD CBAVD with renal abnormalities In 12 21% of cases, CBAVD is associated with malformations or agenesis of the upper urinary tract and in almost all cases, no CFTR mutation can be detected (Anguiano et al., 1992; Gervais et al., 1993; Augarten et al., 1994; Casals et al., 1995; Mickle et al., 1995; Schlegel et al., 1996; Dork et al., 1997; de la Taille et al., 1998; Claustres et al., 2000; McCallum et al., 2001). It is thought that CBAVD with concomitant renal malformations probably does not represent a genital form of CF but rather a distinct clinical entity, termed URA/CBAVD by McCallum et al. (2001). These investigators found no significant differences between the two groups (CF/CBAVD and URA/CBAVD) of patients in terms of physical, laboratory and radiographic findings of the reproductive derivatives as well as in fertilization and pregnancy rates, which was also observed by Robert et al. (2002); by contrast, the percentage of men with a pelvic kidney was 10-fold higher in the URA/CBAVD cohort than the CF/CBAVD group. The fact that the phenotypic outcome of the renal portion of the mesonephric duct is so different between the two groups strengthens the hypothesis that URA/CBAVD and CF/CBAVD have a different genetic basis. During embryonic development, the vas deferens, seminal vesicle, ejaculatory duct and distal two-thirds of the epididymis develop from the reproductive part of the mesonephric duct, while its ureteric part is needed to induce normal renal development. The head of the epididymis (which is present in men with CBAVD or CF) and the testis arise from the genital ridge. The physical separation between the two mesonephric duct derivatives (seminal and renal) occurs by week 7 of gestation (Oates and Amos, 1994). It has been postulated that damage at a very early stage in embryo development (before 7 weeks) will lead to abnormal development of the entire mesonephric duct with its two derivatives, resulting in URA/CBAVD (Hall and Oates, 1993; McCallum et al., 2001), or URA/CUAVD (Donohue et al., 1989). By contrast, the genetic defect in CBAVD-CFTR appears to affect the embryo after the division of the mesonephric parts in the seventh week of gestation, so that only the seminal tract will be altered. A few number of patients with CBAVD and URA have now been reported to be heterozygous for a CFTR mutation (Mak et al., 1999; Casals et al., 2000), the significance of which is unclear as it might be pure coincidence considering the high frequency of CFTR mutations and 5T carriers in the general population and the low numbers of URA/CBAVD patients investigated. Further family studies are required in both the CBAVD populations and in males with URA to determine the proportion of cases with genetic causes, the mode of inheritance and the penetrance of genetic factors in CBAVD and nephrogenesis. CBAVD with discordant familial segregation of alleles Family studies are another approach to study the possibility of genetic heterogeneity. In families with CBAVD linked to CFTR mutations, brothers with CBAVD should inherit the same CFTR genes from their parents, whereas fertile male siblings should have at least one CFTR gene different from their CBAVD brothers. In families with CBAVD not linked to CFTR defects, the inheritance of the CFTR genes should be random with respect to the CBAVD phenotype. Familial segregation of CFTR genes can be assessed by studying polymorphisms within and around the CFTR gene: the combination of alleles at the polymorphic sites (haplotypes) allows one to follow transmission of CFTR genes from parents to children in the pedigrees. Three families with no identified CFTR mutations have been reported in which either the brothers with CBAVD inherited two different alleles from one of their parents (Rave-Harel et al., 1995) or in which fertile brothers inherited the same CFTR alleles as their brothers with CBAVD (Mercier et al., 1995). Such a discordance between phenotypes and marker haplotypes suggested that CBAVD was not caused by two mutated CFTR alleles but was rather the result of mutations at other hypothetic loci (Mercier et al., 1995; Rave-Harel et al., 1995). However, such cases could also be explained by CFTR unidentified mutations that have a more complex pattern of penetrance, and it would have been interesting to re-analyse these three families with the 5T allele and its surrounding modifiers. Moreover, whether CBAVD was associated with renal anomalies, or not, was not mentioned in the initial reports. Nevertheless, 15 20% of patients with CBAVD who have now been extensively screened and do not have CFTR mutations or the 5T allele may indicate a subpopulation of men with other genes responsible for the reproductive anomaly. CBAVD pathogenesis The mechanisms by which CFTR mutations contribute to the pathogenesis of CBAVD are still not fully understood. CBAVD may result from a morphogenic defect during development or from obstruction by abnormal secretions. Timing of CFTR damage in Wolffian derivatives development CFTR mrna is highly expressed in the epithelium of the head of the epididymis by 18 weeks gestation (Tizzano et al., 1993, 1994; Trezise et al., 1993), which suggests that CFTR may play a role in the early development of the reproductive tissues. An initial report has described a fetus with CF that showed extensive fibrosis and regression of the epididymis (Harris and Coleman, 1989). Recently the relationship between CFTR mutations and the congenital absence of the uterus and vagine (CAUV) condition which affects 1 in 5000 females, was examined upon the rationale that the embryological development of the Müllerian ducts directly depends on the prior normal development of the Wolffian ducts (Timmreck et al., 2003). Samples from 25 patients with CAUV were tested for the 33 most common CFTR mutations including the 5T allele, and the data suggested that it is unlikely for CFTR mutations to cause CUAV in females (Timmreck et al., 2003). Finding that CFTR mutations are

18 associated with 80% of cases of CBAVD, a Wolffian duct anomaly, but are not associated with CAUV, a Müllerian duct anomaly, provides further evidence on the timing of CFTR damage in CBAVD. The effects of the CFTR mutations on the Wolffian duct derivatives must occur after the ninth week of embryological development, at a time when the Wolffian and Müllerian ducts have completely separated and are developing independently. CBAVD as the result of progressive regression Several findings favour the hypothesis that, in contrast with CBAVD not linked to CFTR where the absence of vas deferens may originate in fetal life as a defect during organogenesis, CFTR mutations do not affect the embryological development of the Wolffian duct. The fact that normal portions of the genital tract are seen more frequently in CF newborns than in older patients (Oppenheimer and Esterly, 1969) suggests that Wolffian derivatives are present at some stages of fetal life but may degenerate later. Gaillard et al. (1997) examined two fetuses with two severe CFTR mutations after abortion at 12 and 18 weeks respectively, and determined that their vas deferens were normal with no obstruction or stenosis. This observation and the high proportion of normal ducts reported in prepubertal male CF patients (Valman and France, 1969) favour the hypothesis of degenerative lesions secondary to luminal obstruction, probably caused by thick secretions, as the cause of definite post-natal ductal regression (Patrizio and Salameh, 1998). The first study describing ultrasonic features in male CF children confirmed intact kidneys and ureters, and suggested the possibility that ductal genital abnormalities progress with time (Blau et al., 2002). The luminal fluid of the epididymis contains a high protein load and has a low flow rate, and the epididymis and the vas deferens form a tortuous ductal system that is several meters long but has a diameter <0.5 mm. It is possible that the absence or dysfunction of CFTR would make these organs very vulnerable to luminal concentration defects, especially in the distal portion of the vas deferens, where CFTR expression level is low (Tizzano et al., 1993, 1994; Trezise et al., 1993), and lead to progressive obstruction and obliteration, as described for other ducts in CF. Gaillard et al. (1997) proposed that the term atresia should be used in cases of CBAVD associated with CF mutations, whereas the term agenesis should be used for cases of CBAVD associated with urogenital abnormalities in which defects occur during organogenesis. CFTR mutations and spermatogenesis It is generally assumed that azoospermia and infertility in men with CBAVD are due simply to anatomical obstacle and that the sperm production is normal. The histological examination of the testis (Silber et al., 1990a) revealed either normal spermatogenesis (Goldstein and Schlossberg, 1988; Okada et al., 1999) or hypospermatogenesis (15 30% in CBAVD, 45% in CUAVD) (Weiske et al., 2000; Meng et al., 2001). Some degree of hypospermatogenesis is expected because absence of the vas deferens acts as an obstruction of the testicular epididymal system (Weiske et al., 2000). It has been suggested that the CFTR protein may play a role not only in the development of structures derived from the Wolffian ducts but also in the process of spermatogenesis or sperm maturation (Trezise et al., 1993). A number of cytological abnormalities preferentially occurring in the spermatid stage have been observed in CF and CBAVD men (Gottlieb et al., 1991). Larriba et al. (1998) studied exon 9 splicing in testicular tissues and found that the presence of the 5T allele (but not the presence of one or two mutations) was associated with a significant reduction in mature (elongated) spermatids per tubule in CBAVD males. A recent study reported that testicular volumes were significantly lower in men with CBAVD without renal abnormalities that carried the 5T allele only (without another CFTR mutation) (Robert et al., 2002). CFTR is expressed in rat spermatids, and may be involved in volume reduction that occurs in spermiogenesis (differentiation of spermatids into spermatozoa) (Wong et al., 1998; Gong et al., 2001). It is possible that CFTR plays an important role in the formation of the epididymal fluid microenvironment through a synergistic effect with the water channel aquaporin (Cheung et al., 2003). Spectrum of diseases caused by CFTR mutations: from classic CF to CFTR-opathies Recent advances in mutation detection technology have led to the extensive screening of patients with monosymptomatic diseases with similarity to CF (Figure 9). CFTR carrier status as predisposing to disease? In addition to classic forms of cystic fibrosis, there is now a growing recognition of non-classic or atypical cases of CF presenting in adolescence or adulthood, that manifest symptoms in only one or two organ systems, such as isolated pancreatitis (Boyle, 2003). Furthermore, there is growing evidence of an association between CF mutations and other diseases including allergic bronchopulmonary aspergillosis (ABPA), chronic sinusitis or idiopathic bronchiectasis (an irreversible dilatation of proximal subsegmental bronchi). CF is an autosomal recessive inherited disorder that results when both alleles carry a disease-causing mutation. Carriers of a single mutation are asymptomatic, suggesting that a 50% level of functional CFTR protein is sufficient. This widely accepted notion has been recently challenged by the detection of an increased proportion of carriers for CF mutations or polyvariant mutants in several conditions, such as rhinosinusitis (Wang et al., 2000), idiopathic pancreatitis, idiopathic bronchiectasis (Casals et al., 2004) or primary sclerosing cholangitis (Girodon et al., 2002; Sheth et al., 2003). It is possible that the reduction of CFTR function by 50% by a heterozygous mutation be not compensated in cases where modifiers (genetic or environmental) play additive effects, and it is now postulated that CFTR mutations may be factors of genetic predisposition for the development of certain diseases. The increasing number of CFTR-associated diseases raises considerable debate about where to draw the line between CF and other diseases with molecular defects in the CFTR gene (Zielenski, 2000; Boyle, 2003). Moreover, recent studies suggest that even coding SNP may be predisposing factors (Steiner et al., 2004). Casals et al. (2004) found a M470 allele in 90% of patients with idiopathic bronchiectasis heterozygous for a CFTR mutation, versus 63% in the general population or 40% in bronchiectasis patients without CFTR 31

19 Figure 9. Spectrum of phenotypes associated with the CFTR gene (from Zielenski, 2000; Stern, 1997; Mickle and Cutting, 2000). The spectrum of diseases caused by mutations in the CFTR gene as well as the criteria useful to distinguish the phenotypes have been recently reviewed (Boyle, 2003). Classic CF: this phenotype represents those individuals with a complete loss of CFTR function. Since severe mutations (including F508del) account for about 92% of mutated chromosomes in this category, roughly 85% of patients (i.e ) have two severe mutations (class 1 3 or 6) in trans, either identical (homozygote) or different (compound heterozygote). They have severe pancreatic exocrine insufficiency (CF-PI) appearing early in life, chronic obstructive pulmonary disease, abnormal concentrations of sweat electrolytes and the males are infertile because of obstructive azoospermia due to the absence of the vas deferens. Fifteen per cent of patients have inherited one severe mutation in trans to one mild, or two mild mutations (class 4 5) and are pancreatic sufficient (CF-PS). Non-classic (atypical) CF: in this group are individuals with a CF phenotype in at least one organ system and normal (<40 mmol/l) or borderline (40 60 mmol/l) sweat chloride values. They are usually CF-PS, have at least one mild mutation and are diagnosed after childhood. Males can occasionally demonstrate fertility (ex. mutation kbC T). An example of non-classic CF is recurrent idiopathic pancreatitis (which occurs in patients with functional pancreatic acinar tissue) where a high frequency of CFTR mutations is observed and 10 20% of affected individuals carry two mutations (Castellani et al., 2001). CFTR-related diseases or CFTR-opathies (Noone and Knowles, 2001): they do not fit the criteria for CF and do not follow a Mendelian inheritance pattern as they are associated with a significantly higherthan-expected frequency of CFTR mutations (10 30%) in the heterozygote state, and are mostly influenced by other genes and/or environmental factors. CFTR mutations and polyvariant mutants (including the 5T allele and several coding SNP) appear as predisposing alleles. Examples include primary sclerosing cholangitis (13 37% of affected individuals carry one CFTR mutation), allergic bronchopulmonary apergillosis (ABPA) (10 30%), chronic rhinosinusitis (7 12%) and idiopathic bronchiectasis (20 36%). CBAVD: a much milder phenotype of only CBAVD without lung or pancreatic disease is associated with mild mutations (missense or partial mis-splicing) in CFTR. Most individuals in this group have normal or borderline sweat chloride values, a significant proportion have mild sinopulmonary disease and nasal ion transport abnormalities. Whether they should be included in the group of non-classic CF or in the group of CFTR-related disease remains a subject of controversy (Boyle, 2003); perhaps they could be separated into their own clinical category. Although the CF genotype does not absolutely predict the phenotype, the phenotype is determined by a gradient of CFTR dysfunction dependent on the type of CFTR mutations as well as organ sensitivity (Stern, 1997; Davis et al., 1996; Mickle and Cutting, 2000). Patients with genotypes resulting in about 1% CFTR activity have classic CF with pulmonary disease, pancreatic insufficiency, CBAVD, and abnormal sweat chloride concentrations. Patients with approximately 5 10% CFTR function are pancreatic sufficient. Those with about 10 25% CFTR function have CBAVD alone and may also be at risk for monosymptomatic diseases such as pancreatitis. There is an overlap between atypical CF and monosymptomatic disorders. 32

20 mutations. Sheth et al. (2003) also observed that 89% of patients with primary sclerosing cholangitis carried a V470 variant. CF with male fertility Not all male CF patients are infertile, a few of them (2 3%) having fathered children (Taussig et al., 1972). Most of these men are compound heterozygous for the splicing mutation kbC T (Stern et al., 1995; Stuhrmann and Dork, 2000). The presence of this mutation is usually associated with mild CF forms with borderline or normal sweat chloride, pancreatic sufficiency and variable male genital phenotype (fertility or infertility). This mutation generates an aberrant 5 splice site deep in intron 19 of CFTR pre-mrna and activates a cryptic 3 splice site, resulting in the insertion of a new exon of 84 bp containing a premature in-frame stop codon. Both correctly spliced and aberrant transcripts are produced simultaneously, the relative proportions of which contribute to the presence or absence of the genital tract (Nissim-Raffinia et al., 2000). Is CBAVD a (mild) form of CF? There is increasing evidence that the majority of patients with isolated CBAVD (i.e. presenting for evaluation of infertility but without pulmonary symptoms) also display features of mild respiratory disease (such as sinusitis, nasal polyps, chronic cough, raised levels of Pseudomonas aeruginosa antibodies) when carefully examined by clinicians with expertise in incomplete forms of CF in adult (Osborne et al., 1993; Durieu et al., 1995; Castellani et al., 1999). They have intermediate or elevated sweat chloride levels but normal chest radiographs and spirometry. Moreover, recent data suggest that, even in patients who do not manifest any clinical respiratory symptoms, early subclinical bacterial pulmonary infection and inflammation may occur. Bronchoalveolar lavage samples revealed the presence of pathogens and inflammatory mediators in the lower airways of subjects with CBAVD who presented with laboratory evidence of mild CF (Gilljam et al., 2004). Only observation for a prolonged period of time (several years) will help determine their risk of developing symptomatic pulmonary disease over time. Some CBAVD patients who have elevated sweat electrolytes and two CF-causing mutations or abnormal nasal potential difference should be considered as having CF according to the recent consensus statement on diagnosis (Rosenstein and Cutting, 1998). However, there has been no consensus to make a diagnosis of CF (a severe, potentially lethal disease, diagnosed in children) in patients with CBAVD (a benign condition diagnosed in adult males presenting with an infertility problem), probably because of the psychological and social impact of such diagnosis. In CBAVD there is a normal sodium transport across the nasal epithelium (in CF, there is an increased sodium absorption), but an impaired chloride conductance that is intermediate between CF heterozygotes and CF patients (Osborne et al., 1993). Perhaps defining the exact value of measuring nasal potential difference in men with CBAVD could help to better understand the consequences of CFTR dysfunction in the airways and anticipate the long-term prognosis (Pradal and Piacentini, 2004). Homozygosity for the 5T allele The 5T allele alone is not sufficient to induce a typical CF; it appears as a typical allele for CBAVD. However, there are occasional reports on patients without CF mutations who are homozygous for the 5T allele and demonstrate clinical lung features suggestive of CF, such as the homozygous for TG12T5 described by Noone et al. (2000). Although it is impossible to claim with certainty that homozygosity for the 5T allele is completely benign, any potential risk appears to be most likely limited to infertility in males. In a report on CF screening at a nationwide scale (Strom et al., 2004), there were 219 patients compound heterozygotes for F508del and 5T; there were 184 females, of whom 177 had no CF symptoms and six with asthma or sinus surgery; there were 35 males, of whom 50% experienced CF-like symptoms including 12 with CBAVD; (34%), three with pulmonary symptoms. Of the 20 remaining males, only 10 had fathered children and none of the 20 had pulmonary disease. CFTR genetic analysis in obstructive azoospermia It is important to make the association with CF clear (Figure 10), as it helps to stress the importance of genetic counselling for the CBAVD couple and also for their relatives. Furthermore, as men with CBAVD may be at risk for expression of CF disease later in life, they will benefit from thorough evaluation and subsequent monitoring of lung health by clinicians with expertise in CF. The biggest obstacle to implementing CF testing is the extreme mutational heterogeneity of the CFTR gene, boasting perhaps the greatest number of catalogued individual inherited nucleotide alterations of any gene yet described, so that the technical demands are very arduous. Because the complete scanning of coding sequences is so technically challenging and costly, an initial assessment of correct clinical diagnosis of CBAVD, CUAVD or BEDO is essential prior to the genetic testing, including assessment of the absence of renal abnormalities. Unlike patients with classic CF, who are more likely to have common, homozygous or compound heterozygous mutations, azoospermic patients are more likely to exhibit rare or private mutations, so that detection with routine CF mutation panels, especially commercial tests, will miss most CFTR mutations (Figures 10 and 11). Although an interesting assay using mass spectrometry and primer oligonucleotide base extension for 100 mutations increased detection (Wang et al., 2002), optimal detection of CFTR mutations in CBAVD can be reached only by the application of costly scanning methods able to detect any subtle sequence change in the coding regions of the gene, such as DGGE (denaturing gradient gel electrophoresis) (Costes et al., 1995; de Meeus et al., 1996; Bienvenu et al., 1997; Claustres et al., 2000), SSCP (single strand conformation analysis) (Dork et al., 1997), HET (multiplex heteroduplex analysis) (Mak et al., 1999), or by sequencing (Danziger et al., 2004). A disadvantage of scanning methods is the finding of mutants or variants of unknown clinical significance. It is likely that some of the CFTR mutants described in patients may not be true disease-causing alterations (Claustres et al., 2004), and, conversely, some of the 250 polymorphisms reported to the CF Consortium (half resulting in amino acid substitutions) may play a role in 33

21 Figure 10. Detection of CFTR mutations in CBAVD. Identifying CFTR mutations allows: (i) establishing a diagnosis CFTR mutations would confirm diagnosis and be an explanation for obstructive azoospermia; (ii) establishing a possible genetic origin one mutated allele will always be inherited from the affected male by the offspring; (iii) clarifying the pattern of inheritance documentation of the specific familial CFTR mutations; (iv) offering adequate genetic counselling if the partner of a male with CBAVD is a healthy carrier of a CF mutation, there is up to 25% risk of CF to all children, and prenatal or preimplantation genetic testing is offered. Genetic tests: genomic DNA extracted from peripheral blood then amplified by PCR is the usual starting material for CFTR mutation analysis. Most laboratories search only a limited number of known (common) mutations (from 30 to 100, including the three common alleles at the Tn site) by using commercially available mutation detection kits. (Girodon et al., 2000). A few laboratories have developed scanning techniques that are able to detect any subtle sequence anomaly in the coding sequence. Genomic DNA scanning of CFTR gene is considered a high-complexity test, as exons with their intron boundaries have to be amplified and analysed. Whole-gene scanning methods include SSCP (single strand conformation analysis), DGGE (denaturing gradient gel electrophoresis) and DHPLC (denaturing high performance liquid chromatography). These methods rely on DNA conformation changes or heteroduplex formation of the PCR product if the sample is different from the wild-type control DNA. If an alteration is detected, the corresponding portion of the gene is sequenced to confirm and identify the sequence change. The reported sensitivity of these methods ranges from 75 80% (SSCP) to 95 98% (DGGE, DHPLC). Most CBAVD mutations are not detected by routine panels designed for common CF mutations: in a sample of 800 French men with CBAVD without renal agenesis, testing for the 50 most common mutations associated with the classic CF phenotype ( CF panel) in addition to assessment for IVS8 5T (allele 5T in the intervening sequence or intron 8) detected 959 of 1600 (60%) CBAVD alleles, with two mutations in 47%, one in 24%, and no mutation in 29% of CBAVD patients. In a subset of 327 CBAVD men who were more extensively re-analysed by using a scanning method (DGGE) that detects unknown mutations scattered all over the gene, 516 of 654 (79%) alleles were identified, with 71 and 16% of patients carrying two mutations or one respectively, and only 13% remaining with out any detectable CFTR abnormality. Of 137 mutations identified in this sample, 105 (76%) would have been missed by using a kit detecting only the 31 most common mutations found in the European population affected with cystic fibrosis in addition to the 5T allele (Claustres et al., 2000). Figure 11. A higher number of TG repeats is predictive of a pathogenic 5T allele. The individual with CBAVD presents the same mutation genotype as his fertile father (F508del in trans with a 5T allele). However, further investigation of the familial segregation of CFTR polymorphisms reveals that he has inherited from his mother a 5T-haplotype different from the one he has inherited from his father. Haplotype TG12 5T is more detrimental than haplotype TG11-T5. 34

22 disease by interfering with splicing signals, causing some degree of mis-splicing of the gene (Steiner et al., 2004). Up to 80 87% of men with isolated CBAVD have at least one CFTR mutation (Anguiano et al., 1992; Patrizio et al., 1993; Oates and Amos, 1994; Mercier et al., 1995; Dork et al., 1997; Claustres et al., 2000). The inability to identify a CFTR mutation in a fraction (13 20%) of CBAVD males without renal abnormality, even after the entire coding region of the CFTR gene is analysed, could indicate a failure to detect all possible CFTR mutations, as the techniques based on genomic DNA scanning do not detect deep mutations within the introns or large deletions of the gene. As in CF, the analysis of mature mrna could reveal the presence of splicing anomalies or complex gene rearrangements (Audrezet et al., 2004). However, it is unlikely that so many CBAVD alleles be caused by such mutations. Alternatively, these cases might represent a distinct clinical aetiology perhaps due to the involvement of other gene (s). Genetic counselling CBAVD males and reproduction The failure to reproduce for men with CBAVD is thought to be primarily due to the lack of normal anatomic transport (because of absent vas deferens) of spermatozoa from the testes. Most patients with CBAVD have normal or subnormal spermatogenesis, and as the caput portion of the epididymis is invariably present, microsurgical sperm retrieval can be undertaken from the epididymis to allow men with azoospermia to father their own biological children. The first pregnancy for a couple in whom the man had CBAVD was reported in 1987 (Silber et al., 1988). The initial IVF cycles yielded poor oocyte fertilization rates. Since 1993, couples have been offered intracytoplasmic sperm injection (ICSI, in which a single sperm is injected into the cytoplasm of a mature oocyte to obtain a viable embryo) with a considerably improved outlook, resulting in many pregnancies with live births (Tournaye et al., 1994; Silber et al., 1995). Although lower fertilization (Patrizio et al., 1993) or embryo implantation (Hirsh et al., 1994) rates have occasionally been reported in couples with CBAVD, the presence of CFTR mutations in men with CBAVD does not seem to affect sperm function during IVF with micromanipulation (Schlegel et al., 1995; Silber et al., 1995). The chance of conception has been estimated as about 31% per cycle with a take-home-baby-rate of 23% (Silber et al., 1990b). A recent meta-analysis of the outcome of ICSI showed that outcome is not affected by whether the retrieved spermatozoon is fresh, frozen, epididymal or testicular, but suggested a trend to lower fertilization rate and higher miscarriage rate in CBAVD CFTR than acquired causes of obstructive azoospermia (Nicopoullos et al., 2004). The first successful PGD performed for a couple with CBAVD, where both partners were heterozygous for mutation F508del, was reported by Liu et al. (1994): three carrier embryos were transferred and a healthy boy was born. The presence of CF or CBAVD mutations in the male partner does not appear so far to significantly compromise in-vitro oocyte fertilization, embryo implantation rates, or the successful delivery of asymptomatic children after PGD (McCallum et al., 2000; Phillipson et al., 2000). Risk of CF, CBAVD or CFTR-opathies for offspring For a couple with CBAVD associated with CFTR defects, planning to have their own genetic children, the risk for both male and female offspring to have CF or related diseases and for male offspring to have CBAVD depends on whether or not the female partner is a carrier, since one mutated allele will always be inherited from the male. As the carrier frequency of CFTR mutations in many Caucasian populations is in the order of 1/22 1/30, it is highly recommended that genetic testing for CFTR mutations be offered to the couple prior to ICSI. The genetic counsellor should determine whether there is a family history of CF and the couple s ethnicity, as this will affect their carrier risk. The genetic of CBAVD is more complex than in CF, as (i) genetic analysis is able to prove but not to exclude the diagnosis of a genital form of CF, and (ii) the risk of CF or CBAVD in the offspring may be unpredictable when rare mutations are identified in the male or the female. The couple should be informed that the test cannot detect all mutations within the gene; therefore, a negative mutation screen reduces, but does not eliminate, the risk of being a carrier (Figure 12). The most plausible phenotypes supposed to result from combinations of CFTR mutant alleles in CBAVD males and their partners have been described (Lissens et al., 1996). If both male and female carry a severe mutation, then the risk of a child with a severe CF is 1/4, but there is also an additional 1/4 risk of mild CF or CBAVD conferred by the inheritance of the severe mutation of the female and the mild of the male. If the CBAVD male is a carrier of one severe mutation, and his partner screens negative, then the risk of having a CF child can be calculated as 1/964 (0.1%) if the sensitivity of the test is 90% [this risk would increase up to 1/324 (0.3%) if the test detects only 70% of mutations in the ethnic background of the partner]. While these couples are at increased risk over the general untested population, they are not suitable candidates for prenatal diagnosis. A proportion of heterozygote females will be missed by using only a CF mutation panel; on the other hand, using a scanning method cannot be offered in all couples and may introduce new dilemmas for genetic counselling, by identifying previously uncharacterized sequence alterations, the clinical significance of which is unclear. Some combinations of CFTR mutations or variations in the couple imply an impossible prediction of the phenotypic consequences for their offspring: a rare missense mutation combined with a severe CF mutation may result in either a normal phenotype, or CBAVD or even CF. Furthermore, incomplete penetrance of CFTR mutation combinations may lower the risk of having a child with CBAVD or CFTRopathies for couples undergoing sperm retrieval and ICSI. In most countries, the risk of CF is the only one that is ethically sufficient to justify assisted reproduction and treatment by prenatal diagnosis (PND, performed on DNA extracted from chorionic villus sampling or amniotic fluid puncture) or PGD. Because these couples have to undergo assisted reproduction, PGD seems an advantage over PND, as the genetic status of the embryos obtained after ICSI with epididymal spermatozoa can be assessed for CFTR mutations before they are replaced in the uterus. PGD is usually performed to avoid the transfer of embryos carrying two severe mutations (Liu et al., 1994), although it may also be performed to avoid the transfer of embryos carrying combinations for which it is impossible to 35

23 Figure 12. Risk calculation and phenotype prediction (risk is calculated with an assumption of a CF carrier frequency of 1 in 25). (i) Female partner is negative for the test: the residual risk for a partner from the general population of being a carrier after a negative screening for a fraction a of CFTR mutations in her ethnic background will be Z = q(1 a)/(1 aq), with q the prior carrier risk (ten Kate, 1990). The risk for this couple for having a CF child is calculated as follows: 1 (he carries a severe mutation) 1/241 (her test is negative, 90% sensitivity) 1/4 = 1/964. (ii) Female partner is a CF carrier: in this case, the inheritance will follow classical Mendelian rules for autosomal recessive inheritance, and the couple has a one in four risk for having a child affected with CF. Background information on calculation above: Because CF is an autosomal recessive disorder, individuals who are heterozygous carriers (frequency around 1/25 in Caucasian populations) are not clinically distinguishable from individuals with two normal alleles. The Hardy Weinberg law states that the sum of the normal allele frequency (p) and the mutant allele frequency (q) equals 1 (p + q = 1). Since the frequency of the CF disease (individuals with two mutant CF alleles) in the population is known (q 2 = 1/2500), the mutant allele frequency, which is the square root of the CF disease frequency, is q = 1 in 50 or Therefore, the CFTR carrier frequency, which is defined as 2pq, is (2 49/50 1/50). The probability for a man to inherit both the 5T allele from one parent and a CFTR gene mutation from the other parent is 1 in 4 or 25%. Therefore, it can be estimated that approximately one male in 1666 should have a CBAVD due to inheritance of a 5T allele in trans of a CF mutation [0.102 (5T carrier frequency) 0.04 (CFTR carrier frequency) 0.25 (chance of inheriting both 5T allele and a mutation) 0.60 (penetrance of the 5T allele) = 0.06% (1/1666)]. 36 predict the phenotype (for instance, couples with only one mutation detected in the CBAVD male and one or two mutations present in the female) (Lissens, 1997; Phillipson et al., 2000). Methods for detecting point mutations are continuously improved, and, recently, two powerful approaches have been demonstrated to detect mutation F508del at the single cell level, the dhplc (denaturing high performance liquid chromatography; Girardet et al., 2003) and the microarray technology (Salvado et al., 2003). A major advantage of the microarray approach is that it will allow testing for the relevant mutations and, simultaneously, monitoring for the two main technical difficulties in PGD, ADO (allele drop-out) and contamination through the detection of SNP that are linked to disease-specific mutations. The clinical examination and follow-up of children born to couples with CBAVD will be essential to understand the variable phenotypic expression of CF gene mutations, particularly in cases of homozygosity for the 5T allele or compound heterozygosity for a severe mutation and the 5T allele or a severe mutation and R117H-7T. The identification of CFTR mutations in a CBAVD male has important implications not only for himself but also for his family. Healthy siblings of a CBAVD male have a 50% chance of being CF carriers. The possibility of testing and counselling should also be offered to family members, who should be informed about their increased risk of having CF children if they have inherited a CF mutation and their partner is also a carrier. If no mutation is found in the male, the presence of an undetected mutation cannot be definitely excluded but the risk for a CF child is probably lower than 1/1000. It is important to know if the CBAVD patient has brothers with or without CBAVD, as the study of familial segregation of CFTR polymorphisms could also contribute to the discrimination of CFTR-linked and unlinked conditions of CBAVD. In couples with male infertility not related to CF, the risk of having children with CF is not different from the risk for couples in the general population. The risk of infertility in their male offspring is unknown. In conclusion, even with the current state of knowledge, genetic counselling of couples with CBAVD remains very difficult. It is time to initiate collaborative studies to update data on the distribution and frequency of CFTR mutants and variants that have been found after exhaustive screening of both CBAVD patients and their partner, and to determine their prevalence in the general population of different ethnic groups. Risk assessment is an essential component of genetic counselling and testing, and we need consensus risk calculations that are adequate in the context of particular CBAVD clinical scenarios encountered in genetics practice. These calculations should take into account not only CFTR genetics but also relevant information resulting from clinical and bioelectric data specific to CBAVD patients. We also need consensus guidelines for the number of mutations and/or the extent of CFTR genetic analysis to be performed, as well as the choice in PND versus PGD to be offered to CBAVD couples.