Spontaneous rescue from cystic fibrosis in a mouse model

Background From the original CftrTgH(neoim)Hgu mutant mouse model with a divergent genetic background (129P2, C57BL/6, MF1) we have generated two inbred CftrTgH(neoim)Hgu mutant strains named CF/1-CftrTgH(neoim)Hgu and CF/3-CftrTgH(neoim)Hgu, which are fertile and show normal growth and lifespan. Initial genome wide scan analysis with microsatellite markers indicated that the two inbred strains differed on the genetic level. In order to further investigate whether these genetic differences have an impact on the disease phenotype of cystic fibrosis we characterised the phenotype of the two inbred strains. Results Reduced amounts, compared to wild type control animals, of correctly spliced Cftr mRNA were detected in the nasal epithelia, lungs and the intestine of both inbred CftrTgH(neoim)Hgu strains, with higher residual amount observed for CF/1-CftrTgH(neoim)Hgu than CF/3-CftrTgH(neoim)Hgu for every investigated tissue. Accordingly the amounts of wild type Cftr protein in the intestine were 9% for CF/1-CftrTgH(neoim)Hgu and 4% for CF/3-CftrTgH(neoim)Hgu. Unlike the apparent strain and/or tissue specific regulation of Cftr mRNA splicing, short circuit current measurements in the respiratory and intestinal epithelium revealed that both strains have ameliorated the basic defect of cystic fibrosis with a presentation of a normal electrophysiology in both tissues. Conclusion Unlike the outbred CftrTgH(neoim)Hgu insertional mouse model, which displayed the electrophysiological defect in the gastrointestinal and respiratory tracts characteristic of cystic fibrosis, both inbred CftrTgH(neoim)Hgu strains have ameliorated the electrophysiological defect. On the basis of these findings both CF/1-CftrTgH(neoim)Hgu and CF/3-CftrTgH(neoim)Hgu offer an excellent model whereby determination of the minimal levels of protein required for the restoration of the basic defect of cystic fibrosis can be studied, along with the modulating factors which may affect this outcome.


Background
Cystic fibrosis (CF) is a severe autosomal recessive disorder characterised by defective chloride transport in epithelial cells and excess mucus secretion. It is caused by mutations in the cystic fibrosis transmembrane conductance regulator gene (CFTR), resulting in defective cAMPdependent chloride conductance [1].
Isolation of the mouse Cftr gene enabled the modelling of the disease in the mouse by gene targeting in embryonic stem cells. Dorin et al [2] described a transgenic mouse model Cftr TgH(neoim)Hgu /Cftr TgH(neoim)Hgu , generated following targeted insertional mutagenesis into exon 10 of the murine Cftr gene in embryonic stem cells. Unlike the Cftr mutants created by gene replacement low levels of wild type Cftr mRNA were produced as a result of exon skipping and aberrant splicing [3]. In the human, CFTR mutations affecting the pre-mRNA splicing of the gene can also generate both aberrant and correct transcripts, the level of which varies among patients and among organs of the same patient [4][5][6][7][8]. Patients carrying such mutations show variability of disease expression, which has been found to be inversely correlated with the levels of correctly spliced transcripts, such that lower levels are associated with severe disease and higher levels with milder phenotype, suggesting a role for splicing regulation as a genetic modifier [9]. The original Cftr TgH(neoim)Hgu /Cftr TgH(neoim)Hgu mouse suffered from only mild intestinal obstruction, but it exhibited typical pathophysiological features of CF in gut, lung and reproductive tract [2,10]. Cftr TgH(neoim)Hgu / Cftr TgH(neoim)Hgu mutant mice displayed the electrophysiological defect in the gastrointestinal and respiratory tract characteristic of CF and could be unequivocally distinguished from their non-cf littermates (Cftr TgH(neoim)Hgu /+ and +/+) on this basis [2,11]. The residual amounts of correctly spliced mRNA apparently led to the production of some functional Cftr protein that was sufficient to ameliorate the otherwise fatal intestinal obstruction but not enough to ameliorate the intestinal electrophysiological disease phenotype [3].
From the original Cftr TgH(neoim)Hgu mutant mouse model [2] with a divergent genetic background we have generated two inbred Cftr TgH(neoim)Hgu mutant lines named CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu using brother × sister mating for more than 30 generations. Phenotypic evaluation of the inbred CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu animals indicated that both strains have ameliorated the basic defect of CF with a presentation of a normal electrophysiology in both the intestinal and nasal epithelium. Unlike leaky splicing mutations in the human the reduced amounts of correctly spliced mRNA and the subsequent low amounts of wild type Cftr protein substantiated for the restoration of the disease phenotype in the two inbred Cftr TgH(neoim)Hgu strains. Based on this finding both CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu offer an excellent model to determine the minimal levels of wild type Cftr required to ensure a normal chloride secretory capacity of the intestine preventing intestinal obstruction.
There do exist a number of case reports in the literature of how a deleterious mutation is partially corrected by the affected patient. First, the interaction of drugs and/or proteins with sequences in the 3' untranslated region may lead to stabilisation of the mRNA transcript [12]. Second, nonsense-associated altered splicing is a putative correctional response that up-regulates alternatively spliced transcripts that have skipped offending premature termination codons. Third, aminoterminal located nonsense mutations may be partially rescued by the expression of the partially functional amino-truncated protein generated from an alternative in-frame translation start site. These attenuation mechanisms have been reported for mild forms of non-Herlitz junctional epidermolysis bullosa [13], the laryngo-onycho-cutaneous syndrome [14] and X-linked adrenal hypoplasia congenita [15]. Our CF mouse model is a prototype of how a genuine lethal genetic predisposition can be spontaneously bypassed or compensated by post-transcriptional rescue. Genome wide microsatellite genotyping CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mice of generation F 26 were typed at a 20 cM spacing with microsatellite markers to assess their genome-wide status of homozygosity and mutual genetic diversity. The criterion for the selection of the first 78 markers was their informativity to differentiate between the three reference strains BALB/c, C57BL/6 and DBA/2. Informative loci between the two inbred CF strains were genotyped with flanking markers in order to map the regions of genetic heterogeneity, bringing this to a total of 105 microsatellite markers.

Generation of the inbred CF/1-Cftr TgH(neoim)Hgu and CF/3-
Sixteen out of the 105 microsatellites were polymorphic between CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu [see Additional File 1] indicating that these two strains are not identical to one another at all loci. Residual heterozygosity was detected for eight microsatellites [see Additional File 1] three of which were segregating in both CF strains. CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu differ from C57BL/6, BALB/c and DBA/2 strains in 58, 52 and 58 of the 105 markers, respectively [see Additional File 1]. In summary, the two CF inbred strains are highly similar to one another differing in eight chromosomal regions, but they deviate significantly from the three common inbred laboratory strains C57BL/6, DBA/2 and BALB/c.

Cftr mRNA analysis
In order to determine whether CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mutant animals produce correctly spliced wild type Cftr mRNA as previously described for their progenitor with a mixed genetic background [2,3] we used RT-PCR spanning exon 9 to exon 13 encompassing the vector insertion site, using wild type MF1 mouse as control. Correctly spliced wild type Cftr mRNA corresponded to a cDNA product of 555 bp in size. Sequencing of the 555 bp PCR products obtained for the CF/1-Cftr TgH(neoim)Hgu mutant mice (ileum) revealed that the observed reduced amounts correspond 100% to wild type Cftr cDNA sequence. Moreover, a fragment of approximately 750 bp was detected in samples of homozygous CF mutants but not in wild type animals. Sequencing of this product demonstrated that it corresponded to the previously described [3] alternatively spliced mRNA encompassing the 206 bp fragment from the targeting vector pIV3.5H spliced between exon 9 and exon 10. However, this alternatively spliced mutant mRNA unlike in the original report [3] was not consistently detected.
The amounts of total and correctly spliced wild type Cftr mRNA in lungs and intestine were determined at different stages of the inbreeding procedure ( Table 1). The total amount of Cftr mRNA did not significantly change over time in lungs and intestine and was about 1-to 4-fold CF/3-Cftr TgH(neoim)Hgu mice (n = 4) demonstrated the presence of fully glycosylated isoform (band C) of Cftr in the small intestine of these mice ( Figure 1). The intensity of the C band of these mice was reduced when compared with the intensity of the C band of the wild type MF1 mice (n = 2). By running individual samples of the CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mutant mice together with serial dilutions (5×, 10×, 20×, 50×, 100×) of membranes from the MF1 wild type mice the amount of mature CFTR (band C) in the two inbred mutant mice was estimated to be 9.2 ± 1.5% of wild type for CF/1-Cftr TgH(neoim)Hgu and 4.2 ± 2.3% of wild type Cftr protein for CF/3-Cftr TgH(neoim)Hgu .
Histological analysis of intestinal sections from mice of generations 27 and 28 revealed focal hypertrophy of goblet cells in the CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu homozygous mutant mice in the ileum ( Figure 2). However, this hypertrophy was not observed in sections obtained from the jejunum in both inbred Cftr TgH(neoim)Hgu strains ( Figure 2).

Cftr immunocytochemistry
In inbred F 28 CF/1-Cftr TgH(neoim)Hgu staining with the R3195 Cftr antibody revealed weak and mainly diffuse signals with some apical labelling in the upper villus region of the ileum, whereas no specific immunoreactive signals could be detected in sections from duodenum and jejunum ( Figure 3). Sections of the intestine (duodenum, jejunum, ileum) of F 27 CF/3-Cftr TgH(neoim)Hgu mice showed substantially weaker staining at the limit of detection.
In the nasal respiratory epithelium CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mice exhibited a higher response after exposure to the sodium channel inhibitor amiloride in comparison with MF1 wild type, but not in comparison with C57BL/6J mice (Figure 5b). The forskolin activated cAMP-triggered chloride secretory response was indistinguishable from wild type C57BL/6J in CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu . The response of MF1 wild type mice was about two fold higher compared to inbred C57BL/6J wild type animals.
The response of CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mice to forskolin was indistinguishable from wild type C57BL/6J in the respiratory epithelia and intermediate between wild type C57BL/6J and MF1 control animals in the intestine. We envisage that during inbreeding modulating mechanisms were selected in these mice that influenced Cftr expression at the level of transcript, protein and function.

Cftr splicing
Both novel inbred Cftr TgH(neoim)Hgu showed the ability to produce correctly spliced Cftr mRNA as previously reported for their outbred progenitor with a mixed genetic background in the respiratory and intestinal epithelia [3]. Interestingly, the total amount of Cftr mRNA did not change during inbreeding, but the levels of correctly spliced Cftr mRNA decreased to different extent in the two mutant inbred strains, with CF/1-Cftr TgH(neoim)Hgu showing higher levels for every investigated tissue. CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu were divergent in 16 out of 105 microsatellite genotypes in the genomewide scan [18], verifying the genetic difference between the two novel Cftr TgH(neoim)Hgu inbred strains. This data suggests that interstrain genetic variation at non-Cftr loci could account for the phenotypic variation at the level of Cftr splicing observed between the two novel mutant strains.
In the intestine of both CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu the levels of Cftr expression are highest in   [19][20][21], or affect tissue-specific control of CFTR expression [22]. Whether the tissue specific levels of Cftr expression depend on and/or are regulated by the levels of such splicing factors in these two strains remains to be investigated. Furthermore, in agreement with the previous study on the original Cftr TgH(neoim)Hgu mice with a mixed genetic background [3] the novel mRNA species that contained a 206 bp exon of vector sequence spliced between exon 9 and exon 10 was occasionally observed in both inbred strains. The 206 bp exon contains an in frame stop at a position less than half way into the message [3]. Such a premature stop codon may lead to exon skipping as has been demonstrated in several human genetic diseases [23][24][25][26] and could hence explain its inconsistent presence in samples from homozygous CF mutant mice.

Cftr activity restoration: Efficient protein processing, trafficking and function
Regardless of the difference in correctly spliced Cftr mRNA levels both CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mice synthesized and processed Cftr protein more efficiently in the ileum than wild type animals. Starting from small amounts of Cftr mRNA with wild type sequence, a higher proportion of protein was adequately folded, passed the ER quality control and was processed to complex-glycosylated ('band C') Cftr isoforms. Apparently the immature ER isoforms were more efficiently stabilised by molecular chaperones and protected from degradation than in normal mice.
The chloride secretory responses normalized to the pool of band C Cftr were 5-and 10-fold higher in CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu intestinal epithelia, respectively, than in wild type murine tissues. Several mechanisms could account for the higher ion flow per unit time and mass of mature Cftr in CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mice. First, short-circuit current measurements are an indicator of transepithelial chloride transport and thus do not directly measure Cftr chloride channel activity. In wild type mice the transepithelial import of chloride is rate limiting, but at the 10-20 fold lower concentrations in our CF/1-Cftr TgH(neoim)Hgu and CF/ 3-Cftr TgH(neoim)Hgu mice the Cftr mediated current is the rate limiting step for chloride ion flow. At this lower density the relatively larger electrochemical driving force for the apical chloride exit will lead to relatively larger chloride flow/channel. In addition, efficient Cftr trafficking from the post-ER compartment to the apical membrane, along with a reduction in the lysosomal degradation may result in a higher proportion of mature protein localised in the apical membrane compared to the subapical compart- ments. Various components have been shown to influence maturation and trafficking of CFTR from the ER/ Golgi compartment to the basolateral and apical compartments, i.e. COPI-coated vesicles [27], PDZ-containing proteins such as Csp [28], EBP50 [29,30], E3KARP [29], CAP70 [31], CAL [32], SNARE proteins [33], BAP31 [34] and numerous Rab proteins [35,36].
Third, the high ratio of forskolin to genistein Clsecretory responses suggests that even though there is a lower estimated channel density in the CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu compared to wild type in the apical membrane, the proportion of active channels, i.e. in an open state, is higher compared to wild type. The open probability of the CFTR channel is influenced by phosphorylation and ATP hydrolysis [37] and by the interaction with other proteins [38]. Phosphorylation of the regulatory domain is accomplished primarily by elevation of cAMP and activation of protein kinase A. Stronger adrenergic activation and higher levels of cAMP will sustain highly phosphorylated, more active CFTR channels. Moreover, since CFTR channel activity increases with the concentration of ATP [39], elevated intracellular ATP will allow for more CFTR-mediated chloride flux. In summary, higher yields of ion flow per unit mass mature Cftr can be achieved by the larger electrochemical driving force for the apical chloride exit and the modulation of a) posttranslational processing and trafficking, b) signalling and c) metabolic rate. The repeated cycles of inbreeding selected animals which were capable of producing sufficient litters and showed no severe CF disease symptoms. Hence, the rescue of CF to an almost normal phenotype in mice which remained genetically CF can be explained by the selection of genetic modifiers and adaptive mechanisms together with modulators operating at the transcriptional level. Our animal model is an extreme example of how a lethal genetic predisposition can be compensated.
Comparison with the human CFTR splice mutations that lead to the simultaneous presence of correctly and aberrantly spliced transcripts cause CF disease due to insufficient levels of the normal mRNA, and hence, functional protein. Among these are mutations leading to inclusion of intronic sequences, such as 3849+10 kb C-T [40], mutations leading to the skipping of CFTR exons, such as 2789+5 G-A [41], and the TG m -T n genotype at the CFTR intron 8 splice acceptor site [42][43][44].
Studies have reported a wide range of normal CFTR transcripts, from less than 4% to 20% [5,42,45] associated with "leaky" splicing mutations in the human. Disease severity also varies among individuals carrying these splicing mutations, ranging from healthy fertile individuals to the full presentation of classic CF. We know compound heterozygous 2789+5 G-A CF siblings of whom one is suffering from lung disease since adolescence, whereas the other enjoys normal daily life without any restrictions of physical activity still by the age of 60. An inverse correlation was found between the level of the correctly spliced mRNA and the severity of the disease [8,45]. In vitro studies in heterologous expression systems have shown that cellular factors such as hnRNP A1, ASF/SF2, SR proteins and TDP-43 that promote exon skipping and/or exon inclusion modulate the relative proportions of correctly to aberrantly spliced mRNA isoforms [21,46]. Moreover splicing is controlled by cis-regulatory enhancer and silencer elements part of which is under promoter control [6,47,48].
The major difference in the genotype -phenotype correlations of 'leaky' splice mutations between human individuals and the CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mouse models resides in the fact that the manifestation of disease in humans is attenuated with increasing amounts of correctly spliced CFTR mRNA whereas the inbreeding of the mice selected for tissue specific modulation of lowered wild-type Cftr mRNA levels.
Under the standardised conditions of the animal laboratory post-transcriptional mechanisms apparently gave rise to the rescue of the CF phenotype and the partial compensation of the basic defect. These yet undescribed factors that led to more efficient processing and higher channel activity of Cftr should also play a role for the modulation of disease in humans. The CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mice are the adequate model to identify these beneficial modifiers that ameliorate the CF phenotype distal of the primary splicing defect.

Conclusion
In this study we describe the phenotypic characterisation of two novel inbred cystic fibrosis mutant strains generated from an outbred colony of the original Cftr TgH(neoim)Hgu founder with a mixed genetic background (129P2, C57BL/6, MF1). Unlike their outbred progenitor low amounts of wild type Cftr protein substantiated for the amelioration of the disease phenotype in both CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu inbred mutant strains. The post-transcriptional rescue of disease-causing mutations is an emerging field of molecular medicine [49,50]. We envisage that both CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu can become excellent models for the identification of those mechanisms responsible of disease compensation and will increase our knowledge about genetic modifiers of inherited diseases in mammals.

Generation of inbred Cftr TgH(neoim)Hgu mutant mice
For the gene targeting and germline transmission to generate the Cftr TgH(neoim)Hgu mice, the embryonal stem cell line E14 was used [2]. E14 originates from the inbred mouse strain 129 (subline 129/Ola-Hsd, abbreviated designation: 129P2) [2]. These cells were injected into blastocysts of the inbred strain C57BL/6J. The resulting chimeric offspring was crossed to C57BL/6J and to the outbred strain MF1 for several times to increase reproduction [2]. By generation F 4 the same Cftr TgH(neoim)Hgu / Cftr TgH(neoim)Hgu male was mated with two Cftr TgH(neoim)Hgu / Cftr TgH(neoim)Hgu female littermates. Two separate litters were obtained and two animals of each litter became the starting population for the establishment of the two individual inbred Cftr TgH(neoim)Hgu lines CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu which were generated by brother-sister mating for now more than 35 generations. Only one breeding pair per generation was utilized for the continuation of the line. In other words, littermates always shared the same ancestors. Cftr genotyping from tail or ear DNA extractions was conducted by Southern blot hybridization of XhoI/SalI-restricted genomic DNA with the cftr intron 10 probe 1.2 H [2].

Genotyping of microsatellites
High molecular weight DNA was isolated from spleen tissue (from ten F 26 CF/1-Cftr TgH(neoim)Hgu , ten F 26 CF/3-Cftr TgH(neoim)Hgu , two BALB/c, two C57BL/6 and two DBA/ 2) [51]. The primers flanking the 105 chosen microsatellite regions were obtained from the MGI database [52]. Microsatellite markers were genotyped in 96 well plates purchased from Greiner, Frickenhausen, pre-coated with 50 ng DNA per well in a Hybaid Thermocycler (Hybaid, Teddington) with a heated lid. One of the two primers per microsatellite was 5'-terminal biotinylated (Invitrogen Ltd). PCR was performed for all microsatellites in a total volume of 30 µl, without oil overlay, using InViTaq polymerase (InViTek, Berlin) according to the instructions by the manufacturer. Following PCR an 8 µl aliquot was transferred to a multiwell plate and allowed to dry overnight at 37°C, dissolved in 10 µl loading buffer (0.2% w/ v xylenecyanol and bromphenolblue in formamide) and denatured for 5 min at 95°C. The PCR products were separated by direct blotting electrophoresis (GATC 1500, MWG Biotech, Ebersberg, Germany) on a denaturing acrylamide gel (4% acrylamide/N,N'-methylenebisacrylamide 29:1 containing 6 M urea in 0.9 M Tris-0.9 M boric acid-0.02 M EDTA buffer) and simultaneously transferred to a Hybond N+membrane (Amersham). Signals were visualised by blocking the membrane in 1.5% (w/v) of blocking reagent in Buffer 1 (100 mM Tris-HCl, 150 mM NaCl, pH 7.5), followed by incubation in diluted solution of anti-biotin alkaline phosphatase conjugate in Buffer 1.
The membrane was further washed three times with 1% Triton X-100 in Buffer 1 and equilibrated for 15 min in assay buffer (100 mM Tris-HCl, 100 mM NaCl, 50 mM MgCl 2 , pH 9.5). The membrane was covered for 5 min with reaction buffer containing 10%(v/v) Sapphire II (Tropix) and 600 µl CDPstar (Tropix) (12.5 mM) in 50 ml assay buffer, followed by rinsing with a solution containing 10% v/v Sapphire II and 60 µl CDPstar (12.5 mM) in 50 ml assay buffer. Signals were exposed to Kodak XA-R films. Evaluation of results was performed as described by Mekus et al [53].

Isolation of RNA and quantitative RT/PCR
All tissue samples from Ztm: MF1 and F 8-10 as well as F 18 CF mice were immediately frozen in liquid nitrogen after extirpation and stored at -70°C until used. Prior to the isolation of RNA, all solutions and plasticware were treated for at least 12 h with 0.04% (by vol.) aqueous diethyl pyrocarbonate solution. RNA was isolated essentially as described by Chomczynski and Sacchi [54]. Briefly, tissue was ground with a pestle in a mortar under liquid nitrogen and then homogenized in denaturing solution (4 M guanidine thiocyanate, 25 mM sodium citrate, 0.5% sarcosyl, 200 mM β-mercaptoethanol). The homogenate was mixed sequentially with 1 M sodium acetate (pH 4), phenol and chloroform/isoamyl alcohol; the mixture was centrifuged and the RNA removed from the upper aqueous phase. Following isopropanol precipitation, the RNA pellet was redissolved in denaturing solution, reprecipitated with isopropanol, washed with ethanol, dried, dissolved in DEPC-treated water and stored at -70°C until used. The quality of the preparation was checked by UV spectrophotometry and ethidium bromide stain of 2 µg RNA separated by 1% agarose gel electrophoresis. The absence of contamination by DNA in the RNA preparation was verified by PCR with oald141N 5'-GGCAAG-GGCATCCTGGCTGCAGA and oald581I 5'-TAACGGGCCAGAACATTGGCATT. These oligonucleotide primers yield a 421-bp product for both murine genomic DNA and reverse-transcribed murine RNA (PCR conditions per cycle: annealing at 62°C, synthesis at 72°C, denaturation at 92°C 60 s each).
Aliquots of the same oligoT-reverse transcribed reaction mix were amplified in parallel in separate tubes in the same blocks of the thermocycler with primer pairs A1-A2, B1-B2 and oald141N-oald583I, respectively. During the exponential phase of the amplification of the respective cDNA aliquots of 10 µl were withdrawn in 2-cycle intervals and separated by 5% PAGE. The dried gel was exposed to a Fuji imaging plate type BAS-IIIs in order to quantify the yield of cDNA product by photoimaging signal. For each RNA preparation, the optimal window of PCR kinetics had been determined for each of the three cDNA sequences in a prior pilot PCR. For quantification of the cftr mRNA contents in the samples, data sets for each murine tissue were normalized to its invariant aldolase mRNA contents in order to adjust to the variable quality of mRNA template in the RNA preparations and assay-to-assay variability of the reverse transcription. This approach was verified by the observation that the repeated preparation of RNA from one tissue specimen showed similar variations in aldolase cDNA kinetics as tissue RNAs from unrelated animals. The quantitative amount of cftr mRNA in the various samples was calculated from the complete data set for one tissue whereby the experimental curves log [cDNA] vs. PCR reaction cycle n were converted into calibration curves cftr RNA molecules vs. PCR cycle for various yields of cDNA [55]. Only those datasets were evaluated that within experimental error had given indistinguishable Cftr cDNA formation kinetics with primer pairs A1-A2 and B1-B2 for reverse-transcribed Ztm:MF1 RNA (positive control). Cftr mRNA contents of each RNA preparation was assayed at least in triplicate.
In case of the analysis of F 32 CF/1-Cftr TgH(neoim)Hgu and F 32 CF/3-Cftr TgH(neoim)Hgu animals the tissues (duodenum, jejunum, ileum, nose) were stored in RNAlater (Ambion) after extirpation and stored at -20°C until use. RNA was extracted using the RNAeasy mini kit (Qiagen) following the instructions of the manufacturer. Extracted RNA was stored at -80°C until use. The amount of aldolase and correctly spliced Cftr mRNA was determined by semiquantitative RT/PCR kinetics. We titrated for the first reaction cycle, when the gel-separated cDNA became visible by ethidium bromide fluorescence during the late exponential phase of PCR [56]. This procedure yielded an estimate of mRNA concentration within half an order of magnitude, whereas the interexperimental error of the quantitative PCR on mRNA contents in F 8-10 and F 18 CF mice that measured the incorporation of labelled nucleotide during the exponential phase of PCR was twofold or less.

Sequencing
Following RT-PCR cDNA products were sequenced by Qiagen GmbH.

CFTR antibody
The rabbit polyclonal antibody R3195 was raised against a thyroglobulin-conjugated 13 amino acid COOH-terminal peptide sequence of rodent CFTR and was affinity purified on a peptide epoxide-activated Sepharose column, eluted with 4.9 M MgCl 2 , dialysed and concentrated. CFTR labelling specificity has been demonstrated previously in western blot and immunocytochemical assays by the loss of immunostaining in tissue specimens from CFTR -/mice [58].

Immunocytochemical analysis
Intestine: Wild type (MF1 and C57BL/6J) and mutant CF/ 1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mice were sacrificed by cervical dislocation, the intestine was dissected and the jejunum, duodenum and ileum were rinsed with ice-cold saline and fixed in 3% (w/v) paraformaldehyde for 16 h, prior to standard paraffin embedding. Sections (5 µm) were deparaffinised, followed by microwave treatment in 0.01 M sodium citrate solution according to Devys et al. [59]. Endogenous peroxidase activity was blocked by a 30 min pre-incubation in 0.1 M PBS, 0.6% (v/v) H 2 O 2 and 0.12% (w/v) sodium azide. Subsequently, sections were incubated with antibody R3195 (1:100) at room temperature for 1.5 h followed by a 60 min incubation with a peroxidase-conjugated secondary antibody. Enzymatic detection of antigen-antibody complexes was achieved by incubation in substrate solution containing H 2 O 2 and 3,3'-diaminobenzidine tetrahydrochloride (DAB; Serva). Finally, the sections were counterstained with haematoxylin. Labelling specificity was verified by incubations without primary antibody. In both cases, background labelling appeared negligible.

Histological analysis
Deparaffinised sections (5 µm) were stained with Alcian blue (1% w/v in 3% v/v acetic acid, Sigma) for 30 min at RT, rinsed in tab water and counterstained with Nuclear fast red (Sigma) for 10 min.
Mouse ileum was excised under hypnorm/diazepam anaesthesia and reverted on a plastic rod. The muscle layer was cut longitudinally using a blunt razor blade and was stripped of fat manually. The stripped tissue was mounted in a holder with the mucosal side up (exposed tissue area 0.2 cm 2 ). After insertion of the holder into the Ussing chamber the tissue was allowed to recover for 10-20 min and to reach a stable baseline. To the serosal side of the ileum glucose (10 mM) and indomethacin (10 mM) were added. After equilibrium, the following compounds were added consecutively to the mucosal (M) or serosal (S) side of the tissue: Forskolin (10 µM, S), genistein (100 µM, M+S) and carbachol (200 µM, S). All compounds were present throughout the experiment.
Mouse nasal epithelium was isolated as described by Grubb et. al. [60]. In brief, the mice were sacrificed by cervical dislocation and the skin of the head was peeled back in order to reach the underlying paired nasal bones. These were removed and the two sheets of the nasal epithelia, separated by the septum, were isolated independently. The sheets of epithelia were mounted immediately between the Ussing chambers (exposed area 1.13 mm 2 ). The chambers were filled with gassed modified Meyler's solution supplemented with glucose (10 mM) and indomethacin (10 mM) and the short circuit measurements were started. After equilibration amiloride (10 µM, M) was added followed by forskolin (10 µM, after stabilization of the current).

Statistics
P-values were determined by using the student's t-test. Pvalues < 0.05 were considered significant.

Authors' contributions
NC performed the genome-wide scan with microsatellite markers, SJ and NC executed RNA isolations and performed RT-PCR experiments, MW carried out all short-circuit current measurements, MD participated in the supervision of the animal breeding, AB performed western blot analysis, HJ carried out the immunocytochemical and histological analysis, HJH designed and supervised all animal breeding, FS set up all PCR-related methodology and quality control and BT revisited the primary genetic and phenotypic murine data gathered at the Hannover site since 1994. HRDJ and BT conceived and supervised the study and participated in the design of experiments and result analysis. All authors contributed to the writing of this manuscript.

Description.
CF/1-Cftr TgH(neoim)Hgu and CF/3-Cftr TgH(neoim)Hgu mice of generation F 26 were typed at a 20 cM spacing with microsatellite markers. The criterion for the selection of the first 78 markers was their informativity to differentiate between the three reference strains BALB/ c, C57BL/6 and DBA/2. Informative loci between the two inbred CF strains were genotyped with flanking markers in order to map the regions of genetic heterogeneity, bringing this to a total of 105 microsatellite markers.