Testing the Cre-mediated genetic switch for the generation of conditional knock-in mice

The Cre-mediated genetic switch combines the ability of Cre recombinase to stably invert or excise a DNA fragment depending upon the orientation of flanking mutant loxP sites. In this work, we have tested this strategy in vivo with the aim to generate two conditional knock-in mice for missense mutations in the Impad1 and Clcn7 genes causing two different skeletal dysplasias. Targeting constructs were generated in which the Impad1 exon 2 and an inverted exon 2* and the Clcn7 exon 7 and an inverted exon 7* containing the point mutations were flanked by mutant loxP sites in a head-to-head orientation. When the Cre recombinase is present, the DNA flanked by the mutant loxP sites is expected to be stably inverted leading to the activation of the mutated exon. The targeting vectors were used to generate heterozygous floxed mice in which inversion of the wild-type with the mutant exon has not occurred yet. To generate knock-in mice, floxed animals were mated to a global Cre-deleter mouse strain for stable inversion and activation of the mutation. Unexpectedly the phenotype of homozygous Impad1 knock-in animals overlaps with the lethal phenotype described previously in Impad1 knock-out mice. Similarly, the phenotype of homozygous Clcn7 floxed mice overlaps with Clcn7 knock-out mice. Expression studies by qPCR and RT-PCR demonstrated that mutant mRNA underwent abnormal splicing leading to the synthesis of non-functional proteins. Thus, the skeletal phenotypes in both murine strains were not caused by the missense mutations, but by aberrant splicing. Our data demonstrate that the Cre mediated genetic switch strategy should be considered cautiously for the generation of conditional knock-in mice.


Introduction
In the era of big data the question for geneticists is often whether a mutation identified by next generation sequencing in a particular gene can explain the clinical phenotype of the patient. Indeed a crucial issue in biomedical research is to convert sequence information, -omics, analytical and clinical data into knowledge about gene function. PLOS  Skeletal dysplasias represent one of the largest classes of birth defects with 436 different disorders that have been clustered in 42 different groups depending on molecular and/or clinical features. More than 350 disease genes have been identified encoding for proteins involved in a wide spectrum of biological functions in cartilage and bone [1,2]. To shed light on the molecular mechanisms of skeletal disorders, mechanistic studies using in vitro and in vivo approaches are necessary to elucidate the role of the disease gene in the disorder.
Cell cultures, or in vitro studies, provide the first important system to study human diseases, preserving the physiology of living cells and enabling the manipulation under controlled laboratory condition. Patient-derived cells including fibroblasts, induced pluripotent stem cells, primary chondrocytes, osteoblasts and osteoclasts are widely used to model congenital disorders of cartilage and bone [3]; however, their use does not always mirror the whole body condition.
Due to the complexity of skeletal pathologies in vivo models represent a need for the field. Recently, Danio rerio (Zebrafish) has become an appealing animal model to study skeletal development as well as to test drug's efficacy [4][5][6], nevertheless nowadays mouse models still represent the gold standard for bone disease modeling since they are anatomically and physiologically close to humans. The mouse genome can be manipulated in many different ways in order to generate models that carry the same disease causing mutation detected in patients [7]. Moreover, patient specific mutations might be helpful to identify genotype-phenotype correlations and differences in the effects of particular mutations.
For more than 20 years the most suitable technique to generate animal models has been based on homologous recombination in murine embryonic stem cells to change the expression of an endogenous gene [8]. The gene-targeting technique has many advantages related to the fact that homologous recombination defines the site of integration and the genetic change in a very specific manner. Gene-targeting can be designed to introduce different genetic modifications such as gene deletions (knock-out), point mutations (knock-in), gene insertions in a certain locus or chromosomal rearrangements.
Knock-out mice can be also generated by the gene trap strategy, which is based on the disruption of an endogenous gene function by the random insertion of an intronic gene-trap cassette [9]. More recently, the genome-editing CRISPR/Cas9 technology has offered a new, quick and cheap way to model genetic disorders [10].
Genetic changes in the germ line might be useful to track the gene function, but may also result in severe developmental consequences complicating or precluding the experimental analysis (i.e. because of embryonic lethality in knock-outs). To overcome these limits related to constitutive expression of the targeting construct, conditional mice expressing the gene modification only at a specific developmental stage or in selected cells have been generated. Different conditional systems are available; among them, the tetracycline (tet) regulatory system and Cre/loxP or Flp/FRT recombination systems are the most widely used [11,12]. Cre and Flp recombinases mediate different effects on their DNA target sequences including excision, duplication, integration, inversion and translocation depending on the orientation of their specific recognition sequence, namely loxP and FRT, respectively.
However, integration and inversion between wild-type loxP sites is inefficient due to activation of a vicious cycle of re-excision through intramolecular recombination. To overcome such limitation the left element/right element (LE/RE) mutant strategy using LE mutant lox carrying mutations in the left-inverted repeat region and RE mutant lox carrying mutations in the right-inverted repeat region can be used [13,14]. Recombination between a LE mutant lox and a RE mutant lox results in the generation of a double mutant lox site with mutations in both ends and a wild-type loxP site. The double mutant lox site is no longer a substrate for Cre recombinase; therefore, the recombination reaction proceeds only in the forward direction. By combining the ability of Cre recombinase to invert or excise a DNA fragment and the use of wild-type and mutant loxP sites, an efficient and reliable Cre-mediated genetic switch has been proposed [15]. Through this strategy expression of a given gene can be turned off, while expression of another one can be simultaneously turned on. This innovative, flexible and powerful approach can be used to easily generate many genetic modifications in a conditional manner [16].
The genetic switch has also been proposed to invert a wild-type exon with a mutated one in order to generate conditional point mutations [14]. In the present work, we have tested this powerful tool to generate two conditional knock-in mice bearing patient specific missense mutations causative of two different skeletal disorders: chondrodysplasia with joint dislocations, gPAPP type and autosomal dominant osteopetrosis type 2 (ADO2).

Preparation of the gene targeting vectors and generation of mutant mice
A vector for the stable inversion mediated by Cre-recombinase was generated by gene synthesis with a combination of lox71 as LE mutant loxP (carrying mutations in the left-inverted repeat region) and loxKR3 as RE mutant loxP (carrying mutations in the right-inverted repeat region) in a head-to-head orientation [13]. Different restriction sites were synthesized upstream, downstream and in between the lox sites, to allow insertion of different cassettes. To test the efficiency of the Cre-mediated irreversible inversion we took advantage of an FRTflanked neomycin resistance cassette cloned between the lox-sites using EcoRV and EcoRI. The resulting vector was transformed in E.coli expressing Cre-recombinase where efficient recombination was confirmed by inversion in vivo.
Subsequently, the lox71/loxKR3 vector was used for the assembly of the Impad1 and Clcn7 conditional knock-in gene targeting constructs. Briefly, the vectors contained homology arms for the Impad1 or Clcn7 genes of about 5 kb and 2.7 kb or 3.7 kb and 2.8 kb, respectively. Between the homologous sequences a duplicated region of the wild-type and mutated exon (exon 2 for Impad1 and exon 7 for Clcn7) were introduced in a head-to-head orientation; exons were flanked by the head-to-head lox71 and loxKR3 sites and separated by an FRTflanked neomycin cassette (Fig 1).
The missense mutation knocked in the Impad1 gene was a c.726G>A transition (NM_177730.4) leading to substitution of p.Asp175>Asn in exon 2. This mutation corresponds to the p.Asp177>Asn mutation detected in a patient with chondrodysplasia with joint dislocations, gPAPP type [17]. In addition close to the mutation two silent mutations, c.716T>C and c.719C>T respectively, were inserted to generate a ClaI restriction site useful for animal genotyping.
The missense mutation knocked in the Clcn7 gene was a g.14365G>A transition (NM_011930.4) leading to substitution of p.Gly213>Arg. This mutation corresponds to the most common human mutation, p.Gly215>Arg, detected in patients with ADO2 [18].

Generation of Impad1 and Clcn7 targeted embryonic stem cells and mice
The two targeting vectors bearing the Impad1 or the Clcn7 mutation were electroporated into 2 × 10 7 C57Bl/6N-based and 129Ola-based ES cells, respectively. After selection with 200 μg/ ml neomycin (G418) for 8 days, about 400 ES cell clones were isolated and analyzed via PCR screening.
In both ES cell lines, positive ES cell clones in which correct homologous recombination has occurred were identified via long range PCR and confirmed by Southern blot analysis. A list of all primers used for PCR analyses and for the synthesis of the Southern blot probe (generated by PCR with primers Neo3-for and Neo4-rev) as well as for animal genotyping are reported in S1 Table. For Impad1 mouse generation, two positive ES cell clones were injected into grey C57Bl/6N derived blastocysts. The resulting chimeras were bred to C57Bl/6N-based Flp-deleter mice (B6gr-Tg(ACTFLPe)9205Dym/NPg) and the resulting black offspring was analyzed by PCR. Several positive mice that lost the FRT-flanked neomycin cassette via Flp-recombination were identified (here called "floxed" mice, Impad1 Flox/WT mice). Deletion of the neomycin cassette was demonstrated by PCR using primers Neo3-for and Neo4-rev (S1 Table); moreover, the presence of the floxed allele was confirmed by the presence of the lox71 site checked by PCR with primers SY08.20 and SY08.21. Impad1 Flox/WT mice were mated to a Cre mouse strain (B6. FVB-Tg(EIIa-cre)C5379Lmgd) to generate the heterozygous knock-in of the p.Asp175>Asn mutation (Impad1 D175N/WT mouse). Correct Cre-mediated switch was checked taking advantage of the insertion of the ClaI restriction site close to the missense mutation. The 480 bp PCR product of the mutant allele obtained using Imp3 and Imp8 primers was digested by ClaI in two restriction fragments of 124 bp and 356 bp, respectively.
For Clcn7 mouse generation, three positive clones were injected into blastocysts from black C57Bl/6N females. The resulting chimeras were bred to C57Bl/6N-based Flp-deleter (B6gr-Tg (ACTFLPe)9205Dym/NPg) mice. The resulting agouti offsprings were then genotyped and several positive mice showing deletion of the neomycin resistance cassette were identified (here called "floxed" mice, Clcn7 Flox/WT mice). These were bred further with each other to generate homozygous floxed mice. For animal genotyping the insertion of the lox71 site, used as marker of the mutant allele, was checked by PCR with primers flanking the insertion site (SY03.11 and SY03.12). Deletion of the neomycin cassette was demonstrated by PCR using primers Neo3-for and Neo4-rev (S1 Table).

Animals
Animals were bred and maintained in community housing (�5 mice/cage, 22˚C) on a 12 h light/dark cycle with free access to water and standard pelleted food. Care and use of mice for this study were in compliance with relevant animal welfare institutional guidelines in agreement with EU Directive 2010/63/EU for animals, the Italian Legislative Decree 4.03.2014, n. 26 and the Swiss animal protection act (TschG). The experimental protocols were approved by the Italian Ministry of Health (Animal protocol n. 844/2017-PR and n. 564/2016-PR). Research staff received appropriate training in animal care. Mice were euthanized by CO 2 inhalation before the disease status was considered severe. For genetically modified animals, the evaluation of the severity was based on daily monitoring of the mice behavior, including evaluation of mobility and body weight.
Euthanasia was always done before the animals reached the endpoint, Impad1 knock-in animals died at birth. The Impad1 knock-out strain, generated by Frederick et al. [19] was provided by "The Jackson Laboratories", Bar Harbor, Maine, USA.

Mouse genotyping
Mice were genotyped by PCR using genomic DNA from mouse tail or ear clips. For animal genotyping of the Impad1 knock-in strain, the mutant allele was detected by digestion with ClaI of a PCR product generated with primers Imp3 and Imp8. The 480 bp PCR product of the mutant allele was digested by ClaI in two restriction fragments of 124 bp and 356 bp, respectively. Genotyping of the Impad1 knock-out strain was performed as described previously [19].
For animal genotyping of the Clcn7 floxed strain SY03.11 and SY03.12 PCR primers were used to genotype homozygous mutant mice from heterozygous and wild-type animals.

X-rays and image analysis
X-ray analyses were performed using a Faxitron MX-20 cabinet X-ray system (Faxitron Bioptics LLC, USA) and X-ray images were digitalized with a Kodak DirectView Elite CR System (Carestream Health Italia, Italy).

Micro computed tomography
Femurs from 3 week-old mice were fixed in 4% formaldehyde for 48 hours and then scanned by a μCT SkyScan 1174 (Bruker Italia, Italy). The scan was performed with 6.4 μm resolution at 50 kV. The Skyscan NRecon software was used to reconstruct images with a modified Feldkamp algorithm. Three-dimensional analysis was carried out employing a Marching Cubes type model with a rendered surface [20]. Threshold values were applied for segmenting trabecular bone. Bone trabecular and cortical variables were determined according to Bouxsein et al [21].

Differential skeletal staining with alcian blue and alizarin red
Skeletal staining of newborn mice were performed with alcian blue and alizarin red [22]. Briefly, pups were skinned, dehydrated in 96% ethanol, defatted in acetone and stained with alcian blue and alizarin red. Muscles were removed with 1% KOH in 20% glycerol and preparations were stored in glycerol. Mice were then photographed using a Leica M165 FC stereomicroscope connected to a Leica DFC425 C digital camera (Leica Microsystems, Italy).

qPCR and alternative splicing analysis
Skin from newborn mice was homogenized in 1 ml of QIAzol Lysis Reagent (QIAGEN, Italy) and total RNA was extracted according to the manufacturer's instructions. Reverse transcription of 1 μg RNA in a final volume of 20 μl was performed using the SuperScript IV First-Strand Synthesis System (Invitrogen, Thermo Fisher Scientific, Italy) according to the manufacturer's instructions.
For expression analyses by RT-qPCR, the QuantiFast Primer Assays with validated primer sets for Impad1 and Gapdh (QIAGEN) were used with the QuantiTect SYBR Green PCR Kit (QIAGEN) according to the manufacturer's protocol. The primer set of Impad1 span exon 2 and 3. Each sample was run in triplicate in 96 well plates in three independent experiments with the MX3000P apparatus (Stratagene, USA). The expression of Impad1 relative to the housekeeping Gapdh gene was obtained by ΔΔCt method.
For alternative splicing detection, total RNA was used for RT-PCR analysis using primers in exon 1 and exon 5 of the Impad1 gene and in exon 6 and 25 of the Clcn7 gene. PCR products were analyzed by 1.5% agarose gels. For DNA sequencing of the spliced forms of Impad1, the PCR products were cloned with the TA cloning kit (Invitrogen) followed by Sanger sequencing.

HPLC analysis of cartilage glycosaminoglycan sulfation
For cartilage disaccharide analysis, cartilage was obtained from the femoral heads of newborn mice by careful dissection under the dissection microscope and glycosaminoglycans (GAGs) were recovered by papain digestion and cetylpyridinium chloride precipitation as previously described [23]. Purified GAGs were digested with 30 mU of both chondroitinase ABC and chondroitinase ACII (Seikagaku Corp., Japan) in 0.1 M sodium acetate, pH 7.35 at 37˚C overnight. Released disaccharides were lyophilized and redissolved in 40 μl of 12.5 mM 2-aminoacridone (Invitrogen) in 85% DMSO/15% acetic acid and incubated for 15 minutes at room temperature in the dark. Then, 40 μl of 1.25 M sodium cyanoborohydride in water was added and the mixture was incubated overnight at 37˚C. Disaccharides were fractioned with a Pron-toSIL HPLC column (4.6 mm × 200 mm, Bischoff Chromatography, Germany) using a linear gradient (0-42% solvent B in 42 min); mobile phases were 0.1 M ammonium acetate, pH 7.0, (Solvent A) and methanol (Solvent B). The analysis was carried out at room temperature with 0.7 ml/min flow rate and the elution profile was monitored by a fluorescence detector (2475 Multy λ Fluorescence Detector, Waters, Italy), with excitation and emission wavelengths of 425 and 525 nm, respectively.

Statistical analysis
Statistical analyses were carried out by the software Prism by GraphPad v7.0 and the type of analysis is stated in the figure legends; all results are presented as mean ± SD.

Generation of the Cre mediated switch allele in mice
In order to generate the two conditional knock-in mice, a combination of two partially mutated lox sites in the head-to-head orientation was used. This strategy guarantees stable inversion of DNA between the lox sites since after Cre recombination a wild-type loxP site and a double mutant lox, no longer recognized by the enzyme, are generated ( Fig 1A). First the combination lox71 and loxKR3, reported previously by Araki et al [13] in a head-to-head orientation was tested for recombination efficiency in a Cre expressing E.coli strain. Five clones were isolated from each transformation and analyzed via restriction analysis with BstBI. All clones showed the expected restriction pattern corresponding to the inverted version of the test vector (Fig 1B).
Two gene targeting vectors were generated based on the lox71/loxKR3 construct. The first for an Impad1 mutation (p.Asp175>Asn) and the latter for a mutation in the Clcn7 (p. Gly213>Arg) (Fig 1C and Fig 2A and 2B).
Homologous recombination was performed in C57Bl/6N-based (Impad1) and 129Olabased (Clcn7) ES cells and confirmed by long range PCR and Southern blot analysis (Fig 2C  and 2D for Impad1 and Fig 2F and 2G for Clcn7). Proper targeted ES cell clones were injected into C57Bl/6N blastocysts and the resulting chimeras were bred to FLP-deleter mice for neomycin cassette deletion. The heterozygous offsprings, here called "floxed" mice (Impad1 Flox/WT mice and Clcn7 Flox/WT mice, respectively), were then confirmed by the presence of the lox71 site (Fig 2E and 2H) and the deletion of the neomycin cassette based on the absence of a 542 bp fragment after PCR with primers Neo3-for and Neo4-rev.
To generate heterozygous Impad1 knock-in animals (Impad1 D175N/WT ), Impad1 Flox/WT mice were mated to a global Cre mouse strain for stable inversion and activation of the mutation.

Phenotypic characterization of the Impad1 knock-in mouse
Impad1 D175N/WT mice did not show any phenotypic alteration by X-ray analysis and visual inspection as human carriers of IMPAD1 mutations and were no further studied.
Homozygous mutant (Impad1 D175N/D175N ) animals died at birth and were smaller compared to wild-type and heterozygous littermates; visual inspection and X-ray analysis demonstrated growth retardation and skeletal defects (Fig 3A and 3B). Mutant animals showed severe hypoplasia of the skeleton; the length of the axial skeleton and of the limbs were reduced. By X-rays and alcian blue and alizarin red skeletal staining, the femur, tibia and fibula were markedly shorter compared to wild-type animals; furthermore, reduced sternal length and diminished rib spacing were evident (Fig 3B and 3D). Moreover, Impad1 D175N/D175N mice showed cleft palate already observed in patients with chondrodysplasia with joint dislocations, gPAPP type (Fig 3C). Surprisingly, the Impad1 D175N/D175N mouse showed the same peculiar phenotype described previously in the Impad1 knock-out mouse [19,24]. Thus, we further investigated the molecular and biochemical basis causing neonatal lethality in our animal model. Impad1 encodes for a Golgi 3'-phosphoadenosine 5'-phosphate phosphatase crucial for macromolecular sulfation. In whole embryo and in the limbs of the Impad1 knock-out mouse, severe undersulfation of chondroitin sulfate proteoglycans was demonstrated compared to wild-type mice [19,24]. For this reason, proteoglycan sulfation was measured by HPLC disaccharide analysis in femoral head cartilage of wild-type and Impad1 D175N/D175N mice and, for comparison, in the Frederick's Impad1 knock-out [19]. Glycosaminoglycans were purified by digestion of femoral head cartilage with papain. Then glycosaminoglycans were digested with chondroitinase ABC and ACII and released disaccharides separated by HPLC after derivatization with a fluorescent tag. A dramatic increase in the relative amount of the chondroitin non- Challenges in the generation of conditional knock-in mice sulfated disaccharide (ΔDi-0S) was observed in newborn Impad1 D175N/D175N mice compared to wild-type animals indicating chondroitin sulfate undersulfation (61 ± 9.2% and 14 ± 1.6% ΔDi-0S, respectively; P < 0.001 n = 3). Interestingly, the extent of proteoglycan undersulfation in mutant mice was similar to homozygous Impad1 knock-out animals (63 ± 1.6% ΔDi-0S, n = 3) (Fig 4).
Since the clinical and biochemical phenotype of Impad1 D175N/D175N animals overlap with Impad1 knock-outs, the expression of Impad1 was measured by qPCR with primers spanning exons 2 and 3. Impad1 mRNA was barely detectable in Impad1 D175N/D175N mice compared with wild-types; Impad1 expression level in heterozygous mice was reduced by half (Fig 5). Sulfation of proteoglycans was determined by HPLC disaccharide analysis after digestion by chondroitinase ABC and ACII of chondroitin sulfate proteoglycans from the femoral head cartilage of wild-type (WT) and Impad1 D175N/D175N mice at birth. In parallel the same analysis was performed also in the Impad1 knock-out (Impad1 -/-) mouse studied by Frederick [19]. The amount of non sulfated disaccharide (ΔDi-0S) relative to the total amount of disaccharides (ΔDi-0S, ΔDi-4S and ΔDi-6S) is significantly increased in mutant mice compared to the wild-types indicating proteoglycan undersulfation. Interestingly, the level of proteoglycan undersulfation in mutants is similar to the Frederick's knock-out mouse. Three mice per group were used; data are reported as mean ± SD (Student's t-test, � p<0.05; ��� p<0.001). https://doi.org/10.1371/journal.pone.0213660.g004 Challenges in the generation of conditional knock-in mice Since in Impad1 D175N/D175N mice no Impad1 transcript was detected by qPCR, we checked whether the modified Impad1 locus impairs correct RNA splicing. Thus, we considered potential alternative splicing of Impad1 exons by RT-PCR using primers Imp7 and Imp12 spanning the coding sequence from exon 1 to 5 (Fig 6). Results showed that in wild-type animals one band, 639 bp long, corresponding to a unique transcript including the 5 coding exons was present. This band was not detected in Impad1 D175N/D175N mice; conversely two different bands, 477 bp and 383 bp long, respectively, were observed (Fig 6A). Sequencing of the two bands demonstrated that the two transcript variants lack exon 2 or exon 2 and exon 3 (Fig 6B) causing, if translated, frameshift mutations leading to premature stop codons p.Ile128ArgfsX 13 and p.Ile128LeufsX5, respectively.
Retrospectively we noted that Impad1 Flox/WT cross breeding resulted in an unusual small litters size. The analysis of some pups found dead at birth revealed their homozygous Impad1 Flox/Flox genotype. Furthermore, their X-ray showed the same severe skeletal phenotype observed in Impad1 D175N/D175N mice (S1 Fig).

Phenotypic characterization of the Clcn7 floxed mice
Since the Impad1 D175N/D175N phenotype was not due to the missense mutation, but by altered splicing of the modified Impad1 allele, and that aberrant splicing was likely responsible for Impad1 Flox/Flox mice unexpected lethal outcome, we checked whether similar modification RT-qPCR on total RNA isolated from skin of wild-type (WT), heterozygous (Impad1 WT/D175N ) and Impad1 D175N/D175N newborn mice was performed with primers spanning exon 2 and exon 3. Expression of Impad1 mRNA normalized to Gapdh is absent in homozygous mutant animals compared to wild-types. Three mice per genotype were used; each sample was run in triplicate and three different experiments were performed. https://doi.org/10.1371/journal.pone.0213660.g005 Challenges in the generation of conditional knock-in mice affected also the Clcn7 Flox/Flox mice. For this reason heterozygous Clcn7 floxed mice (Clcn7 Flox/WT ) were mated together to observe the phenotype in homozygous floxed mice (Clcn7 Flox/Flox ). Since these mice were not mated yet to a Cre expressing murine strain, inversion of the mutated exon, bearing the p.Gly213>Arg mutation, did not occurr yet and thus no phenotype should have been observed. The Clcn7 Flox/Flox mouse presented the phenotypic features of autosomal recessive osteopetrosis. In particular the Clcn7 Flox/Flox displayed a phenotype resembling Clcn7 knock-out mice [25], with reduced survival (Fig 7) and a normal Mendelian ratio of genotypes at birth. As expected, no significant changes in survival was found in the heterozygous Clcn7 Flox/WT mice (Fig 7).
Moreover, Clcn7 Flox/Flox mice were shorter than heterozygous and wild-type littermates ( Fig  8A) and showed no tooth eruption when compared with wild-type littermates (Fig 8B). Both features are clinical signs of autosomal recessive osteopetrosis.
After gross evaluation, femoral bones were analyzed by X-ray and μCT that revealed the 3D architecture of the bones and confirmed a severe osteopetrotic phenotype in Clcn7 Flox/Flox mice (Fig 9), with an extremely high bone volume/tissue volume (72 ± 4.3% in Clcn7 Flox/Flox and 13 ± 3.2% in wild-type littermates).
Altogether, these results revealed that the Clcn7 Flox/Flox mouse did not show the expected phenotype, but rather recapitulated the loss-of-function model suggesting that the genomic modification resulted in a knock-out allele.
We checked whether the modified Clcn7 locus impairs correct RNA splicing as observed in mutant Impad1. Thus, we considered potential alternative splicing of Clcn7 exons by RT-PCR using primers Clcn7Ex6 and Clcn7Ex25 spanning the coding sequence from exon 6 to 25 ( Fig  10). In wild-type animals one band, 1683 bp long, corresponding to a unique transcript including 20 coding exons was present; conversely a band 1216 bp long, lacking exon 7-11 was observed in Clcn7 Flox/Flox mice (Fig 10). This should result in an in-frame deletion of 129 amino acid residues.

Discussion
We have tested a Cre-mediated genetic switch combining the ability of Cre recombinase to invert a DNA fragment, depending upon the orientation of the flanking loxP sites, and the use of wild-type and mutant loxP sites in order to make recombination irreversible. The notion was that Cre recombinase-mediated inversion would place the exon bearing the missense mutation into a position where it would be spliced properly, while the inverted wild-type exon would be spliced out.
We have used this strategy to generate two conditional knock-in mice for the Impad1 and Clcn7 genes. Mutations in these genes cause in humans two different skeletal disorders: chondrodysplasia with joint dislocations gPAPP type and autosomal dominant or recessive osteopetrosis, respectively. Impad1 encodes for a Golgi resident adenosine 3',5'-bisphosphate phosphatase crucial for macromolecular sulfation [19,24], while the Clcn7 gene encodes for the ClC7 protein, a proton/chloride antiporter important in lysosomal acidification [26]. In bone, the ClC7 is expressed especially by osteoclasts and it has a pivotal role in bone resorption, charge-balancing the acidified osteoclast resorption lacuna. The missense mutation knocked-in the murine Impad1 was the p.Asp175>Asn corresponding to the p.Asp177>Asn detected previously in a patient [17]. The mutation knocked-in the murine Clc7 was the p. Gly213>Arg described previously in a patient with ADO2 [27,28].
Both mouse models generated through this strategy resulted in an unexpected severe skeletal phenotype. Heterozygous floxed Impad1 (Impad1 Flox/WT ) mice were mated to a Cre deleter mouse strain in order to generate Impad1 D175N/WT animals. Impad1 D175N/D175N mice died at RT-PCR using primers Imp7 and Imp12 that amplify a region spanning exon 1 to exon 5 was performed from skin total RNA of wild-type and Impad1 D175N/D175N newborn mice and analysed by 1.5% agarose gel. In wild-type animals one band, 639 bp long, corresponding to the correctly spliced Impad1 transcript including the 5 coding exons is present. This band is not detected in mutant mice, conversely two different bands, 477 bp and 383 bp long, respectively are observed. (B) Sequencing of the two bands demonstrates that the two transcript variants lack exon 2 or both exon 2 and exon 3. birth with severe hypoplasia of the skeleton and at the biochemical level cartilage proteoglycans were dramatically undersulfated. The phenotype overlapped with the lethal phenotype  https://doi.org/10.1371/journal.pone.0213660.g008 described previously in Impad1 knock-out mice generated by a gene trap approach [19,24]. Expression studies by qPCR and RT-PCR demonstrated that the mutant Impad1 mRNA underwent abnormal splicing with loss of exon 2 or exons 2 and 3; no mutant full length mRNA spanning exons 1-5 as in normal Impad1 mRNA was detected. This may result in the activation of non-sense mediated decay or anyway in the synthesis of transcripts encoding for a non-functional enzyme since the active site of the phosphatase is encoded by exon 2.
A skeletal phenotype was observed also in homozygous Clcn7 Floxed/Floxed mice before mating to Cre deleter mice, when the wild-type exon 7 was into a position where it would be spliced properly and thus a wild-type Clcn7 mRNA would have been generated. The skeletal phenotype observed in these mice overlaps with the Clcn7 knock-out reported previously [25]; in fact also in this murine model abnormal splicing causing exons skipping occurred, generating a knock-out allele, likely because of the wild-type and mutant exon 7 in opposite orientation in the mutant Clcn7.
Thus in both animal models the skeletal phenotype parallels the phenotypes of their knockout mice suggesting that the skeletal defects were not due to the missense mutations, but to aberrant splicing causing the synthesis of non-functional proteins. It is tempting to speculate that the two exons in opposite orientation generate a hairpin loop leading to exon skipping.
A similar strategy was devised previously in order to generate a conditional knock-in of the cAMP response element binding protein (CBP) [14]. To achieve the Cre recombinase-mediated mutation CBP Tyr658Ala , a targeting construct containing the wild-type exon 5 of the CBP and a mutated exon 5 (p.Tyr658>Ala) in an inverted orientation was generated. These Challenges in the generation of conditional knock-in mice sequences were flanked by two mutated loxP sites, which were positioned in a head-to-head orientation. The authors demonstrated that in ES cells the genetic switch worked properly when cells were transfected with a plasmid expressing Cre-recombinase, but unfortunately, no data were available regarding the expression at the mRNA or protein level. The generation and further characterization of transgenic animals from these ES cells has never been reported to date.
In conclusion, the Cre mediated genetic switch strategy has paved the way to the engineering of sophisticated genetic modifications including conditional point mutations, conditional rescue, conditional gene replacement and recombinase mediated cassette exchange which have been used successfully in vitro. However, it is worth noting that modifying the genome of eukaryotic cells to generate Cre mediated switch alleles may have some drawbacks. The RT-PCR using primers that amplify a region spanning exon 6 to exon 25 was performed from bone total RNA of wild-type (Clcn7 WT/WT ) and Clcn7 Flox/Flox mice and analysed by 1.5% agarose gel. In wild-type animals one band, 1683 bp long, corresponding to the correctly spliced Clcn7 transcript including 20 coding exons is present. This band is not detected in mutant mice, conversely a 1216 bp long band lacking exon 7-11 is observed. https://doi.org/10.1371/journal.pone.0213660.g010 Challenges in the generation of conditional knock-in mice repetition of endogenous splicing sites in the antisense orientation may also induce the occurrence of aberrantly spliced mRNA. In any case, it is probably safe to reduce DNA repetitions as much as possible and to test for functionality in transiently transfected cells when possible, or in ES cell clones before further proceeding to blastocyst injections.
Supporting information S1 Table. PCR primers used in this study. (DOCX) S1 Fig. X-rays