A new mouse mutant with cleavage-resistant versican and isoform-specific versican mutants demonstrate that proteolysis at the Glu441-Ala442 peptide bond in the V1 isoform is essential for interdigital web regression

Highlights • • A novel Vcan mouse allele, VcanAA, has ADAMTS protease-resistant versican.• • VcanAA/AA mice are viable and develop soft tissue-syndactyly (STS)• • VcanAA/AA STS is rendered more severe in combination with Adamts20Bt/Bt.• • Mice lacking the versican GAGβ domain, but not the GAGα domain, also have STS.• • The versican GAGβ proteolytic fragment versikine is necessary for web regression.


Introduction
Extracellular matrix (ECM) proteolysis is necessary for tissue remodeling during mammalian development. For example, branching morphogenesis, cardiac valve leaflet formation, endochondral ossification and interdigital web regression each require precisely timed and spatially controlled tissue remodeling, which was demonstrated by impairment of these processes in mice with engineered mutations in specific proteases [1][2][3][4][5][6]. A number of secreted and cellsurface metalloproteases are implicated in ECM remodeling [7]. Because protease function is realized through substrate cleavage, identifying the most impactful cleavages will elucidate potentially important biological pathways. However, since few proteases have a single substrate, an inherent challenge in interpreting protease-deficient phenotypes is determining which of many possible reduced proteolytic cleavages are responsible for an observed phenotype. A corollary to this statement is that the phenotype may result from reduced proteolysis of several substrates, not just a single one, requiring stringent analysis of each substrate to determine the precise contribution of each. The most rigorous validation, indeed a proof, of the importance of cleavage of a substrate would be partial or complete recapitulation of a phenotype observed in a protease-deficient mutant by a cleavage-resistant substrate [8]. However, such experiments are infrequently undertaken because of the high level of risk involved, since many proteases may attack a protein and a protease could cleave the protein at many sites. This risk reduces the likelihood of definitive insights that could be obtained from rendering a single site resistant to proteolysis. Therefore, defining the significance of specific cleavages in substrates has generally not reached the highest levels of experimental rigor. As a result, the functional landscape of proteolysis during embryonic development, physiological processes and diseases is incompletely annotated.
Versican is a large, aggregating chondroitin sulfate (CS) proteoglycan which is a major constituent of ECM, especially provisional ECM [9]. It is essential for survival beyond 10 days of gestation in the mouse and is required specifically for early development of the entire circulatory system, i.e., the heart, blood vessels and blood cells [10][11]. It forms networks in ECM via two main mechanisms, i.e., by binding hyaluronan through its Nterminal G1-domain and by interacting with numerous ECM components including fibronectin, fibrillins, fibulin-1 and À2 and tenascin-R and -W via its C-terminal G3 domain (reviewed in [12][13]).CS chains are covalently attached to large, alternatively spliced core protein domains between the G1 and G3 domains, named GAGa (encoded by Vcan exon 7) and GAGb (encoded by Vcan exon 8) [14]. Alternative splicing of these exons generates distinct versican variants: V0 (containing both GAGa and GAGb domains), V1 (containing only the GAGb domain), V2 (containing only the GAGa domain) and V3 (containing neither GAG-bearing domain). Because of its abundance in embryonic tissues and crucial role in embryogenesis, proteolytic turnover of versican has elicited considerable interest [12].
The evidence linking ADAMTS proteases to versican processing in interdigit webs, and the importance of versican proteolysis in this context is compelling. Adamts5, Adamts9 and Adamts20 mRNAs are co-expressed at heightened levels in interdigital web mesenchyme just prior to web regression [4]. Moreover, reduced anti-DPEAAE staining of interdigital webs during the period during which regression occurs was evident in combined Adamts5 + Adamts20 mutant mice, which consistently develop soft-tissue syndactyly (STS) and in mice with limb-specific Adamts9 deletion [4,23]. Impaired interdigital web regression in these mutants was associated with persistence of versican-rich ECM and a subsequent failure of apoptosis [4,23]. The N-terminal G1 domaincontaining fragment (versikine), was found to induce cell death during interdigital web regression in Adamts5 + Adamts20 mutants, and deletion of one Vcan allele in Adamts20 bt/bt mice led to more severe STS, suggesting that versican itself, albeit after proteolysis, could be essential for web regression [4]. Taken together, the findings suggest that cooperative versican processing by ADAMTS proteases generates a critical level of versikine to promote apoptosis of interdigital mesenchymal cells; conversely, when versican proteolysis is reduced below this hypothetical critical threshold, interdigit cells may be resistant to BMP-induced apoptosis [4].
Here, we report the generation and analysis of mice with versican resistant to cleavage at the E 441 -A 442 ADAMTS site, using a technically different approach than that recently utilized to generate another cleavage-resistant Vcan allele (designated Vcan(R/R) in the homozygous state) [33]. Our previously reported experiments had demonstrated that substitution of E 441 with A (Ala) reduced, but did not eliminate versican cleavage by ADAMTS5 in vitro and had revealed a cryptic ADAMTS5 cleavage site at E 438 -A 439 (variant V1 sequence enumeration) [34]. Replacement of both glutamic acid residues with alanine, i.e., E 438 to A + E 441 to A eliminated ADAMTS5 processing in this region of the versican core protein [34]. With this preliminary evidence of specific point mutations that could abrogate versican processing, we used homologous recombination in mouse embryonic stem cells to replace E 438 and E 441 with A and generate transgenic mice with cleavage-resistant versican, an allele designated as Vcan AA . Here, we used this new allele to rigorously test the biological relevance of versican proteolysis at these two sites in the context of web regression and along with analysis of web regression in Vcan exon 7-or exon 8specific mutants, we provide new insights on the role of ADAMTS proteases and versican in this process.

Results
Homologous recombination of a targeting construct bearing the E 438 + E 441 mutations in ES cells successfully introduced the mutations into the Vcan locus (Fig. 1A, B). Both hemizygous and homozygous mutant mice were viable and fertile. qRT-PCR demonstrated that transcription of the four major Vcan splice isoforms was unaffected (Fig. 1C, D). Furthermore, RNA in situ hybridization demonstrated comparable distribution and intensity of Vcan mRNA using specific exon 7 and exon 8 probes, with each probe showing overlapping expression in wild-type and Vcan AA/AA mutant embryos ( Fig. 2A). Survival of Vcan AA/AA mice indicated absence of the cardiac, vascular and hematopoietic developmental defects seen in Vcan hdf/hdf mutants [11,[35][36]. Thus, the specific core protein mutations that were introduced did not interfere with these essential functions of versican. Immunostaining using anti-GAGa antibody suggested comparable staining in wild-type and Vcan AA/AA embryos (Fig. 2B). Weaker GAGb immunostaining (with a commercial anti-GAGb antibody) was seen in Vcan AA/AA embryos, since the mutagenized Glu residues occurred within the immunogenic sequence of this antibody (Fig. 2B). Another GAGb antibody, anti-VC [34], and anti-DPEAAE were unreactive in the mutant mice, which was expected, since the mutated sequence DPAAAA alters epitopes of each antibody (Fig. 2C,D) [34]. Specifically, since the peptide epitope of anti-VC, V 436 -PKDPEAAEARRGQ 445 (human versican V1 isoform sequence enumeration; the residues forming " the ADAMTS scissile bonds are in bold) is centered on the cleavage sites [34] it was expected to be nonreactive in Vcan AA/AA mice, unlike the commercial versican antibody, anti-GAGb, which is generated to a much longer polypeptide containing the anti-VC epitope and remains weakly reactive (Fig. 2B, D). We therefore used a neo-epitope antibody that detects the new N-terminus, 442 ARRGQV, formed by cleavage at the GAGb site, since this sequence was unaltered by targeted mutagenesis (Fig. 2C). In sections from Vcan AA/AA mice, anti-ARRGQV reactivity was also lost, indicating that the engineered mutations abrogated cleavage as intended (Fig. 2D). To investigate cleavage at an ADAMTS processing site previously described in the human versican GAGa domain (NIVSFE 405 ) [16] but uncharacterized in mice, we generated a new neo-epitope antibody, anti-NIVNSE, against the corresponding mouse sequence. The antibody reacted with the immunogen peptide with Cterminal residue E 405 , which was shown to be crucial for immunoreactivity, but not when E 405 was succeeded by another residue (i.e., equivalent to the intact, uncleaved peptide sequence) (Supplemental Fig. 1). Anti-NIVNSE staining was present in both wild-type and mutant embryos (Fig. 2E). To explore whether abrogation of the processing sites would replicate ADAMTS-deficient phenotypes, Vcan AA/+ mice were intercrossed at the 3rd (N3) and 9th (N9) generation of outcrosses into The versican antibodies anti-VC (whose peptide epitope is centered on the cleavage site, unlike anti-versican GAGb, which is generated to a longer spanning polypeptide) and anti-DPEAAE showed staining in sections from wild-type E14.5 embryos (Wt), but not Vcan AA/AA embryos, since the introduced mutations alter the epitopes of these antibodies. In contrast, anti-ARRGQF, which reacts with an N-terminal epitope exposed by cleavage at E 441 -A 442 reacted similarly to anti-DPEAAE in Wt embryos, but also gave no staining in Vcan AA/AA embryos, demonstrating lack of cleavage at this site. In contrast, GAGa staining detected by anti-NIVNSE was present in both Wt and Vcan AA/AA embryos, since the E 405 -A 406 cleavage site was unmodified. Scale bars in A-E = 1 mm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) the C57BL/6 strain, with the majority of studies being undertaken at N3. Variably penetrant white spotting, a phenotype seen in Adamts20 and Adamts9 mutants [20,23,27,29], was initially seen in a small number of Vcan AA/+ and Vcan AA/AA mice, but disappeared upon further outcrossing to the C57BL/6 strain. In regard to other ADAMTS protease mutant phenotypes, Vcan AA/AA mice lacked cleft palate, a phenotype associated with reduced versican proteolysis in the palatal shelves of mice with combined reduction/inactivation of Adamts9 and Adamts20 [23,[26][27][28]. Versican cleavage at the E 441 -A 442 site occurs during ovarian follicular maturation and ovulation, and was previously attributed to ADAMTS1, since Adamts1 mutant mice had impairment of both processes [37][38][39]. Vcan AA/AA female mice were fertile, suggesting no impact on ovulation and delivered pups without experiencing dystocia, a phenotype reported in mice with smooth muscle-specific conditional inactivation of Adamts9 that was also accompanied by versican accumulation [32].
We undertook detailed analysis of interdigital web regression in Vcan AA/AA mice, since this is the best characterized of the various defects associated with reduced versican processing. An observed low penetrance of hindlimb STS in Vcan AA/+ mice was increased in Vcan AA/AA mice, but was nevertheless lower than that seen in Adamts20 Bt/Bt mice (Fig. 3A). Introducing one Vcan AA/+ allele in Adamts20 Bt/Bt mice doubled the penetrance, and STS was fully penetrant in Vcan AA/AA ;Adamts20 Bt/Bt mice, affecting the webs between digits 2 and 3 and digits 3 and 4 (Fig. 3A). Forelimb STS was prevalent in Adamts20 Bt/Bt mice, as previously shown [4], whereas hindlimb STS was twice as common as forelimb STS in Vcan AA/AA mice (Fig. 3B). Interestingly, Vcan AA/AA ,Adamts20 Bt/Bt mice had roughly equivalent and near-complete penetrance of forelimb and hindlimb STS (Fig. 3B). The grade of STS, i.e., the extent of the syndactylous web along the length of the digit also progressed correspondingly in association with these genotypes (Fig. 3C, D).
To ascertain the impact of the mutations on versican turnover, we stained for versican and its cleavage products in wild-type, Vcan AA/AA , Adamts20 Bt/Bt and Vcan AA/AA ;Adamts20 Bt/Bt hindlimb autopods at 13.5 days of gestation. At this developmental stage, web regression is ongoing and incomplete so that it is not possible to anticipate which of the mutant webs will eventually fail to regress. Nevertheless, staining with both GAGa and GAGb antibodies appeared to be stronger in each of the mutant autopods compared to wild-type autopods (Fig. 4A). As expected, we saw no anti-DPEAAE staining in any mutant with Vcan AA/AA alleles (Fig. 4B). DPEAAE staining was detectable in the Adamts20 Bt/Bt webs (Fig. 4B), consistent with previous work showing that multiple ADAMTS proteases mediate versican proteolysis at the E 441 -A 442 site [4]. Anti-ARRGQR gave no staining in interdigits from embryos having the Vcan AA/AA genotype, demonstrating abrogation of cleavage after mutagenesis of Glu 338 and Glu 441 (Fig. 4B). Because we observed more intense GAGa staining in Adamts20 Bt/Bt mice (Fig. 4A) suggesting that ADAMTS20 might participate in versican cleavage in this core protein domain, we stained autopods of the various genotypes with anti-NIVNSE and found greatly reduced staining in Adamts20 Bt/Bt autopods (Fig. 4C). Finally, consistent with the prior finding of reduced apoptosis in syndactylous autopods from ADAMTS mutants, we observed reduced immunostaining for cleaved caspase-3, an apoptosis marker, in Vcan AA/AA , Adamts20 Bt/Bt and Vcan AA/AA ;Adamts20 Bt/Bt autopods (Fig. 4D).
Because prior work had demonstrated that versikine induced cell death in ADAMTS-deficient webs [4], we genetically interrogated the role of versican itself in interdigital web regression by analysis of the limbs of mutants specifically lacking either the GAGa domain [40] or GAGb domain (Burin Des Roziers and Valleix, manuscript in preparation), which had not been previously done. In contrast to Vcan hdf/hdf mutant mice [11], which lack all versican isoforms, the isoform-specific Vcan mutants survive past birth and into adulthood [40] [and Burin Des Roziers and Valleix, manuscript in preparation]. STS did not occur in the GAGa homozygous mutants which consistently showed fully separated digits, but was seen in all GAGb homozygous mutants, specifically, 100% (98/98) of GAGb domain homozygous mutant mice (Fig. 5 A, B). It affected forelimbs in 77% (76/98) of mice and hindlimbs in 86% (85/98) of mice, typically involving the 2-3 and 3-4 interdigit web (Fig. 5 B, C).

Discussion:
In a prior in vitro analysis of versican cleavage by ADAMTS5, we demonstrated that mutation of the P1 and P3 glutamic acid residues (according to cleavage site nomenclature of Schechter and Berger [41]in the versican V1 core protein eliminated versican proteolysis at E 441 -A 442 [34]. On the C-terminal side of the cleavage site, we found that elimination of two proximate GAG chains also Anti-ARRG staining is present in wild-type mice but also in Adamts20 Bt/Bt mice, consistent with expression of other ADAMTS proteases that can compensate for the absence of ADAMTS20, contributing to cleavage, but is absent in mice homozygous for the Vcan mutation. Anti-NIVNSE staining is absent in the Adamts20 Bt/Bt autopod, but evident in the Vcan mutant, where the GAGa processing site is intact. Reduced cleaved caspase staining is evident in the presence of mutant versican. Scale bars in A-D = 200 lm. reduced processing, likely via removing a putative binding site for the ADAMTS5 ancillary domain [34]. These findings suggested two alternatives for generating a mutant versican core protein that would be cleavage-resistant at the E 441 -A 442 site. We chose to mutate E 438 and E 441 in the versican core protein in mice since mutation of the residues for GAG chain attachment and elimination of GAG chains could affect the intrinsic properties of versican. In contrast to our approach, Islam et al recently created a cleavage-resistant versican mouse mutant by leaving the DPEAAE sequence intact, but replacing the P1 0 to P3 0 residues (A 442 R 443 R 444 ) by NVY [33]. Despite the intact DPEAAE sequence, Vcan(R/R) mice lacked anti-DPEAAE staining, demonstrating lack of cleavage [33]. Although anti-DPEAAE can no longer be used to detect versican cleavage in Vcan AA/AA mutants, lack of anti-ARRGQF reactivity in Vcan AA/AA embryos strongly supports elimination of cleavage at the E 441 -A 442 site. Thus, each of these mouse mutants appears to have successfully abrogated embryo-wide proteolysis of versican at the E 441 -A 442 site. Accordingly, hindlimb STS occurred in both Vcan AA/AA and Vcan(R/R) mice [33], which carry different mutations engineered using different approaches, i.e., by homologous recombination in ES cells and CRISPR-Cas9 mutagenesis respectively. In contrast to Vcan(R/R) mice [33], Vcan AA/AA mutants did not show lethality or other severe phenotypes. A detailed characterization of the Vcan AA/AA mice suggested that the inserted mutations did not interfere with Vcan expression or isoform prevalence.
We focused on interdigital web regression during the embryonic period as a classic example of developmental tissue sculpting where versican proteolysis was previously suggested to have a crucial role [4]. The present findings provide novel insights on the role of versican and ADAMTS proteases in web regression. First, occurrence of partially penetrant STS in Vcan AA/AA mutants suggests that versican processing at the E 441 -A 442 site contributes to, but does not alone suffice for complete web regression. However, concurrent Adamts20 deletion in Vcan AA/AA mutants led to more highly penetrant STS. If ADAMTS20, previously shown to cleave versican at the E 441 -A 442 site [20], cleaved it nowhere else, then its deletion should not have increased STS penetrance and severity. Therefore, we deduced that ADAMTS20 may cleave versican at one or more additional sites in the versican core protein or has one or more additional substrates in the ECM whose proteolysis is necessary for web regression. Vcan AA/AA webs that were also lacking ADAMTS20 showed increased staining with a GAGa-specific antibody, indicating higher levels of V0/V2 versican, and we observed reduced anti-NIVNSE staining in these webs. Thus, web regression is associated with cleavage of versican at previously defined sites of ADAMTS proteolysis in both the GAGa and GAGb domains. Whether ADAMTS20 cleaves versican at additional sites, or cleaves other ECM molecules to enable web regression remains unresolved for now and will be the subject of future studies.
The finding that GAGb mutant webs but not GAGa mutant webs consistently have defective web regression is consistent with prior work identifying a pro-apoptotic role for versikine, a GAGb-derived fragment, during web regression [4]. Furthermore, Adamts20 Bt/Bt mice haploinsufficient for a Vcan null allele (i.e., Vcan hdf ) had more highly penetrant soft-tissue syndactyly than Adamt-s20 Bt/Bt mice, implying that a versican fragment was necessary for web regression [4]. Taken together, these findings strongly support an active role for versican in web regression, not merely as an ADAMTS substrate that is cleared during web regression, but as an essential contributor to this process as the source of versikine. Since the prior experiments that first suggested this possibility [4] used the Vcan hdf mutant allele that lacked all isoforms, the present analysis now confirms that it is the GAGb domain from which versikine is generated that is crucial. The homozygous GAGb mutants however have a much higher penetrance of soft-tissue syndactyly than the Vcan AA/AA mutants, suggesting that mutating a single cleavage site may not be sufficient to prevent versican proteolysis and that other cleavage sites may exist in the versican core protein that could generate versikine-like activity. This possibility is potentially supported by the higher penetrance resulting from combining the Vcan AA/AA and Adamts20 Bt/Bt mutants. The data we present also suggests that although GAGa cleavage occurs during web regression, it is not causally involved in the process. However, it is not possible to precisely assess the relative contribution of versikine activity vs versican clearance during web regression. GAGb proteolysis is undoubtedly essential, and perhaps even an obligate initiating event; however, which regions or motifs of versikine are critical and their underlying mechanisms are undefined. That the GAGa mutant did not develop syndactyly suggests that versikine activity may not reside in the G1 domain, which is common to both versikine and GHAP, but it could reside in the N-terminal segment of the GAGb domain that constitutes the C-terminal region of versikine.
Mutagenesis of substrates to ascertain the significance of site-specific proteolytic cleavages has been previously used in only a few instances. For example, mice expressing collagen with an introduced mutation that abrogates the classic collagen I cleavage site (Col1a1 (r/r) mice) provided vital information about the role of collagen turnover in bone, wound healing, atherosclerosis and post-partum uterine remodeling [42][43][44][45][46][47][48]. Mice with a mutation that rendered aggrecan resistant to the action of ADAMTS proteases at the E 373 -A 374 site were protected from surgically induced osteoarthritis and inflammatory arthritis, indicating that ADAMTS proteases were major instruments of articular cartilage breakdown in arthritis [49]. Mice with cleavage-resistant reelin have demonstrated the importance of reelin processing by ADAMTS2 and ADAMTS3 in cerebral cortical and hippocampal development [50].
Genetic combination of cleavage-resistant substrate and protease mutants as reported here is rarely used, but can offer new insights. Conceptually, if a protease (e.g., ADAMTS20) has only one substrate (e.g., versican), which it cleaves at a single site (e.g., E 441 -A 442 ), then genetically combining the cleavage-site mutant and protease mutant should not worsen the phenotype of the cleavage-site mutant. In the only other instance of a similar mutant inter-cross we are aware of, a combination of the Col1a1 (r/r) mouse mutant and Mmp2 -/mice led to more severe phenotypes than either transgene alone, highlighting collagen I cleavage at an alternative site by MMP2 [51]. In this regard, we are currently investigating other sites of ADAMTS proteolysis in the versican core protein.
In the future, Vcan AA mutant mice can be investigated more thoroughly for additional morphogenetic roles, such as in the heart, craniofacial development and neural development, or in adult physiological and disease roles. Vcan (R/R) mice demonstrated accelerated wound healing attributed in part to greater myofibroblast activation [52], consistent with prior work showing that pericellular versican accumulation in dermal fibroblasts, achieved either by genetic inactivation of Adamts5 or versican V1 overexpression led to enhanced TGFb signaling and fibroblast-tomyofibroblast transition [18,53]. Furthermore, combining the Vcan AA mutant with other ADAMTS mouse mutants, such as was done here using Adamts20 Bt , may well provide additional insights on the role of versican and ADAMTS processing of versican in the future.

Methods
Targeted mutagenesis of Vcan. The Ensembl transcript Vcan-001 ENSMUST00000109546 (gene ENSMUSG00000021614) was used as the reference for design of a construct targeting the single Vcan locus on chromosome 13. Exon 8 of the 15 Vcan exons generating this transcript encodes the entire GAGb domain. A~9.2 kb region was subcloned from a C57BL/6 BAC clone (RP23-265 J12) into plasmid vector pSP72 (Promega, Madison, WI), containing an ampicillin selection cassette to construct the targeting vector. The point mutations were engineered by PCR mutagenesis. The engineered mutations in exon 8 altered the sequence from 5´GAAGCTGCAGAA3( translated as E 438 AAE 441 ) to 5Ǵ CAGCTGCAGCA3´(translated as A 438 AAA 441 ) and were located 269 nt downstream of the 5 0 end of exon 8. The PCR fragment carrying the mutation was then inserted into the vector using a homologous recombination-based method. The targeting vector contained a 5 0 -homology arm of 3.0 kb juxtaposed to a FRT-LoxP-flanked selfexcising neomycin resistance cassette (IV-UBS-FRT-LoxP flanked Neomycin cassette). This cassette was inserted 203 bp upstream of the intron 7-exon 8 splice site. The 3 0 homology arm extended 5.7 kb downstream of the introduced mutations. The total size of the targeting construct (including pSP72 vector backbone, Neo cassette and a Diphtheria toxin cassette for negative selection) was 16.2 kb. The targeting vector construction was confirmed by restriction mapping and Sanger sequencing. 10 lg of the targeting vector linearized using NotI was electroporated into HF4 (129/SvEv Â C57BL/6) (FLP hybrid) embryonic stem cells (ESC). After selection with G418, surviving ESC clones were expanded for PCR analysis to identify recombinant ESC. Neo was excised by FlpE recombinase during ESC clone expansion leaving behind a 118 nt FRT-loxP tandem site footprint. 4 positive clones identified as positive by Southern blot and PCR screening were selected for further expansion. For Southern blot, genomic DNA was digested with NheI, electrophoresed on a 0.8% gel and hybridized to a 398 bp probe external to the 5 0 homology arm as shown. DNA from the HF4 strain was used as a wild-type control. Real-time quantitative PCR showed a single copy of the transgenic vector integrated in ES cells.
Targeted ESC from clone 151 were microinjected into C57BL/6 blastocysts. Resulting chimeras with a high percentage agouti coat color were mated to C57BL/6 wt mice to generate germline Neo deleted mice. 4 heterozygous mice (3 female,1 male) from one high percentage chimera derived from clone 151 were intercrossed to determine the effect of the transgene. Deletion of Neo and FlpE and the presence of the inserted mutations in these mice were confirmed by Sanger sequencing of PCR-amplified genomic DNA using appropriate primers. Outbreeding to the C57Bl/6 strain was undertaken for a further nine generations and mice were analyzed after the third and ninth intercross for the impact of the transgene on interdigital web regression. Mice were carefully examined for the incidence, extent and severity of STS as previously described [4]. Penetrance of STS was defined as a persistent web of any severity in at least one interdigit of a mouse [4].
Other mouse transgenes: The splice variantspecific versican GAGa mutant mouse strain Vcan (tm1Zim) carries a translational stop codon preceded by an ER-retention signal in the GAGaencoding exon 7. In consequence, versican isoforms V0 and V2 are absent in homozygous animals, while the isoforms V1 and V3 are normally expressed [40]. Mice lacking the GAGb domain were generated by deleting the entire Vcan exon 8 by CRISPR-Cas 9 gene editing and will be described elsewhere (Burin Des Roziers, C. and Valleix,S., manuscript in preparation). qPCR analysis of the different Vcan transcript isoforms coupled with RNASeq analysis, and immunohistochemistry analysis of GAGb homozygous mutant tissue has shown an absence of the V0/V1 transcript isoforms, a marked increase in V2/V3 transcript isoforms and absence of GAGb immunostaining (Burin Des Roziers, C. and Valleix,S., manuscript in preparation).
Antibodies, ELISA and Western blotting. The generation of anti-442 ARRGQF, to the new Nterminus generated by cleavage at E 441 -A 442 was previously described [54]. In accordance with wellcharacterized methods for generation of neoepitope antibodies [55], rabbits were immunized with the peptide immunogen CGGNIVNSE 405 covalently coupled to keyhole limpet hemocyanin. Immune sera were purified by affinity chromatography using the immobilized peptide immunogen as the ligand and tested by ELISA against the same peptide. To determine the specificity of anti-NIVNSE for the neoepitope but not the bridging (intact) peptide sequence and for determining the critical requirements for antibody binding, we utilized ELISA against the immunogen peptides and a number of variant peptides. Peptides (4 lg/ml) were adsorbed to Nunc MaxiSorb plates and reacted with increasing dilutions of the respective antibodies. Anti-DPEAAE 441 , anti-GAGa and anti-VC were previously described [16][17]34].

DECLARATION OF COMPETING INTEREST
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.