Regulated splicing of the fibronectin EDA exon is essential for proper skin wound healing and normal lifespan

Fibronectins (FNs)^* are high molecular weight glycoproteins that are present on many cell surfaces, in extracellular fluids, connective tissues, and basement membranes. They interact with other ECM proteins and cellular ligands, such as glycosaminoglycans, collagen, fibrin, and integrins (Kornblihtt and Gutman, 1988; Hynes, 1990). FNs play key roles in the adhesive and migratory behavior of cells related to fundamental processes such as embryogenesis, malignancy, hemostasis, wound healing, host defense, and maintenance of tissue integrity (Hynes, 1990; ffrench-Constant, 1995; Kornblihtt et al., 1996). FN molecules are a mixture of several protein types that differ in both their primary structure and posttranslational modifications. The amino acid sequence variations are the consequence of the alternative processing of a single primary transcript at three sites: extra domain B (EDB) or extra type III homology B (also called EIII-B or EDII), extra domain A (EDA [EIII-A or EDI]) and type III homologies connecting segment (IIICS; V region in rat). The alternative splicing process is cell type-, developmentally, and age regulated. The EDA and EDB are single exons coding for single type III repeats that are included or excluded from the FN mRNA by exon skipping (Kornblihtt et al., 1984; Gutman and Kornblihtt, 1987; Schwarzbauer et al., 1987; Zardi et al., 1987). The IIICS region undergoes a more complex pattern of splicing which is species dependent (Schwarzbauer et al., 1983). In humans, over 20 main different isoforms are generated after mRNA processing, whereas in rats and mice, 12 variants are produced (ffrench-Constant, 1995; Kornblihtt et al., 1996). Two major forms of FN exist, plasma (pFN) and cellular (cFN), having similar but not identical polypeptides of ∼220–240 kD. pFN is a soluble dimeric form that lacks both the EDA and EDB domains. It is synthesized by hepatocytes and secreted into the bloodstream. cFN is a dimeric or cross-linked multimeric form containing the EDA and EDB domains at variable proportions. It is made by fibroblasts, epithelial, and other cell types, and is deposited as fibrils in the ECM (Hynes, 1990; ffrench-Constant, 1995; Kornblihtt et al., 1996).

The molecular mechanisms involved in processing the FN transcript are well understood for both the EDA and EDB exons (Mardon et al., 1987; Lavigueur et al., 1993; Caputi et al., 1994; Du et al., 1997; Lim and Sharp, 1998; Muro et al., 1999). The mouse EDA exon is regulated in a similar manner to its human counterpart and optimization of the 5′ and 3′ splice junctions of the exon leads to constitutive splicing in NIH3T3 cells (Muro et al., 1998).

Despite the fact that both FN primary structure and FN alternative splicing were described almost two decades ago (Kornblihtt et al., 1983; Petersen et al., 1983; Schwarzbauer et al., 1983; Kornblihtt et al., 1984; Zardi et al., 1987), many aspects of the function of FN isoforms remain elusive. In vivo studies in chicken, rat, and Xenopus showed the increased inclusion of EDA and EDB exons in the FN mRNA derived from embryos during development (ffrench-Constant and Hynes, 1989; Oyama et al., 1989; Pagani et al., 1991; DeSimone et al., 1992). Once development is complete, the inclusion of EDA and EDB decreases in a wide range of tissues. In fact, EDA⁻ and EDB⁻ forms are the most abundant forms in adult humans and rats (Magnuson et al., 1991; Pagani et al., 1991). This exclusion is cell type specific and differs in extent between the two exons. Quantitative mRNA studies in these tissues showed that the EDB exon is excluded more often than the EDA exon from the FN mRNA (ffrench-Constant and Hynes, 1989; Pagani et al., 1991; Caputi et al., 1995). It has been postulated that the inclusion of the spliced regions could alter the conformation of the RGD sequence, which is the main cell-binding site, via modification of its neighboring modules (ffrench-Constant, 1995) or by the alteration of the global conformation of the FN molecule (Manabe et al., 1997). In both cases, the alteration might affect the strength of the central cell–binding domain interaction with integrins. Since integrins α₉β₁ and α₄β₁ were recently described as the cellular receptors for the EDA segment, another possibility is that cell adhesion can be directly regulated by alternative splicing (Liao et al., 2002). Other functions proposed for the EDA segment are as follows: wound healing (Clark et al., 1983; ffrench-Constant et al., 1989); matrix assembly (Guan et al., 1990); dimer formation (Peters et al., 1990); secretion (Wang et al., 1991); cell adhesion (Xia and Culp, 1995); cell differentiation (Jarnagin et al., 1994); tissue injury and inflammation (Satoi et al., 1999; Okamura et al., 2001); and cell cycle progression and mitogenic signal transduction (Manabe et al., 1999). Because most of the above studies were done using in vitro and cell culture systems, the in vivo role of the EDA segment still remains obscure. The use of mouse models could provide the best approach to understand the in vivo function of the EDA exon. However, the in vivo study of the function of the different protein isoforms is complex due to the simultaneous presence of more than one protein form at a given specific developmental time and tissue. Gene targeting in mice has allowed the specific deletion of alternatively spliced exons in different protein systems such as α₆ integrin, γ-aminobutyrate receptor, dopamine D2 receptor, FN, Pax6, and Stat3 (Gimond et al., 1998; Homanics et al., 1999; Usiello et al., 2000; Wang et al., 2000; Fukuda et al., 2002; Singh et al., 2002; Yoo et al., 2002). Thus, it represents the best approach to understand the in vivo function of the EDA exon.

Taking advantage of our detailed knowledge of the elements involved in EDA splicing regulation (Mardon et al., 1987; Caputi et al., 1994; Muro et al., 1998, 1999), we have designed a novel approach to study the in vivo function of protein isoforms coded by genes that undergo alternative splicing without modifying the coding sequence of the FN gene. Using gene targeting, we generated a mouse strain containing the EDA allele with optimized splice sites at both splicing junctions (EDA⁺ allele), with loxP sites located in the adjacent introns. By mating this mouse strain with a “deleter” CRE-recombinase expressing mouse, an EDA-null allele was obtained (EDA⁻ allele). Here, we show that FN mRNA produced by homozygous EDA^+/+ mice contained constitutive inclusion of the EDA exon devoid of developmental and tissue-specific regulation. Mice lacking EDA splicing regulation (EDA^+/+ and EDA^−/−) were viable and phenotypically similar to the wild-type mice. However, the mouse strain producing EDA⁻ FN showed an abnormal cutaneous skin wound healing. On the other hand, mice having the EDA exon constitutively included in the FN mRNA showed normal wound healing but had a striking decrease in the levels of FN in most of the organs analyzed from adult mice. Furthermore, and most importantly, both mouse strains had a significantly shorter lifespan than EDA^wt/wt animals, suggesting that the presence of both isoforms was necessary for efficient long-term maintenance of biological functions.

The study of the in vivo function of protein isoforms generated by alternatively spliced genes is generally limited by the fact that the different isoforms coexist in a temporal and spatial manner in the same tissue of the organisms. Therefore, the availability of animal models having a controlled expression of the different forms is crucial for these studies.

We have previously reported that optimization of the 5′ and 3′ splicing sites of the mouse EDA exon from the FN gene leads to constitutive exon inclusion in NIH3T3 cells (Muro et al., 1998). These results coupled with the power of homologous recombination technology in mouse embryonic stem (ES) cells enabled us to study the biological significance of protein isoforms generated by alternative splicing (for a detailed description of the rationale see Online supplemental material). Furthermore, to date, there is no report of optimization of splice sites leading to constitutive splicing in vivo of a natural alternatively spliced exon.

To generate an FN allele devoid of alternative splicing in the EDA exon, the wild-type EDA exon in the FN gene was replaced with a “floxed” EDA exon having both the 5′ and 3′ optimized splice sites described in Fig. S1 (available). After electroporation of 129 Sv/J ES cells with the targeting construct (Fig. 1 A), G418-resistant clones were selected and analyzed by Southern blot hybridization. A second step of recombination by transient expression of the CRE-recombinase was performed to eliminate the Neo-TK cassette from the EDA-flanking intron. Southern blot analysis and sequencing of the PCR-amplified genomic EDA region confirmed the introduction of the optimized 3′ and 5′ splice sites and the loxP sites flanking the EDA exon in one of the FN alleles. Hereafter, this allele will be called the EDA⁺ allele. The heterozygous EDA^+/wt ES cells were used to obtain male chimeras that transmitted the EDA⁺ allele to the germline. Heterozygous EDA^+/wt mice were mated with CMV-CRE-transgenic mice to obtain the EDA-null allele (henceforth, called the EDA⁻ allele) after in vivo deletion of the loxP-flanked EDA exon. Homozygous mice were obtained by interbreeding heterozygous mice, and screening was performed by Southern blot analysis (Fig. 1 B) or by PCR of tail DNA biopsies (not depicted).

The homozygous EDA^+/+ and EDA^−/− mice after birth exhibited no apparent developmental abnormalities compared with wild-type or heterozygous littermates; they reached adulthood and were fertile. In addition, no differences were detected in the organs from homozygous EDA^+/+ and EDA^−/− mice by gross histological and pathological analysis (unpublished data). The mutated alleles segregated at the expected Mendelian frequency, and litter size and survival rate of newborn mice (measured until weaning of the litter) were similar to that of EDA^wt/wt animals. No differences were observed neither in the shape of the embryos nor in the number of somites of embryos derived from timed matings. The detailed data are shown in the Online supplemental material section.

The inclusion of the EDA exon within the FN mRNA is subjected to a tight tissue-specific regulation. The most dramatic example is seen in liver, where pFN is synthesized and the EDA exon is completely excluded from the mRNA in the adult (Tamkun and Hynes, 1983; Kornblihtt et al., 1984).

Radioactive RT-PCR was used to analyze most of the tissues and organs in all three genotypes (EDA^wt/wt, EDA^+/+, and EDA^−/−). An example of the EDA splicing pattern in representative tissues such as liver, brain, and kidney is shown in Fig. 1 C. These results were confirmed by RNase protection analysis of the same samples (Fig. S2 available). We observed that the EDA⁺ form was never present in the samples from EDA^−/− mice and the EDA⁺ product from EDA^+/+ mice was present in all tissues. Both RNase protection and RT-PCR analysis of the FN mRNAs showed that there is 99–100% inclusion of the EDA exon in all tissues from EDA^+/+ mice, whereas there is 100% EDA exon exclusion in all tissues belonging to the EDA^−/− mice. This confirmed that the rationale of genetic modification was successfully fulfilled in the real in vivo situation.

Next, we examined the expression of FN mRNA, protein isoforms and ECM formation in MEF prepared from EDA^wt/wt, EDA^+/+, and EDA^−/− animals. RT-PCR analysis confirmed the constitutive splicing-in of the EDA exon in the EDA^+/+ and its absence in the EDA^−/− MEFs (Fig. 2 A). Analysis of protein extracts by Western blot using polyclonal anti-FN and monoclonal anti-EDA antibodies showed normal protein levels in both the EDA^+/+ and EDA^−/− MEF (Fig. 2 B). The anti-EDA antibody detected the EDA⁺ FN isoform only in EDA^+/+ and EDA^wt/wt MEF (Fig. 2 C).

Immunofluorescence analysis using a confocal microscope showed the presence of a normal FN-ECM in EDA^+/+ and EDA^−/− MEF when compared with EDA^wt/wt MEF (Fig. 2 D). The signal obtained with the anti-EDA antibody colocalized with that observed with the anti-FN antibody.

Next, we examined whether the modifications in the natural alternative splicing pattern of FN had some consequences in other cellular proteins or FN itself. Consequently, protein extracts from different tissues of the different genotypes were prepared from 3-mo-old mice and analyzed by SDS-PAGE. No evident difference was observed at least at the sensitivity level of Coomassie blue staining of 5–17% gradient gels (unpublished data). On the other hand, a Western blot analysis using anti-FN polyclonal antibody showed a striking decrease in FN levels in most of the tissues analyzed from EDA^+/+ mice (Fig. 3 A). This decrease was also evident in EDA^+/wt and EDA^+/− mice, both bearing one EDA⁺ allele. The observed differences were not due to a different protein load of the gels, as evidenced by the Coomassie blue staining and by the incubation of the same membranes with an anti–β-tubulin mAb (Fig. 3 B). A significant decrease in FN amounts was also observed in the plasma and serum samples prepared from EDA^+/+ animals when compared with those from EDA^wt/wt mice (Fig. 3 C). Western blot using an anti-EDA antibody confirmed the presence of EDA containing FN in the EDA^+/+ plasma and its complete absence in the EDA^−/− sample (Fig. 3 C).

It was considered important to analyze if the FN level differences observed in the different tissues were associated to changes in the levels of the specific EDA receptor α₉β₁-integrin (Liao et al., 2002). We observed no differences in α₉- and β₁-integrin levels among EDA^wt/wt, EDA^+/+, and EDA^−/− mice by Western blot analysis in liver and brain (Fig. 3 D). However, it remains to be seen if the localization pattern of the integrin receptors or their signaling have been affected in the mutant mice.

On the other hand, to determine whether the observed decrease in FN protein levels in the EDA^+/+ mice was correlated to a decrease in the mRNA levels, we performed Northern blot analysis of total RNA isolated from the tissues having the higher decrease in FN levels (brain and heart) from 3-mo-old male mice. No significant variations were observed in RNA samples from the brain and heart of EDA^+/+ mice (Fig. S3 available).

To determine whether the mechanisms that produced the FN decrease were developmentally regulated, we performed a Western blot analysis of samples prepared from mice of different ages. The decrease in FN levels was not present during embryonic development. It started after birth in most of the tissues, and arrived to a plateau in 2-mo-old mice (Fig. 4). In some tissues, the FN levels in EDA^+/+ mice were ∼10–20% of the amount found in EDA^wt/wt mice (heart, blood, brain, kidney, and diaphragm), whereas the liver showed a lesser decrease in FN levels in EDA^+/+ mice, arriving to 60–70% of the amount present in EDA^wt/wt mice.

In an attempt to determine alterations in the FN-degradative mechanisms, we performed gelatin zymography and observed no variations in plasma and other tissues (brain and heart) in the levels of matrix metalloproteinases (MMPs), enzymes known to degrade FN (Sternlicht and Werb, 2001). Moreover, no differences were observed in the degradation of in vitro translated recombinant fragments of FN with or without the EDA exon that were incubated with tissue extracts (unpublished data). Western blot analysis of tissue extracts using a polyclonal anti-FN antibody did not reveal the presence of extra bands or any smaller degradation product of EDA^+/+ tissues, neither with a 6% nor a 12% SDS-PAGE gel (unpublished data). Additionally, we were unable to reproduce in vitro the reduction in FN levels after culturing heart fibroblasts of adult mice (from 2-, 4-, and 6-mo-old animals), as the cell extracts from EDA^+/+ mice showed similar levels of FN to those of the EDA^wt/wt fibroblasts (unpublished data).

These experiments showed that the striking reduction in FN protein levels in EDA^+/+ mice occurred after birth. The decrease in FN was not related to the mRNA levels present in each tissue or to a detectable EDA⁺-dependent degradation, suggesting that alternative mechanisms should be responsible for the decrease of tissue FN in the EDA^+/+ animals.

The FN present in the skin of wild-type animals does not contain the EDA exon. However, the FN-EDA⁺ form is thought to participate actively in the reepithelization process (Clark et al., 1983; ffrench-Constant et al., 1989). To explore the role of the EDA exon in the wound healing process, we performed full thickness excisional skin wounds in EDA^wt/wt, EDA^−/−, and EDA^+/+ mice. The wound healing process was analyzed at 1, 3, 5, 7, 10, and 14 d after wounding. Hematoxylin-eosin staining of tissues from full thickness wounds during the early phase of reepithelization (days 1 and 3) did not show any differences in reepithelization and neovascularization (unpublished data). During late phase (days 5 and 7), the reepithelization and formation of granulation tissue were indistinguishable between EDA^+/+ and EDA^wt/wt mice (Figs. 5 and 6). However, the abnormal healing process was observed in the EDA^−/− mice. In fact, at day 5, the newly formed epidermis completely covered the wound area in the EDA^−/− mice but the border between the new epidermis and the granulation tissue was not sharply defined as in the control or EDA^+/+ mice. Moreover, the granulation tissue of the EDA^−/− wounds showed less compact cells with edematous-like areas below the interface between the epidermis and the derma granulomatous reaction (Fig. 5, bottom).

At day 7, the EDA^−/− wounds were clearly different from those from EDA^wt/wt and EDA^+/+ mice (Fig. 6 A, bottom). In fact, the EDA^−/− wounds showed ulcerative processes at the newly formed epidermal region level, resulting in a delay in the reepithelization. Infiltration of inflammatory cells such as polymorphonuclear leukocytes and macrophages were detected by microscopic analysis of the edematous-like and ulcerative regions, and a higher cell proliferative activity was observed by in vivo BrdU labeling in these regions (Fig. 6 B, bottom right). On the contrary, neither EDA^wt/wt nor EDA^+/+ wounds presented a high number of BrdU-positive cells in the area below the newly formed epidermal region (Fig. 6 B, bottom left and center). The status of the basal lamina in the EDA^−/− skin was similar to that of the EDA^wt/wt, as seen by both the Azan-Mallory coloration and the specific staining of laminin in the same sections using a rabbit polyclonal antilaminin antibody. Cell death and apoptosis-specific analysis were performed by TUNEL and caspase-3 assays of the wound sections, respectively, and showed no significant differences in the number of positive cells between the different genotypes (Fig. S4 available). Scoring of the skin healing process of three independent experiments (eight 3-mo-old mice per genotype in each experiment) for the presence of ulceration in the wounds showed that EDA^−/− mice developed them 2.9 times more frequently than the EDA^wt/wt mice (63% of EDA^−/− vs. 22% of EDA^wt/wt, P < 0.0002 by both the Fisher Exact test and the χ² test) at d 7 after wounding.

A limited number of 11-mo-old mice (five EDA^wt/wt and six EDA^−/−mice) that were subjected to the same wound healing protocol showed an even higher proportion in the frequency of wound ulceration in the EDA^−/− mice (83.3% vs. 20%, EDA^−/− and EDA^wt/wt; respectively. P ≤ 0.01 by both the Fisher Exact test and the χ² test).

To determine whether integrin α₉β₁ levels were changed, we performed a Western blot analysis of untreated skin samples and observed similar levels of integrin α₉β₁ among all three genotypes (Fig. 3 E).

The in vivo studies performed on rats show a critical decrease in EDA⁺/EDA⁻ ratio after birth, with an additional gradual decrease of the ratio in some tissues that is correlated with the age of the animal (Magnuson et al., 1991; Pagani et al., 1991). We addressed the question if in EDA^+/+ and EDA^−/− mice, having no regulation of EDA splicing, the presence or absence of the EDA exon could have an effect on their lifespan. For this purpose, we performed a long-term experiment using 139 mice (EDA^wt/wt, n = 39; EDA^+/+, n = 47; and EDA^−/− n = 53) that were caged separately throughout the experiment and the time of animal death was recorded. The Kaplan-Meier graphical representation of the data is shown in Fig. 7. It can be seen that up to 15-mo-old mice from all the genotypes did not show any difference in survival. However, from thereon, there was a decline in the survival rate in EDA^+/+ and EDA^−/− mice. At 30 mo, only seven and eight EDA^+/+ and EDA^−/− mice were alive, (10 and 12.5% survival, respectively), whereas 21 EDA^wt/wt mice survived (53.8% survival). Log-rank test statistical analysis indicated that both EDA^+/+ and EDA^−/− mice lifespan curves were statistically significant when compared with EDA^wt/wt mice with P ≤ 0.0005. Postmortem external examination showed that 20–25% of the EDA^−/− mice had different skin hyperproliferative and spontaneous ulcerative lesions on their mid-dorsal skin. In case of the EDA^wt/wt and EDA^+/+ mice, <5% had any kind of skin lesions. Autopsies were performed on naturally deceased mice of all three genotypes. Liver tumors, intestinal obstruction, prostatic necrosis, and pulmonary edema were the most frequently observed features, but there was no significant difference in these pathologies of the internal organs between the three genotypes.

The first evidence of the in vivo importance of FN was demonstrated by R. Hynes's group who showed that FN-null mice die during embryonic development (George et al., 1993). The early embryo lethality makes it difficult to study the biological functions of FN in the in vivo models. Therefore, one of the best methodologies to study the function of FN is by selective modification of the gene. Along with this line, Sakai et al. (2001) described a mouse model lacking pFN which showed increased neuronal apoptosis and larger infarction areas after transient focal cerebral ischemia. In an attempt to elucidate the role of the EDB exon, mice lacking this exon were also produced, but showed no obvious abnormalities in vivo (Fukuda et al., 2002).

We have investigated the in vivo role of the EDA exon by constructing mouse strains unable to carry out EDA exon alternative splicing. The EDA^+/+ strain showing constitutive expression of the EDA exon in all tissues was obtained without any modification of the coding sequences and gene structure. This novel approach avoided the difficulties found in a previous attempt to produce animals devoid of alternative splicing of the EDB exon in the FN gene where the modification of the gene structure through the introduction of a selection cassette within the introns produced an FN-null allele (Georges-Labouesse et al., 1996). A similar phenomenon has recently been observed in other genes (Lewandoski, 2001). None of these negative effects was observed in our mutant mice, indicating that the insertion of loxP sites within the flanking introns did not modify the normal pre-mRNA transcription and processing.

Moreover, we observed no obvious embryonic mortality nor malformation in any of the homozygous mutant mice analyzed (EDA^+/+ and EDA^−/− animals), suggesting that the alternative splicing regulation of the EDA exon is not essential for the developmental processes or that it could be compensated by other gene products. The absence of any obvious developmental abnormality in EDA^+/+ and EDA^−/− mice suggests that the information and postulated roles for the EDA exon obtained by using recombinant FN fragments (Introduction) cannot be extended automatically to the in vivo functions where more complex mechanisms are acting.

The observation of decreased levels of FN in adult EDA^+/+ mice, but not in embryos and very young mice (<1 mo old) suggests that different control mechanisms of FN levels are acting before and after puberty. This decrease could be due to one of the following explanations or a combination of them: (1) a decrease in FN mRNA levels; (2) a decrease in the secretion rate of FN; or (3) an increase in extracellular degradation of FN.

A series of experiments have been performed to test which of these mechanisms, if any, was acting in the EDA^+/+ mice. Northern blot experiments showed that the mRNA levels were not directly related to FN protein levels (Figs. 3 and 4 and Fig. S3). Protease levels were similar to those of EDA^wt/wt, no additional FN degradation products were seen in EDA^+/+ tissue extracts, and we were unable to see a decrease of FN levels in in vitro cultured cells derived from adult organs of EDA^+/+mice. The failure to identify the precise cause of the low FN-EDA⁺ levels suggests that an atypical mechanism might be acting. One possibility is that a specific intracellular trafficking mechanism is able to modulate the levels of EDA⁺ FN forms in adult animals. This mechanism was suggested for polarized secretion of FN forms in airway epithelial cells (Wang et al., 1991). It is interesting to note that Schwarzbauer et al (1989) have shown that the IIICS-0/IIICS-0 dimers were not secreted. It is possible that a similar control in the secretion mechanism could also be present for the EDA⁺ subunits.

However, none of these effects was obvious in cultured fibroblasts from embryos and adult animals, and in the early stages of development of mice. In fact, the decrease in the amount of FN was evident only after 2–3 mo of age, which is after sexual development. Alternatively, a down-regulation of specific tissue inhibitors of MMPs in adult animals would have as its consequence an increase in activity of specific MMPs, a subtle change that direct gelatin zymography analysis will not detect.

Initially, up to 3 d after wounding, there were no differences in the healing process between the three genotypes (EDA^wt/wt, EDA^−/−, and EDA^+/+). This seems to be consistent with the fact that in the first days after wounding, deposited FN is mainly derived from plasma (Clark et al., 1983; ffrench-Constant et al., 1989). It was shown that normal wound healing could be helped by the presence of cFN (ffrench-Constant et al., 1989; Sakai et al., 2001). In fact, in mice devoid of pFN, normal wound healing was reported (Sakai et al., 2001). cFN is produced in situ 3 d after wounding (Clark et al., 1983; ffrench-Constant et al., 1989; Sakai et al., 2001) and a thin layer of cFN deposited on the surface of the wound immediately after injury in pFN-null mice (Sakai et al., 2001). Expression of cFN by the cells found at the base and edges of the wound just beneath the epidermis was observed, and the presence of cFN preceded the migration of epithelial cells that eventually covered the wound (ffrench-Constant et al., 1989). Previous studies have also shown that FN appears in the provisional matrix beneath the migrating epidermis coordinating with the FN-receptor expression on migrating epidermal cells (Clark et al., 1982, 1983; Grinnell, 1984). These data correlate perfectly with our results showing that after that period, despite the formation of granulation tissue and neovascularization, we observed abnormal reepithelization in EDA^−/− mice. The granulation tissue of the EDA^−/− wounds displayed edematous regions that occur before the ulcerative process. The subsequent epidermal ulceration was accompanied by a proliferative stimulus with the influx of polymorphonuclear leukocytes and macrophages, and a delay in the healing of the wounds. Possible reasons for these phenomena could be related to defects in the basal lamina, a change in integrin receptor levels, and/or a defect in the skin keratinocytes, fibroblasts, or immune system. The basal lamina looked similar in EDA^−/− mice compared with EDA^wt/wt mice by both Azan-Mallory coloration and the specific immunostaining of laminin. However, we cannot rule out minor differences in one of the other specific proteins that are present in the basal lamina. Moreover, similar levels of α₉β₁ integrin, one of the EDA receptors (Liao et al., 2002), were observed in skin of the three genotypes. Preliminary attempts to define differences between the EDA^wt/wt and EDA^−/− genotypes in immune system cells (T/B lymphocytes ratio and delayed hypersensitivity) and in the characteristics of fibroblasts (in vitro cell migration, cell adhesion, and duplication time) failed to show significant differences. However, it remains to be seen if the localization pattern of the integrin receptors or their signaling have been affected in the mutant mice.

The increase in ulceration in the EDA^−/− wounds could be the result of a higher rate of cell death in the skin of the mice during wound closure. This does not seem to be the case as the number of positive cells between the EDA^wt/wt, EDA^−/−, and EDA^+/+ wounds in the granulation tissue were similar in the area devoid of ulceration by both TUNEL and caspase-3 assay.

On the basis of these considerations and the data described in Results, we conclude that FN containing the EDA segment plays an essential role in the organization of the granulation tissue and probably in epidermal cell migration, either directly or by interaction with other ECM components.

It has been shown that deficiency of SH2-containing inositol-5-phosphatase, lysosomal acid lipase, and DNA topoisomerase IIIβ had no consequences in embryonic development, but shortens the lifespan of the animals (Helgason et al., 1998; Du et al., 2001; Kwan and Wang, 2001). The shortened lifespan could be due to the changes in biological processes, which require the presence of the investigated protein during aging.

In the case of the FN mice, we showed that the absence of regulation of the EDA exon affects the lifespan of mutant animals. In normal animals, a dramatic change in the EDA⁺/EDA⁻ ratio is observed after birth and a gradual decrease of EDA inclusion is observed in the tissues that increases with the age of the animal (Magnuson et al., 1991; Pagani et al., 1991). Alternative splicing of FN is regulated at both the extracellular (ffrench-Constant, 1995; Kornblihtt et al., 1996; Boyle et al., 2000) and the molecular levels (Mardon et al., 1987; Lavigueur et al., 1993; Caputi et al., 1994; Muro et al., 1999). The mutant EDA^+/+ and EDA^−/− mice have no regulation of splicing of the EDA exon. It was possible that some biological processes that required specific EDA⁺/EDA⁻ ratios of FN were altered because of the absence of regulation of EDA splicing. In fact, this was shown in this paper for the specific case of cutaneous wound healing in the EDA^−/− mice. Furthermore, the low healing efficiency was exacerbated in older mice. In a similar fashion, other maintenance and tissue repair mechanisms could be affected, and the cumulative lower efficiency may result in a decrease in the lifespan of mutant animals. In the specific case of the EDA^+/+ mice, due to the dramatic reduction in FN levels in most organs, it remains unclear whether the shortened lifespan seen in EDA^+/+ mice was due to a change in FN quantity and/or quality. Although no significant differences have been reported on the lifespan of c129 and C57BL/6 (Smith et al., 1973; Goodrick, 1975; Kunstyr and Leuenberger, 1975), we cannot formally rule out that the differences in chromosome background at the FN EDA locus (C57BL/6 for the EDA^wt/wt and c129 with one or two loxP sites for EDA^−/− and EDA^+/+, respectively) could be responsible for the observed differences in lifespan.

In summary, we generated mice devoid of regulated splicing at the EDA exon of the FN gene. Surprisingly, both mouse strains showed that the alternative splicing at the EDA exon was dispensable during embryonic development, but it was necessary for a normal lifespan. Moreover, mice unable to incorporate the EDA exon in the FN protein displayed abnormal skin wound healing. On the other hand, mice having constitutive inclusion of the EDA exon showed an important decrease in the FN levels in all tissues.

The results presented here are important for three reasons: first, we presented a novel method to study the function of protein isoforms; second, we added valuable information regarding the function of the FN isoforms; and third, we showed that changes in the regulation of splicing are responsible for subtle changes in the phenotype that may be present, but undetected, in human pathologies.

C57BL/6 mice were bred at the International Centre for Genetic Engineering and Biotechnology animal house or purchased from Harlan, Italy. All mice were housed in rooms at 25°C with 12-h light/dark cycles. Mice were fed with standard food and water ad libitum. In all experiments, mice were males and age-matched, except in the aging experiment in which both males and females were used. Blood was drawn by puncture of the internal maxillary artery.

A 13-kbp genomic clone containing the EDA exon region of the murine FN gene was isolated from a 129/Sv λ DashII library (a gift of G. Friedrich, Fred Hutchinson Cancer Research Center, Seattle, WA). The exon–intron structure and restriction map of this fragment was determined by restriction enzyme mapping followed by Southern blot analysis using individual exons as probes, which were PCR amplified from the human FN cDNA. Both the 5′ and 3′ splice sites of the EDA exon were modified by PCR-directed mutagenesis as described previously (Muro et al., 1998). In brief, the 5′ splice site was mutated to the 5′ consensus (5′CAGgtaagt3′) and the 3′ splice site was replaced by that of the constitutively spliced-in second exon of the apolipoprotein AI gene, which matches exactly the 3′ consensus (see Online supplemental material).

Our targeting construct consisted of 11.5 kbp of mouse genomic DNA. Neomycin (Neo) and diphtheria toxin (DTA) genes were used as positive and negative selection markers, respectively, for the first round of recombination. Three loxP sites were included in the targeting construct flanking the Neo-Thymidine Kinase (TK) cassette and the EDA exon.

The targeting construct was electroporated into ES cells, and 350 G418-resistant clones were genotyped by Southern blot hybridization of NcoI–NdeI-digested DNA using the 5′ external probe (Fig. 1 A). Eight recombinant clones were obtained, and three of them underwent a second step of recombination by transient expression of the CRE recombinase (by transient transfection of the pBS185 plasmid; a gift of B. Sauer, Stowers Institute for Medical Research, Kansas City, MO) and selection in the presence of gancyclovir. Two recombinant clones having the deleted Neo-TK cassette and the loxP-flanked EDA were selected and amplified. More detailed Southern blot analysis and sequencing of the PCR-amplified genomic EDA region of two recombinant clones was done to confirm the introduction of the optimized 3′ and 5′ splice sites and the loxP sites flanking the EDA exon in one of the FN alleles. Both clones were used in order to obtain chimeras. Male chimeras obtained by microinjection of C57BL/6 blastocyst were mated to C57BL/6 females in order to generate heterozygous animals (F1 mice).

The F1 offsprings were analyzed for the transmission of the modified EDA exon within the FN gene. HindIII-digested DNA from tail biopsies from the agouti progeny was genotyped by Southern blot using the EDA + 1 internal probe (Fig. 1, A and B). F1 heterozygous offsprings were mated with C57BL/6 mice to obtain F2 generation, which were then mated to each other to produce EDA^+/+ mice (inclusion of the modified EDA exon in both the FN alleles). F1 EDA^+/wt mice were mated with a CRE-recombinase transgenic mice (on a C57BL/6 background) to obtain EDA^−/wt progeny (F2) that were subsequently mated to obtain EDA^−/− animals.

Total cellular RNA from various tissues was prepared by the guanidinium thiocyanate method (Chomczynski and Sacchi, 1987), quantitated by measuring absorbance at 260 and 280 nm, and RNA integrity was confirmed by running the samples on 1% agarose formaldehyde gels. For embryonic RNA samples, embryos were dissected under a stereomicroscope (model SMZ-10; Nikon) and tissues were pooled from five to six embryos (13.5 d p.c.). RNA was prepared using RNAZOL TM B (Biotecx Lab) according to the manufacturer's instructions.

10–20 μg of total RNA from different tissues was denatured and run on a 1.2% agarose formaldehyde gel, blotted onto a nylon membrane (Hybond-N; Amersham Biosciences), and stained with a methylene blue solution to verify the efficiency of the transfer. The membrane was prehybridized in ExpressHyb^TM hybridization solution (CLONTECH Laboratories, Inc.) and hybridized with the radioactive-labeled probe. It was exposed overnight using BIOMAX MS films (Kodak), and quantification of the radioactive signal was done with a phosphorimager (Canberra-Packard). FN cDNAs were analyzed as described in the Materials and methods section of the online supplemental material.

Adult mouse tissues were surgically removed, homogenized in lysis buffer (150 mM NaCl, 50 mM Tris-HCl, pH 8, 1% wt/vol NP-40, and 2× complete protease inhibitor cocktail [Boehringer] with a polytron [Ultra-Turrax T8; Boehringer]), and analyzed by SDS-PAGE after protein determination using the Bradford assay (protein assay kit; Bio-Rad Laboratories).

In the case of embryos, tissues from five to six embryos per genotype were surgically removed with the help of a stereomicroscope and pooled. The protein extracts were prepared by sonication in 150 μl of lysis buffer.

50 μg protein extracts were run on a 6%, 12%, and a gradient of 5–17% SDS-PAGE gels. The proteins were blotted onto a nitrocellulose membrane, incubated with the first and secondary antibodies, and the reaction was developed using the ECL system (Amersham Biosciences). The following primary antibodies were used: anti–human FN goat polyclonal antibody (Sigma-Aldrich), anti–human EDA mouse mAb 3E2 (Sigma-Aldrich), anti–α₉-integrin rabbit polyclonal (a gift from D. Sheppard, University of San Francisco, San Francisco, CA), anti–β1-integrin rabbit polyclonal (Santa Cruz Biotechnology, Inc.), and the anti–β-tubulin mAb E7 (Developmental Studies Hybridoma Bank, University of Iowa) antibodies. Densitometry analysis was performed by using the Versa-Doc Analyzer (Bio-Rad Laboratories) with Quantity One software (Version 4; Bio-Rad Laboratories).

Embryonic fibroblasts (13.5 d) from EDA^wt/wt, EDA^+/+, and EDA^−/− mice were cultured on coverslips for 24 h. Cells were washed and fixed with 4% vol/vol PFA in PBS, permeabilized with 0.5% vol/vol Triton X-100 in PBS, and blocked. Cells were incubated with anti–human FN goat polyclonal antibody and in parallel with goat normal serum or with the mAb 3E2, and incubated with fluorescein-labeled rabbit anti–goat IgG (DakoCytomation) or with rhodamine-labeled rabbit anti–mouse IgG (DakoCytomation), respectively. Negative controls were performed with cells incubated directly with the fluorescent antibody. The analysis of the immunofluorescence was done by using a confocal laser scanning microscope (model Axiovert 100 M; Carl Zeiss MicroImaging, Inc.).

Full thickness skin wounds were done in 2–3-mo-old male mice (n = 12 per genotype) in two independent experiments. Genotypes were blind to the researcher and observer. In brief, animals previously anesthetized with Avertin, were shaved, swabbed with 70% ethanol and two circular, full thickness excisional wounds (5-mm diam, separated by 10 mm) were created on their backs, by excising the skin and panniculus carnosus. The mice were housed individually. For macroscopic analysis, mice were monitored daily and wounds were photographed and measured at each time point (0, 3, 5, 7, 10, and 14 d) to analyze the healing process.

For histology of wounds, six 3-mo-old male mice from each genotype were killed at each time point (1, 3, 5, 7, 10, and 14 d), and their wounds were harvested by complete excision including the 3 mm of the epithelial margins. Wounds were immediately fixed in 2% formaldehyde in PBS containing 0.02% NP-40 at 4°C for 3 h. They were placed in 2% formaldehyde in PBS, left overnight in the cold, and paraffin embedded. 3–5-mm sections from the middle of the wounds were stained with hematoxylin-eosin and analyzed by light microscopy. The analysis of the wound tissue at day 7 was performed in groups of eight mice per genotype in three independent experiments showing similar results. For basal lamina immunohistochemistry of the wounds, we used the polyclonal antilaminin antibody from Biogenex as described by the manufacturers.

To allow the analysis of the cell proliferation in the wounds, the nucleoside analogue BrdU (250 mg/kg of body weight) was injected i.p. 2 h before the animals were killed (eight animals per each genotype, day 7 after wounding). Samples were prepared as described in the previous paragraph. Incorporation of i.p.-injected BrdU into DNA of replicating cells was analyzed using anti-BrdU mAbs conjugated with alkaline phosphatase as described by the manufacturer (BrdU-labeling and detection kit II; Roche). The day 7 time point with BrdU labeling was also performed with six 11-mo-old male mice per genotype. Representative data are shown in the figures.

Data analyses were performed with the use of the software package GraphPad PRISM, version 2.0. Data were summarized with the mean as a measure of central tendency and the standard error as a measure of dispersion. The differences in proportions between the wild-type and mutant mice were assessed by using the χ² and the Fisher Exact test. A p-value of 0.05 was chosen as the limit of statistical significance. Log-rank test statistical analysis was used to evaluate the survival curves.

The online supplemental material section contains the rationale used for the optimization of splice sites and RNAse protection analysis of liver samples from control and mutant mice (see Figs. S1 and S2, respectively). The experimental analysis of allele frequencies, offspring, and embryos is also described. Northern blot analysis and TUNEL and caspase-3 analysis of skin wounds are shown in Figs. S3 and S4, respectively.

We thank Dr. Pierre Chambon (Institut de Génétique et de Biologie, Strasbourg, France) for the CRE-transgenic mice; G. Faulkner for critical reading of the manuscript; and Giancarlo Lunazzi and Mauro Sturnega for animal care.