Mice with defects in HB-EGF ectodomain shedding show severe developmental abnormalities

Heparin-binding EGF-like growth factor (HB-EGF), a member of the EGF family, binds the EGF receptor (EGFR) and ErbB4 to initiate signaling (Higashiyama et al., 1991; Elenius et al., 1997). Like other EGF family members (for review see Massague and Pandiella, 1993), HB-EGF is first synthesized as a membrane-anchored form (proHB-EGF), and then the soluble form (sHB-EGF) is released from the cell surface by ectodomain shedding (Goishi et al., 1995). sHB-EGF is a diffusible factor with potent mitogenic and chemoattractant activities for a number of cell types. ProHB-EGF forms a complex with other membrane proteins at the cell–cell contact site and transduces biological signals to neighboring cells in a nondiffusible manner (for review see Iwamoto and Mekada, 2000). Thus, in addition to being a precursor of sHB-EGF, proHB-EGF is thought to be a biologically active molecule itself. Several studies in vitro have revealed that a number of signaling molecules control the ectodomain shedding of proHB-EGF (Izumi et al., 1998; Umata et al., 2001; Takenobu et al., 2003), implying that strict control of ectodomain shedding is critical for HB-EGF function.

Recently, we demonstrated that HB-EGF is critical for proper heart development and function by the analyses of HB-EGF null mice (Iwamoto et al., 2003). However, it remains unclear which forms of HB-EGF are necessary for these process. The relative roles of either proHB-EGF or sHB-EGF and the significance of the control of ectodomain shedding in vivo have yet to be determined. To address these issues, we generated two kinds of mutant mice, by targeted gene replacement, that express either an uncleavable (HB^uc) or a transmembrane domain–truncated form (HB^Δtm) of proHB-EGF. Analysis of these mutant lines indicates that proHB-EGF shedding is essential in vivo and that this process must be controlled.

An uncleavable form of proHB-EGF was generated by creating double point mutations, L148S/P149T, in the juxtamembrane domain (designated HB^uc) (Fig. 1 A). As shown previously in cases of each single point mutation (Hirata et al., 2001), HB-UC (a product of HB^uc) was also resistant to ectodomain shedding in response to various shedding-inducing stimuli, while the other biological properties of HB-UC were similar to those of WT proHB-EGF (Fig. S1, A–C, and supplemental Results, available at http://www.jcb.org/cgi/content/full/jcb.200307035/DC1). To assess the biological significance of proHB-EGF ectodomain shedding, we created mutant mice expressing HB-UC instead of WT proHB-EGF by targeted replacement of the proHB-EGF gene with HB^uc cDNA (Fig. S2, supplemental Results, and supplemental Materials and methods, available at http://www.jcb.org/cgi/content/full/jcb.200307035/DC1). Homozygous mice (HB^uc/uc) were born at the predicted Mendelian frequency. Northern blotting of the transcripts obtained from adult mice indicated that the WT and HB^uc alleles were expressed equally in heart, lung, and kidney (Fig. 1 B).

We examined whether HB-UC was resistant to ectodomain shedding in mice (Fig. 1 C). We used adult mouse skin samples obtained from HB^uc/uc mice and HB^lox/lox mice as a control. In HB^lox/lox mice, the proHB-EGF locus was replaced with WT proHB-EGF cDNA (Iwamoto et al., 2003), making HB^lox/lox mice more suitable controls than WT mice. In adult mouse skin, HB-EGF protein bands were hardly detected in samples obtained from adult HB^uc/uc and HB^lox/lox mice. All-trans retinoic acid (tRA) is known to induce HB-EGF expression, followed by epidermal hyperplasia (Xiao et al., 1999). When tRA was applied to skin on the backs of mice, proHB-EGF protein was induced and clearly detected in samples from HB^lox/lox and HB^uc/uc mice by using an antibody against the proHB-EGF ectodomain. Under these conditions, the band corresponding to sHB-EGF was also detected in HB^lox/lox mice, but not in HB^uc/uc mice. An antibody recognizing the cytoplasmic domain of proHB-EGF did not detect this sHB-EGF band (unpublished data), confirming that this band corresponds to secreted HB-EGF. When tRA-induced skin was further treated with O-tetradecanoylphorbol-13-acetate (TPA), the sHB-EGF band appeared more intensely in HB^lox/lox mice. However, the sHB-EGF band was not detected in samples from tRA + TPA–treated HB^uc/uc mice. These results indicate that ectodomain shedding of proHB-EGF is severely impaired in HB^uc/uc mice.

Recently, we showed that HB-EGF null mice (HB^del/del) have defects in cardiac chamber dilation and cardiac valve malformation (Iwamoto et al., 2003). Therefore, we analyzed the heart phenotype in HB^uc/uc mice. Autopsies of HB^uc/uc mice revealed massive enlargement of the heart. Histological analysis showed that wall thickness was reduced, accompanied by sporadic fibrosis in 12-wk-old HB^uc/uc mice (Fig. 2, A–D). These phenotypes are highly similar to those observed in HB-EGF null mice (Iwamoto et al., 2003) and ErbB2 conditional knockout mice (Crone et al., 2002; Ozcelik et al., 2002).

Transthoracic echocardiography indicated marked dilation and poor left ventricular contraction in HB^uc/uc mice (Fig. 2, E and F). Dilation could be detected in 4-wk-old mice, though the physical activity level and appearance was indistinguishable between WT and HB^uc/uc mice. Mutant mice also had reduced cardiac wall movement, and ventricular fractional shortening (FS), a representative measure of systolic function, was greatly reduced.

In addition to the ventricular chamber abnormality, heart valve malformation was also observed in HB^uc/uc mice as in HB-EGF null mice (Iwamoto et al., 2003; Jackson et al., 2003) and EGFR knockout mice (Chen et al., 2000). Histological analysis of E17.5 embryonic HB^uc/uc hearts showed enlarged semilunar (aortic and pulmonic) and atrioventricular (mitral and tricuspid) valves (Fig. 3, A–D). Scoring of cardiac valve size in E17.5 hearts revealed enlarged semilunar and atrioventricular valves in HB^uc/uc mice (Fig. 3 E). No overt abnormality was observed in HB^lox/lox heart chambers or valves (Iwamoto et al., 2003), indicating that heart abnormalities in HB^uc/uc mice were not due to a nonspecific effect of cDNA knock-in.

The similar heart defects displayed in HB^del/del and HB^uc/uc mice indicate that the process of ectodomain shedding is essential for HB-EGF function in normal cardiac valve development and cardiomyocyte function, and that soluble HB-EGF is required for these processes. We also examined whether proHB-EGF has any in vivo physiological role aside from acting as the sHB-EGF precursor. Interestingly, HB^del/del mice displayed a shorter life span than HB^uc/uc mice. Half of HB^uc/uc mice survive over 18 wk, while >60% of HB^del/del mice died in the first postnatal week. These differences between HB^del/del and HB^uc/uc mice suggest that proHB-EGF may function in unidentified developmental processes. Further comparison of the phenotypes between HB^del/del and HB^uc/uc mice may help us answer this question.

Although studies of HB^uc/uc mice indicated that ectodomain shedding of proHB-EGF and release of sHB-EGF is necessary for proper HB-EGF function in vivo, the physiological importance of the control of proHB-EGF ectodomain shedding remained unclear. To address this issue, we prepared another mouse mutant that only expresses sHB-EGF. The transmembrane domain–truncated mutant (HB^Δtm) was generated by insertion of a stop codon between Leu¹⁴⁸ and Pro¹⁴⁹ (Fig. 4 A), the major site for proHB-EGF processing (Higashiyama et al., 1992). Mitogenic activity was similar between HB-ΔTM (product of HB^Δtm) and WT sHB-EGF derived from proHB-EGF shedding, but HB-ΔTM is secreted at much higher levels than WT sHB-EGF (Fig. S1, D–G, and supplemental Results). The targeting construct for HB^Δtm is similar to that for HB^uc. One allele of the proHB-EGF gene in ES cells was replaced with HB^Δtm through homologous recombination (Fig. S2, supplemental Results, and supplemental Materials and methods). Chimeric mice carrying HB^Δtm were generated from these ES clones. Most chimeric founders exhibited abnormally small bodies and thickened skin (Fig. 4, B–D). The majority of these mice died before or during the neonatal stage (unpublished data). A few mice, however, survived and were fertile. Most of the resulting F1 heterozygotes carrying the HB^Δtm allele also died before or during the neonatal stage (unpublished data), displaying a more severe phenotype than that seen in chimeric mice (Fig. 4 E).

We confirmed the expression of the HB^Δtm cDNA in chimeric mice by RT-PCR of skin and heart specimens (Fig. 4 F). Transcripts derived from the WT proHB-EGF allele were detected in both WT and chimeric tissues, while those derived from the HB^Δtm allele were only identified in chimeric mice. Expression of HB^Δtm was also confirmed at the protein level using anti–HB-EGF antibody, specific for the extracellular domain of proHB-EGF. In heterozygote skin, significant quantities of soluble HB-EGF were detected, whereas sHB-EGF protein was only barely detectable in WT skin (Fig. 4 G). The quantities of proHB-EGF with a molecular mass of ∼19 kD, detected by an anti–HB-EGF antibody specific for the cytoplasmic domain, displayed a concomitant decrease in heterozygote skin relative to the levels seen in wild-type skin (Fig. 4 G).

The most obvious phenotype of HB^Δtm-carrying mice, distinguishing them from WT mice, was the presence of abnormal elephant-like skin (Figs. 4, B–E). HB^Δtm mice have short lives; thus to further study the epidermal hyperplasia, back skin was transplanted from P3 WT or HB^Δtm chimeric mice onto the backs of Balb/c nu/nu mice for 2 wk. This transplantation method also allowed for the specific examination of the effects of HB^Δtm in the skin in an animal with otherwise normal expression and processing of HB-EGF. Histological examination revealed severe epidermal hyperplasia, accompanied by the presence of large, disorganized, hair follicle–like structures (Fig. 5 A). Immunohistochemistry of epidermal marker proteins revealed perturbed differentiation and proliferation of keratinocytes (Fig. 5, B–F). Keratin 5, normally expressed by both mitotically active, basal layer keratinocytes and hair follicles, was detected in the suprabasal epidermis of chimeric mice (Fig. 5 D). Keratin 1, normally expressed by differentiating keratinocytes in the suprabasal epidermis, but not by the cells of the hair follicle (Heid et al., 1988), was expressed in the hair follicle–like structures of chimeric mice (Fig. 5 B). Additionally, keratin 10, a differentiating keratinocyte marker, was down-regulated in chimeric mice (Fig. 5 C). Keratin 6, expressed predominantly in hair follicle cells, and at lower levels in the suprabasal epidermis, was detected throughout the epidermis of chimeric mice (Fig. 5 E), consistent with the induction of keratin 6 expression by epidermal hyperproliferation (Werner et al., 1993). The expression of Ki-67, a nuclear mitotic marker, was also increased in the basal layer of chimeric skin compared with levels seen in WT (Fig. 5 F), indicating that HB^Δtm expression accelerates basal layer keratinocyte proliferation. No overt abnormality was observed in the skin area surrounding the transplanted skin, suggesting that the abnormality in the transplanted skin is due to the action of HB-ΔTM in a manner of autocrine or paracrine in limited distance, rather than paracrine in long distance.

Morphological abnormalities were also found in the heart of HB^Δtm mice. Histological specimens showed ventricular hypertrophy in hearts from HB^Δtm E16.5 embryos (Fig. 5 G) and P0 mice (Fig. 5 I). HB-EGF is expressed in cardiomyocytes (Iwamoto et al., 2003), consistent with the present results. The cardiac muscle fibers in chimeric mice were loose, with enlarged nuclei (Fig. 5, H and J). These results suggest that expression of HB^Δtm reduces cardiomyocyte differentiation or terminates proliferation. These cardiac developmental abnormalities may be the primary cause of early death in HB^Δtm mice.

The phenotype observed in HB^Δtm mice is likely due to dysregulated secretion of sHB-EGF. In normal conditions, a portion of proHB-EGF molecules are converted to sHB-EGF, but the majority of proHB-EGF molecules on the cell surface seems to be internalized without shedding (Goishi et al., 1995). In the case of HB^Δtm, most synthesized molecules would be secreted without shedding, resulting in oversecretion of sHB-EGF even though the native HB-EGF promoter regulates gene expression. Therefore, dysregulated secretion of sHB-EGF would result in severe developmental abnormalities. The present study thus confirms the notion that ectodomain shedding of proHB-EGF must be strictly controlled in vivo.

One question regarding the truncated HB^Δtm mutant concerns whether the observed hyperplasia might be due to an intracrine rather than a paracrine mechanism. A transmembrane domain–truncated EGF mutant was reported to activate EGFR in an intracrine manner, as a result of interaction with EGFR within cytoplasmic vesicles, before the molecules reached the cell surface (Wiley et al., 1998). However, this was not the case for HB-EGF. We have performed ex vivo transfection of HB^Δtm cDNA into mouse embryonic skin (Fig. S3 and supplemental Results, available at http://www.jcb.org/cgi/content/full/jcb.200307035/DC1). Transfection of HB^Δtm, but not WT proHB-EGF, resulted in embryonic epidermal hyperplasia. CRM197, a protein that specifically inhibits the mitogenic activity of HB-EGF (Mitamura et al., 1995), inhibited HB-ΔTM–induced hyperplasia. As CRM197 is membrane impermeable, hyperplasia induced by transfection with HB^Δtm must be mediated by secreted HB-ΔTM in a paracrine manner.

In conclusion, proHB-EGF shedding and the strict control of this process are essential for the function of this growth factor. Not only HB-EGF, but also other EGF family growth factors and cytokines are synthesized as membrane-anchored forms. The present study suggests that the strict control of ectodomain shedding is essential for the physiological function of these membrane-anchored ligands. This study also indicates that posttranslational regulation, in addition to transcriptional control, is crucial for the function of membrane-anchored growth factors.

Total RNA was isolated from tissues using ISOGEN (Nippon Gene), according to the manufacturer's instructions. Details of the hybridization procedure are shown in the supplemental Materials and methods (available at http://www.jcb.org/cgi/content/full/jcb.200307035/DC1).

For the detection of HB-EGF and HB-UC in adult skin, back skin from 11-wk-old HB^lox/lox and HB^uc/uc mice was treated with 800 nmol of tRA for 3 d. 2 h before specimen isolation, the back skin was additionally treated with 8 nmol of TPA. The isolated full back skin was homogenized in lysis buffer (Iwamoto et al., 2003). For detection of HB-ΔTM in neonatal skin, skin from P1 HB^Δtm heterozygous mice or WT littermates was homogenized in lysis buffer. Details of the immunoblotting procedure using these lysate samples are shown in the supplemental Materials and methods (available at http://www.jcb.org/cgi/content/full/jcb.200307035/DC1).

Mouse hearts were fixed by perfusion with 4% paraformaldehyde, dehydrated, and embedded in paraffin. 4-μm sections were stained with either hematoxylin-eosin or Azan-Mallory. For immunohistochemical analysis of transplanted skin, specimens of full-thickness skin (2–3 cm²) were transplanted onto the backs of Balb/c nu/nu mice (8 wk old) and then fixed with adhesive bandages for 1 wk. 2 wk after transplantation, skin specimens were isolated from the grafted skin and subjected to immunohistochemical analysis. Information of the used antibodies for immunohistochemistry, microscopy, and image processing is shown in the supplemental Materials and methods (available at http://www.jcb.org/cgi/content/full/jcb.200307035/DC1).

Transthoracic echocardiograph was performed with a cardiac ultrasound recorder (SONOS 5500; Hewlett-Packard) with a 15-MHz transducer, as described previously (Iwamoto et al., 2003).

RNA was isolated from tissues of P14 mice using TRIzol reagent (Invitrogen). Reverse transcription was performed using a reverse transcriptase, ReverTra Dash (TOYOBO). Primer sets used in PCR analyses are shown in the supplemental Materials and methods (available at http://www.jcb.org/cgi/content/full/jcb.200307035/DC1).

The supplemental material is available at http://www.jcb.org/cgi/content/full/jcb.200307035/DC1. Fig. S1 shows the characterizations of HB-UC and HB-ΔTM. Fig. S2 shows the targeting construct of HB^uc and HB^Δtm and genotypic analyses. Fig. S3 shows ex vivo transfection of HB^Δtm cDNA into mouse embryonic skin. Supplemental Results, Materials and methods, and References are also presented.

We thank I. Ishimatsu, M. Hamaoka, and T. Yoneda for technical assistance.

This work was supported by the Research for the Future Program of the Japan Society for the Promotion of Science (97L00303 for E. Mekada) and by Grants-in-Aid from the Ministry of Education, Culture, Sports, Science, and Technology (12215152 and 14032202 for E. Mekada and 12680705 for R. Iwamoto).