Accumulation of type IV collagen in dilated ER leads to apoptosis in Hsp47-knockout mouse embryos via induction of CHOP

Collagen is the most abundant protein in mammals. It plays an essential role in the function of the extracellular matrix (ECM) and the lack of its expression or function has serious consequences (Kadler, 1995; Kadler et al., 1996). Twenty-seven types of collagen have been identified so far. Of these, the disruption of collagen types I, II and III is known to generate serious defects (Schnieke et al., 1983; Lohler et al., 1984; Li et al., 1995; Liu et al., 1997). The 47-kDa heat-shock protein (Hsp47) is a collagen-specific molecular chaperone (Nagata and Hosokawa, 1996; Nagata, 2003) that recognizes the Pro-Arg-Gly sequence in the Gly-X-Y repeats that constitute the triple helices of collagen (Koide et al., 2000; Koide et al., 2002). Hsp47 shows weak but significant amino acid sequence similarity with the serine proteinase inhibitor (serpin) family of proteins (Hirayoshi et al., 1991). Hsp47 is the only heat-inducible protein in the endoplasmic reticulum (ER) in mammalian cells, and it transiently binds to pro-collagen in the ER (Tasab et al., 2000; Saga et al., 1987). In vitro binding analyses have shown that Hsp47 can bind collagen types I-V (Nagata and Yamada, 1996; Natsume, 1994). In addition, yeast two-hybrid screening assays revealed that Hsp47 can bind to other collagenous proteins through their triple helical regions (Sato et al., 2002).

In the initial steps of collagen maturation, three alpha chains of newly synthesized pro-collagen assemble at the C-terminus and they fold into a triple helix from the C- to N-termini in a zipper-like fashion (Engel and Prockop, 1991; McLaughlin and Bulleid, 1998). After binding to pro-collagen in the ER, Hsp47 is co-transported with pro-collagen from the ER to the cis-Golgi or ER-Golgi intermediate compartment, where it dissociates from pro-collagen in a pH-dependent manner (Saga et al., 1987; Satoh et al., 1996). Hsp47 is then thought to return to the ER through the effect of the Arg-Asp-Glu-Leu (RDEL) sequence at its C-terminal end. This sequence is similar to Lys-Asp-Glu-Leu (KDEL), which is the ER retention signal (Sauk et al., 1998). Following or during secretion into the extracellular matrix, the N- and C-propeptides of the procollagens are cleaved off, which generates mature collagens in the extracellular matrix (Prockop et al., 1998; Cal et al., 2001). Consistent with the role of Hsp47 in collagen maturation, the expression level of Hsp47 is closely correlated with that of collagen in normal tissues (Levio et al., 1980; Masuda et al., 1994; Masuda et al., 1998; Moriyama et al., 1998). It is also known that Hsp47 is upregulated in the fibrotic legions of liver cirrhosis, glomerulosclerosis and keloids (Masuda et al., 1994; Moriyama et al., 1998; Naitoh et al., 2001), and that inhibition of Hsp47 expression by administering antisense Hsp47 RNA can suppress the accumulation of collagen in glomerulonephritis (Sunamoto et al., 1998).

We previously produced Hsp47-knockout (Hsp47^–/–) mice and found by silver staining that their basement membrane (BM) at 9.5 days postcoitus (dpc) exhibit no or very weak staining (Nagai et al., 2000). Moreover, electron microscopy revealed abnormal discontinuous morphology in these BMs (Nagai et al., 2000). Various abnormalities that are probably due to the lack of BMs were also observed in tissues and organs of the knockout embryos at 10.5 dpc, including the rupture of endothelial cell layers of blood vessels. Consequently, Hsp47^–/– embryos cannot survive beyond 11.5 dpc. The Hsp47^–/– fibroblastic cell lines derived from the knockout embryos were found to produce incorrectly folded type I collagen. These observations indicate that Hsp47 is essential for the production of type I collagen, and that the abnormal BM in the Hsp47^–/– embryos may be caused by the defective molecular maturation of type IV collagen, as this is a major component of the BM. However, there is no direct evidence of the relationship between the abnormal BM in Hsp47^–/– embryos and the function of Hsp47 in type IV collagen maturation.

Here, we examined the distribution of type IV collagen and other components that localize in the BM, including laminin and nidogen, in Hsp47^–/– embryos. These analyses revealed that unlike the other BM components, type IV collagen does not localize in the BM of the knockout embryos: it accumulates in the ER. We also report here that immediately prior to the death of the Hsp47^–/– embryos, they undergo apoptosis that occurs concomitantly with the upregulation of CHOP. We discuss the role of Hsp47 in type IV collagen maturation in vivo, and the role of type IV collagen in the maintenance of BM structures during development. We also discuss the biological significance of the induction of CHOP during the apoptosis of Hsp47^–/– embryos.

The following antibodies were used: mouse monoclonal antibodies specific for Hsp47 (Stressgen Biotechnologies, Victoria, Canada) and entactin/nidogen-1 (Neomarkers, CA, USA) and rabbit polyclonal antibodies specific for mouse collagen type IV (Chemicon, Temecula, CA, USA) and laminin (LSL, Tokyo, Japan). The anti-laminin antibody recognizing alpha 1, beta 1 and gamma 1 chains was affinity-purified by using mouse engelbreth-holm-swarm sarcoma laminin (Sigma, St. Louis, MO, USA) that was immobilized on CNBr-Sepharose (Amersham Biosciences, Buckinghamshire, England). Goat anti-rabbit Ig (Biosource International, Camarillo, CA, USA), anti-mouse Ig (Biosource International) and anti-rat IgG (Cappel, Darham, UK) conjugated with FITC or rhodamine were used as secondary antibodies.

Mutant embryos and littermate controls at 9.5 and 10.5 dpc were fixed in 0.1 M phosphate buffer, pH 7.4 (PB) containing 4% paraformaldehyde at room temperature for 2 hours. The fixed embryos were then embedded in Tissue-Tek OCT compound (Sakura Finetechnical, Tokyo, Japan), snap frozen, and stored at –80°C. 4 μm cryosections were prepared using a freezing microtome, mounted on poly-L-lysine-coated cover glass and dried. After nonspecific binding was blocked with 1.5% horse normal serum in PBS for 30 minutes, the sections were incubated with the appropriate primary antibodies in PBS containing 1.5% normal horse serum for 1 hour. Following incubation with secondary antibodies conjugated with FITC or rhodamine, the localization of the antibodies was visualized by using a fluorescent microscope.

The pre-embedding silver enhancement immunogold method was performed for immunoelectron microscopy as described (Mandai et al., 1997), with a slight modification. Briefly, mouse embryos were fixed in 4% paraformaldehyde in PB for 1 hour. 6 μm cryosections were prepared and incubated in blocking solution (PB containing 0.005% saponin, 10% bovine serum albumin, 10% normal goat serum and 0.1% coldwater fish skin gelatin) for 30 minutes. The sections were then treated with rabbit antibodies against type IV collagen or laminin in blocking solution overnight and then washed five times in PB containing 0.005% saponin for 10 minutes. After incubation with goat anti-rabbit IgG conjugated with colloidal gold (1.4 nm diameter) in the blocking solution for 2 hours, the sections were washed five times with PB for 10 minutes and fixed with 1% glutaraldehyde in PB for 10 minutes. Following several washes, the gold labeling was intensified by using a silver enhancement kit for 6 minutes at 20°C in the dark. After washing in distilled water, the sections were post-fixed in 0.5% OsO₄ for 90 minutes at 4°C, washed in distilled water, incubated with 50% ethanol for 10 minutes and stained with 2% uranyl acetate in 70% ethanol for 2 hours. The sections were further dehydrated with a graded series of ethanol and embedded in epoxy resin. Ultra-thin sections were prepared and doubly stained with uranyl acetate and lead citrate.

The in situ visualization of DNA fragmentation was carried out with embryo sections by using In Situ Cell Death Detection Kit (Roche Diagnostics, Basel, Switzerland) according to the manufacturer's instructions.

The embryos were lysed in lysis buffer (150 mM NaCl, 10 mM Tris-HCl pH 8.0, 10 mM EDTA and 0.1% SDS) supplemented with 0.5mg/ml proteinase K and incubated at 55°C overnight. Following centrifugation, the supernatant was recovered and DNA was isolated by phenol:chloroform extraction followed by ethanol precipitation. The precipitated DNA was dissolved in TE and DNA fragmentation was analysed by 1.8% agarose gel electrophoresis followed by ethidium bromide staining.

Embryos were lysed by shearing with syringes and their total RNA was isolated by acid phenol extraction using TRIzol (Invitrogen, Carlsbad, CA, USA). Equal amounts of total RNA (1 μg) were converted to first-strand cDNA by using reverse transcriptase and specific cDNA was amplified by PCR with Taq DNA polymerase and the following oligonucleotide primers: Hsp47, 5′-AAGACCAGGCGGTGGAGAACATCC-3′ and 5′-TCTCGCATCTTGTCTCCCTTGGGC-3′; CHOP, 5′-TCATACACCACCACACCTGAAAGC-3′ and 5′-AATGTACCGTCTATGTGDAAGCCG-3′; β-actin, 5′-GTTCGCCATGGATGACGATATCGC-3′ and 5′-AGGTGGACAGTGAGGCCAGGATGG-3′.

PCR consisting of 24 cycles for β-actin, 30 cycles for CHOP and 37 cycles for Hsp47 was then performed. Each cycle consisted of 95°C for 30 seconds, 62°C for 30 seconds and 72°C for 1.5 minutes. The PCR products were analysed by 1.8% agarose gel electrophoresis followed by ethidium bromide staining.

Quantitative RT-PCR was carried out as follows. Total RNAs (1 μg) were converted to first-strand cDNA by First-strand cDNA Synthesis System for Quantitative RT-PCR (Marligen Biosciences, Nunc, Denmark) and amplified by PCR reaction in SYBER Green PCR Master Mix (Applied Biosystems, Foster City, CA) using the specific primer sets of CHOP and β-actin. Fluorescence from DNA-SYBER Green complex was recorded by ABI PRISM 7700 Sequence Detection System (Applied Biosystems) throughout the PCR reaction, and the level of CHOP mRNA relative to β-actin mRNA was determined according to the manufacturer's instructions.

Type IV collagen and laminin are the major components of the BM and are linked together with nidogen (Mayer et al., 1998). Previously, we reported that the BM of Hsp47^–/– embryos exhibits a discontinuously disrupted abnormal morphology (Nagai et al., 2000). To investigate the molecular reason for these morphological abnormalities, we first examined sections of Hsp47-knockout embryos and littermate wild-type controls at 9.5 dpc by immunohistochemistry (Fig. 1). Hsp47 was detected in the epithelial cells and fibroblasts of normal embryos (Fig. 1C), which is consistent with the known expression patterns of collagen types IV and I in those cells, respectively. In contrast, expression of Hsp47 was undetectable in the Hsp47^–/– embryos (Fig. 1D), which confirms the complete disruption of Hsp47 in these mice. Type IV collagen, a putative substrate of the Hsp47 chaperone, was not detectable in the BM regions of any tissues of the Hsp47^–/– embryos (Fig. 1F), whereas the same antibody clearly stained the BM of wild-type embryos (Fig. 1E).

Unlike type IV collagen, laminin, another major component of the BM, was present in the BM of Hsp47^–/– embryos at 9.5 dpc (Fig. 2B) indicating that the localization of laminin is not affected by the disruption of the Hsp47 gene. Nidogen-1, a supporting component of the BM, was also shown to localize normally in Hsp47^–/– embryos (Fig. 2C,D). The distribution of fibronectin, a major component of the ECM, was also not affected by the absence of the Hsp47 gene (Fig. 2E,F), which further confirms that the effect of Hsp47 disruption is collagen specific. Thus, Hsp47 is required for the accumulation of type IV collagen in the BM, and the localization of other BM components, including laminin and nidogen-1, is not affected by the absence of type IV collagen.

As the Hsp47-deficient embryos die just after 10.5 dpc (Nagai et al., 2000), we next examined the distribution of the BM components in 10.5 dpc Hsp47^–/– embryos. In the wild-type embryos, the epithelial cells and fibroblasts of the neural tube epithelium and their supporting mesenchymal tissues, respectively, were clearly stained with the anti-Hsp47 antibody (Fig. 3C). In contrast, no staining was observed in Hsp47^–/– embryos (Fig. 3D). As with the 9.5 dpc embryos, the BM in 10.5 dpc Hsp47^–/– embryos was not stained with the anti-collagen IV antibody (Fig. 3F), which is in contrast to the strong staining of the BM in the wild-type embryos (Fig. 3E). Intriguingly, the anti-laminin antibody staining patterns of the BM in the 10.5 dpc Hsp47^–/– embryos now exhibited fragmented patterns (Fig. 4B), whereas normal continuous staining of the BM was observed in the 10.5 dpc wild-type embryos (Fig. 4A). Similar staining patterns were obtained with the antibody against nidogen-1 (Fig. 4C,D). These staining patterns were in marked contrast to those of the 9.5 dpc embryos (compare Figs 2 and 4). The distribution of fibronectin was again not affected by the absence of Hsp47 in 10.5 dpc embryos (Fig. 4E,F). Thus, the continuous sheet-like structure of the BM cannot be maintained in an intact form in the absence of type IV collagen at 10.5 dpc, although the initial formation of the BM itself does not always require the presence of type IV collagen.

Electron microscopy with immunogold labeling confirmed that type IV collagen is not present in the BM of 9.5 dpc Hsp47^–/– embryos (Fig. 5B), unlike the BM of wild-type embryos (Fig. 5A). Immunoelectron microscopy also confirmed that laminin is present in the BM of Hsp47^–/– and wild-type embryos at 9.5 dpc (Fig. 5C,D). Thus, no type IV collagen is accumulated in the BM in the absence of Hsp47, despite the fact that laminin is present.

Hsp47^–/– embryos at 9.5 dpc have been reported to synthesize type IV collagen at the same level as 10.5 dpc wild-type embryos (Nagai et al., 2000). Thus, the absence of type IV collagen in the BM of Hsp47^–/– embryos is due either to the accumulation of the type IV collagen within the collagen-producing cell through the lack of its secretion, or it is degraded during or after its secretion. To test these possibilities, we analysed the epithelial cells that lie adjacent to the BM by immunoelectron microscopy. Most of the ERs of the epithelial cells of Hsp47^–/– embryos were found to be dilated compared to those in wild-type embryos, and intriguingly, a significant amount of type IV collagen was localized in these dilated ERs (Fig. 6A,B). In contrast, type IV collagen was rarely detected in the ERs of wild-type embryos. Thus, in Hsp47^–/– cells, type IV collagen is not secreted into the ECM, but it accumulates in the ER.

The disruption of Hsp47 severely retards the growth of the embryo as determined by measuring the head-to-tail length (Fig. 7) and the number of somites (Nagai et al., 2000) and Hsp47^–/– embryos never survive beyond 11.5 dpc (Nagai et al., 2000). To investigate the mechanisms underlying the embryonic lethality of Hsp47 disruption, we analysed the DNA fragmentation, which served as an indicator of apoptotic cell death, in the wild-type and mutant embryos. DNA extracted from Hsp47^–/–, Hsp47^+/– and Hsp47^+/+ embryos from the same litter were analysed by agarose gel electrophoresis (Fig. 8). Although no DNA fragmentation was observed in the embryos of any genotype at 9.5 dpc, the DNA ladders that are typical of apoptosis were detected in Hsp47^–/– embryos at 10.5 dpc. In contrast, these ladders were not observed in heterozygotic or wild-type embryos. TUNEL staining also revealed that DNA fragmentation only occurred in 10.5 dpc Hsp47^–/– embryos (Fig. 9). Thus, apoptotic cell death occurred in Hsp47^–/– embryos prior to their death.

The accumulation of unfolded proteins in the ER is known to cause ER stress that results in the cells activating the so-called unfolded protein response (UPR) pathway. Moreover, it is known that the ER can become dilated when secretion-incompetent materials accumulate within it (Umebayashi et al., 2001). In addition, strong ER stress induces the expression of CHOP (Barone et al., 1994; Fleming et al., 1997; Zinszner, 1998; Jousse et al., 1999), which triggers a pathway that leads to apoptosis. As the ER of Hsp47^–/– mouse embryos exhibited a dilated morphology and an accumulation of type IV collagen, we examined the levels of CHOP mRNA in wild-type and Hsp47^–/– mouse embryos by RT-PCR (Fig. 10). At 10.5-11.0 dpc, the Hsp47^–/– embryos showed significantly higher CHOP mRNA levels compared to wild-type embryos, although at 9.5 dpc the CHOP mRNAs of the two genotypes did not significantly differ. The strong upregulation of CHOP mRNA in the Hsp47^–/– embryos at 10.5 dpc, but not at 9.5 dpc, was confirmed by quantitative real-time PCR (Fig. 11). The elevated level of CHOP mRNA is probably responsible for the high level of apoptosis observed in the Hsp47^–/– mouse embryos.

We previously demonstrated that Hsp47 plays an essential role in the production of mature type I collagen by mouse embryos, and that the type I collagen secreted by cultured embryonic fibroblasts is not correctly folded into the triple helices in the absence of Hsp47 (Nagai et al., 2000). In the present study, type IV collagen was also shown to require Hsp47 for its maturation because type IV collagen was not detected in the BM of Hsp47^–/– embryos (Figs 1, 3 and 5). Considered together with the fact that Hsp47 recognizes type lV collagen in vitro (Natsume et al., 1994), these observations clearly demonstrate the indispensable role of Hsp47 as a molecular chaperone of not just collagen type I but also type IV. This view is consistent with the previous observation that Hsp47 preferably binds the Pro-Arg-Gly sequence in Pro-Pro-Gly repeats in vitro (Koide et al., 2002) because type IV collagen has these sequences in abundance.

BMs play important roles in the morphogenesis of tissues during embryonic development. BMs separate epithelial (or endothelial) cells from mesenchymal tissues and maintain the function and morphology of epithelial cells. In the BM, laminin and type IV collagen form meshes that are linked by nidogen (Carlin et al., 1981; Timpl et al., 1983). We found here that laminin and nidogen-1 accumulate normally in the BM of 9.5 dpc Hsp47^–/– embryos, despite the fact that the normal accumulation of type IV collagen in the BM does not occur (Figs 2 and 5). Thus, the accumulation of laminin and nidogen-1 in the BM does not seem to require type IV collagen. This is consistent with the previous observation that type IV collagen-free BM exists in Caenorhabditis elegans (Gupta et al., 1997). However, laminin is necessary for BM formation, as LAMC1-knockout mouse embryos cannot produce BM structures and die earlier than 5.5 dpc (Smyth et al., 1999).

More importantly, we demonstrated in the present study that the BM becomes fractured at 10.5 dpc in the absence of type IV collagen (Fig. 4). These observations indicate that type IV collagen is essential for the maintenance of the continuous sheet-like structures of the BM. Type IV collagen therefore seems to be important for providing the BM with the strength that is needed to cope with the growing tensile stresses imposed on the BM during the rapid morphological changes that take place during development.

Immunoelectron microscopic analysis of epithelial cells also revealed that type IV collagen accumulates within dilated ERs in the absence of Hsp47 (Fig. 6). Type I collagen secreted from Hsp47^–/– cultured fibroblasts has been reported to be sensitive to protease digestion (Nagai et al., 2000), which indicates that the triple helix structure of the secreted type I collagen molecules is incompletely formed. Recently, we observed that type IV collagen secreted from Hsp47^–/– embryonic stem cells also consists of an incomplete triple helix that is sensitive to protease digestion (Matsuoka et al., 2004). Thus, the type IV collagen that accumulates in the ER in the Hsp47-knockout embryos is probably misfolded and/or aggregated. Even if some of these proteins were secreted from the cells, the incompletely folded type IV collagen may be degraded by proteinases in the ECM. The accumulation of unfolded or misfolded proteins in the ER is known to activate the ER stress response that induces the expression of molecular chaperones and folding enzymes in the ER (Kaufman, 1999; Mori, 2000; Urano et al., 2000). ER stress also induces apoptosis that serves to remove severely damaged cells (Zinszner, 1998; Nakagawa et al., 2000; Iwawaki et al., 2001). In the present study, laddered DNA fragmentation typical of apoptotic cell death was observed in Hsp47^–/– embryos at 10.5 dpc but not at 9.5 dpc (Fig. 8). This was also detected by TUNEL staining (Fig. 9). These observations indicate that apoptotic cell death occurs in Hsp47^–/– embryos at 10.5 dpc just prior to the death of the embryos.

We also demonstrated that the mRNA levels of CHOP are elevated in Hsp47^–/– embryos at 10.5-11.0 dpc (Figs 10 and 11). CHOP is known to be a transcription factor that is strongly upregulated by ER stress (Barone et al., 1994; Fleming et al., 1997; Jousse et al., 1999) and that plays a crucial role in ER stress-induced apoptosis (Zinszner, 1998; Nakagawa et al., 2000; Iwawaki et al., 2001; Oyadomari, 2001). Thus, our present observations suggest that the accumulation of unfolded collagen in the ER induces apoptosis through the stimulation of CHOP expression. This is similar to the observation of the pancreatic β cells of mice that bear a C96Y mutation in their insulin gene: these cells die by apoptosis through the induction of CHOP (Oyadomari et al., 2002) after the accumulation of misfolded insulin in the ER stimulates the UPR pathway (Nozaki et al., 2004). However, the cell types that are affected in the insulin mutant are much more restricted than those in the Hsp47^–/– embryos. It is worthwhile to note that we cannot exclude the possibility that CHOP might be induced by hypoxia or by impairment of nutrient exchange from the placenta, because of the disruption of the vasculature in the Hsp47-disrupted mouse (Nagai et al., 2000) or the disruption of the BM in the placenta, respectively. Further studies could be carried out by crossing knockout Hsp47 and knockout CHOP mice (Oyadomari et al., 2001) to determine the role of CHOP in the death of Hsp47^–/– embryos.

While this manuscript was in preparation, it was reported that the knockout mouse of type IV collagen α1 and α2 chains exhibits lethality at 10.5-11.5 dpc (Poschl et al., 2004). These authors also reported that although BM-like structures were present in the Col4α1/2-knockout embryos, their ultrastructure was abnormal. The authors proposed that type IV collagen is important for the maintenance of the integrity of the BM but is dispensable for the formation of the BM. These observations are entirely consistent with our observations of the Hsp47^–/– embryo, and strongly support the idea that the phenotype of the Hsp47^–/– embryos is primarily due to the absence of functional type IV collagen. It is noteworthy that the Hsp47^–/– embryos die slightly earlier than the Col4α1/2-knockout embryos, as this suggests that the defects in BM structure and related functions are more severe in the Hsp47^–/– embryos. This difference may be due to absence of other collagens in the BM and/or supporting connective tissues in the Hsp47^–/– embryos, as Hsp47 recognizes many different types of collagen in vitro (Natsume et al., 1994). It is also conceivable that in addition to the defects in BM structures, the accumulation of misfolded type IV collagen in the ER may induce the earlier apoptosis-driven death of the Hsp47^–/– embryos, because the ER stress due to the accumulation of misfolded collagen would not be expected in the case of the Col4α1/2-knockout mice. Thus, the observations of Poschl and colleagues (Poschl et al., 2004) are consistent with our conclusion that Hsp47 plays an essential role in type IV collagen maturation and that type IV collagen is important for the maintenance of BM structures.

We thank T. Matsushita for helpful comments on our immunohistochemical experiments and M. Kadota for assistance with animal care.