Abnormal Expression of Collagen IV in Lens Activates Unfolded Protein Response Resulting in Cataract*

Human diseases caused by mutations in extracellular matrix genes are often associated with an increased risk of cataract and lens capsular rupture. However, the underlying mechanisms of cataract pathogenesis in these conditions are still unknown. Using two different mouse models, we show that the accumulation of collagen chains in the secretory pathway activates the stress signaling pathway termed unfolded protein response (UPR). Transgenic mice expressing ectopic Col4a3 and Col4a4 genes in the lens exhibited activation of IRE1, ATF6, and PERK associated with expansion of the endoplasmic reticulum and attenuation of general protein translation. The expression of the transgenes had adverse effects on lens fiber cell differentiation and eventually induced cell death in a group of transgenic fiber cells. In Col4a1+/Δex40 mutant mice, the accumulation of mutant chains also caused low levels of UPR activation. However, cell death was not induced in mutant lenses, suggesting that low levels of UPR activation are not proapoptotic. Collectively, the results provide in vivo evidence for a role of UPR in cataract formation in response to accumulation of terminally unfolded proteins in the endoplasmic reticulum.

mutations in extracellular matrix (ECM) 4 genes are also often associated with an increased risk of cataract. Stickler and Marshall syndromes are two disorders caused by mutations in the COL2A1 gene that are associated with the early onset of distinctive cataracts (10,11). Alport syndrome, caused by mutations in either the COL4A3, COL4A4, or COL4A5 genes, is also associated with lens capsule abnormalities and cataract formation (12)(13)(14). Humans carrying mutations in the COL4A1 locus often exhibit lens abnormalities and cataracts along with porencephaly and sporadic intracerebral hemorrhage (15)(16)(17)(18)(19). To date, approximately 13 independent mutations in the mouse Col4a1 locus and three independent mutations in mouse Col4a2 locus have been found to cause vacuolar cataract and lens abnormalities in mice (19 -21). However, the underlying mechanisms of cataract pathogenesis resulting from these collagen mutations are still unknown.
In other tissues, mutations in genes encoding secretory pathway proteins have been found to cause endoplasmic reticulum (ER) stress and subsequent activation of the unfolded protein response (UPR), a set of evolutionarily conserved signaling pathways activated upon ER stress (22)(23)(24)(25)(26)(27)(28). UPR has been implicated in the pathogenesis of many conformational diseases, such as Alzheimer disease, Parkinson disease, and diabetes, and is being investigated in many others (29 -33). UPR pathways are activated following accumulation of unfolded proteins in the ER lumen and attempt to relieve the stress by 1) up-regulating the ER folding capacity through increasing the levels of ER-resident molecular chaperones and expansion of the ER, 2) reducing the demand on the ER through attenuation of protein synthesis, and 3) increasing the clearance of unfolded proteins from the ER through up-regulation of ER-associated degradation (24,34,35). However, if these mechanisms cannot relieve the stress, the UPR pathway activates apoptosis (36 -38). Mammalian UPR is mediated by three ER-resident transmembrane proteins, IRE1, PERK (PKR-like ER kinase), and ATF6, whose combined activation alters transcriptional and translational programs and induces profound changes in cellular processes, such as cell growth, differentiation, and survival (38 -40). Notably, UPR pathways are induced in the lens in response to oxidative stress; however, the relative contributions of UPR and oxidative damage to lens pathogenesis are difficult to ascertain (41,42).
Here we test the hypothesis that the presence of unfolded proteins within the lens secretory pathway results in UPR activation, disrupts lens differentiation and/or lens cell survival, and contributes to cataract pathogenesis. We first investigated whether the chronic production of unassembled collagen IV ␣ chains in the lens results in ER stress, the activation of UPR pathways, and the cellular changes leading to cataract formation. We then determined that UPR pathways were also activated in the lenses from Col4a1 mutant mice, suggesting that UPR induction may be generally important in the pathogenesis of cataracts associated with the mutation of ECM genes.

EXPERIMENTAL PROCEDURES
Generation of Transgenic and Mutant Mice-All experiments using transgenic animals were approved by the University of Delaware Institutional Animal Care and Use Committee. The ␦en␣A plasmid containing the hybrid ␦en/␣A-crystallin promoter, the rabbit ␤-globin intron, and the human growth hormone poly(A) signal was a gift from Lixing Reneker (University of Missouri, Columbia, MO) (43). The mouse cDNAs for Col4a3 and Col4a4 were created as described previously (44). After the cDNA for either Col4a3 or Col4a4 was placed between the rabbit ␤-globin intron and human growth hormone poly(A) signal in ␦en␣A, the fragment containing the promoter, cDNA, intron, and poly(A) signal was liberated from the backbone plasmid by restriction enzyme digestion and gel-purified. This fragment was microinjected into the pronuclei of FVB/N fertilized eggs by the University of Delaware transgenic mouse facility as described previously (45). The genotyping of the transgenic mice is described in the supplemental material.
Col4a1 ϩ/⌬ex40 mice were bred at the University of California, San Francisco, animal facility and genotyped as previously described (19). All experiments using mutant animals were approved by the University of California Institutional Animal Care and Use Committee.
Embryos used in the study are staged by designating the day that the vaginal plug was observed in the dam as E0.5. Postnatal mice were staged by designating the day of birth as P0.
Morphological Analysis-For gross documentation of alterations in lens structure, lenses were isolated under a dissecting microscope and photographed under dark field optics. For histological analysis, eyes (postnatal mice) or heads (embryos) were isolated and immediately fixed in Pen-Fix (Richard Allan Scientific, Kalamazoo, MI) for 2 h prior to paraffin embedding. Serial 6-m sections were cut and mounted on slides. Selected slides were then stained by hematoxylin and eosin by standard methods and visualized by a light microscope.
Scanning Electron Microscopy-Scanning electron microscopy was performed as described by Duncan et al. with some minor modifications (46). The protocol is described in detail in the supplemental material.
Immunolocalization-All antibodies used in this study are described in the supplemental material.
Immunohistochemistry-Paraffin sections, prepared as above for hematoxylin and eosin, were deparaffinized, incu-bated with rabbit anti-bovine ␤-crystallin or rabbit anti-bovine ␥-crystallin antibodies (gifts of Samuel Zigler, NEI, National Institutes of Health, Bethesda, MD). Antibody signal was then detected with the horseradish peroxidase Dako Envision kit (Dako Laboratories, Carpinteria, CA).
Immunofluorescence-All immunofluorescence experiments were performed as described previously (47). A detailed protocol is described in the supplemental material.
Western Blotting-Lenses were isolated and immediately homogenized with 0.1 ml of ice-cold lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with Halt protease and phosphatase inhibitor mixture (Thermo Scientific, Rockford, IL). The insoluble material was removed by centrifugation at 12,000 ϫ g for 30 min. Final protein concentrations were determined using a Bio-Rad protein assay kit according to the manufacturer's specifications. Forty micrograms of total protein was resolved by SDS-polyacrylamide gel and transferred onto supported nitrocellulose membranes (Bio-Rad). The protein blots were blocked with SuperBlock T20 blocking buffer (Thermo Scientific) overnight at 4°C and incubated with the primary antibody in the same blocking buffer for 2 h. After incubation with secondary antibodies conjugated with horseradish peroxidase (Calbiochem) for 1 h at room temperature, the signals were detected using an ECL detection kit (Amersham Biosciences). The intensity of the signal for collagen IV chains were quantified using Photoshop histogram analysis.
RNA Preparation and RT-PCR-Total RNA from lenses was extracted using the SV total RNA isolation kit (Promega). For RT-PCR, 20 ng of total RNA was used for each reaction. The primers used are Xbp1 Fwd (5Ј-GAA CCA GGA GTT AAG AAC ACG-3Ј) and Xbp1 Rev (5Ј-AGG CAA CAG TGT CAG AGT CC-3Ј) for Xbp1 splicing and CHOP Fwd (5Ј-CAC TAC TCT TGA CCC TGC GT-3Ј) and CHOP Rev (5Ј-GGA GAG ACA GAC AGG AGG TGA-3Ј) for CHOP expression.
TUNEL Assay-Sections were prepared in the same way as in immunofluorescence studies. Frozen sections were then fixed in ice cold 1:1 acetone-methanol for 20 min, washed in 1ϫ phosphate-buffered saline, and blocked for 20 min in equilibration buffer containing 1ϫ TE (10 mM Tris, 1 mM EDTA, pH 8.0), 1ϫ terminal transferase buffer, and 1.67 mM CoCl 2 at room temperature (Roche Applied Science). Then excess equilibration buffer was absorbed with a KimWipe, and reaction buffer containing 0.005 mM ChromaTide Alexa Fluor 488-5-dUTP (Molecular Probes, Inc., Eugene, OR), 0.01 mM dATP (both diluted in equilibration buffer), and 1 l of TdT enzyme (Roche Applied Science) was added onto the sections and incubated at 37°C for 60 -90 min. Sections without TdT enzyme were used as negative controls. Slides were soaked in 1ϫ SSC for 15 min at room temperature to stop the reaction, washed in 1ϫ phosphate-buffered saline, and stained with Draq5 diluted 1:2000 in 1ϫ phosphate-buffered saline for 5 min. Slides were rinsed briefly in phosphate-buffered saline, mounted, and visualized using a Zeiss LSM 510 confocal microscope (Carl Zeiss Inc., Göttingen, Germany).
Protein Translation Assay-Lenses were isolated and immediately placed in a 6-well dish with Medium 199 (Mediatech Inc., Manassas, VA) supplemented with a tritium ( 3 H)-labeled amino acid mixture (catalog number ART0328, American Radiolabeled Chemicals Inc., St. Louis, MO) to a final concentration of 20 Ci/ml. Lenses were incubated for at least 6 h in a humidified atmosphere of 5% CO 2 and 95% air at 37°C, as described previously (48). Medium was removed, and lenses were washed three times with non-radioactive Medium 199. Thirty microliters of lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with halt protease and phosphatase inhibitor mixture (Thermo Scientific) was added, and lenses were lysed using a mortar and pestle. The radioactivity incorporated into the lenses were determined as previously described (49). Briefly, 10 l of the cell lysate was spotted on a Whatman glass microfiber filter (catalog number 1822 021). The filter then was washed with boiling trichloroacetic acid for 5 min and twice more with cold trichloroacetic acid (5 min each). The filter was rinsed briefly in ice-cold 95% ethanol and then in ice-cold acetone and air-dried. The radioactivity remaining on the dried filter was measured with a liquid scintillation counter. To calculate the radioactivity per cubic millimeter volume, the radii of the lenses were determined from lens photographs using Image J, and the lens volume was estimated using the formula, 4/3⅐⅐r 3 . Student's t test was performed for statistical analysis.

Overexpression of Col4a3 or Col4a4 in the Embryonic Lens
Results in Microphthalmia and Cataract-All six known Col4 genes, (a1-a6), are expressed in the adult mouse lens; however, Col4a3 and Col4a4 are not expressed in the embryonic lens, resulting in the absence of the COL4(A3.A4.A5):(A3.A4.A5) heterotrimeric network from the lens capsule until 2-3 weeks postnatal (50). Because the assembly of the COL4(A3.A4.A5) protomer requires the presence of all three subunits, we hypothesized that ectopic expression of either normal Col4a3 or Col4a4 alone in the embryonic lens will result in the retention of unassembled single COL4A3 or COL4A4 chains in lens cells inducing the unfolded protein response. Thus, we generated transgenic mice ectopically expressing either Col4a3 or Col4a4 in all lens cells starting at the late lens pit stage (E11-E11.5) using the ␦en/␣A-crystallin hybrid promoter (Fig. 1A). Two independent founder transgenic mice were derived from each transgene. Both Col4a4 lines (A4-501 and A4-520) and one of the Col4a3 lines (A3-21) had affected lenses, whereas the remaining Col4a3 line, A3-9, had clear lenses (data not shown). The affected transgenic lenses exhibit variable phenotype in lens size and transparency; however, lenses from all three affected lines were significantly smaller and had obvious opacities compared with the normal lens (Fig. 1B).
Transgenic Lenses Produce High Levels of Exogenous Collagen Chains That Accumulate in Lens Fiber Cells-To determine when the expression of the transgenes starts, we used antibodies specific to COL4A3 and COL4A4 chains. Both lines did not have detectable transgene expression up to E12.5 (data not shown). At E12.5, we detected high expression of the Col4a3 transgene in A3-21 lens fiber cells ( Fig. 2A, top), whereas the A3-9 line, which has clear lenses, did not have detectable Col4a3 expression ( Fig. 2A, bottom), suggesting a correlation between the cataract phenotype and the expression of the transgene. When overexposed, a slight amount of COL4A3 was also detected around the lens, suggesting the secretion of a small amount of the transgene product ( Fig. 2A, middle). Similarly, in the affected A4-520 line, we detected prominent expression of the Col4a4 transgene in lens fiber cells starting from E12.5 (Fig. 2B, top), whereas there was no detectable Col4a4 expression in the WT lens at this age (Fig. 2B, bottom). When overexposed, a weak fluorescent signal was also detected around the lens, suggesting the secretion of some of the transgene (Fig. 2B, middle). By Western blotting using a pan-collagen IV antibody, we confirmed that the affected transgenic lines (A4-520 and A3-21) have higher levels of soluble collagen IV chains in the lens at E14.5 (Fig. 2C). Quantification of the signal on the membrane suggested that the increase in soluble collagen IV chains is ϳ35% for the A3-520 line and 48% for the A3-21 line. Notably, we did not detect abnormal expression of Col4a4 in the Col4a3-overexpressing A3-21 line or vice versa, suggesting that the overexpression of one chain does not cause abnormal expression of the other (data not shown). These results suggested that the overexpressed collagen chains were unable to exit the ER and instead accumulate within lens fiber cells.
Lenses Overexpressing Collagen IV Chains Have Defective Fiber Cells-We then analyzed how the intracellular accumulation of collagen IV interferes with lens development and differentiation. Most of the lens is composed of concentric layers of elongated fiber cells that are formed by the continuous differentiation of epithelial cells at the lens equator. Because there is no turnover of lens fiber cells, the steady addition of new fiber cells throughout life concentrates the older fibers in the center of the lens (51,52). We analyzed transgenic embryonic lenses using hematoxylin and eosin staining starting from E11.5. In both transgenic lines, primary fiber cell differentiation occurred normally as in WT lenses (data not shown), and elongating transgenic fiber cells reached to the epithelium (Fig. 3, B and C). However, by E15.5, A3-21 fiber cells lost contact with the epithelium, resulting in the formation of a gap between the epithelium and fiber cells (Fig. 3F). Similarly, by E16.5, A4-520 fiber cells detached from the epithelium (Fig. 3H). This phenotype suggested that transgenic lenses have a fiber cell elongation defect, where the fiber cells expressing the transgene cannot keep up with the growing lens. In lenses from newborn mice, the gap between epithelium and fiber cells is mostly closed, suggesting that the elongation defect is partially relieved; however, fiber cells still look abnormal and form a disorganized central mass surrounded by the secondary fibers (Fig. 3, K and L). A closer examination of the adult transgenic fiber cells by scanning electron microscopy revealed that these fiber cells are irregularly shaped and poorly aligned (Fig. 4, bottom left) compared with the uniform hexagonal cells of the normal lens (Fig. 4, top left). Also, although normal fiber cells have regularly arrayed ball-and-socket-like lateral membrane interdigitations (Fig. 4, top right), transgenic fiber cells have an abnormal lateral membrane structure, resulting in the distortion of the tightly packed fiber cell array (Fig. 4, bottom right). Differentiating transgenic fiber cells also retain their nuclei and are strongly stained by the lectin concanavalin A, which binds to carbohydrate moieties in the ER (53,54), and prohibitin, a marker of the inner mitochondrial membrane, suggesting that the degradation of nuclei and organelles that occurs during normal lens fiber cell differentiation is arrested (supplemental Fig. 1). These results showed that the accumulation of the transgene-derived proteins in lens fiber cells interferes with fiber cell elongation during early embryogenesis and results in altered fiber cell structure and differentiation.
Trangenic Lenses Do Not Have a Major Defect in General Secretion-Next, we tested whether the lens defects seen in transgenic lenses resulted from a general secretion defect caused by the accumulation of high levels of exogenous collagen chains in the secretory pathway. Immunostainings for panlaminin (data not shown) and the endogenous collagen ␣ chains (A1, A2, A5, and A6) showed that they are secreted normally, forming a capsule around the lens (Fig. 5). Thus, we concluded that the transgenic lenses are not suffering a major global defect in the secretory pathway, suggesting that the pathogenesis in these lenses is mostly due to the presence of unfolded proteins in the secretory pathway.
The Accumulation of Transgenes in the Lens Causes Immunoglobulin-binding Protein (BiP) Up-regulation and Extension of the Endoplasmic Reticulum-One known cellular consequence of protein accumulation in the secretory pathway is the activation of the unfolded protein response. UPR activation changes both transcriptional and translational programs and can have a major effect on cell biology (34,55,56). BiP is a major chaperone found in the ER, and its up-regulation has been used as a sensitive indicator of UPR activation (57,58). BiP immunostaining showed that the affected A4-520 and A3-21 lines have drastically elevated levels of BiP expression at postnatal day 2 (Fig. 6A). This BiP up-regulation starts at E12.5 and coincides with the accumulation of collagen chains in lens fiber cells (Fig.  6B). Western blot analysis confirmed the increase in BiP protein levels at E14.5 and E16.5 transgenic lenses (Fig. 6C). One of the hallmarks of UPR is the activation of factors that induce ER membrane production, resulting in ER expansion (59 -62). Transgenic fiber cells do have strong concanavalin A label-FIGURE 2. Exogenous COL4A3 or COL4A4 chains accumulate to high levels in affected transgenic fiber cells. A, a chain-specific antibody against COL4A3 was used to detect the expression in the affected A3-21 (top and middle) and the unaffected A3-9 (bottom) lenses. At E12.5, Col4a3 transgene is highly expressed in the affected A3-21 fiber cells, whereas there is no detectable expression in the unaffected A3-9 lenses at this age. B, a chain-specific antibody against COL4A4 was used to detect its expression in the affected A4-520 (top and middle) and the WT (bottom) lenses. At E12.5, Col4a4 transgene is highly expressed in the affected A4-520 fiber cells, whereas there is no detectable expression in the wild type lenses at this age. The gain for the middle and bottom panels in both A and B was increased ϳ25% to obtain an overexposed image showing possible secretion of exogenous collagen chains. The arrowheads indicate the detection of secreted collagen chains around the lens. Bars, 77 m; e, epithelium; f, fiber cells. Red, collagen IV chain; blue, DNA. C, expression of collagen IV chains in E14.5 lenses was detected by Western blotting using a pan-collagen IV antibody. The affected A4-520 and A3-21 lenses have elevated levels of soluble collagen IV chains. DECEMBER 18, 2009 • VOLUME 284 • NUMBER 51

Unfolded Protein Response and Cataract
ing around the nuclei of lens fiber cells and throughout the cytoplasm, suggesting that the ER in transgenic fiber cells has been expanded (Fig. 6D). Because these results suggested that the transgenic lenses are experiencing the unfolded protein response, we proceeded to identify which pathways are activated.
The IRE1/Xbp1 Pathway Is Activated in the Transgenic Lenses-The UPR pathway that has been conserved from yeast to mammals is transduced by IRE1 (inositol-requiring enzyme 1), an atypical type I transmembrane protein consisting of an ER luminal dimerization domain as well as cytosolic kinase and endoribonuclease domains (63). Activated IRE1 removes a 26-base intron from the Xbp1 mRNA, introducing a frameshift and an alternative C terminus to create a potent transcription factor (64 -66). Immunostaining using an XBP1 antibody that detects both spliced and unspliced forms of the XBP1 showed that total levels of XBP1 protein start to increase in transgenic lenses by E12.5 (Fig. 7A). By RT-PCR analysis using primers designed to differentiate between spliced and unspliced forms, we also detected a relative increase in spliced Xbp1 mRNA compared with unspliced in the affected transgenic lenses starting from E12.5 (Fig. 7B). Furthermore, Western blot analysis demonstrated that the protein generated from the spliced Xbp1 mRNA (54 kDa) is only found in the transgenic lenses, whereas protein generated from unspliced Xbp1 mRNA (33 kDa) is found in both WT and transgenics, indicating the ER stressinduced production of the XBP1 transcription factor (Fig. 7C). Notably, control lenses also start up-regulating XBP1(U) protein between E14.5 and E16.5. Although unspliced Xbp1 has been shown to bind and regulate spliced Xbp1 function (67), other functions of this protein are still unclear, and its role in normal lens development and differentiation needs to be investigated.
ATF6 Pathway Is Activated in the Transgenic Lenses-Another sensor activated in mammalian UPR is ATF6 (activating transcription factor 6), which, upon ER stress, progresses to the Golgi, where it is cleaved by S1P and S2P proteases, liberating its cytosolic domain as a soluble transcription factor (68 -70). There are two ATF6 genes in mammals, ATF6␣ and ATF6␤, that appear to be regulated by the same mechanism. The staining of lenses with an ATF6␣/␤ antibody, which detects both intact and cleaved (nuclear) forms, showed that total ATF6␣/␤ greatly increases in transgenic lens fiber cells by E12.5 (Fig. 8A). By Western blotting, we showed that at E14.5, pro-ATF6␣/␤ is cleaved into a ϳ50-kDa protein in transgenic lenses, suggesting its activation (Fig. 8B). Notably, pro-ATF6␣/␤ was detected as a   doublet, which possibly corresponded to fully glycosylated and partially glycosylated forms of ATF6␣/␤ that have previously been shown to arise upon UPR induction (71). Interestingly, at E16.5, both transgenic and normal lenses had cleaved (nuclear) ATF6␣/␤ as the major form, suggesting that ATF6 is also being activated during normal lens development and differentiation between E14.5 and E16.5 (Fig. 8B).

PERK Pathway Is Activated in the Transgenic Lenses-
The third UPR pathway is induced by PERK, an ER transmembrane protein, whose luminal domain senses ER stress and cytoplasmic domain directly phosphorylates eukaryotic initiation factor 2 (eIF2␣) on serine 51, leading to its inactivation. This results in the inhibition of translational initiation for most cellular mRNAs (72)(73)(74). We used an antibody specific to phosphorylated PERK to detect active PERK expression. Interestingly, phospho-PERK was also present in the apical tips of normal fiber cells (Fig. 9A, top); however, its levels were increased at the transgenic fiber cell tips that lost contact with the epithelium (Fig. 9A, bottom). Immunostaining using an antibody specific to phosphorylated eIF2␣ showed that there is an increase in the amount of phosphorylated eIF2␣ in the transgenic fiber cells (Fig. 9B). By Western blotting, we confirmed the increase in phosphorylated eIF2␣ levels in transgenic lenses starting from E14.5, whereas total eIF2␣ levels are constant, suggesting the possibility of a global translational attenuation in the transgenic lenses (Fig. 9C). Notably, in control lenses, total eIF2␣ levels FIGURE 6. The endoplasmic reticulum in affected transgenic lenses is greatly extended with an up-regulation of the major ER chaperone BiP. A, the expression of BiP (red) in 2-day-old WT (left), A4-520 (middle), and A3-21 (right) lenses detected by immunofluorescence. The two affected transgenic lines (A4-520 and A3-21) show much higher levels of BiP expression in lens fiber cells compared with WT lens. B, the expression of BiP (red) in WT (top) and A4-520 (bottom) lenses detected by immunofluorescence. In the A4-520 lens, BiP expression starts to up-regulate at E12.5. Bars, 77 m; e, epithelium; f, fiber cells. C, the expression of BiP in WT and A4-520 lenses detected by Western blotting. BiP protein is found at higher levels in E14.5 transgenic lenses and remains elevated after birth. The asterisks indicate the smaller 72 kDa band detected by BiP antibody. This band was previously proposed to be a degradation product of wild type BiP (95). WT lenses also express two BiP bands upon longer exposure (not shown). D, endoplasmic reticulum was probed in WT (left) and A4-520 (right) lenses using concanavalin A (green). Note that the affected A4-520 fiber cells have much higher staining around the nuclei and throughout the cytoplasm. Bars, 40 m; e, epithelium; f, fiber cells. In all panels, blue labels DNA. ␤2-Microglobulin was used as a control. C, immunoblot analysis of XBP1 protein in WT and A4-520 lenses. Note that the protein produced from spliced Xbp1 is only found in the affected A4-520 lens, whereas protein from unspliced Xbp1 is produced both in WT and A4-520 lenses. Actin was used as a loading control. also start increasing between E14.5 and E16.5, with a slight increase in phospho-eIF2␣ levels.
Attenuation of protein translation could have a dramatic effect on lens fiber cell differentiation because one of the characteristics of epithelial to fiber cell differentiation is elevated levels of protein synthesis (75,76). Particularly, fiber cells synthesize proteins called crystallins, which make up about 90% of the soluble protein content of lens fibers, generating the high refractive index needed for lens transparency and function (77,78). Immunostaining with pan-␤and ␥-crystallin antibodies indicated that their expression is decreased in the transgenic lenses at E16.5 (Fig. 10A). When we compared BiP, ␥-crystallin, and actin levels in 1% of lens soluble proteins between normal and transgenic lenses by Western blotting, BiP levels were much higher, and ␥-crystallin levels were reduced in transgenic lenses, whereas actin levels were similar between the two samples, confirming the apparent decrease in ␥-crystallin levels seen by immunohistochemistry (Fig. 10B). In order to test whether the decrease in crystallin synthesis is due to a global defect in protein translation, we measured the protein translation rate in intact lenses by incubating freshly isolated E17.5 lenses in cell culture medium supplemented with a tritiumlabeled amino acid mixture. When we compared the amount of radioactivity incorporated in newly synthesized proteins within the same lens volume, transgenic lenses incorporated significantly less radioactivity in newly synthesized proteins over the course of the incubation. This indicated that transgenic lenses had a reduced protein translation rate (Fig. 10C).  ATF6. B, immunoblot analysis of ATF6 protein in WT and A4-520 lenses. Note that at E14.5, ATF6 protein is not detectable in WT lenses, but both forms (full-length/ER-resident and cleaved/ nuclear) of ATF6 are highly expressed in A4-520 lenses. Actin was used as a loading control. ATF6(N), nuclear/cleaved form of ATF6.

FIGURE 9. PERK/eIF2␣ pathway is activated in transgenic lenses.
A, expression of phospho-PERK (red) in WT (top) and A4-520 (bottom) lenses detected by immunofluorescence. Note that phospho-PERK levels are increased at the apical tips of A4-520 transgenic fiber cells at E16.5. The arrowheads indicate the phospho-PERK detected at the apical tips of normal fiber cells (top) and transgenic fiber cells that lost contact with the epithelium (bottom). B, expression of phospho-eIF2␣ (red) in WT (top) and A4-520 (bottom) lenses detected by immunofluorescence. Note that phospho-eIF2␣ levels are initially up-regulated in A4-520 lenses at E12.5 and continue to increase with time. Bars, 77 m; e, epithelium; f, fiber cells. In all panels, blue labels DNA. C, immunoblot analysis of eIF2␣ protein in WT and A4-520 lenses. Note that at E14.5, both total and phospho-eIF2␣ are not detectable in the WT lens, but they are present at high levels in A4-520 lenses. Also note that total eIF2␣ levels are similar in WT and A4-520 lenses at E16.5 and P2, whereas phospho-eIF2␣ levels are elevated in the A4-520 lens. Actin was used as a loading control.
PERK-mediated phosphorylation of eIF2␣ also results in the enhanced translation of ATF4, a b-ZIP transcription factor of the cyclic AMP-response element-binding protein family (79,80). We detected intense cytoplasmic ATF4 staining in both normal and transgenic fiber cells at E12.5 (Fig. 11A, top); however, this expression was greatly reduced at E14.5 lenses (Fig.  11A, bottom). Although the cytoplasmic expression was reduced, in transgenic lenses, ATF4 continued to be expressed by a group of fiber cells and became strongly nuclear, suggesting its involvement in transcriptional activation of UPR-responsive genes (Fig. 11A, bottom right). By Western blotting using an ATF4 antibody, we detected two major bands. The lower band existed in both normal and transgenic lenses, whereas the 50 kDa band was only present in the transgenic lenses (Fig. 11B). This 50-kDa ATF4 has been previously reported to be induced in a variety of cancer cells upon hypoxia (81). An important UPR-specific target gene of ATF4 is CHOP/GADD153, a b-ZIP transcription factor highly up-regulated during UPR (82,83). By RT-PCR, we found that Chop transcript levels are increased in transgenic lenses starting from E14.5 (supplemental Fig. 2). By immunostaining, we detected strong nuclear CHOP expression in some of the transgenic fiber cells starting from E14.5, whereas there was no CHOP expression in the normal lenses (Fig. 11C). By Western blotting using an antibody against CHOP, we detected two major bands between 25 and 30 kDa, whereas CHOP is reported to be a 29-kDA protein. The larger band was only found in transgenic lenses, indicating the UPRspecific production of the CHOP transcription factor (Fig.  11B). Although another possible outcome of the activated PERK pathway is autophagy (84), we found no evidence of this by looking at the conversion of LC3 by Western blotting and immunostaining (data not shown).
Apoptosis Is Induced in the E18.5 Transgenic Lenses-Because chronic or severe UPR activation can initiate apoptosis, we investigated whether there is increased cell death in the transgenic lenses. Immunostaining of lenses with an antibody specific for the cleaved form of caspase 3 did not reveal any apparent apoptotic cells prior to E18.5. However, at E18.5, a few central fiber cells of transgenic lenses positively stained for cleaved caspase 3 (Fig. 12, A and B). We also detected TUNELpositive central fiber cells in E18.5 transgenic lenses (Fig. 12, E  and F), whereas there were no TUNEL-positive cells in normal lens sections at this age (Fig. 12, C and D) except at the tunica vasculosa lentis, a vascular network formed by hyaloid vessels that is regressing at this time via coordinated endothelial apoptosis (85). We also assayed caspase 12 activation because it is specifically localized on the cytoplasmic side of the ER and has been proposed as one of the initiator caspases in UPRinduced death in mice (86). By Western blotting, we found that pro-caspase 12 is cleaved into shorter fragments in transgenics, consistent with the induction of cell death in transgenic lenses (Fig. 12G).
The Activation of UPR Sensors Is Also Seen in the Col4a1 Mutants-Because mutations in ECM genes are known to cause cataract, we analyzed the activation of UPR pathways in a cataractous mouse model carrying mutations in an endogenous ECM gene that would let us assess the contribution of UPR activation to cataractogenesis associated with physiological lev- A, expression of ␥ (left) and ␤ crystallins (right) in WT (top) and A4-520 lenses (bottom) detected by immunohistochemistry. At E16.5, A4-520 lenses have less crystallin expression as compared with WT lenses. Bars, 330 m. c, cornea; e, epithelium; f, fiber cells; r, retina. Blue, DNA; brown, crystallins. B, 1% of lens soluble proteins were subjected to SDS-PAGE, and the expression of BiP, ␤-crystallins, and actin was detected by immunoblotting. In 1% of lens soluble proteins, compared with WT lenses, A4-520 lenses have much higher levels of BiP protein and lower levels of ␤-crystallins, whereas actin levels are similar. C, protein translation rate was measured in WT and A4-520 lenses by incubating freshly isolated E17.5 lenses in Medium 199 supplemented with 3 H-labeled amino acid mixture. The y axis shows the radioactivity measured in dpm/ml 3 of lens volume. The bars represent the average radioactivity measured in WT and A4-520 lenses. Statistical analysis was done using Student's t test. p Ͻ 0.0001. DECEMBER 18, 2009 • VOLUME 284 • NUMBER 51 els of mutant protein accumulation. Col4a1 mutant mice that carry a mutation in exon 40 of Col4a1 develop porencephaly and cataractous lenses (19,87). Immunostaining using an anti-body specific for COL4A1 showed that the mutant COL4A1 chains are accumulating inside the lens cells. The strongest accumulation was detected in lens epithelial cells (Fig. 13E), but some retention was also seen in fiber cells (Fig. 13G). Notably, COL4A2 also appeared to be partially retained in lens epithelial cells, indicating co-accumulation with abnormal COL4A1 in the secretory pathway (Fig. 13H). We then showed that this accumulation of collagen chains activates UPR pathways but to a much lesser degree when compared with the transgenic mice. Mutant lenses had up-regulation of BiP and ATF6 in both the mutant epithelium and fiber cells (Fig. 14, A and B). Total XBP1 FIGURE 11. ATF4 and its target CHOP are expressed by the transgenic fiber cells. A, expression of ATF4 (red) in WT (left) and A4-520 (right) lenses detected by immunofluorescence. At E12.5, ATF4 is present in the cytoplasm of both normal and transgenic lenses. However, at E14.5, transgenic fiber cells exhibit elevated levels of nuclear ATF4 in fiber cells. The ϫ40 magnification images (40ϫ) (x and y) on the right show the boxed regions from transgenic lenses. B, immunoblot analysis of ATF4 and CHOP expression. ATF4 was detected as two bands. The lower band was abundantly produced in both normal and transgenic lenses, whereas the 50 kDa band (marked with an asterisk) was found only in the transgenics. The 29-kDa CHOP (marked with an asterisk) was only found in the transgenic lenses. C, expression of CHOP (red) in WT (left) and A4-520 (right) lenses detected by immunofluorescence. At E12.5, there is a very weak CHOP expression in transgenic fiber cells. At E14.5, a group of fiber cells express CHOP strongly in their nuclei. The ϫ40 magnification images (40ϫ) (x and y) on the right show the boxed regions from transgenic lenses. Bars,77m; e, epithelium; f, fiber cells. In all panels, blue labels DNA.

FIGURE 12. Apoptosis is induced in some transgenic fiber cells at E18.5.
A and B, cleaved caspase 3 was detected in a group of central fiber cells by immunofluorescence. A, an antibody specific to the cleaved form of caspase 3 (red) was used to detect the activation of this caspase in E18.5 transgenic lenses. Note the red staining in A4-520 transgenic fiber cells indicated with arrows. B, cleaved caspase 3 staining (red) without the nuclear counterstain. C-E, apoptotic cells were detected by TUNEL staining. C, TUNEL staining (green) of E18.5 WT lenses. Note that no WT lens cells are found to be TUNEL-positive, whereas the blood vessels behind the lens (tunica vasculosa lentis) are positively stained. D, TUNEL staining (green) of WT lenses without the nuclear stain. E, TUNEL staining (green) of E18.5 A4-520 lenses. A group of central A4-520 fiber cells (indicated by arrows) are stained TUNEL-positive. F, TUNEL staining (green) of A4-520 lenses without the nuclear stain. In all panels, blue labels DNA. Bars, 77 m. e, epithelial cells, f, fiber cells; tvl, tunica vasculosa lentis. G, activation of caspase-12 in E18.5 transgenic lenses examined by Western blot analysis. Asterisks, active forms of caspase-12 detected with anti-caspase-12 antibody. levels were increased in some parts of the epithelium (Fig. 14C), and RT-PCR analysis showed that this Xbp1 was being spliced in the mutant lenses (Fig. 14D). We also detected the presence of phosphorylated eIF2␣ in a few central fiber cells (Fig. 14E). However, the increased nuclear expression of ATF4 or CHOP induction was not detected in the mutant lenses, whereas these lenses were also negative for cleaved caspase 3 staining (data not shown), suggesting that there is no UPR-mediated cell death in these mutant lenses. These results suggest that the accumulation of mutant chains in the secretory pathway at physiological levels can also activate UPR pathways in the lens, and this activation may contribute to the cataract pathogenesis seen in several ECM disorders.

DISCUSSION
It has been previously proposed that UPR can be induced in the lens by oxidative insults, and UPR activation could influence cataractogenesis caused by oxidative stress (41). However, the impact of UPR activation on lens cell fate and/or differentiation programs has not been previously studied in vivo. Here we investigated the contribution of UPR induced by unfolded protein accumulation in the ER to cataractogenesis both in an induced transgenic and physiologically relevant disease model. The transgenic lenses studied express elevated levels of normal COL4A3 or COL4A4 chains that cannot assemble alone into heterotrimers and thus accumulate within the secretory pathway (Fig. 2). In contrast, Col4a1 ϩ/ex⌬40 mice, which have been described as an animal model of hereditary porencephaly, do not efficiently assemble the COL4(A1.A1.A2) heterotrimer, leading to intracellular retention of both COL4A1 and COL4A2 (Fig. 13). We have found that both in transgenic and mutant lenses, the accumulation of collagen chains in the lens secretory pathway activates the unfolded protein response.
In transgenic lenses, this activation interfered with several aspects of lens fiber cell differentiation resulting in distorted fiber cell structure and function (Fig. 15). The differentiation of epithelial cells to fiber cells is a well defined developmental pathway and depends on a highly coordinated set of events, including withdrawal from cell cycle, extensive cell elongation, storage of crystallin proteins, and degradation of nuclei and organelles (51). The altered fiber cells experiencing ER stress, on the other hand, exhibited retention of the nuclei and organelles, defective fiber cell elongation, and reduced crystallin synthesis. The detrimental effects of UPR activation in the FIGURE 13. Mutant COL4A1 accumulates in Col4a1 ؉/ex⌬40 lenses along with its binding partner COL4A2. Chain-specific antibodies against COL4A1, laminin, and COL4A2 were used to detect these proteins in WT (left) and Col4a1 ϩ/ex⌬40 lenses (right). At E18.5, COL4A1 (A-C, red), laminin (B and C, green), and COL4A2 (D, red) are secreted into the lens capsule of normal lenses. In mutant lenses, COL4A1 accumulates inside mutant Col4a1 ϩ/ex⌬40 lens epithelial (E and F, red) and fiber cells (G, red). Similarly, COL4A2 also partially accumulates inside mutant lens cells (H). The arrowheads indicate the epithelial cells that retain collagen chains in mutant lenses. Bars, 77 m. c, lens capsule; e, epithelium; f, fiber cells; tz, transition zone. Blue, DNA; red, collagen chains; green, pan-laminin.  DECEMBER 18, 2009 • VOLUME 284 • NUMBER 51 lens were probably mediated by both activation of UPR-specific transcription factors and UPR-induced global translational attenuation. The production of ATF6(N) and XBP1(S) could alter fiber cell transcriptional programs by up-regulating genes involved in ER folding and ER biogenesis in lens fiber cells that would normally undergo programmed degradation of nuclei and cytoplasmic organelles (59,62,88). On the other hand, PERK-mediated translational attenuation could interfere with lens fiber cell differentiation through lowering the synthesis of crystallins and other fiber cell-specific proteins that are needed for fiber cell differentiation. The active PERK pathway has been shown to be essential for survival of cells upon ER stress because it aims at decreasing the load on the already challenged ER (79). However, because epithelial to fiber differentiation involves large amounts of protein synthesis, translational attenuation in the lens might cause adverse effects upon differentiation. Notably, a major defect that we observed was in elongation of lens fiber cells during embryogenesis (Fig. 3), which could be related to the insufficient production of fiber cell-specific membrane or other elongation-associated proteins (89). Although fiber cells experiencing UPR showed major structural defects, we did not detect the induction of cell death in the transgenic lens cells prior to E18.5; however, at this time, we detected a group of apoptotic fiber cells in the center of the lens (Fig. 12). This suggests that the induction of cell death is a later response in UPR signaling restricted to a subset of cells that probably experienced prolonged UPR activation because the center of the lens is composed of the oldest fiber cells (90). In contrast, we could not detect cell death in Col4a1 ϩ/⌬ex40 lenses. Because mutant lenses have lower levels of collagen accumulation in the lens, apoptosis appears to only occur upon severe UPR activation (Fig. 15).

Unfolded Protein Response and Cataract
However, the transgenic mouse data propose that early UPR responses, such as translational inhibition, can disrupt the highly coordinated events of fiber cell maturation, leading to defective lens physiology and function. Thus, although the lower activation levels of UPR sensors in Col4a1 ϩ/⌬ex40 lenses make the contribution of UPR to cataractogenesis less clear in this case, the activation of UPR pathways in the lens could be subtle but still be detrimental to lens structure even without the induction of cell death. Tsang et al. (27) showed similar results with transgenic mice expressing mutant collagen X in hypertrophic chondrocytes, where the resulting UPR activation did not induce cell death but altered chondrocyte differentiation and function, causing chondroplasia.
Humans carrying COL4A1 mutations develop cataracts; however, the underlying mechanism has not been identified (17,18). In both the human patients and mutant mice, the eye phenotype shows great variations, suggesting the influence of multiple factors (19,21). However, the absence of a phenotype in mice heterozygous for a null allele of Col4a1 and Col4a2 suggests that the mutant protein is required to cause the disease (91). Numerous other ECM disorders also have associated lens defects, but the etiology of the cataracts is largely undetermined. For instance, cataracts and/or lenticonus are reported in Alport syndrome caused by mutations in either COL4A3, COL4A4, or COL4A5 (92), and it has been reported that the lens epithelial cells from cataractous lenses were irregular in shape and had a dilated ER and Golgi consistent with UPR activation (93,94). Similar to the findings with mice null for Col4a1 and Col4a2, we found that lenses from Col4a3 null mice are transparent and do not show any gene expression changes by microarray analysis, suggesting that the presence of mutant chains is needed to cause the cataract pathology. 5 Moreover, the clinical heterogeneity of the cataracts associated with porencephaly and Alport syndrome could be associated with variation in the extent of UPR activation, because the level of UPR activation is directly related to the strength and/or duration of the ER stress experienced, and it may vary between different individuals, leading to different clinical outcomes. Thus, the possibility that the activation of UPR pathways upon the accumulation of mutant collagen chains within the ER contributes to the cataract pathology in these disorders is an attractive theory. However the evidence for a direct role of UPR in the cataracts seen in humans and mice harboring mutations in collagen genes is still missing, and further research is needed to reveal a better understanding of the role of UPR in lens 5 B. P. Danysh and M. K. Duncan, unpublished data. FIGURE 15. UPR is activated in the lenses of transgenic and mutant mice upon accumulation of collagen chains. In transgenics, exogenous expression of Col4a3 or Col4a4 genes in the embryonic lens causes high levels of collagen chain accumulation inside the transgenic fiber cells and results in immediate activation of all three known UPR sensors. In Col4a1 ϩ/ex⌬40 mice, mutant COL4A1 chains cannot properly assemble and accumulate in lens cells along with COL4A2 and activate all three UPR sensors. In transgenics, the translational attenuation induced upon eIF2␣ phosphorylation negatively affects the synthesis and accumulation of major lens proteins, crystallins. The phosphorylation of eIF2␣ also results in selective translation of ATF4 and induction of its target CHOP in transgenic lenses, whereas in mutants, this response has not been detected (marked with an asterisk). Transgenic lens fiber cells experiencing UPR activation exhibited defective fiber cell elongation around E15.5 and cell death later in E18.5. In contrast, cell death was not observed in mutants (marked with an asterisk). Overall, the data suggest that all UPR responses could contribute to the lens pathology observed, including inhibition of denucleation, disruption of lens ultrastructure, opacity, and microphthalmia.
pathology. Nevertheless, our results support the proposition that UPR can be a common cataractogenic mechanism in disorders resulting from abnormal ECM molecule production by the lens.