Essential roles of grp94 in gut homeostasis viachaperoning canonical Wnt pathwayBei Liua, Matthew Staronb, Feng Honga, Bill X. Wua,c, Shaoli Sund, Crystal Moralesa, Craig E. Crossonc,Stephen Tomlinsona, Ingyu Kime, Dianqing Wue, and Zihai Lia,1

aDepartment of Microbiology and Immunology, Hollings Cancer Center, Medical University of South Carolina, Charleston, SC 29425; bDepartment ofImmunobiology, Yale University School of Medicine, New Haven, CT 06520; cStorm Eye Institute, Medical University of South Carolina, Charleston, SC 29425;dDepartment of Pathology and Laboratory Medicine, Medical University of South Carolina, Charleston, SC 29425; and eDepartment of Pharmacology, YaleUniversity School of Medicine, New Haven, CT 06520

Edited by Arthur L. Horwich, Yale University School of Medicine, New Haven, CT, and approved March 13, 2013 (received for review February 14, 2013)

Increasing evidence points to a role for the protein quality controlin the endoplasmic reticulum (ER) in maintaining intestinal homeo-stasis. However, the specific role for general ER chaperones in thisprocess remains unknown. Herein, we report that a major ER heatshock protein grp94 interacts with MesD, a critical chaperone forthe Wnt coreceptor low-density lipoprotein receptor-related pro-tein 6 (LRP6). Without grp94, LRP6 fails to export from the ER tothe cell surface, resulting in a profound loss of canonical Wnt sig-naling. The significance of this finding is demonstrated in vivo inthat grp94 loss causes a rapid and profound compromise in intesti-nal homeostasis with gut-intrinsic defect in the proliferation of in-testinal crypts, compromise of nuclear β-catenin translocation, lossof crypt-villus structure, and impaired barrier function. Taken to-gether, our work has uncovered the role of grp94 in chaperoningLRP6-MesD in coordinating intestinal homeostasis, placing canoni-cal Wnt-signaling pathway under the direct regulation of the gen-eral protein quality control machinery in the ER.

Heat shock protein (HSP) grp94 (1), also known as gp96 (2),encoded by HSP90b1 (3), is the endoplasmic reticulum (ER)

paralog of cytosolic HSP90. Like other HSPs, grp94 is induced bythe accumulation of misfolded proteins (4) and binds and hydrol-izes ATP (5–7). As the most abundant protein in the ER lumen, itis phylogenically conserved (8) and ubiquitously expressed in allnucleated cells. An important function of the ER is the steady-statefolding of nascent polypeptides into their mature tertiary/quater-nary structures and the assembly of large multimeric proteincomplexes in the secretory pathway. Recent work demonstratesthat grp94 is the critical chaperone for multiple Toll-like receptors(TLRs) and integrins (9–13) and that it participates in the unfoldedprotein response (UPR) (14). Without grp94, most integrins andTLRs are unable to fold properly and, thus, fail to exit the ER andtraffic to their proper “post-ER” compartment. Misfolded proteinsare actively retained in the ER as a part of the ER quality-controlprocess, and their accumulation in the ER can result in ER stressand activation of the UPR. Up to 10% of cytosolic proteins aredependent on HSP90 for folding (15); it is, therefore, unlikely thatgrp94 clients are limited to TLRs and integrins.The intestinal epithelium is continually replenished through

the proliferation and differentiation of intestinal stem cells thatreside within the intestinal crypts (16). Canonical Wnt signalingthrough the surface receptor, Frizzled, and its coreceptor low-density lipoprotein receptor-related protein 5 (LRP5) or LRP6,is instrumental for gut homeostasis (16). LRP5 and LRP6 arehighly homologous to each other, and their function in the ca-nonical Wnt pathway is redundant in response to some Wntligands but not so to others (17–21). Mesoderm development(MesD), an ER chaperone, was recently shown to be necessaryfor surface expression of LRP5/6 and Wnt signaling (22). In-terestingly, like grp94 and MesD, LRP5 and -6 are also necessaryfor mesoderm formation and gastrulation in mice (20), suggest-ing a possible overlap among these proteins/pathways in earlydevelopment. However, the functional connection between

grp94 and Wnt pathway has not been reported. Furthermore,a direct role for MesD or LRP5/6 in controlling adult intestinalhomeostasis has not been demonstrated because of lack of ap-propriate genetic models.Here, we report that grp94 interacts with MesD, which is

necessary for proper cell surface expression of LRP6 and ca-nonical Wnt signaling. Consistent with a role for grp94 in Wntsignaling, proliferation in intestinal crypts and intestinal integrityin mice are critically dependent on grp94.

Resultsgrp94 Interacts with MesD and Chaperones LRP6. In an effort toexpand the client network of grp94, we took an unbiased approachto immunoprecipitate grp94 clientale complex from a preleukemiaB-cell line, 14.GFP (9). We then resolved the complex on SDS/PAGE and stained with Coomassie blue.We subjected the proteinbands that were specifically associated with grp94 pull-down totrypsin digestion, followed by tandemmass spectrometry (MS/MS)to identify grp94-associated proteins. We unexpectedly discoveredMesD as a grp94-interacting molecule (Fig. 1A). MesD, so namedafter its critical role in mesoderm formation during early em-bryogenesis, has been shown to be a critical chaperone for thesurface expression of LRP5/6 (22). To confirm the interactionbetween grp94 andMesD, we FLAG-taggedMesD and expressedMesD-FLAG in HEK293 cells, followed by a coimmunoprecipi-tation study. Indeed, a strong interaction between the two wasdemonstrated (Fig. 1 B and C). Similarly, we found that grp94interacts with LRP6 (Fig. 1D), which is a coreceptor for the cellsurface Wnt receptor Frizzled and is required for canonical Wntsignaling (23). After maturation and surface expression, LRP6undergoes γ-secretase–dependent regulated intramembrane pro-teolysis (RIP) to liberate its extracellular and intracellular domain(24). Indeed, in wild-type (WT) mouse embryonic fibroblasts(MEFs) (Fig. 1E), LRP6 undergoes RIP to release its extracel-lular domain (Fig. 1F). There are two forms of the full-lengthLRP6 inWT cells: Endoglycosidase H (Endo H)-resistant surfaceLRP6 and Endo H-sensitive LRP6 in the ER (Fig. 1F). However,we found that in grp94 knockout (KO) cells, LRP6 does not un-dergo RIP or transport to the cell surface, as indicated by thesensitivity of LRP6 to Endo H in grp94 KO cells (Fig. 1F). Infurther support of this conclusion, surface biotinylation followedby avidin pull-down and immunoblot failed to detect LRP6 on thesurface of grp94 KO cells (Fig. 1G), although the same methoddetected similar amounts of transferrin receptor and Wnt re-ceptor frizzled-4 (Fzd4) from WT and KO cells (Fig. 1 H and I).Finally, we tested WT and grp94 KO cells for their abilities to

Author contributions: B.L., M.S., F.H., D.W., and Z.L. designed research; B.L., M.S., F.H.,B.X.W., S.S., C.M., and I.K. performed research; B.L., M.S., F.H., B.X.W., S.S., C.M., C.E.C.,I.K., D.W., and Z.L. analyzed data; and B.L., M.S., S.T., and Z.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: zihai@musc.edu.

www.pnas.org/cgi/doi/10.1073/pnas.1302933110 PNAS | April 23, 2013 | vol. 110 | no. 17 | 6877–6882

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

respond to a known Wnt ligand, Wnt-3a. Axin2 is a negativeregulator of the Wnt-signaling pathway and is up-regulated inresponse to Wnt ligand. As expected, Wnt-3a treatment led to

a dose-dependent up-regulation of Axin2 mRNA in WT cells butnot in KO cells (Fig. 1J).To determine the roles of grp94 in LRP6-MesD interaction,

we expressed the both LRP6 and MesD in WT or grp94knockdown HEK293 cells, followed by examination of thecomplex formation between the two. We found that knockingdown grp94 resulted in a significant reduction of LRP6-MesDcomplex (Fig. 2).

Conditional Deletion of grp94 Results in Loss of Intestinal Homeostasisand Reduced β-Catenin Nuclear Translocation. Intestinal homeostasisis critically dependent on the Wnt/β-catenin signaling pathway(25). For example, genetic deletion of β-catenin (26), LRP5/6 (27),and transcription factor 4 (Tcf4) (28) in mice results in profoundloss of intestinal villus structure and death. If grp94 participatesin Wnt coreceptor biogenesis, we would expect to see a similarintestinal phenotype in grp94 KO mice. We addressed thishypothesis using tamoxifen-inducible knockout strategy bycrossing Hsp90b1flox/flox mice with a Rosa26ERT-cre mouse (29).Comparing with WT mice, we found that the KO mice experi-enced rapid weight loss, diarrhea, and hematochezia and ulti-mately death 12–14 d post-tamoxifen injection (PTI) in 100% ofmore than 200 mice studied (Fig. 3A). Gross pathological exam-ination revealed significant intestinal dilatation, edema and hem-orrhage, fecal obstruction, and thickening of the intestinal wall(Fig. 3B). Likely because of differences in the turnover rate ofepithelium and the villus length (30), disease was often morepronounced in the terminal ileum, although the duodenum andjejunum were also affected to varying degrees without significantpathology in the colon. Like other Wnt-pathway KO mice, thepathology of the large bowel came later and was not easily ob-served because of death of the mice. The histopathological anal-ysis revealed marked necrosis, neutrophil infiltration, and frankloss of intestinal villi and crypts in the ileum of grp94 KO mice(Fig. 3C). The pathology was apparent as further demonstrated bythe pathology score to combine the degree of architectural lossand neutrophil infiltration (Fig. 3D). Importantly, loss of the gutepithelium is not limited to enterocytes. We observed near-totalloss of Paneth cells by lysozyme immunohistochemistry (Fig. 3E).By BrdU pulsing, we found that there was significant reduction ofBrdU uptake in the KO crypts (Fig. 3F), consistent with growtharrest and loss of β-catenin signaling (26). Remarkably, grp94 KOileum had an almost complete loss of full-length LRP6 proteinexpression 12 d PTI (Fig. 3G), likely reflecting the increasedsensitivity of unfolded LRP6 to protease-rich environment in thegut. The expression of E-cadherin, an important structural proteinin the intestine (31), was not affected (Fig. 4 A and B). As ex-pected, ultrastructural analysis of the grp94-null ileum revealed thecomplete loss of microvilli and Paneth cells (Fig. 4C). The severecompromise of the gut integrity was also evident from the obser-vation of loss of gap junction protein ZO1 (Fig. 4D) and absenceof Goblet cells by Alcian blue stain (Fig. 4E). Consequently,

Fig. 1. grp94 interacts with MesD and is a critical molecular chaperone forLRP6. (A) Identification ofMesD as a grp94-interacting protein.MS/MS analysisof grp94-interacting proteins identified MesD. The sequence in red reflectsregions authenticated by MS. (B) Immunoblot of grp94 and MesD-FLAG fol-lowing immunoprecipitation of MesD-FLAG from HEK293 cell lysates. (C) Im-munoblot of grp94 and MesD-FLAG following immunoprecipitation of grp94from HEK293 cell lysates. Immunoprecipitation with isotype control antibody(ISO) is also shown. Expression of grp94 and MesD in whole cell lysate (WCL) isindicated. Data are representative of two independent experiments. (D) Im-munoblot of LRP6 and grp94 following immunoprecipitation of grp94 fromlysates of WT or grp94 KO MEFs. (E) Immunoblot analysis of grp94 from WTand grp94 KO MEF lysates. (F) Expression of endogenous LRP6 and its sensi-tivity to N-glycase Endo H and peptide N glycosidase F (PNGaseF) in WT andgrp94 KO MEFs. Asterisk indicates cleaved LRP6. β-Actin served as a loadingcontrol. (G) WT or grp94 KO MEFs (duplicates) were subjected to surfacebiotinylation and pull-down with avidin beads, followed by immunoblot forcell surface LRP6. Total LRP6 was also blotted as a control. (H) Immunoblot ofvarious proteins from the total lysates ofWT and grp94MEFs. After incubationof WT and grp94 mutant cells with EZ-Link Sulfo-NHS-SS-Biotin, lysates fromcells were examined by immunoblot analysis for LRP6, transferrin receptor(Tfr), grp94, Fzd4, and tubulin. (I) Cell surface biotinylation of WT and grp94KO MEFs. Lysates from cells treating with EZ-Link Sulfo-NHS-SS-Biotin wereimmunoprecipitated with NeutrAvidin beads and harvested. The pull-down–biotinylated membrane surface proteins were analyzed by immunoblotting.(J) Expression of Wnt target gene, Axin2, in response to Wnt stimulation. WTand grp94 KOMEFswere stimulated withWnt-3a for 24 h, and fold expressionof the Axin2 relative to unstimulated cells was measured by quantitative PCR.Data are representative of more than two independent experiments. Fig. 2. Knockdown of grp94 compromises LRP6-MesD interaction. LRP6-myc

and MesD-FLAG were expressed in WT or grp94 knockdown (KD) HEK293cells. Expression of indicated proteins in the whole-cell lysate of WT and KDcells was determined by immunoblot (IB). IP of LRP6-myc or MesD was thenperformed, followed by SDS/PAGE and IB for MesD-FLAG, grp94, and LRP6-myc. Two independent experiments were performed with similar findings.

6878 | www.pnas.org/cgi/doi/10.1073/pnas.1302933110 Liu et al.

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

systemic dissemination of bacteria to the blood, liver, andmesenteric lymph node was evident with KO but not WT mice(Fig. 4F).To determine whether loss of canonical Wnt signaling is re-

sponsible for the observed gut phenotype, we subjected grp94 KOmice to treatment with glycogen synthase kinase 3 beta (GSK3β)inhibitor TWS119 (32). Inhibition of GSK3β is expected to lib-erate β-catenin from its destruction complex and activate down-stream Wnt target genes in a Wnt receptor/LRP6-independentmanner. We found indeed that TWS119 treatment rescued thegut pathology in KO mice, which correlated with increased cryptcell proliferation (indexed by Ki67) and restored β-catenin nu-clear translocation in the intestinal epithelial cells (Fig. 5A).TWS119 treatment significantly improved the pathology score(Fig. 5B) and resulted in restoration of the Paneth cells in thecrypt (Fig. 5C).

Intestine-Specific Knockout of grp94 Recapitulates the Gut Pathology.grp94 deletion in the hematopoietic system can cause gran-ulocytosis and defective lymphopoiesis (12). To rule out thepossibility that KO immune system contribute indirectly to the

development of the gut pathology, we reconstituted grp94 KOrecipients with WT bone marrow (12). Over 90% chimerism ofthe gut-associated lymphoid tissues was confirmed by congenicmarker (Fig. 6A). We found that WT hematopoietic system wasunable to rescue the KO gut phenotype (Fig. 6B). Moreover,administration of broad-spectrum antibiotics, which effectivelyeliminated over 90% of the gut flora (33), failed to alter theseverity or kinetics of the intestinal disease (Fig. 6C), suggestingthat grp94 plays a critical and gut-intrinsic role in gut homeo-stasis via regulation of the canonical Wnt-signaling pathway. Tofurther address this hypothesis, we crossed Hsp90b1flox/flox micewith a Villincre transgenic mouse (34) to allow for gut-specificdeletion of grp94. Canonical Wnt pathway in the gut is de-pendent on the transcriptional activation complex between nu-clear β-catenin and Tcf4, to transactivate downstream genes.Tcf4 plays indispensable roles in maintaining the crypt stem cellsof the small intestine (28). Similar to Tcf4 KO mice, we foundthat gut-specific deletion of grp94 does not affect embryogenesis.The KO mice were born at expected Mendelian ratios (Fig. 6D).However, the intestinal epithelium-specific deletion of grp94 wasassociated with postnatal death of mice (Fig. 6D). The only 2

Fig. 3. Deletion of grp94 in mouse results in compromise of gut homeo-stasis and loss of Wnt coreceptor LRP6. (A) Percent weight changes in WTand grp94 KO mice 12 d PTI. (B) Gross pathology showing bowel dilatation(∼three times of the diameter of the WT mice), obstruction, edema, andhemorrhage in the small intestine of grp94 KO mice (one representativemouse over more than 120 is shown). (C) H&E staining of gut sections ofgrp94 WT and grp94 KO mice 12 d PTI. The KO mice have fewer and shortervilli, especially in the ileum, which shows almost complete loss of villi withmarked reduction of crypt. Multiple experiments (more than 30) were donewith similar findings. (Scale bar: 100 μm.) (D) Quantification of gut pathol-ogy of KO (n = 5) and WT (n = 6) mice 12 d PTI. *P < 0.05. (E) Lysozyme stainof Paneth cells in the crypts of the ileum section. There was loss of crypt-villus structure and absence of Paneth cells in the KO ileum. (F) Immuno-fluorescence for BrdU and DAPI in the ileum of untreated (UT) and 12-dtamoxifen-treated mice. (Scale bar: 100 μm.) Comparing to the ileum ofuntreated mice, the tamoxifen-treated mice show significant loss of villi andcrypts with markedly decreased BrdU uptake in the crypt. (G) Immunoblotfor endogenous LRP6 at day 0 (WT mice) and day 12 (grp94 KO mice). grp94and β-actin (loading control) expression is shown.

Fig. 4. Loss of intestinal barrier function and bacterial translocation in grp94KO mice. (A) Normal expression of E-cadherin in ileum of grp94 KO mice.Immunofluorescence microscopy for E-cadherin (green) on day 0 (untreated)(WT) and day 12 (D12) grp94 KO mice. DAPI staining is shown (blue). (Scalebar: 100 μm.) (B) Western blot analysis of E-cadherin in colon and ileum ofWTand KOmice. grp94 and β-actin (loading control) expression is shown. Data arerepresentative of multiple mice per group and experiments. (C) Ultrastructuralanalysis of small intestine of grp94 KO mice. Hsp90b1flox/wtR26RERT-cre (WT)and Hsp90b1flox/foxR26RERT-cre (KO) mice were treated for 12 d with tamoxi-fen. Distal ileum was dissected, carefully rinsed with PBS, and immediatelyfixed in 4% paraformaldehyde in PBS at 4 °C before processing for trans-mission electron microscope. LD, lipid droplets; MV, microvilli; P, Paneth cells.Loss of MV and Paneth cells can be readily seen. Data are representative offour mice per group. (D) Loss of gap junction protein Z01 in the KO ileum byimmunofluorescence. (Scale bar: 100 μm.) (E) grp94 deletion leads to significantreduction of Goblet cells in all mice examined. (F) Presence of bacteria in cul-tured tissue lysate from mesenteric lymph nodes (MLN), liver (LIV), and blood(BLD) of grp94KOmice. Representative images from each condition are shown.Numbers underneath these images represent average number of colonies afterovernight culture from six WT and five KO mice, followed by the frequency ofmice with positive bacteria culture in parenthesis.

Liu et al. PNAS | April 23, 2013 | vol. 110 | no. 17 | 6879

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

survived mice out of 12 expected KO mice did so because of onlypartial knockdown of grp94. No mice with complete KO of grp94survived. Examination of the newborn mice euthanized beforedeath demonstrated the concurrent failure of β-catenin nucleartranslocation (Fig. 6E). Histological analysis demonstrated de-creased number of villi, significant reduction of cells in the cryptregions between the villi, decreased mitosis figures, and frequent“lifting” of the epithelial layer, which is reminiscent of Tcf4 KOmice (28). The complete phenocopy of gut-specific grp94 KOmice with Tcf4 KO mice led us to further conclude that the gutpathology associated with grp94 deletion is attributable to gut-intrinsic loss of Wnt signaling.

DiscussionHerein, we report the surprising and indispensable role for a majorER-resident molecular chaperone, grp94, in chaperoning LRP6and in intestinal homeostasis. grp94 binds to MesD and LRP6, andit plays a critical role for the maturation and surface expression ofLRP6. The essential roles of grp94 in LRP6 expression andfunction in canonical Wnt pathway are established both in vitroand in vivo.We demonstrated that grp94 interact with both LRP6 and the

ER resident “chaperone” MesD (22). MesD does not appear toplay any other role in ER function besides its unique function inpromoting LRP5/6 surface expression. Interestingly, grp94 alsoappears to play an important role in mesoderm formation duringdevelopment (35). Thus, our data strongly indicate that full LRP6maturation and cell surface expression is dependent on the co-ordinated action of both grp94 and MesD. The conditional grp94

KO mouse now provides a unique experimental system to studythe roles of LRPs in the biology of intestinal homeostasis in bothphysiological and pathological conditions.The proliferation and differentiation of the intestinal epithe-

lium are tightly controlled by the Wnt/β-catenin pathway. Theloss of enterocytes/villi and crypts in grp94 KO intestine closelyresembled that of mouse models of gut-specific deletion ofβ-catenin (26) and LRP5/6 (27). Both β-catenin and LRP5/6conditional KO mice developed severe intestinal pathology ∼4 dafter deletion. grp94 KO mice developed pathology around 10 dPTI, reflecting the requirement for 1 additional week to deletegrp94. Our focus on LRP5/6 also came from two other obser-vations: that grp94 was found to complex with MesD and thatLRP6 has some structural similarities to integrins, a knownfamily of grp94-dependent client proteins (11). We found, in-deed, that LRP6 expression in the intestine was significantlycompromised in the absence of grp94. More importantly, LRP6in grp94 KO cells is unable to acquire Endo H resistance andfails to export to cell surface, indicating that LRP6 is trapped inthe ER in the absence of grp94. We also demonstrated thatgrp94 is required for optimal interaction between LRP6 andMesD. Further study is necessary to understand the precise

Fig. 5. β-Catenin activation rescues the intestinal pathology in grp94 KOmice. (A) After 10 d of treatment with DMSO or tamoxifen (TAM) with orwithout TWS119 (TWS), ileum tissues were stained for grp94, β-catenin, andKi67 by immunohistochemistry. Shown also is an H&E-stained section. (Scalebar: 100 μm.) The marked ileum injury (mucosal ulceration, loss of villi, anddecreased crypts) in TAM mice is reversed on H&E stain, and the immuno-histochemical staining pattern of indicated proteins are restored to thatof DMSO-treated control mice with TWS119 treatment (TAM+TWS). (B)Quantification of gut pathology of KO mice treated with indicated con-ditions (n = 5 per group). *P < 0.05. (C) Restoration of crypt-villus structureand appearance of Paneth cells (arrows) in the crypts of the ileum section ofTAM-treated KO mice with TWS119.

Fig. 6. Gut-intrinsic roles of grp94 in gut homeostasis. (A) Flow-cytometricanalysis of donor cells based on congenic marker CD45.2. (B) H&E staining ofileum sections of WT→WT and WT→KO bone marrow chimeric mice 12 d PTI.(Scale bar: 10 μm.) (C) H&E staining of ileum sections of grp94WT and KOmicetreated with antibiotics before deletion of grp94 15 d PTD. Data are repre-sentative of two independent experiments with multiple mice (n > 5) pergroup. (Scale bar: 25 μm.) (D) Loss of intestine-specific grp94 KO mice duringthe postnatal period. Villin-cre+/−Hsp90b1flox/WT mice were crossed with Villin-cre−/−Hsp90b1flox/flox mice. The genotypes of newborn mice, as well as that ofthe postweaning adult mice, were determined. The observed frequency ofVillin-cre+/−Hsp90b1flox/flox adult mice was significantly reduced from theexpected value of 25%. *P < 0.05. (E) H&E and immunohistochemistry stainingof grp94 and β-catenin of ileum sections from the villincreHsp90b1flox/flox andthe control villincreHsp90b1flox/WT mice. (Scale bar: 25 μm.) H&E stain of theileum section of the KO newborn mice shows shorter villi with decreased cellnumber in intervillus crypt region (arrows) and mitosis, indicating less pro-liferation activity. Immunohistochemistry stain reveals complete loss of grp94expression in KO ileum with concurrent loss of β-catenin nuclear translocationcompared with WT mice. A representative image from four mice is shown.

6880 | www.pnas.org/cgi/doi/10.1073/pnas.1302933110 Liu et al.

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

molecular mechanism of grp94-MesD-LRP interaction and itsregulation during steady-state and pathological conditions.grp94 is an endoplasmic reticulum resident HSP and a master

chaperone for TLRs and integrins (8–11, 13, 36). We also con-sidered the possibility that loss of grp94-dependent client pro-teins TLRs and/or integrins could partially contribute to thedevelopment of the intestinal pathology in grp94 KO mice.However, mice that lack the TLR-signaling molecules myeloiddifferentiation primary response 88 (MyD88) or TIR-domain-containing adapter-inducing interferon beta (TRIF), althoughmore sensitive to experimental colitis, do not develop sponta-neous enteritis (37). Although loss of β1 integrin in the intestineresults in the development of colitis (38), β1 integrin expressionis not dependent on grp94 (12). Moreover, many of the integrinα-chains that pair with β1 are not dependent on grp94 either,with the exception of α2 (11). α2 knockout mice do not developgut disease, and although α2 is expressed in the mouse intestine,the specific function of α2 in the gut is unknown. Neither TLRsnor grp94-dependent integrins have been directly linked to thegut homeostasis intrinsically. Most importantly, we demonstratedthe complete phenocopy of gut-specific grp94 KO mice with Tcf4KO mice and that the gut phenotype associated with grp94 losscan be rescued by activation of canonical Wnt-signaling pathwayvia GSK3β inhibitor. Thus, the pathogenesis of gut diseases ingrp94 KO mice is primarily attributable to the loss of gut-in-trinsic canonical Wnt pathway, rather than the lack of otherfunctional clients of grp94.In summary, we have uncovered a function of a major ER lu-

minal HSP grp94 in chaperoning LRP6 and Wnt signaling. Giventhe importance for Wnt in tumorigenesis and other aspects ofdevelopment (25, 39), a previously unappreciated and critical rolefor grp94 in the Wnt signaling may have fundamental implicationsin linking the general ER protein homeostasis with Wnt signaling.Elucidation of the precise mechanism of grp94 in chaperoningLRP6 will also facilitate grp94-targeted therapeutics for cancer.

MethodsMice and Genotyping. Conditional Hsp90b1-deficient mice were describedpreviously (10, 11, 40). villincre mice were purchased from The Jackson Labo-ratory. Animal use was approved by the Medical University of South CarolinaAnimal Care Committee.

Plasmids. grp94 constructs in MigR were described previously (10). HumanLRP6-myc was a generous gift from Cristof Neihrs (Deutsches Krebsfor-shungszentrum, Heidelberg) and Yonghe Li (Southern Research Institute,Birmingham, AL). Mouse MesD-FLAG and mouse LRP6-myc plasmids werea generous gift from Janet Lighthouse and Bernadette Hoeldner (StonyBrook University, Stony Brook, NY). Site-directed mutagenesis was per-formed on parental grp94/MigR using a commercially available kit (Stra-tagene) and verified by sequencing. Transient transfection of HEK293 cellswas performed using 5–10 μg of plasmid DNA and Lipofectamine 2000. Cellswere analyzed 48 h after transfection.

Cell Lines. HEK293 cells were originally obtained from the ATCC. WT grp94and tamoxifen-inducible KO MEFs were generated and immortalized usingtransduction of large-T antigen retrovirus. After cloning, MEF cells were eithertreated with ethanol (vehicle) or 1 mM hydroxy-tamoxifen (Sigma-Aldrich) togenerate grp94-null MEFs. MEF cells were maintained as WT or grp94 KO.

Antibodies. Most antibodies were purchased from Sigma-Aldrich except theantibodies against the following antigens: Lysozyme (Abcam), β-cat (CellSignaling), grp94 (Enzo Life Sciences), Ki67 (Nova Biologicals), ZO1 (Abcam),BrdU (Invitrogen), transferrin (Invitrogen), and LRP6 (Cell Signaling).

Tamoxifen-Inducible Deletion of grp94 and Model of Acute Intestinal Diseaseand Bone Marrow Chimeras. grp94 KO mice and bone marrow chimeras weredescribed previously (12). Tamoxifen citrate (Sigma-Aldrich) was dissolved at10 mg/mL in peanut or corn oil and further diluted to 1 mg/mL for injections.Low-dose tamoxifen (5 μg/g body weight) was injected intraperitoneally toHsp90b1flox/wtR26RERT-cre (WT), Hsp90b1flox/flox, or Hsp90b1flox/foxR26RERT-cre

(KO) mice for 0–12 consecutive days. For rescue experiment, GSK3β inhibitor

TWS119 (Cayman Chemical) was administered intraperitoneally after day 3at 15 mg/kg body weight daily for 7 d (32).

Antibiotic Treatment and Commensal Depletion. Mice were treated with 1 g/Lampicillin (Sigma), 1 g/L neomycin (Sigma), 1 g/L metronidazole (Sigma), and0.5 g/L vancomycin (RPI Corp) in the drinking water for 1 mo. Commensaldepletion was verified by fecal plating on LB agar without antibiotics (33).Water was changed every third day, and antibiotic treatment was continuedthroughout tamoxifen treatment. Tamoxifen was administered to mice at5 μg/g body weight as above. Mice were killed at day 12, or their survival wastracked indefinitely. Bacterial load from mice was quantified by inoculating300 μL of diluted blood (1:10 in PBS), homogenate of liver (50 mg/mL), ormesenteric lymph node (25 mg/mL) to modified trypticase soy agar (TSA II)plates with 5% (vol/vol) sheep blood (BD Biosciences), followed by overnightculture at 37 °C and counting of bacterial colonies.

Reverse Transcription and Quantitative PCR. Approximately 1-cm segments ofintestinal tissue were snap frozen and stored at −80 °C. RNAwas extracted byTRIzol (Gibco) method. cDNA was made by reverse-transcription PCR usingSuperScript polymerase II (Invitrogen). Quantitative PCR was performed us-ing Sybr Green (Applied Biosystems) method using the primer sets. Expres-sion level was calculated using the formula 2−ΔCT and multiplied by a factorof 10. β-Actin was used as an internal control.

Histopathology, Immunofluorescence, and Immunohistochemistry. Approxi-mately 1-cm segments of tissue were embedded in OCT freezing medium(Thermo Scientific) and immediately frozen on dry ice or fixed in 4%formaldehyde/PBS, rehydrated in 30% sucrose/PBS, before embedding in OCTmedium. Samples were stored at −80 °C until sectioning. Five to 7-μm sec-tions were cut on a cryostat onto poly-L lysine-coated slides (Sigma). Forhistological examination, slides were immediately stained with hematoxylin/eosin (H&E). Gut pathology score is defined by summation of two parameterscores to achieve a range of 0–6: inflammation (0, no inflammation; 1, neu-trophil rarely discernible; 2, presence of neutrophils in every high powerfield; 3, collection of >10 neutrophils in any given area); tissue destruction(0, normal structure; 1, villus length shortened by 50%; 2, flattened mucosawith loss of villi; 3, evidence of submucosa ulceration). For staining Gobletcells, fixed tissue was stained with Alcian blue solution for 20 min at roomtemperature and then rinsed thoroughly in tap water, followed by coun-terstain with Kernechtrot’s nuclear fast red solution for 5 min. After rinsingin distilled water, the tissue was dehydrated and mounted. For immuno-fluorescence, slides were fixed in ice-cold acetone. Alternatively, for grp94intracellular staining, slides were fixed in 4% formalin/PBS for 20 min andpermeabilized with ice-cold methanol for 20 min. Slides were washed in PBS;Fc receptor was blocked and/or slides were blocked in 10% goat serum/PBSfor 30 min. Primary antibodies were diluted in 2% BSA/PBS, and slides werestained for 45 min. After washing in PBS, slides were stained with secondaryantibody (or directly conjugated antibodies) diluted in 2% BSA/PBS for30 min. Slides were washed again and counterstained with nuclear stain,DAPI. For BrdU staining, fixed-tissue samples were cut into 10-μm sectionsand fixed a second time in 10% formalin/PBS for 30 min. After washing,slides were treated in 2 M HCl/H2O for 20 min at 37 °C. Slides were rinsed inPBS and neutralized in two washes in 0.1 M sodium borate buffer (pH 8.5).After thorough washing, slides were stained with anti-BrdU antibody (MoBU-1or ZBU-30; 1:25; Invitrogen).

Electron Microscopy. Tissues were dissected and fixed in 4% buffered para-formaldehyde, osmicated, stained in block with uranyl acetate, dehydrated,and embedded in Poly/Bed resin. Thick 1-μm sections were cut on a Leica EMUC7 ultramicrotome and examined in the light microscope. Thin 70- to80-nm sections were cut and collected on 200 Cu/Rh mesh grids, stained withuranyl acetate and lead citrate, and observed in a Hitachi H-7650 trans-mission electron microscope.

BrdU Pulse.Mice were injected intraperitoneally with BrdU (1 mg) in PBS 24 hbefore euthanasia.

Protein Extraction, Immunoprecipitation, and Western Blot. Protein extraction,immunoprecipitation, and Western blot were carried out as described pre-viously (10). Briefly, HEK293 or MEF cells were harvested by trypsin-EDTA or5 μM EDTA, subjected to dithiobis (succinimidyl propionate) (Thermo Scientific)for 30 min at room temperature, washed in PBS, and lysed on ice in radio-immunoprecipitation assay (RIPA) lysis buffer plus protease inhibitor mixture(Sigma-Aldrich). Nuclear-free protein lysate was quantified by Bradford assay,

Liu et al. PNAS | April 23, 2013 | vol. 110 | no. 17 | 6881

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

and an equal amount of lysate was incubated with antibodies or isotypecontrol as indicated. grp94 was immunoprecipitated using 9G10 antibody(Stressgen) and Protein-G beads (Pierce), and MesD-FLAG was immunopreci-pitated using anti-FLAG antibody (BioM2; Sigma) and Neutravidin beads(Thermo Scientific).

In-Gel Digestion and Nano–LC-MS/MS Analysis of grp94-Associated Proteins.grp94 immunoprecipitates were resolved on SDS/PAGE. In-gel digestion ofunique bands was performed using trypsin, followed by microcapillary HPLCand LC-MS/MS analysis on the linear trap quadrupole. The obtained MS/MSdata were subjected to database searches of mouse proteome using SEQUESTsoftware, as described (41).

Cell Surface Biotinylation. After washing cells three times with 1× PBS CM[10 mM potassium phosphate (pH 7.5), 140 mM NaCl, 0.1 mM CaCl2, and1 mM MgCl2], cells were incubated for 30 min with the 1× PBS CM con-taining 0.5 mg/mL of EZ-Link Sulfo-NHS-SS-Biotin (Thermo Fisher Scientific;catalog no. 21331) on ice with rocking. The biotinylation reactions werethen incubated with 1× PBS CM containing 50 mM NH4Cl on ice for 5 min toquench free biotin. Then, the cells were washed three times with 1× PBS CMand lysed in lysis buffer containing 1.25% Triton X-100, 0.25% SDS, 50 mMTris·HCl (pH 8.0), 150 mMNaCl, 5 mM EDTA, 5 mg/mL iodoacetamide, 10 μg/mLAPMSF, and protease inhibitor mixture (Roche). Whole-cell lysates werecentrifuged at 15,000 × g at 4 °C for 15 min, and the supernatant was collectedand incubated with NeutrAvidin beads (Thermo Fisher Scientific; catalog no.29200) at 4 °C overnight while rotating. The NeutrAvidin beads were centri-fuged at 1,000 × g at 4 °C for 3 min and washed with wash buffer [0.5% TritonX-100, 0.1% SDS, 50 mM Tris·HCl (pH 8.0), 150 mM NaCl, 5 mM EDTA). The

proteins in NeutrAvidin beads were eluted with SDS sample buffer and sub-jected to Western blot analysis.

Wnt Signaling. WT and KO MEF cells were seeded at 1 × 105 cells per well ina 12-well plate the night before. The next day, cells were either stimulatedwith Wnt 3a (Peprotech) at the indicated dose or left unstimulated. Quan-titative PCR for Axin2 was performed using the following primers: 5′-TGACTCTCCTTCCAGATCCCA-3′ (forward) and 5′-TGCCCACACTAGGCTGACA-3′(reverse). Relative fold expression was calculated by the ΔΔCT method relativeto β-actin control using the following formula: 2−ΔΔCT.

ELISA. Serum was collected at the time of euthanasia. TNFα, IL-6, and IL-12p40ELISAs (BD Bioscience) were performed according to themanufacturer’s protocol.

Statistics. Error bar represents standard deviation. The Student’s t test or χ2

test was used to determine whether the difference between two groups wasstatistically significant (P < 0.05).

ACKNOWLEDGMENTS. We thank the past and present members of ourlaboratories for their input throughout the course of this work. B.L. is aNational Institutes of Health (NIH) KL2 scholar and is supported by the SouthCarolina Clinical and Translational Research Institute at the Medical Universityof South Carolina (NIH Grants KL2RR029880 and UL1RR029882). The work wasalso supported, in part, by the Flow Cytometry and Cell Sorting SharedResource, Hollings Cancer Center, Medical University of South Carolina (NIHGrant P30 CA138313). Z.L., C.E.C., S.T., and D.W. are supported by NIH grants.Z.L. is the Abney Chair Remembering Sally Abney Rose in Stem Cell Biologyand Therapy.

1. Lee AS, Delegeane A, Scharff D (1981) Highly conserved glucose-regulated protein inhamster and chicken cells: Preliminary characterization of its cDNA clone. Proc NatlAcad Sci USA 78(8):4922–4925.

2. Srivastava PK, DeLeo AB, Old LJ (1986) Tumor rejection antigens of chemically inducedsarcomas of inbred mice. Proc Natl Acad Sci USA 83(10):3407–3411.

3. Chen B, Piel WH, Gui L, Bruford E, Monteiro A (2005) The HSP90 family of genes inthe human genome: Insights into their divergence and evolution. Genomics 86(6):627–637.

4. Kozutsumi Y, Segal M, Normington K, Gething MJ, Sambrook J (1988) The presence ofmalfolded proteins in the endoplasmic reticulum signals the induction of glucose-regulated proteins. Nature 332(6163):462–464.

5. Li Z, Srivastava PK (1993) Tumor rejection antigen gp96/grp94 is an ATPase: Im-plications for protein folding and antigen presentation. EMBO J 12(8):3143–3151.

6. Frey S, Leskovar A, Reinstein J, Buchner J (2007) The ATPase cycle of the endoplasmicchaperone Grp94. J Biol Chem 282(49):35612–35620.

7. Dollins DE, Warren JJ, Immormino RM, Gewirth DT (2007) Structures of GRP94-nucleotide complexes reveal mechanistic differences between the hsp90 chaperones.Mol Cell 28(1):41–56.

8. Morales C, Wu S, Yang Y, Hao B, Li Z (2009) Drosophila glycoprotein 93 Is an orthologof mammalian heat shock protein gp96 (grp94, HSP90b1, HSPC4) and retains disulfidebond-independent chaperone function for TLRs and integrins. J Immunol 183(8):5121–5128.

9. Randow F, Seed B (2001) Endoplasmic reticulum chaperone gp96 is required forinnate immunity but not cell viability. Nat Cell Biol 3(10):891–896.

10. Yang Y, et al. (2007) Heat shock protein gp96 is a master chaperone for toll-like re-ceptors and is important in the innate function of macrophages. Immunity 26(2):215–226.

11. Liu B, Li Z (2008) Endoplasmic reticulum HSP90b1 (gp96, grp94) optimizes B-cellfunction via chaperoning integrin and TLR but not immunoglobulin. Blood 112(4):1223–1230.

12. Staron M, et al. (2010) gp96, an endoplasmic reticulum master chaperone for in-tegrins and Toll-like receptors, selectively regulates early T and B lymphopoiesis.Blood 115(12):2380–2390.

13. Liu B, et al. (2010) Folding of Toll-like receptors by the HSP90 paralogue gp96 requiresa substrate-specific cochaperone. Nat Commun 1:79.

14. Christianson JC, Shaler TA, Tyler RE, Kopito RR (2008) OS-9 and GRP94 deliver mutantalpha1-antitrypsin to the Hrd1-SEL1L ubiquitin ligase complex for ERAD. Nat Cell Biol10(3):272–282.

15. Zhao R, et al. (2005) Navigating the chaperone network: An integrative map ofphysical and genetic interactions mediated by the hsp90 chaperone. Cell 120(5):715–727.

16. van der Flier LG, Clevers H (2009) Stem cells, self-renewal, and differentiation in theintestinal epithelium. Annu Rev Physiol 71:241–260.

17. Tamai K, et al. (2000) LDL-receptor-related proteins in Wnt signal transduction.Nature 407(6803):530–535.

18. He X, Semenov M, Tamai K, Zeng X (2004) LDL receptor-related proteins 5 and 6 inWnt/beta-catenin signaling: Arrows point the way. Development 131(8):1663–1677.

19. Goel S, et al. (2012) Both LRP5 and LRP6 receptors are required to respond to phys-iological Wnt ligands in mammary epithelial cells and fibroblasts. J Biol Chem 287(20):16454–16466.

20. Kelly OG, Pinson KI, Skarnes WC (2004) The Wnt co-receptors Lrp5 and Lrp6 are es-sential for gastrulation in mice. Development 131(12):2803–2815.

21. van Amerongen R, Berns A (2006) Knockout mouse models to study Wnt signaltransduction. Trends Genet 22(12):678–689.

22. Hsieh JC, et al. (2003) Mesd encodes an LRP5/6 chaperone essential for specification ofmouse embryonic polarity. Cell 112(3):355–367.

23. Haegebarth A, Clevers H (2009) Wnt signaling, lgr5, and stem cells in the intestine andskin. Am J Pathol 174(3):715–721.

24. Mi K, Johnson GV (2007) Regulated proteolytic processing of LRP6 results in release ofits intracellular domain. J Neurochem 101(2):517–529.

25. Clevers H (2006) Wnt/beta-catenin signaling in development and disease. Cell 127(3):469–480.

26. Fevr T, Robine S, Louvard D, Huelsken J (2007) Wnt/beta-catenin is essential for in-testinal homeostasis and maintenance of intestinal stem cells. Mol Cell Biol 27(21):7551–7559.

27. Zhong Z, Baker JJ, Zylstra-Diegel CR, Williams BO (2012) Lrp5 and Lrp6 play com-pensatory roles in mouse intestinal development. J Cell Biochem 113(1):31–38.

28. Korinek V, et al. (1998) Depletion of epithelial stem-cell compartments in the smallintestine of mice lacking Tcf-4. Nat Genet 19(4):379–383.

29. Ventura A, et al. (2007) Restoration of p53 function leads to tumour regression invivo. Nature 445(7128):661–665.

30. Creamer B, Shorter RG, Bamforth J (1961) The turnover and shedding of epithelialcells. I. The turnover in the gastro-intestinal tract. Gut 2:110–118.

31. Nelson WJ, Nusse R (2004) Convergence of Wnt, beta-catenin, and cadherin path-ways. Science 303(5663):1483–1487.

32. Gattinoni L, et al. (2009) Wnt signaling arrests effector T cell differentiation andgenerates CD8+ memory stem cells. Nat Med 15(7):808–813.

33. Liu B, et al. (2006) TLR4 up-regulation at protein or gene level is pathogenic for lupus-like autoimmune disease. J Immunol 177(10):6880–6888.

34. Madison BB, et al. (2002) Cis elements of the villin gene control expression inrestricted domains of the vertical (crypt) and horizontal (duodenum, cecum) axes ofthe intestine. J Biol Chem 277(36):33275–33283.

35. Wanderling S, et al. (2007) GRP94 is essential for mesoderm induction and muscledevelopment because it regulates insulin-like growth factor secretion. Mol Biol Cell18(10):3764–3775.

36. Wu S, et al. (2012) The molecular chaperone gp96/GRP94 interacts with Toll-likereceptors and integrins via its C-terminal hydrophobic domain. J Biol Chem 287(9):6735–6742.

37. Rakoff-Nahoum S, Paglino J, Eslami-Varzaneh F, Edberg S, Medzhitov R (2004)Recognition of commensal microflora by toll-like receptors is required for intestinalhomeostasis. Cell 118(2):229–241.

38. Jones RG, et al. (2006) Conditional deletion of beta1 integrins in the intestinal epi-thelium causes a loss of Hedgehog expression, intestinal hyperplasia, and earlypostnatal lethality. J Cell Biol 175(3):505–514.

39. Klaus A, Birchmeier W (2008) Wnt signalling and its impact on development andcancer. Nat Rev Cancer 8(5):387–398.

40. Staron M, et al. (2011) Heat-shock protein gp96/grp94 is an essential chaperone forthe platelet glycoprotein Ib-IX-V complex. Blood 117(26):7136–7144.

41. Wu L, et al. (2007) Global survey of human T leukemic cells by integrating proteomicsand transcriptomics profiling. Mol Cell Proteomics 6(8):1343–1353.

6882 | www.pnas.org/cgi/doi/10.1073/pnas.1302933110 Liu et al.

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1