Activation of goblet-cell stress sensor IRE1β is controlled by the mucin chaperone AGR2

Abstract

Intestinal goblet cells are secretory cells specialized in the production of mucins, and as such are challenged by the need for efficient protein folding. Goblet cells express Inositol-Requiring Enzyme-1β (IRE1β), a unique sensor in the unfolded protein response (UPR), which is part of an adaptive mechanism that regulates the demands of mucin production and secretion. However, how IRE1β activity is tuned to mucus folding load remains unknown. We identified the disulfide isomerase and mucin chaperone AGR2 as a goblet cell-specific protein that crucially regulates IRE1β-, but not IRE1α-mediated signaling. AGR2 binding to IRE1β disrupts IRE1β oligomerization, thereby blocking its downstream endonuclease activity. Depletion of endogenous AGR2 from goblet cells induces spontaneous IRE1β activation, suggesting that alterations in AGR2 availability in the endoplasmic reticulum set the threshold for IRE1β activation. We found that AGR2 mutants lacking their catalytic cysteine, or displaying the disease-associated mutation H117Y, were no longer able to dampen IRE1β activity. Collectively, these results demonstrate that AGR2 is a central chaperone regulating the goblet cell UPR by acting as a rheostat of IRE1β endonuclease activity.

Keywords: AGR2; Goblet Cells; IRE1β; Mucus Homeostasis; Unfolded Protein Response (UPR).

Figure 1. IRE1β activity is attenuated upon overexpression in goblet cells.

(A) RT-qPCR analysis of ERN2 (IRE1β) and ERN1 (IRE1α) transcript expression in human cell lines. N = 3 culture dishes were sampled and transcript levels are shown relative to the expression in LS174T parental cells. Error bars show SEM. (B) Schematic representation of cell lines used for further studies. First, ERN1−/− clones of the LS174T and Calu-1 parental lines were established by CRISPR/Cas9. Then, ERN1−/− cells were transduced with a TetOn module for doxycycline-controlled IRE1β-FLAG expression. (C) Photographs showing the phenotype of cultures overexpressing IRE1β-FLAG. Left panels show untreated cultures, middle panels show cultures that received 1 μg/ml doxycycline for 72 h (and thus express IRE1β-FLAG), cultures in the right panels received 1 μg/ml doxycycline and 1 μM IRE1 endonuclease inhibitor 4μ8C. Scale bar represents 200 µm. Representative of three independent experiments. (D) Western blot verifying transgene expression in the cell lines used in (C). All remaining adherent cells after 24 h of transgene induction were collected and lysates were probed for FLAG-IRE1β expression via immunoblot. Tubulin was used as a loading control. Multiplicity of Infection (MOI) indicates the theoretical number of viral particles added. (E, F) RT-qPCR analysis of XBP1S/T (E) and BLOC1S1 transcript levels (F) after 24 h of transgene induction. Representative of three independent experiments with three replicates per condition. Error bars show SEM. (E) bottom picture shows XBP1 splicing in the same samples assayed by conventional PCR, representative of two independent experiments. The fast-migrating band is the spliced XBP1 transcript, and the slow-migrating band the XBP1 unspliced transcript. Source data are available online for this figure.

Figure 2. The mucin chaperone AGR2 is a goblet cell-specific interactor of IRE1β.

(A) Verification of transgene expression and successful immunoprecipitation of FLAG-tagged IRE1 in the samples analyzed by MS in (B, C). Lysates were probed for IRE1-FLAG expression using anti-FLAG, and actin was used as a loading control. (B) Proteins with a log₂FC enrichment of >2 and log₁₀Adj P value of >2 for IRE1α-FLAG and IRE1β-FLAG immunoprecipitation (IP) compared to control cells. The Venn diagram shows the number of proteins that were detected uniquely associated with one of the two IRE1 paralogues or that were commonly identified with both IRE1 paralogues. (C) Volcano plot depicting the cutoff criteria and significantly enriched proteins in each IP. X axis shows the log₂ fold change of the measured peptide intensities of a given protein in the control condition over the IRE1 IP condition. Y axis shows the FDR corrected P value obtained by two-sample t test in Perseus. (D) Confirmation of specific interaction between AGR2 and IRE1β, but not IRE1α. IRE1 proteins were tagged with an Avi-tag that is specifically biotinylated upon BirA co-expression. The biotinylated Avi-tag was precipitated using streptavidin beads. For control conditions, BirA was omitted. Blots were probed for co-precipitation of AGR2 and streptavidin to detect Avi-tag-biotinylated IRE1. Tubulin was used as a loading control. Non-specific signal is indicated with an asterisk. Representative of two independent experiments. (E) Confirmation of the AGR2-IRE1β interaction in murine tissue. Colons were isolated and digested, and IP was performed using anti-AGR2. Agr2-deficient mice were used as a negative control to assess whether IRE1β binds specifically to the antibody/bead complex. IP samples were probed for IRE1β co-precipitation via immunoblot. Tubulin was used as a loading control. Representative of two independent experiments. Source data are available online for this figure.

Figure 3. Co-expression of AGR2 dampens endonuclease activity of IRE1β.

Calu-1^{ERN1-/-IRE1βFLAG-DOX} cells were transduced with a constitutive AGR2 transgene (“Calu-1^AGR2“). Cells denoted as “Calu-1” are the original Calu-1^{ERN1-/-IRE1βFLAG-DOX} cells. (A) RT-qPCR analysis of XBP1S/T transcript levels after 24 h of transgene induction using 1 μg/ml doxycycline. Bottom picture shows XBP1 splicing in the same samples assayed by conventional PCR. Representative of four independent experiments with three replicate wells per condition. Error bars show SEM. (B) RT-qPCR analysis of BLOC1S1 transcript levels after 24 h of transgene induction using 1 μg/ml doxycycline. Representative of five independent experiments with three replicates per condition. Error bars show SEM. (C) Photographs showing the phenotype of cultures overexpressing IRE1β-FLAG with and without exogenous expression of AGR2. Left panels show untreated cultures, middle panels show cultures that received 1 μg/ml doxycycline for 72 h, and cultures in the right panels received 1 μg/ml doxycycline and 1 μM 4μ8C. Scale bar represents 100 µm. Representative of three independent experiments. (D) Quantification of cell death in cultures overexpressing IRE1β-FLAG with and without exogenously added AGR2 after 48 h. All cells in the culture dish were stained with Annexin V and Live/Dead stain and analyzed by flow cytometry. All single and double positive cells were considered as dead cells. Representative of two independent experiments with three replicate wells per condition. Error bars show SEM. (E) Analysis of IRE1β-FLAG and AGR2 expression in the cell lines used for (C). Only cells that remained attached in the dish were collected and lysates were probed for IRE1β expression using anti-FLAG and AGR2 expression using anti-AGR2. Tubulin was used as a loading control. (F) RT-qPCR validation of AGR2 knockdown efficiency in LS174T^{ERN1-/-IRE1βFLAG-DOX} cells. NTC is a non-targeting control pool of siRNA’s, #1 and #3 are siRNA’s targeting AGR2. Representative of three independent experiments with n = 3 replicate wells per condition. Error bars show SEM. (G) Log₂ fold changes in XBP1 splicing after AGR2 partial knockdown and/or treatment with 4μ8C or DMSO (vehicle). Splicing is shown as a log₂ fold change of XBP1S mRNA over the NTC/vehicle-treated cells. Representative of three independent experiments with n = 3 replicate wells per condition. Error bars show SEM. (H) Western blot confirmation of (F, G). Proteins were extracted after 72 hours and probed for XBP1S, AGR2 and tubulin expression. Representative of two independent replicates. (I) Quantification of XPP1S protein levels normalized to tubulin in n = 2 experiments represented in (H). Source data are available online for this figure.

Figure 4. AGR2 blocks IRE1β activity through disruption of IRE1β oligomers.

(A) Schematic overview of gel filtration experiments in (B, C). (B) msfGFP fluorescence measured during elution of HEK293T lysates overexpressing IRE1β in the absence of AGR2 (black dotted trace) or in the presence of AGR2 (orange and purple traces indicating different ratios of transfected AGR2:IRE1β plasmid). Top scale represents the approximate elution profile and expected MW of protein standards. Bottom drawings indicate expected oligomerization status based on protein standards and the previously obtained elution profile (Grey et al, 2020). Each trace shows a single chromatography run. (C) IRE1β-FLAG expression in fractions collected after gel filtration of protein lysates from Calu-1^{ERN1-/-IRE1βFLAG-DOX} cells, in the absence or presence of additional AGR2 expression. Line graph shows quantification of band intensities from the gel of three independent replicates. Error bar shows SEM. (D) Schematic representation of competition IP experiments in (E, F). IRE1β is expressed with an Avi-tag or FLAG-tag in equimolar amounts. After biotinylation of the Avi-tag by BirA, both the Avi-tag and FLAG-tag will be detected after streptavidin IP if dimers have been formed. If addition of another protein (e.g., AGR2) would block this process, a loss of signal is expected. (E) Competition IP showing loss of dimer formation upon co-expression of AGR2. Samples were immunoblotted with anti-AGR2, anti-FLAG and Streptavidin. Tubulin was used as a loading control in input samples. Representative of three independent experiments. Non-specific signal is indicated with an asterisk. (F) Competition IP demonstrating concentration-dependent loss of dimer formation upon increasing AGR2 co-expression. Samples were immunoblotted with anti-AGR2, anti-FLAG and Streptavidin. Tubulin was used as a loading control in input samples. Representative of two independent experiments. Source data are available online for this figure.

Figure 5. The catalytic-dead C81S and disease-causing H117Y mutations in AGR2 abrogate its ability to bind and inhibit IRE1β activity.

(A) The structure of AGR2 (pdb: 2LNS) visualized in PyMol with the relevant mutations indicated. Purple and grey cartoons depict two AGR2 molecules and their dimer structure (Patel et al, 2013), specific residues are represented as ball-and-sticks. (B) Schematic overview of competition IP using AGR2 mutants. (C) Competition IP showing loss of dimer inhibition using C81S and H117Y AGR2 mutants. Samples were immunoblotted with anti-AGR2, anti-FLAG-IRE1β and Streptavidin. Tubulin was used as a loading control in input samples. Representative of three independent experiments. (D) Quantification of IP band intensities normalized over corresponding input sample on western blots from (C), with each data point representing one experiment. Error bars show SEM. (E) Photographs showing Calu-1^{ERN1-/-IRE1βFLAG-DOX} cultures, transduced with a constitutive AGR2 transgene (wild-type or the indicated mutants). IRE1β-FLAG overexpression was induced with 1 μg/ml doxycycline for 72 h. Scale bar represents 200 µm. Representative of two independent experiments. (F) Quantification of cell death in cell lines from (E) after 72 h of transgene expression. All cells in the culture dish were stained with Annexin V and Live/Dead stain and analyzed by flow cytometry. All single and double positive cells were considered as dead cells. Representative of two independent experiments with n = 2 culture dishes. (G) RT-qPCR analysis of XBP1S/T transcript levels after 24 h of transgene induction using 1 μg/ml doxycycline. Representative of two independent experiments with three replicates per condition. Error bars show SEM. (H) RT-qPCR analysis of BLOC1S1 transcript levels after 24 h of transgene induction using 1 μg/ml doxycycline. Representative of two independent experiments with three replicates per condition. Error bars show SEM. Source data are available online for this figure.

Figure 6. Proposed mechanism of IRE1β regulation by AGR2.

In steady-state conditions, AGR2 is bound to most IRE1β molecules and under these conditions, IRE1β is mainly present in the inactive, monomeric forms. As a result, overall IRE1β activity will be low. In conditions where an activating trigger is present (possibly unfolded MUC2 polypeptides, though this remains to be demonstrated), AGR2 is released from IRE1β in favor of binding other AGR2 chaperone substrates. As a result, IRE1β is released, activated and overall IRE1β activity will be high. In case of AGR2^C81S and AGR2^H117Y, interaction with IRE1β is disrupted leading to spontaneous IRE1β dimerization triggering its activity.

Figure EV1. Validation of LS174T^{ERN1-/-IRE1βFLAG-DOX} and Calu-1 ^{ERN1-/-IRE1βFLAG-DOX} model systems.

(A) IRE1β expression in cell lines. Proteins were extracted and probed for IRE1β expression via immunoblot. Tubulin was used as a loading control. (B) Photographs showing the phenotype of cultures overexpressing IRE1β-FLAG in IRE1α wild-type (ERN1^+/+) and IRE1α deficient (ERN1^−/−) cells. Images of IRE1α deficient (ERN1^−/_) cultures are the same images as shown in Fig. 1C. Left panels show untreated cultures, middle panels show cultures treated with 1 μg/ml doxycycline for 72 hours, right panels show cultures treated with both 1 μg/ml doxycycline and 1 μM IRE1 endonuclease inhibitor 4μ8C. Scale bar represents 200 µm. (C) Quantification of IRE1β transgene expression over time in Calu-1^{ERN1-/-IRE1βFLAG-DOX} and LS174T^{ERN1-/-IRE1βFLAG-DOX} cells by western blot. Cultures were treated with 1 μg/ml doxycycline for the indicated times, protein lysates were probed for IRE1β-FLAG expression and tubulin as a loading control.

Figure EV2. AGR2 expression is restricted to colon epithelial cell lines and affects Calu-1^{ERN1-/-IRE1βFLAG-DOX} phenotype.

(A) AGR2 (purple) and ERN2 (blue insert, same data as shown in Fig. 1A) transcript expression in cell lines assayed by RT-qPCR. N = 3 culture dishes were sampled and AGR2 and ERN2 expression is shown relative to the expression detected in LS174T parental cells. Error bars show SEM. Bottom picture shows protein expression by western blot. Protein lysates were probed for AGR2 and tubulin was used as a loading control. (B) HSPA5 (purple) and ERN1 (blue insert, same data as shown in Fig. 1A) transcript expression in cell lines assayed by RT-qPCR. N = 3 culture dishes were sampled and HSPA5 and ERN2 expression is shown relative to the expression detected in LS174T parental cells. Error bars show SEM. Bottom picture shows protein expression by western blot. Protein lysates were probed for BiP and tubulin was used as a loading control. (C) IRE1β-FLAG transgene expression over time by western blot in cell lysates derived from Calu-1^{ERN1-/-IRE1βFLAG-DOX} co-expressing ER-targeted BirA as a control protein (left, orange), or AGR2 (right, purple). Cells received 1 μg/ml doxycycline to induce expression of IRE1β. Protein lysates were prepared at the indicated times and probed for IRE1β-FLAG expression, XBP1S and tubulin as a loading control. (D) Quantification of IRE1β-FLAG and XPP1S protein levels normalized to tubulin in three replicate experiments represented in (C). Error bars represent SEM. (E) Gating strategy to assess cell death. Doublets were gated out and dead cells were gated on via Annexin V and Live/Dead positive staining. All cells staining positive for a single, or both cell death markers were considered dead (red gate).

Figure EV3. Inverse relationship between AGR2 IP and IRE1β complex formation.

Quantification of immunoblots in Fig. 4F (n = 2). IRE1β-FLAG band intensity was normalized to input expression levels. Band intensity values are expressed as a fraction of the maximal intensity that was obtained in each respective experiment.

Figure EV4. AGR2 expression in stably transduced cell lines.

AGR2 expression in cultures analyzed in Fig. 5D,E. Tubulin was used as a loading control.

Figure EV5. AGR2 is predicted to bind a flexible loop region in the IRE1β luminal domain.

(A) BLAST alignment of the regions containing cysteines in human IRE1α and IRE1β. Green square indicates the sole conserved cysteine in IRE1α and IRE1β luminal domain, orange squares show cysteines present in only one of the paralogues. (B) Highest scoring AlphaFold2-Multimer model (pTM score = 0.662), modeled using IRE1β residues 35-377 (Uniprot Q76MJ5) and AGR2 residues 41-175 (Uniprot O95994). The IRE1β luminal domain is shown in orange and AGR2 in grey. Labels indicate the highlighted green residues. (C) Predicted aligned error (PAE) plot for the model shown in B.