IL23R (Interleukin 23 Receptor) Variants Protective against Inflammatory Bowel Diseases (IBD) Display Loss of Function due to Impaired Protein Stability and Intracellular Trafficking*

Genome-wide association studies as well as murine models have shown that the interleukin 23 receptor (IL23R) pathway plays a pivotal role in chronic inflammatory diseases such as Crohn disease (CD), ulcerative colitis, psoriasis, and type 1 diabetes. Genome-wide association studies and targeted re-sequencing studies have revealed the presence of multiple potentially causal variants of the IL23R. Specifically the G149R, V362I, and R381Q IL23Rα chain variants are linked to protection against the development of Crohn disease and ulcerative colitis in humans. Moreover, the exact mechanism of action of these receptor variants has not been elucidated. We show that all three of these IL23Rα variants cause a reduction in IL23 receptor activation-mediated phosphorylation of the signal-transducing activator of transcription 3 (STAT3) and phosphorylation of signal transducing activator of transcription 4 (STAT4). The reduction in signaling is due to lower levels of cell surface receptor expression. For G149R, the receptor retention in the endoplasmic reticulum is due to an impairment of receptor maturation, whereas the R381Q and V362I variants have reduced protein stability. Finally, we demonstrate that the endogenous expression of IL23Rα protein from V362I and R381Q variants in human lymphoblastoid cell lines exhibited lower expression levels relative to susceptibility alleles. Our results suggest a convergent cause of IL23Rα variant protection against chronic inflammatory disease.

Genome-wide association studies as well as murine models have shown that the interleukin 23 receptor (IL23R) pathway plays a pivotal role in chronic inflammatory diseases such as Crohn disease (CD), ulcerative colitis, psoriasis, and type 1 diabetes. Genome-wide association studies and targeted re-sequencing studies have revealed the presence of multiple potentially causal variants of the IL23R. Specifically the G149R, V362I, and R381Q IL23R␣ chain variants are linked to protection against the development of Crohn disease and ulcerative colitis in humans. Moreover, the exact mechanism of action of these receptor variants has not been elucidated. We show that all three of these IL23R␣ variants cause a reduction in IL23 receptor activation-mediated phosphorylation of the signaltransducing activator of transcription 3 (STAT3) and phosphorylation of signal transducing activator of transcription 4 (STAT4). The reduction in signaling is due to lower levels of cell surface receptor expression. For G149R, the receptor retention in the endoplasmic reticulum is due to an impairment of receptor maturation, whereas the R381Q and V362I variants have reduced protein stability. Finally, we demonstrate that the endogenous expression of IL23R␣ protein from V362I and R381Q variants in human lymphoblastoid cell lines exhibited lower expression levels relative to susceptibility alleles. Our results suggest a convergent cause of IL23R␣ variant protection against chronic inflammatory disease.
Interleukin 23 receptors (IL23Rs) have been implicated in chronic inflammatory diseases through their role in initiating the differentiation of helper T cells (Th17). The Th17 pathway is important to control acute microbial infections (1)(2)(3), and deregulation of the pathway results in inflammatory bowel disease (IBD) 2 (4 -6). IL23R is expressed in specific subsets of T cells as well as other immune cells (7). Cells that express the IL23R polypeptide chain do not express IL12R␤2, which is also a partner for IL12R␤1 forming the interleukin 12 receptor (IL12R) (7). Recently, evidence has been presented that IL23R in type 3 innate lymphoid cells is involved in development of colitis through IL23 signaling (8). It is thus essential to understand how IL23R␣-protective variants affect IL23 signaling as evidence to support development of therapeutic strategies for IBD. To avoid confusion due to the nomenclature that denotes the IL23R heterodimer receptor from the individual chains, in this report the chain distinct to IL23R will be denoted as IL23R␣ and the heterodimer denoted as IL23R. IL23R signals through the interaction between two receptor chains, IL23R␣ and IL12R␤1, and together signals by binding IL23 cytokine (9,10). IL23R␣ constitutively interacts with Janus kinase 2 (JAK2). Binding of IL23 by IL23R results in activation of JAK2 and subsequent phosphorylation of the IL23R followed by recruitment of STAT3 and STAT4 and their phosphorylation (9). Phosphorylated STAT3 and STAT4 both form homodimers that translocate to the nucleus to transcribe proinflammatory cytokines. The increase in transcription of proinflammatory cytokines leads to differentiation of CD4ϩ T cells into Th17 cells resulting in inflammation (11,12). Multiple evidence suggests that IL23R may play a pivotal role in chronic inflammation (13)(14)(15)(16).
Genome-wide association studies have identified potential biological pathways that are responsible for disease establish-ment and pathogenesis (17)(18)(19)(20), and the IL23R gene was reported to show strong association (susceptibility or protection) with IBD (18,(21)(22)(23). IL23R-protective coding variants such as R381Q were identified by genome-wide association studies (24), whereas G149R and V362I were identified by targeted re-sequencing (20). Four notable studies were performed on IL23R variants discovered in genome-wide association studies to gain insight on the role of these variants. One study reported that a risk variant for CD and ulcerative colitis located in the 3Ј-untranslated region lost its binding to microRNA, Let7f, and thereby its deregulation led to an increase in mRNA and protein levels of IL23R␣. This suggests that the deregulation of IL23R␣ expression could lead to a hypersensitivity of the IL23R and resultant overstimulation of the IL23R-activated signaling pathway (25). In contrast, an IL23R␣ variant that has been studied in some detail is R381Q and is predicted to be a protective variant. Three studies reported that this variant results in a loss of function because cells expressing this variant show reduced levels of phosphorylated STAT3 and consequently reduced production of proinflammatory cytokines with reduced Th17 effector functions (26 -28).
In this study we sought to identify the biochemical mechanisms by which IL23R variants may provide protection against IBD. The IL23R variants are positioned within distinct domains of the folded receptor. R381Q is found in the C-terminal cytoplasmic portion, V362I is in the transmembrane region, and G149R is in the extracellular domain of the receptor, suggesting that these variants will have distinct biochemical properties. We show here that IL23R␣ R381Q and V362I variants have lower protein stability leading to reduced expression levels, whereas the G149R variant is retained in the endoplasmic reticulum (ER) as unfolded polypeptides.

Experimental Procedures
Antibodies and Reagents-IL23 was purchased from R&D Systems, and IL23R antibody was purchased from Abcam. Antibody to IL12R␤1 was purchased from Santa Cruz Biotechnology, and antibody toward the FLAG tag was purchased from Sigma. Antibodies to detect ER stress proteins, phosphorylated STAT3 (Tyr-705), STAT3, phosphorylated STAT4, STAT4, and ␣-tubulin, purchased from Cell Signaling Technology. Transfection reagents such as X-treme gene transfection reagent were purchased from Roche Applied Science, and the TransIT-HeLa MONSTER transfection kit was purchased from Mirus Bio. Chemical compounds such as cycloheximide, brefeldin A, and tunicamycin were purchased from Sigma. Reagents for glycosidase experiments such as Endo H, peptide N-glycosidase F, and neuraminidase were purchased from New England BioLabs. Nuc Blue live cell stain (Hoechst 33342), Cell light ER-RFP (BacMam 2.0), and MTT were purchased from Molecular Probes by Life Technologies. Alexa Fluor 647 mouse IgG was purchased from Molecular Probes by Life Technologies.
Plasmid Construction-cDNAs coding for IL23R␣ and IL12R␤1 were amplified from pCDNA3.1IL23R␣ and pCDNA3. 1IL12R␤1 (the Dr. Rioux laboratory). To generate IL23R␣-protective variants, pCDNA3.1IL23R␣ was used as template, and the QuikChange method was performed as described by the manufacturer of the QuikChange II site-directed mutagenesis kit (Agilent Technologies) to create pCDNA3.1IL23R␣R381Q, -G149R, and -V362I. To account for transfection efficiency and to approximate the expression levels of the receptors, we decided to insert fluorescent proteins Venus YFP (vYFP) and mPlum (mRFP) downstream of an IRES sequence found in the pLpC binary vector as described previously (29). The pCDNA3.1 carrying the cDNA to IL23R␣ and variants was used as the template to PCR-amplify and subclone into PacI/ ClaI restriction sites found pLpC ϩ Venus retroviral vector to generate pLpCIL23R␣ϩvYFP, pLpCIL23R␣R381QϩvYFP, pLpCIL23R␣G149RϩvYFP, pLpCIL23R␣V362IϩvYFP. To construct pLpCIL12R␤1ϩmPlum, the pCDNA3.1IL12R␤1 was used as a template to PCR-amplify and subclone into PacI/ClaI restriction sites of pLpC ϩmPlum vector. These vectors were used to generate stable cell lines in HEK293.
To generate N-terminal tagging of IL23R␣ with 3ϫFLAG tags, first, 5Ј-phosphorylated oligonucleotides that correspond to the signal peptide of IL23R␣ was annealed followed by subcloning into the BamHI/PacI restriction site of pLpCϩvYFP vector to create pLpCSPϩvYFP. Next, the cDNA corresponding to IL23R␣ and its variants were PCR-amplified from their corresponding pCDNA3.1 templates using primers specific for 3ϫFLAG and IL23R␣ and subcloned into PacI/ClaI restriction sites found in pLpCSPϩVenus vector.
For generation of C-terminal vYFP tagging of IL23R␣ (pCDNA3.1IL23R␣-Venus) and its variants, first, oligonucleotides that encode the 2ϫ GGGGS sequence were 5Ј-phosphorylated and annealed followed by ligation into NotI/XhoI restriction sites of pCDNA3.1 vector to create the pCDNA3.110aa construct. The coding sequence of IL23R␣ common and protective variants were PCR-amplified followed by subcloning into NheI/NotI restriction sites of the pCDNA3.110aa construct to create the pcDNA3.1IL23R␣10aa and the variant derivative constructs. Finally, the pCDNA3.1IL23R␣ and variants constructs were digested using XhoI/XbaI restriction enzymes followed by ligation of DNA sequence coding for Venus fluorescent protein that was PCR-amplified from pLpCϩVenus construct.
Cell Culture-Stable cell lines were generated from HEK293 cells using pLpC retroviral vector as described elsewhere (31). Both HEK293 cell lines and HeLa cells were maintained in DMEM supplemented with 10% FBS, penicillin (100 units/ ml), and streptomycin (100 units/ml), and for propagation of stable cell lines, puromycin at 2.5 g/ml was added. Human lymphoblastoid cell lines were obtained from the NIDDK, National Institutes of Health Central Repository (www.niddkrepository.org) and had been generated by the NIDDK Inflammatory Bowel Disease Genetics Consortium (IBDGC) study. These lymphoblastoid cell lines were maintained in RPMI with 20% FBS, penicillin (100 units/ml), and streptomycin (100 units/ml). All cell lines were propagated at 37°C and 5% CO 2 .
Rluc PCA-Rluc PCA was performed as described previously (30). HEK293 cells grown in 6-well plates were transfected with pCDNA3.1 constructs encoding IL23R␣ Rluc F [2] fusion and IL12R␤1 Rluc F [1] fusions. After 24 h of transfection, cells were harvested, washed with PBS, and resuspended in 500 l of DMEM minus phenol red. Approximately 100,000 cells were transferred to 96-well white-walled plates (Corning), and after the addition of benzyl-coelenterazine (5 M, Nanolight) to the cells; bioluminescence was monitored by using the LMaxII 384 luminometer (Molecular Devices). For detection of receptor activation by IL23, the cells were treated with IL23 per 96-well for 10 min before the addition of benzyl-coelenterazine and measurement of luminescence.
Western Blotting of Cell Lysates-Generally, after cells were harvested, the cell pellet was lysed in 250 mM Tris-HCl, pH 6.8, 5% SDS containing protease inhibitors (Roche Applied Science). The cell lysate was boiled for 10 min and centrifuged to remove any traces of cell debris, and the supernatant was used to quantify total protein concentration by BCA (Pierce). Equal amounts of proteins were separated on SDS-PAGE after Western blotting using antibodies specific to the protein of interest. The membrane was blocked with 5% skim milk in TBS-T (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Tween 20) and probed with primary and the corresponding secondary antibodies (anti-mouse-HRP and anti rabbit-HRP) in 5% skim milk in TBS-T. In the case for STAT4 and pSTAT4 antibodies, both the primary and secondary antibodies were diluted in TBST with 5% BSA. After both primary antibody (1 h at room temperature or overnight at 4°C) and secondary antibody incubation step (1 h at room temperature), the blots were washed three times with TBS-T followed by detection using the chemiluminescent detection reagent (Luminata Forte, Millipore). This procedure was followed for analysis of cell lysates from all cell lines used in this study (HEK293 and HeLa; NIDDK) unless otherwise stated. For stability experiments, 500 M cycloheximide was added to HEK293 cells that stably expressed IL23R␣ and its variants at ϳ80% confluency for 0, 0.5, 1, 2, and 4 h. Cell lysates from the stability experiments were analyzed by Western blotting as described here using IL23R␣ antibody at 1/1000 dilution. Similarly, brefeldin A (BFA; 1 g/ml) or tunicamycin (1 g/ml) treatment of HEK293 cells stably expressing IL23R␣ and its variants for 5 h at 37°C with 5% CO 2 was followed by preparation of cell lysates and analyzed by Western blotting as described here. For analysis of the rate of IL23R␣ synthesis, cycloheximide was added to cells at 500 M as stated before followed by incubation at 37°C with 5% CO 2 for 4 h. After this, the cells were washed with PBS to remove cycloheximide, and then the cells were incubated for the indicated times at 37°C with 5% CO 2 . After specified time points, (0, 30, 60, 120, and 240 min) the cells were harvested, and the cell lysates were subjected to SDS-PAGE and Western blotting for IL23R␣.
Glycosidase Digestion of IL23R␣-HEK293 cells stably expressing IL23R␣ and its variants were lysed with Nonidet buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40) containing 0.5% SDS, 0.04 M DTT, and protease inhibitors as described elsewhere (33). Briefly, the lysate was boiled for 10 min, and the supernatant containing soluble proteins (40 g of total protein) were digested with either Endo H (1000 units) or with peptide N-glycosidase F (1000 units) and neuraminidase (100 units) at 37°C for 16 h. SDS-PAGE and Western blotting of the digested products were conducted as described above.
Immunofluorescence and Confocal Microscopy-For all microscopy, HEK293 cells were seeded onto 96-well black, clear glass-bottom plates (MGB101-1-2-LG, Matrical Bioscience) followed by transfection using X-treme GENE reagent. Plas-mids constructs harboring N-terminally 3ϫFLAG-tagged IL23R␣ and its variants (e.g. pLpCSP-3ϫFLAGIL23R␣ϩvYFP) were transfected, and the following day the cells were prepared for immunofluorescence as described elsewhere (34,35). The cells were washed with ice-cold PBS followed by gentle fixation by 15 min of incubation on ice with 1.25% formaldehyde. After removal of the formaldehyde and washing with PBS, the cells were blocked with 5% goat serum in PBS for 30 min. After the blocking step, the cells were incubated with FLAG-tag antibody (mouse, 1/100 dilution) diluted in 3% skimmed milk for 2 h at room temperature or overnight at 4°C. The cells were washed with PBS followed by incubation with Alexa Fluor 647 goat anti-mouse (1/100 dilution) diluted in 3% skimmed milk for 1 h at room temperature. The cells were then washed with PBS followed by confocal imaging using an INCell 6000 confocal microscope (GE Healthcare). Cells were imaged in PBS with NucBlue stain (2 drops/ml of PBS) for nuclear staining. For detection of the nucleus, cells were excited at 405 nm, and emission was measured at 455/50 nm; to detect membrane bound receptors, cells were excited at 642 nm, and emission was measured at 706.5/70 nm. To control for transfection and to normalize for expression level of IL23R␣ and its variants, vYFP fluorescence was imaged at 488-nm excitation and 525/20-nm emission wavelengths. Images were processed using Metamorph software (version 7.5.3, MDS Analytical Technologies). Images were thresholded followed by creation of binary masks. The binary masks were used in integrated morphometry analysis to quantify the total average fluorescence intensity of the population of cells in the field of view.
For detection of IL23R␣-vYFP fusions, HEK293 cells were transfected with pCDNA3.1IL23R␣-vYFP or its variants, and 6 h after transfection cells were incubated with BacMam 2.0 ER-RFP viruses (Life Technologies), which encode the ER signal peptide fused to the red fluorescent protein (RFP) to measure co-localization between IL23R␣ and ER. The cells were maintained in DMEM minus phenol red, NucBlue stain was added (2 drops/ml), and image acquisition was performed using INCell6000 (GE Healthcare) as stated above. To detect the RFP, cells were irradiated at 561-nm excitation and detected at 605/52-nm emission. Image analysis was performed using Metamorph software (version 7.5.3, MDS Analytical Technologies). To quantify the level of IL23R␣ retained in the ER, the fluorescent images from ER localization was thresholded, and binary masks were generated to identify the ER, which were then used in integrated morphometry analysis to measure the fluorescent intensities from vYFP fluorescent proteins that are C-terminally tagged to the IL23R␣. The ratios between the total IL23R␣ levels and ER retained were calculated.

IL23R Quaternary Structure and Activation Is Similar to Erythropoietin Receptor (EpoR)-IL23R
is grouped under cytokine class I receptor superfamily, which also includes EpoR. Both receptors contain the WSXWS motif in the extracellular N-terminal domain, a single transmembrane hydrophobic domain, and a cytoplasmic C-terminal domain. Binding of Epo was shown to induce a scissor-like conformational change at the interface of the EpoR dimer that bring together JAK2 kinases associated with the EpoR cytosolic domains to allow for cross-phosphorylation and activation (Fig. 1A) (36,37). IL23R is composed of IL23R␣ and IL12R␤1. Heterodimerization of these two chains is essential to IL23R activation by IL23, and we hypothesize that the receptor is activated in the same manner as EpoR. To test this hypothesis, we used a Rluc PCA as a molecular ruler to measure heterodimer dynamics (Fig. 1A, middle and bottom) (36). Briefly, we fused cDNA coding for the IL23R␣ and IL12R␤1 chains to those coding for Rluc complementary N-and C-terminal fragments separated by sequences coding for 5, 10, or 20 amino acid (20 aa) peptide linkers composed of the 5-aa repeats of GGGGS (Fig. 1A). Our reasoning was that linkers of 5 aa would not be long enough for the Rluc fragments to associate and fold in the inactive form of the receptor dimer, where the C termini of the receptor chains are separated by ϳ70 Å, but would be separated by ϳ40 Å in the active dimer (Fig. 1A, middle; assuming the length of a peptide bond is ϳ4 Å, two 5-aa peptides could traverse a maximum distance of ϳ28 Å). Rluc PCA signal for the 5-aa linker was increased 4-fold after IL23 treatment but not with constructs with 10-or 20-aa linkers, consistent with what we observed with EpoR ( Fig. 1B) (36). We also showed that IL23R␣ and IL12R␤1 interact with their cognate kinases JAK2 and TYK2, respectively (Fig. 1C). Finally and consistent with the EpoR activation model, when JAK2 and TYK2 were expressed as fusions to the Rluc PCA reporter fragments along with their respective IL23R␣ and IL12R␤1 binding partners, the Rluc signal increased in cells treated with IL23 ( Fig. 1D) (36).
IL23R Activation Is Reduced in Protective Variants-Having established that IL23R complex structure and activation is similar to that of EpoR, we now assessed the effects of sequence variants of IL23R␣ on IL23-induced receptor conformation change. IL23R␣-protective variants are located at three distinct sites: the G149R change is found in the extracellular domain of the receptor, whereas the V362I change is located within the transmembrane domain, and the R381Q change is located at the C terminus of the receptor five amino acids from the transmembrane domain. We found that the IL23R␣ variants retained their ability to interact with IL12R␤1; however, the extent to which they are induced by IL23 was reduced compared with the common variant (Fig. 1E). Finally, interactions of IL23R␣ to IL12R␤1, JAK2, and to TYK2 were comparable for all variants (Fig. 1, F, G, and H).
Protective Variants of IL23R␣ All Result in Reduced IL23Ractivated STAT3 and STAT4 Phosphorylation-STAT3 is a primary effector of IL23R signaling, and to investigate IL23-dependent STAT3 phosphorylation, we co-transfected HeLa cells with a plasmid encoding the cDNA for IL23R␣ or its variants together with a plasmid encoding the cDNA for IL12R␤1 (32). After serum starvation, we measured IL23-induced STAT3 phosphorylation (pSTAT3). Cells expressing the protective R381Q, G149R, and V362I variants showed reduced pSTAT3 levels compared with cells expressing the common variant (Fig.  2, A and B). Transfection of HeLa cells with only IL12R␤1 or together with IL23R␣, but without IL23, did not show significant STAT3 phosphorylation (Fig. 3A). Western blotting with antibodies specific to both IL23R␣ and IL12R␤1 confirmed expression of both receptor subunits (Fig. 3B). Furthermore, we investigated the effects of IL23R variants on IL23-mediated STAT4 signaling as several studies have demonstrated that IL23R subunits activate STAT4 in response to IL23 (9,10,38). To this end we co-transfected HeLa cells with cDNA of STAT4 with plasmid encoding the cDNA for IL23R␣ or its variants together with a plasmid encoding the cDNA for IL12R␤1. The same experimental setup was followed as stated above for STAT3 signaling. We found that the variants show reduced STAT4 phosphorylation levels compared with the common variant (Fig. 2, E and F). The reduction in signaling corroborates well with the reduced activation in protective variants by IL23 (Fig. 1E). We, therefore, turned to questions of receptor maturation and expression.  (36,62). The extracellular domain exists in a conformation that keeps the intracellular domains and associated JAK2 and TYK2 separated from each other by ϳ70 Å. Upon binding to IL23, the extracellular dimer is reorganized and brings the intracellular cytoplasmic domains within ϳ 40 Å of each other, allowing for interaction between and activation of their associated kinases. For identification of IL23R activation mechanism, the extracellular and transmembrane domains of IL23R and IL12R␤1 are fused to Renilla luciferase (Rluc) complementary N-and C-terminal fragments, Rluc F[1] and F [2], via flexible linkers consisting of 1, 2, or 4 five-amino acid repeats to generate the following: IL23R and IL12R␤1 fused to 5-, 10-, and 20-amino acid linkers that are in turn fused to the Rluc fragments. HEK293 cells transfected with these receptor fusions and PCA were detected by measuring the luminescence. When the receptors are fused via a 5-aa linker, folding of the Rluc reporter protein from the fragments only occurs for the IL23 hormone-bound active form of the receptor dimer. If receptors fused to Rluc fragments via 10-or 20-aa linkers, Rluc reporter protein can fold from the fragments in either inactive or active dimer conformations. The structural model of Renilla luciferase was rendered using information from 2pSj.pdb (63) using PyMOL 1.3. B, Rluc enzyme assay detected the IL23R activation in HEK293 cells. Cells were co-transfected with IL23R␣ and IL12R␤1 fused to Rluc PCA fragments via various linker lengths. The next day the cells were incubated with and without 100 ng of IL23 cytokine followed by measurement of Rluc enzyme activity using benzyl coelenterazine as the substrate and detection using a luminometer. The signal is expressed as relative light units (RLU). C, Rluc PCA reveals the interaction between IL23R␣ and their subunits and their associated kinases (JAK2 and TYK2). HEK293T cells were transfected with pCDNA3.1 harboring the cDNA of full-length IL23R receptor subunits and the kinases JAK2 and TYK2 fused to Rluc fragments to detect the interaction as measured by PCA (RLU) of Rluc. D, JAK2 and TYK2 association in the presence of IL23 binding to IL23R was detected by Rluc PCA. Rluc fragments were fused to JAK2 and TYK2 and were co-transfected in HEK293T cells stably expressing IL23R subunits followed by measuring Rluc luminescence. Statistical significance denoted by the asterisk (*) was calculated by Student's t test, where p Ͻ 0.05. E, IL23R␣ variants show the reduction in activation by IL23. IL23R␣, its, variants and IL12R␤1 were fused to Rluc fragments to detect activation by IL23 (10 ng/ml), and luminescence was measured as in B. The change in RLU between cells not treated and treated with IL23 was calculated, and activation was thus observed as a function of change in RLU. Statistical significance denoted by the asterisk (*) was calculated by ANOVA, where p Ͻ 0.01.

IL23R␣, R381Q, and G149R Variants Perturb Receptor Maturation in HEK293
Cells-Both IL23R␣ and IL12R␤1 migrated as two bands (Fig. 3B), consistent with the persistent presence of early and mature glycosylated forms of these receptors (39,40). One possible explanation for reduced activity of IL23R variants is that they could be defective in maturation. We thus examined the maturation of all IL23R variant polypeptides in HEK293 cells. We tested for mature versus early glycosylation marks of IL23R␣ in stably transfected HEK293 cells. We found that the low molecular mass signal of IL23R␣ was sensitive to Endo H, which cleaves N-linked glycopeptides, but the high molecular mass band was Endo H-resistant (Fig. 4A). However, treatment with peptide N-glycosidase and neuraminidase, glycosidases that together hydrolyze all types of glycosides, removed both the high and low molecular mass bands. These results suggested that both bands corresponded to glycosylated species of the IL23R␣, but the lower molecular mass form corresponded to immature receptor, and the high molecular mass corresponded to mature receptor.
To investigate whether there are any differences in maturation between the IL23R␣ common and disease protective variants, we performed quantitative Western blotting of cell lysates from cells that transiently expressed each variant, and we determined the ratio of mature to immature receptors (Fig. 4, B and C). R381Q and G149R were reduced in their mature-to-immature receptor ratio to ϳ50 and 30%, respectively. In contrast, the V362I variant showed a similar ratio to the common variant, although the protein level was much reduced (discussed further below), requiring that a higher amount of V362I cDNA had to be transfected to achieve a similar level of receptor expression as that of the common variant (Fig. 4B). Thus both R381Q and G149R displayed impairment in maturation.
IL23R␣ R381Q and G149R Show Reduced Surface Expression and Impairment in ER to Golgi Trafficking-We next asked whether the impairment in maturation of R381Q and G149R FIGURE 2. IL23R␣ protective variants R381Q, G149R, and V362I display reduction in IL23R signaling. A, HeLa cells were co-transfected with plasmids carrying IL23R␣ or variants and IL12R␤1. After washing with PBS to remove media and overnight growth in serum-free medium, transfected cells were incubated with or without IL23 (15 ng/ml) for 30 min. Cell lysates were subjected to SDS-PAGE followed by Western blotting with antibodies specific to phospho-STAT3 (pSTAT3) and STAT3. Representative blots from three independent experiments are shown here. B, corresponding densitometry of the levels of pSTAT3 in IL23R␣ and IL23R␣ variants obtained from Western blots above. The signal of pSTAT3 detected on Western blot was measured as pixels using ImageJ 1.46r (National Institutes of Health) and were further normalized to levels of total STAT3. Data and S.D. are representative of three independent experiments. Statistical significance calculated by ANOVA is denoted by the asterisk (*), where p Ͻ 0.01 is compared with IL23R␣; however, R381Q was significantly different from V362I, with a p Ͻ 0.05. C, HeLa cells were co-transfected with plasmids carrying STAT4, IL23R␣ or variants, and IL12R␤1. As stated for STAT3 experiments (A), the cell lysates were subjected to SDS-PAGE followed by Western blotting with antibodies specific to phospho-STAT4 (pSTAT4) and STAT4. D, corresponding densitometry of the levels of pSTAT4 in IL23R␣ and IL23R␣ variants obtained from Western blots of C. The signal of pSTAT4 detected on Western blot was measured as pixels using ImageJ 1.46r (National Institutes of Health) and were further normalized to levels of total STAT4. Data and S.D. are representative of three independent experiments. Statistical significance calculated by ANOVA is denoted by the asterisk (*), where p Ͻ 0.01 is compared with IL23R␣; in addition, R381Q was significantly different from G149R and V362I with a p Ͻ 0.01. EV denotes empty vector. FIGURE 3. IL23-induced STAT3 phosphorylation is dependent on the presence of both IL23R␣ and IL12R␤1 and IL23 cytokine. A, HeLa cells were co-transfected with plasmids carrying either IL12R␤1 or IL12R␤1 together with IL23R␣. After washing with PBS to remove medium and overnight growth in serum free medium, transfected cells were incubated with or without IL23 (15 ng/ml) for 30 min. Cell lysates were subjected to SDS-PAGE followed by Western blotting with antibodies specific to phospho-STAT3 (pSTAT3) and tubulin. EV denotes empty vector. B, HeLa cells were co-transfected with plasmids carrying IL23R␣ or variants and IL12R␤1. After washing with PBS to remove medium and overnight growth in serum free medium, transfected cells were incubated with or without IL23 (15 ng/ml) for 30 min. Cell lysates were subjected to SDS-PAGE followed by Western blotting with antibodies specific to IL23R␣ and IL12R␤1.
variants could result in decreased cell surface expression. IL23R␣ common and protective variant surface expression was quantified by immunofluorescence on HEK293 cells expressing FLAG-tag IL23R␣. The 3ϫFLAG sequence was fused to the N terminus of the IL23R␣ receptor and its variants as there are no effective commercial antibodies available. Using antibody against the FLAG tag, surface-expressed IL23R␣ were imaged by confocal microscopy and quantified by normalizing for total receptor expression (Fig. 5A). We found that relative to the common variant of IL23R␣, both R381Q and G149R were reduced in surface expression by ϳ60 and 30%, respectively, consistent with the maturation data (Fig. 5B).
It is generally accepted that immaturely glycosylated forms of proteins are mostly retained in the ER/cis-Golgi, whereas mature proteins (both N-and O-glycosylated) are found in post-Golgi and other compartments (40). We tested for ER retention of IL23R␣ common and protective variants fused to vYFP (41) (C terminus to receptor) and quantified expression in the ER using confocal microscopy. Co-localization with ER (ER retention signal-mRFP) with IL23R␣-vYFP revealed that a fraction of the receptors are retained within the ER (Fig. 5C), and quantification of ER-retained IL23R␣ revealed that variants R381Q and G149R display higher retention in the ER than the common variant (Fig. 5D). These findings further confirmed that R381Q and G149R maturation is impaired in ER-Golgi trafficking.
The glycosidase tests strongly suggest that ER retention is reflected in failed completion of glycosylation. It is, however, also possible that the R381Q and G149R variants also had immature O-glycosylation. To test for differences in O-glycosylation, HEK293 cells expressing stable IL23R␣ and its variants were treated with BFA, which fuses the ER and Golgi membranes, thereby allowing for the partial redistribution of enzymes such as O-GlcNAc transferase found in the Golgi membranes to the ER (42). BFA treatment resulted in intermediate molecular weight products that migrated between the mature and immature IL23R␣ receptor, the novel species of signal named aberrantly glycosylated receptor (Fig. 6). The ability of all the IL23R␣ variants showing this aberrantly glycosylation confirms that these variants are able to undergo O-glycosylation when they interact with the necessary enzymes. Alternatively, proteins are also retained in the ER when they do not pass through the quality check by the ER resident chaperones (43).
IL23R␣ G149R Expressing HEK293 Cells Display an Increase in Unfolded Protein Response-To determine whether the IL23R␣-protective variants cause disturbances to the homeostasis of the ER, we probed the cell lysates from HEK293 cell lines stably expressing IL23R␣ and its variants for accumulation of proteins involved in the unfolded protein response (43). Expression levels of Ero1L␣ (endoplasmic reticulum oxidoreductase ␣), CCAT/enhancer binding protein homologous protein (CHOP), PDI (protein disulfide isomerase), BiP, IRE1␣ (inositol-requiring enzyme 1 ␣), and calnexin were analyzed by Western blotting. Increased levels of ER chaperone immunoglobulin-binding protein (BiP) was identified in the G149R variant (Fig. 7A). BiP induction in the absence of CHOP, which is associated with induction of cell death (43), suggests that the ER stress induced with the G149R variant is most likely due to protein misfolding/accumulation of unfolded polypeptides. To confirm that the viability of cells expressing IL23R␣-protective variants was not affected, we assayed for mitochondrial dehydrogenase activity in the IL23R␣ common and protective variants using the MTT assay. The IL23R␣ common and protective variant-expressing cells exhibited similar levels of cell viability (Fig. 7B). Therefore, G149R variant is severely affected in receptor maturation, and its accumulation in ER increases the induction of BiP without causing cell death (44).
IL23R␣ R381Q and V362I Show Reduction in Stability-The stability of IL23R␣ and its variants was assessed by pulse-chase experiments after cycloheximide inhibition of de novo protein synthesis and detection of protein at different times by Western blotting (Fig. 8A). There is a reduction in both mature and immature IL23R␣ over time, but R381Q and V362I degrade more rapidly (Fig. 8, A and B). IL23R␣ had a half-life of ϳ187 min, and R381Q and V362I had a half-life of 16 and 72 min, respectively (Fig. 8C). Having measured the turnover of IL23R␣ common and protective variants, we turned to investigating the de novo synthesis of the receptors. Cells were treated for 4 h with cycloheximide, during which a large amount of IL23R␣ was degraded, after which cycloheximide was removed by washing the cells with PBS. This was followed by monitoring the rate of mature and immature IL23R␣ synthesis at various time points by Western blot using anti-IL23R␣ antibodies (Fig.  9A). Interestingly, the rate of de novo mature receptor synthesis was markedly reduced for R381Q and G149R variants compared with rates for the common and V362I variants (Fig. 9, A  and B). The rate of synthesis of mature versus immature IL23R␣ receptor was 1:1 for the common and V362I-protective variant, whereas this ratio was severely reduced for the protective variants R381Q and G149R (Fig. 9C). These results provide further evidence that trafficking is affected in the protective variants R381Q and G149R; hence, there are fewer receptors at the cell surface. The V362I variant, however, showed a higher turnover rate than that of the common variant even if the rate of synthesis of mature versus immature receptor was similar for both. Thus, cells expressing the V362I variant appeared to have lower levels of both mature and immature forms of the receptor. To assess whether this holds true for endogenous expression of the common IL23R␣ variant, as well as R381Q and V362I variants, we used lymphoblastoid cell lines derived from individuals where these haplotypes were present. Western blotting of cell lysates from these cell lines revealed that R381Q (homozygote and heterozygote) and V362I (homozygote) show a reduction in expression of the receptor at basal levels (Fig. 10). Taken together, these data demonstrate that both transfected and endogenous protein expression of IL23R␣-protective variants leads to a reduced expression relative to the common variant, suggesting a loss of function for protective variants.

Discussion
Our results suggest that all protective coding variants of IL23R␣ result in reduced cell surface expression of mature receptors due either to reduced stability and or impaired trafficking from ER. In all three cases reduced surface expression resulted in reduced IL23-mediated signaling. Previous studies showed that the IL23R␣ R381Q variant displayed reduced signaling; however, mechanistic details of loss of activity were not explored (26 -28). Our results are consistent with these findings. We now extend this observation to two new coding variants of IL23R␣, all described as protective variants in inflammatory bowel disease. Moreover, we provide clear mechanistic details of loss of activity for the three protective variants, all of which result in reduced cell surface expression of mature recep- , to identify the nucleus and antibody toward FLAG tag, was used to localize IL23R␣. Overlay images (Merge) from Hoechst and anti-FLAG staining is shown to demonstrate IL23R␣ localization. Scale bar, 10 m. B, mean fluorescent intensities from the corresponding confocal images were used to quantify IL23R␣, which were normalized to the total expression of IL23R␣ and its variants using levels of vYFP fluorescent protein as described under "Experimental Procedures." Different fields of images were quantified, and S.E. values are n ϭ 3. Statistical significance calculated by ANOVA is denoted by asterisk (*), where p Ͻ 0.01 compared with IL23R␣.R381Q was significantly different from G149R and V362I, and G149R is significantly different from V362I with a p Ͻ 0.01. C, localization of IL23R␣ variant-vYFP fusions; green, IL23R␣; red, ER. Hoeschst staining was used to identify the nucleus, and cell light ER-RFP was used to identify ER. Scale bar, 10 m. D, ratio of total fluorescence from v-YFP-tagged IL23R␣ and ER-localized v-YFP-tagged IL23R␣ is calculated as described under "Experimental Procedures." On average, 10 cells were quantified for each variant, and the S.E. was calculated. Statistical significance calculated by ANOVA is denoted by the asterisk (*) compared with IL23R␣, where p Ͻ 0.05; in addition, R381Q was significantly different from G149R, with a p Ͻ 0.05, and G149R was significantly different from V362I, with a p Ͻ 0.01. FIGURE 6. BFA treatment of HEK293 cells stably expressing IL23R␣ and variants results in partial O-glycosylation. HEK293 cells expressing IL23R␣ and its variants when exposed to BFA (1 g/ml) for 5 h is shown. Cell lysates from two independent experiments were subjected to SDS-PAGE, and Western blotting with antibodies to IL23R␣ show that the immature receptors in both wild type and variants are able to be partially O-glycosylated when the responsible enzymes found in Golgi are available to them. Arrows indicate the mature (M IL23R␣) and immature (IM IL23R␣) and the aberrant IL23R␣ (abIL23R␣) that is partially O-glycosylated.

Loss of Function for IL23R-protective Variants
tors due either to reduced stability and or impaired trafficking from ER. In all three cases, reduced surface expression resulted in reduced IL23-mediated signaling, such that all protective variants negatively impact IL-23 signaling.
The mutations encoding R381Q, G149R, and V362I are located in different regions of the IL23R␣ receptor structure close to the transmembrane domain (C terminus of the receptor), extracellular region, and transmembrane domain, respectively. It is expected, therefore, that although the consequences of the mutations are identical, the mechanisms of reduced expression are different. We observed that the variants R381Q and G149R, unlike V362I, displayed a reduction in the mature/ immature receptor ratio, and both R381Q and G149R variants were reduced in surface expression where signaling via IL23 is initiated and G149R is impaired in ER to Golgi trafficking. Similar observations have been made in other studies with variants of proteins or receptors that undergo maturation from ER to Golgi and were affected in their arrival at their destined location such as the missense mutation in very low density lipoprotein receptor, cystic fibrosis transmembrane receptor, and amyloid precursor protein (APP) (44 -50). For instance, the results obtained here are reminiscent of the triple variant of APP that displayed impairment in APP maturation and trafficking from ER-Golgi. As reported for IL23R␣ G149R variant, the APP triple variant also showed an increase in the level of BiP, as it was retained in the ER as misfolded polypeptide without inducing cell death. In addition, it was reduced in protein stability (44). However, unlike the APP triple variant, the IL23R␣ G149R variant was not reduced in its stability. In fact, our cycloheximide studies revealed that the R381Q variant and to a lesser extent the V362I variant were unstable. The lower levels of expression of these two receptor variants in endogenously expressing cell lines further support this notion. In summary, the protective effect by the variants R381Q and V362I is due to protein instability leading to lower levels of total receptor expression coupled with a reduction in signaling. The V362I variant is unique in that the ratio of mature-to-immature protein levels is comparable with that of the common variant. However, the V362I variant exhibits a reduced half-life and presents with lower levels of endogenous receptor expression in NIDDK V362I homozygous cells.
A question that still remains is: Why would the G149R variant be unfolded in the ER but its stability not affected? A possible explanation for the observed ER retention of the G149R variant may reside in the observation that G149R creates a cleavage site, RX(L/V)X, where X is any amino acid, for PCSK-SkI-1, a convertase that is found to reside in the trans Golgi (51,52). It is possible that the G149R variant is translocated to the Golgi, but the variant receptor may get cleaved at the extracellular domain by the convertases in the Golgi, resulting in an increase in receptor truncations and degradation.
In summary, the R381Q variant displays a higher turnover rate resulting in lower levels of receptor expression at the cell surface. This may be attributed to the fact that residue 381 of the IL23R␣ chain is the fifth amino acid after the transmembrane domain. It is generally understood that if residues in this region are positively charged (positive inside rule), the proteins do not get translocated but do get inserted into the membrane (53,54), and so mutations here may affect membrane protein insertion. The G149R variant primarily remains in the ER and does not get translocated to the plasma membrane surface, perhaps due to changes in its structure and/or due to its being a substrate for SKI-1 convertase in the Golgi. In the future, it would be interesting to obtain crystal structures of the IL23R␣ receptor, which may provide some insight into how the R381Q affects protein insertion. Finally, the V362I substitution is located within the transmembrane domain of the receptor. It is possible that the added bulk of a methyl group to the side chain at this position may disrupt shape complementarity and, therefore, destabilize the interaction between IL23R␣ and IL12R␤1 receptor transmembrane domains but maintain peptide chain rigidity due to branching at the ␤-carbon as has been shown to be essential for other transmembrane protein-protein interactions (55). Thus the three protective variants, although found at distinct regions of the receptor, through different mechanisms may be expressed at lower levels at the plasma membrane sur- FIGURE 7. Unfolded protein response is up-regulated in cells expressing the IL23R␣ G149R variant. A, HEK293 cells stably expressing IL23R␣ and its variants were harvested, and cell lysates were electrophoresed on SDS-PAGE followed by Western blot. Antibodies specific to ER stress and unfolded protein response markers indicate that BiP abundance is higher in IL23R␣ G149R compared with the common variant, suggesting that IL23R␣ G149R is retained in the ER. BiP abundance could also be induced in HEK293 cells expressing IL23R␣ when these cells are treated with tunicamycin (1 g/ml) for 5 h. B, the viability of cells expressing IL23R␣ variants is not affected. HEK293 cells stably expressing IL23R␣ and variants were plated in 96-well plate (20,000 cells) in triplicate. The next day the cells were assayed for levels of mitochondrial dehydrogenase activity by using MTT (64). Data and error bars represent the S.D. from three independent experiments. Half-life of IL23R␣ and variants was calculated by using GraphPad Prism 6.0. Error values were S.D. where n ϭ 3. Statistical significance calculated by ANOVA is denoted by the asterisk (*) compared with IL23R␣ where p Ͻ 0.01; in addition, R381Q was significantly different from G149R and V362I with a p Ͻ 0.01, and G149R was significantly different from V362I with a p Ͻ 0.01. The error values are S.E. where n ϭ 3. C, ratio of the rate for IL23R␣ and variants between their mature and immature receptor synthesis is represented here. Error values are S.D. where n ϭ 3. Statistical significance calculated by ANOVA is denoted by the asterisk (*) compared with IL23R␣, where p Ͻ 0.01. In addition, R381Q was significantly different from V362I, with a p Ͻ 0.01, and G149R was significantly different from V362I, with a p Ͻ 0.01. face, resulting in receptor-mediated signaling and thereby play a protective role in IBD.
Our results would appear to be consistent with the fact that antibodies against IL23 hormone itself present a relevant therapeutic strategy for treating IBD (56). We would argue that any strategy that directly inhibits IL23 receptor signaling could be of even greater potential therapeutic value. Targeted antagonists of IL23 receptor subunit assembly or competitive inhibitors of IL23R could prove superior therapeutics, as unlike IL23 antibodies, the number of receptors is likely much lower than that of IL23 hormone, and thus therapeutic levels of IL23Racting agents would be lower and more effective. For instance, compounds that target protein-protein interaction for development of therapeutics has been successful in a number of cases (57,58). As examples, competitors of IL2-IL2R␣ and BcL-2/ BcL-xL interactions have been identified using fragment based approach, and the compounds bind to a groove that locks the conformation of the protein and prevents its interaction with its natural partners (57). Alternatively, we have reported an allosteric peptide inhibitor of IL23R function that, in contrast to anti-p40 IL23 hormone subunit, selectively inhibits IL-23induced inflammation in three in vivo mouse models (59). In addition, development of structure-specific synthetic antibodies could prove another alternative to developing functionally antagonistic or competitive inhibitors of IL23R function (60).
Finally our observations that the IL23R appears to be activated in a similar manner to erythropoietin receptor provides a mechanistic target for therapeutic development, and the Rluc PCA we report here or a version based on our latest infrared fluorescent protein PCA can provide an unambiguous readout for screening any class of molecules for antagonists of IL23R receptor function (29,30,36,61). Inhibition of IL-23 signaling may thus have a positive impact in preventing disease onset.