Generic membrane-spanning features endow IRE1α with responsiveness to membrane aberrancy

Experimental platform to study membrane aberrancy–mediated IRE1α signaling

To explore the role of IRE1α in responding to membrane aberrancy, we exploited a CHO-K1–derived cell line (S21) containing both a C/EBP homologous protein (CHOP)::green fluorescent protein (GFP) transcriptional reporter of the UPR PERK-dependent branch (Novoa et al., 2001) and an mRNA-based XBP1s::Turquoise reporter responsive to IRE1α’s sequence-specific RNase activity (based on Iwawaki et al., 2004). Flow cytometry analysis of cells treated with either tunicamycin or 2-deoxy-d-glucose (2DG; to perturb protein-folding homeostasis in the ER) or palmitate (a saturated fatty acid, to promote membrane aberrancy) showed that in these cells, all three agents activated both strands of the UPR in a concentration-dependent manner. In most experiments, activation of the XBP1s::Turquoise reporter did not saturate, even at the highest concentrations of palmitate that were compatible with cell viability (Figure 1). Nonetheless, the ascending limb of the concentration–response curve was deemed broad enough to enable study of the IRE1α branch of the UPR.

To explore the role of IRE1α’s TMD in responding to membrane aberrancy, we first used CRISPR/Cas9-mediated gene editing to introduce frameshifting mutations into the TMD-encoding region of the two chromosomal copies of Ern1. As expected, the resulting inactivation of IRE1α in the ∆TM12 clone led to selective loss of the XBP1s::Turquoise signal in stressed cells and only a modest attenuation of the mostly PERK-dependent CHOP::GFP signal (Figure 2A and Supplemental Figure S2, A–C). Retargeting the mutant Ern1 locus of ∆TM12 clone with CRISPR/Cas9 and a repair template encoding the wild-type TMD restored stress-dependent activation of XBP1s::Turquoise on a discernible fraction of the cells (Figure 2B, left). “Rescued” cells were segregated by fluorescence-activated cell sorting (FACS) of the XBP1s::Turquoise⁺ population (their recovery as viable clonogenic cells was greatly facilitated by exploiting the reversible action of 2DG) and subsequently analyzed either as an expanded polyclonal pool or as individual rescued clones (Figure 2B, right).

Functional consequences of sequence modifications of the endogenous IRE1α’s TMD

The aforementioned experimental system was used to explore features of IRE1α’s highly conserved TMD (Supplemental Figure S2D) that might be relevant to recognizing membrane aberrancy. Offering a repair template that encoded either a wild-type or a sequence-scrambled (similar to that used in Volmer et al., 2013) IRE1α TMD to the retargeted ∆TM12 cells generated isogenic cells that differed only in the sequence of the portion of the Ern1 gene encoding IRE1α TMD (Figure 3A). Levels of IRE1α expression varied in retargeted, rescued, ∆TM12 clones (Figure 3B). This variation likely reflected the combined effects of gene dosage and properties of the rescued allele. Through genotypic analysis, we confined the downstream studies to clones that had a single detectable allele (encoding either a wild-type or a scrambled TMD). However, with the tools available to us, we were unable to distinguish between cells having two rescued Ern1 alleles and cells having one rescued allele in-trans to a large deletion that was not detected in the fragment-based genotypic analysis (see Materials and Methods). Thus clone TM-WT-16 is likely homozygous for wild-type Ern1 allele (having wild-type levels of IRE1α protein and a wild-type response to stress), whereas clone TM-WT-22 is likely heterozygous for wild-type and null Ern1 alleles. Zygosity of the scrambled TMD clones (SC-4 and SC-8) cannot be guessed at; however, assuming that the subtle sequence differences between the wild-type and scrambled repair template had no effect on the relative recovery of cells with one rescued Ern1 allele and with two rescued Ern1 alleles, the lower levels of IRE1α protein observed in the polyclonal pool of cells rescued with the scrambled allele suggests that the latter encoded a protein that accumulated to lower levels in cells (Figure 3B, left).

Despite the lower levels of expression of the scrambled IRE1α, both wild-type and scrambled TMD-rescuing alleles gave rise to similar profiles of activity in cells treated with tunicamycin, 2DG, and, remarkably, palmitate. This was evident both by comparing polyclonal populations of ∆TM12 cells rescued with the wild-type and scrambled TMD–encoding repair templates (Figure 3C and Supplemental Figure S3A) and by studying the features of individual clones (Figure 3D and Supplemental Figure S3B). These findings speak against a stringent requirement for the arrangement of the side chains of the TMD in IRE1α function.

Covino et al. (2016) described features of the TMD of the yeast ER-tethered transcription factor Mga2p that endow it with sensitivity to changes in lipid packing wrought by altering acyl-chain saturation. They drew attention to the role of bulky aromatic residues, such as tryptophan and phenylalanine, and to the presence of proline residues in favoring alternative conformations of the Mga2p TMD dimer in more loosely packed and more ordered membranes constituted of unsaturated and saturated lipids, respectively. These features of Mga2p contribute to lipid homeostasis in yeast by coupling expression of OLE1, the gene encoding a fatty acid desaturase, to changes in membrane lipid saturation (Covino et al., 2016). Furthermore, Russ and Engelman (2000) pointed out the importance of an interrupted diglycine motif in homodimerizing TMD.

IRE1α has a conserved tryptophan, phenylalanine, tyrosine, and proline in its TMD (Supplemental Figure S2D), as well as an interrupted glycine and alanine motif whose presence was unaffected by scrambling (Figure 3A). To determine whether these residues contributed to sensing of membrane aberrancy by IRE1α, we reconstituted the endogenous locus of the nullizygous ∆TM12 clone with mutant versions of Ern1 encoding IRE1α proteins lacking these key features (Figure 4, A and B). Although the mutants varied in their ability to activate the XBP1s::Turquoise reporter, none exhibited a selective defect in response to lipid aberrancy (Figure 4, C and D, and Supplemental Figure S4). Even a mutant in which all of the predicted membrane-spanning residues had been replaced by leucines, IRE1α^TM>L, retained a hierarchy of response to palmitate, 2DG, and tunicamycin that resembled the wild type. For all mutants tested, endurance of the response to membrane aberrancy was noted at levels of expressions that either matched or were lower than those of the wild-type IRE1α. This was especially conspicuous in the IRE1α^TM>L, which reproducibly gave the weakest signal in the immunoprecipitation-immunoblot (Figure 4B).

The region immediately N-terminal to the IRE1α TMD is predicted to form an amphipathic extension of the membrane-spanning helix, a feature conserved in the yeast Ire1p (Figure 5A and Supplemental Figure S5, A and B). It has been proposed that hydrophobic side chains of residues from one face of this predicted amphipathic helix are inserted into the lipid bilayer, constraining the TMD in a manner that favors dimerization in an ordered bilayer (Halbleib et al., 2017). To examine the role of these residues in the responsiveness of IRE1α to lipid aberrancy, we reconstituted the endogenous locus with mutant versions of Ern1 encoding IRE1α proteins in which the amphipathic features of this region were disrupted (this was archived by exploiting Ern1^∆LD15, a signaling-defective allele of IRE1α described later). Of note, both the IRE1α^V437R and IRE1α^L441R mutant proteins retained wild-type levels of inducibility by palmitate, 2DG, and tunicamycin (Figure 5, B and C, and Supplemental Figure S5C). Although we have not examined this aspect experimentally, the elevated basal activity of the XBP1s::Turquoise reporter reproducibly noted in the IRE1α^L441R mutant cells might be explained by the loss of a repressive contact with the Sec61 complex (Sundaram et al., 2017).

The IRE1α^Y161A mutation selectively enfeebles the response to unfolded protein stress, exposing IRE1α’s ability to sense membrane aberrancy

The experimental observations presented so far speak against an important role for the side chains of specific TMD residues (and TMD-proximal residues of the predicted N-terminal amphipathic helix) in endowing IRE1α with the ability to recognize membrane aberrancy imposed by culturing cells in a range of concentrations of palmitate. However, as noted in the Introduction, manipulation of membrane lipid composition has pleiotropic effects that might include induction of unfolded protein stress in the ER lumen. The observed indifference of IRE1α to mutations in its TMD might therefore reflect dominance of the unfolded protein stress over membrane aberrancy in the forces that act on IRE1α in palmitate-exposed CHO-K1 cells. Our efforts to address this possibility by selectively buffering the consequences of protein misfolding with a chemical chaperone, 4-phenyl butyrate (4-PBA), were frustrated by the fact that at the concentrations required to attenuate UPR induction by tunicamycin, 4-PBA nonspecifically suppressed unrelated stress signaling. Therefore we sought to create an IRE1α allele that was se­lectively compromised in its ability to respond to unfolded protein stress over its response to membrane aberrancy.

Using CRISPR/Cas9, we created a CHO-K1 cell line with an in-frame deletion allele of Ern1 encompassing the entire coding sequence of the stress-sensing luminal domain of IRE1α (in-trans to a larger deletion that disrupts the coding sequence, Ern1^∆LD15/null; Figure 6, A and B, and Supplemental Figure S6, A–C). Overexpression of a transgene encoding a similar protein imparted selective responsiveness to palmitate loading upon Ern1-deleted mouse cells (Volmer et al., 2013). However, although detectable, the IRE1α^∆LD protein encoded by the mutant Ern1^∆LD15 locus failed to activate the XBP1s::Turquoise reporter in palmitate-treated cells (Supplemental Figure S6D; the minimal increase in XBP1s::Turquoise of palmitate-treated ∆LD15 cells likely reflects a process unrelated to IRE1α-mediated splicing, as it was also observed in the ∆TM12 cells lacking all IRE1α function [Supplemental Figure S2C]). This observation is consistent with a role for luminal domain–mediated dimerization in the responsiveness of IRE1α to membrane aberrancy when expressed at physiological levels, a requirement that is bypassed by overexpression. Retargeting the Ern1^∆LD15 locus with CRISPR/Cas9 while offering a human cDNA (“minigene”)-based repair template (encompassing exons 2–12 and encoding the wild-type luminal domain of human IRE1α) restored IRE1α activity in tunicamycin-, 2DG-, and palmitate-treated cells (Figure 6C and Supplemental Figure S6C). Of importance, when a repair template encoding an IRE1α^Y161A mutation (corresponding to yeast Ire1p^F285A; Credle et al., 2005; Zhou et al., 2006) was used in place of the wild type, activation of XBP1s::Turquoise by tunicamycin or 2DG was lost, but a measure of response to palmitate loading reproducibly remained (Figure 6, C and D, and Supplemental Figure S6, E–G). The residual responsiveness of IRE1α^Y161A to palmitate was clearly a feature of the mutant luminal domain, as it was absent from the parental ∆LD15 cells (Supplemental Figure S6E). These findings indicate that the Y161A mutation in IRE1α drives a wedge between the responsiveness to unfolded proteins and membrane aberrancy at levels of expression that are lower than the wild type. The mechanism by which the Y161A mutation interferes with IRE1α’s responsiveness to unfolded protein stress remains to be determined; nonetheless, this point mutation provides a platform for selectively testing the effect of secondary mutations on IRE1α’s ability to recognize membrane aberrancy.

Membrane aberrancy is robustly sensed by TMD-mutant IRE1α

Building on this finding, we retargeted the Ern1^∆LD15 locus with CRISPR/Cas9 while offering a human cDNA (“minigene”)-based repair template encoding the IRE1α^Y161A mutation alongside the compound IRE1α^{V437R; L441R}, an IRE1α^∆403-441 deletion (encompassing the predicted amphipathic helix), or IRE1α^TM>L, in which the TMD was replaced with a string of leucines (Figure 7, A and B). In the context of the IRE1α^Y161A modification, which biases IRE1α away from sensing unfolded proteins and allows the response to lipid aberrancy to dominate, all three mutants (IRE1α^{V437R; L441R}, IRE1α^∆403-441, and IRE1α^TM>L) retained their ability to activate XBP1s::Turquoise reporter in palmitate-treated cells at levels similar to those of the reference IRE1α^Y161A mutant (Figure 7, C and D). Replacing the TMD residues of IRE1α with leucine (IRE1α^TM>L) sensitized IRE1α^Y161A to membrane aberrancy imposed by palmitate and to a lesser degree to tunicamycin (Figure 7, D and E). This feature of the compound Ern1^Y161A;TM>L mutation, which was reproducibly observed in two independently isolated clones (Supplemental Figure S7), was all the more striking, given that the encoded compound mutant IRE1α^Y161A;TM>L protein is expressed at lower levels than IRE1α^Y161A, which serves as a reference in this experiment (Figure 7A).

The experiments noted here point to an unanticipated robustness of IRE1α’s ability to respond to lipid aberrancy by promoting the unconventional splicing of the XBP1s::Turquoise reporter. This was further examined by studying the response of endogenous XBP1 and its target genes in cells expressing the aforementioned IRE1α derivatives as their only source of IRE1α protein. The Y161A mutation selectively enfeebled the response to unfolded proteins also when assessed by IRE1α-mediated XBP1 splicing, although the defect was slightly less complete than that observed by the flow cytometry–based measurement of XBP1s::Turquoise activity (compare Figure 8A with Figure 6, C and D). In the context of the IRE1α^Y161A mutation, both IRE1α-dependent splicing of endogenous XBP1 mRNA (Figure 8, A and B, and Supplemental Figure S8) and the expression of two well-validated IRE1α-dependent UPR target genes (Dnajb9/ERdJ4 and Sec61a1; Shoulders et al., 2013; Adamson et al., 2016) were unaffected by secondary IRE1α^{V437R; L441R}, IRE1α^∆403-441, or IRE1α^TM>L mutations in palmitate-treated compound mutant cells (Figure 8C).

Plasmid construction

A combination of PCR-based manipulations, restriction digests, and site-directed mutagenesis procedures was used to create DNA constructs. Supplemental Table S1 lists the plasmids used, their lab names, description, and published reference where available.

Preparation of palmitate–bovine serum albumin complex

Palmitate–bovine serum albumin (BSA) complex was prepared by modification of a published method (Spector, 1986). Briefly, a 20 mM solution of sodium palmitate (P9767; Sigma-Aldrich, United Kingdom) in 0.01 M NaOH was incubated at 70°C for 30 min. Prewarmed palmitate solution was added dropwise to 20% fatty acid–free BSA (A6003; Sigma-Aldrich) in phosphate-buffered saline (PBS) at a 6.6:1 palmitate-to-BSA molar ratio. The resultant palmitate–BSA complex solution was filtered, aliquoted, and stored at −20°C.

Cell culture and treatment

CHO-K1–based cell lines were maintained in regular medium consisting of Nutrient Mixture F-12 Ham (Sigma-Aldrich), 10% fetal calf serum (FCS; FetalClone II, Lot ABB214492; Hyclone-GE Healthcare Life Sciences, South Logan, UT), 2 mM l-glutamine (Sigma-Aldrich), and 1× penicillin/streptomycin (Sigma-Aldrich) at 37°C with 5% CO₂. HEK293T cells were cultured in DMEM (Sigma-Aldrich) supplemented as described.

Cells were treated with drugs at the following final concentrations: 0.080–2.0 µg/ml tunicamycin (Melford, United Kingdom), 0.13–16 mM 2DG; Acros Organics, Belgium), 0.025–0.63 mM palmitate (as a BSA complex; see earlier description), and 6-8 μg/ml puromycin (Calbiochem, La Jolla, CA). All drugs were first diluted in fresh, prewarmed medium and then applied to the cells by medium exchange. For palmitate treatment, 1% FCS medium was used.

Generation of the genome edited cells by CRISPR/Cas9

Flow cytometry analysis

S21 cells and their derivatives were plated at a density of 1 × 10⁵ cells per well on 12-well plates. On the next day, the culture medium was replaced with 1 ml of fresh medium, and cells were treated with indicated compounds for 24 h. Immediately before analysis, the cells were washed with PBS and collected in PBS containing 4 mM EDTA. Single-cell fluorescence signals (10,000 cells/sample) were measured by a dual-channel flow cytometry with an LSRFortessa cell analyzer (Becton Dickinson, Franklin Lakes, NJ). GFP (excitation laser 488 nm, filter 530/30) and Turquoise (modified cyan fluorescent protein; excitation laser 405 nm, filter 450/50) signals were detected. FlowJo software (FlowJo, Ashland, OR) was used to analyze the data.

Immunoprecipitation and immunoblot

Membrane-associated proteins were prepared from digitonin-permeabilized cells as previously described (Le Gall et al., 2004), with some modifications. Three 10-cm dishes of confluent cells were washed with PBS, harvested with PBS containing 4 mM EDTA, centrifuged at 400 × g for 5 min at 4°C, and washed with HCN buffer (50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid [HEPES], pH 7.5, 150 mM NaCl, 2 mM CaCl₂). Cells were then resuspended in 1.5 ml of HCN buffer containing 0.09% (wt/vol) digitonin (300410; Calbiochem) and incubated on ice for 10 min. The digitonin-permeabilized cells were centrifuged (1000 × g for 5 min at 4°C), washed twice with HNE buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM ethylene glycol tetraacetic acid), lysed in a lysis buffer (1% Triton X-100, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 10% glycerol, 2 mM phenylmethylsulfonyl fluoride, 10 µg/ml aprotinin, 4 µg/µl pepstatin, 4 µM leupeptin, 10 mM sodium pyrophosphate, 100 mM NaF, 17.5 mM β-glycerophosphate) on ice for 5 min. The lysates were cleared for 10 min at 21,000 × g at 4°C. Bio-Rad protein assay reagent (BioRad, Richmond, CA) was used to determine the protein concentrations of lysates, followed by normalization. Supplemental Figure S10 presents an experimental analysis of the extractability of wild-type and mutant IRE1α.

For immunoprecipitation of IRE1α protein, 20 µl of protein A–Sepharose (Invitrogen) and 1 µl of anti-IRE1α rabbit polyclonal antibody (NY200; Bertolotti et al., 2000) were added to cell lysates and rotated overnight at 4˚C. The beads were washed in 0.6 ml of lysis buffer three times. The beads were mixed with 40 µl of 2 × SDS-sample buffer and incubated at 70˚C for 10 min. After spinning down of beads, supernatant was collected and subjected to immunoblot analysis.

A 20-µl amount of immunoprecipitated samples or 20–40 µg of protein of lysates was subjected to 7% SDS–PAGE and transferred onto Immobilon-P polyvinylidene fluoride membrane (Merck Millipore, Bedford, MA). Membranes were blocked with 5% (wt/vol) dried skimmed milk in TBS (25 mM Tris-HCl, pH 7.5, 150 mM NaCl) and incubated with primary antibodies, followed by IRDye fluorescently labeled secondary antibodies (LiCor, Lincoln, NE). The membranes were scanned with an Odyssey near-infrared imager (LiCor). Primary antibodies and antisera against IRE1α (NY200; Bertolotti et al., 2000), calnexin (ab22595; Abcam, Cambridge, United Kingdom), and hamster BiP (Avezov et al., 2013) were used.

XBP1 splicing assay and quantitative real-time PCR

Total RNA was isolated from cells by acid guanidinium thiocyanate-phenol-chloroform extraction using RNA STAT 60 (Amsbio, Lake Forest, CA) and isopropanol precipitation. A 2-µg amount of RNA was reverse transcribed by a reverse transcriptase RevertAid (ThermoFisher Scientific, Grand Island, NY) with oligo(dT)18 primer.

For XBP1 splicing assay, unspliced and spliced Xbp1 cDNA fragments were amplified by PCR using primers 5 and 1470 and separated by 4% agarose gel containing SYTO60 (Invitrogen). Gels were visualized and quantified using an Odyssey near-infrared imager. The percentage of XBP1 splicing was calculated using the equation

where I(XBP1u), I (XBP1s), and I (XBP1^HD) indicate band intensities of unspliced XBP1, spliced XBP1, and a heteroduplex of unspliced and spliced XBP1, respectively.

Quantitative PCR analysis was performed using Power SYBR Green PCR Master Mix (Applied Biosystems) according to the manu­facturer’s instruction on a 7900HT Fast Real-Time PCR system (Applied Biosystems). Oligo DNAs in Supplemental Table S2 were used for PCR: 1929 and 1930 for Dnajb9, 1931 and 1932 for Sec61a1, and 1487 and 1488 for Ppia. Relative quantities of amplified PCR products were determined using SDS 2.4.1 software (Applied Biosystems) and normalized to Ppia values.