Endoplasmic reticulum stress activates human IRE1α through reversible assembly of inactive dimers into small oligomers

Abstract

Protein folding homeostasis in the endoplasmic reticulum (ER) is regulated by a signaling network, termed the unfolded protein response (UPR). Inositol-requiring enzyme 1 (IRE1) is an ER membrane-resident kinase/RNase that mediates signal transmission in the most evolutionarily conserved branch of the UPR. Dimerization and/or higher-order oligomerization of IRE1 are thought to be important for its activation mechanism, yet the actual oligomeric states of inactive, active, and attenuated mammalian IRE1 complexes remain unknown. We developed an automated two-color single-molecule tracking approach to dissect the oligomerization of tagged endogenous human IRE1 in live cells. In contrast to previous models, our data indicate that IRE1 exists as a constitutive homodimer at baseline and assembles into small oligomers upon ER stress. We demonstrate that the formation of inactive dimers and stress-dependent oligomers is fully governed by IRE1's lumenal domain. Phosphorylation of IRE1's kinase domain occurs more slowly than oligomerization and is retained after oligomers disassemble back into dimers. Our findings suggest that assembly of IRE1 dimers into larger oligomers specifically enables trans-autophosphorylation, which in turn drives IRE1's RNase activity.

Keywords: IRE1; UPR; cell biology; endoplasmic reticulum; human; molecular biophysics; single-molecule; stress signaling; structural biology.

Figure 1.. Endogenously tagged IRE1α is fully active despite not forming large clusters.

(A) Schematic representation of IRE1 with a C-terminal HaloTag, the construct used for tagging IRE1 at the endogenous locus. IF1^L and IF2^L refer to the primary dimerization and oligomerization interfaces of the lumenal domain, respectively. (B) RT-PCR analysis of stress-dependent XBP1 mRNA splicing in WT U-2 OS cells, IRE1 knock-out (KO) U-2 OS cells, and U-2 OS cells in which IRE1 has been fully edited with a C-terminal HaloTag. Tm indicates treatment with 5 μg/ml tunicamycin. (C) Immunoblot of UPR activation in response to 5 μg /ml tunicamycin (left) and 100 nM thapsigargin (right) treatments in the three cell lines shown in panel B. (D) Maximum intensity projections of representative spinning-disk confocal images of live cells expressing endogenously tagged IRE1-HaloTag, labeled with the JF549 dye. Regions shown with yellow boxes are enlarged below. (E) Same as D, except the cells have been treated with 5 μg/ml tunicamycin for 5 hr.

Figure 1—figure supplement 1.. Comparison of high- and low-expression clones.

(A) Immunoblot showing IRE1 expression levels and UPR activation in WT U-2 OS cells, IRE1 KO U-2 OS cells, partial KO cells used as the parental cell line for generating HaloTag knock-ins, and two clones of endogenously labeled HaloTag (with high and low IRE1 expression levels). Note the shift in protein size due to the addition of the HaloTag and the absence of a WT IRE1 band in the two clones on the right. (B) Flow cytometry analysis of the low-and high-expressing clones shown in panel A. Cells were labeled with 5 nM JF549-HaloTag dye for 1 hr prior to the start of the flow cytometry experiment. Note the unimodal intensity distributions of both clones, ruling out the possibility that the lower-expressing clone simply contains a bimodal mixture of low- and high-expressing cells. Error bars represent 95% confidence intervals.

Figure 1—figure supplement 2.. qPCR analysis of of XBP1 splicing, RIDD, and RIDDLE activity in IRE1-HaloTag cells.

Relative abundance of XBP1u (A), XBP1s (B), DGAT2 (C), BCAM (D) and TGOLN2 (E) mRNA. HPRT1 mRNA expression was used as housekeeping gene. All values are normalized to the amount of mRNA present in unstressed WT cells.

Figure 1—figure supplement 3.. Examples of stress-induced clustering of IRE1-HaloTag in the context of overexpression.

Representative confocal microscopy images of IRE1 KO U-2 OS cells transiently transfected with IRE1-HaloTag (the same construct used throughout this study) and treated with 5 µg/ml tunicamycin for 4 hr. Expression levels following transient transfections are highly variable, and cells expressing visibly high amounts of IRE1-HaloTag readily form clusters detectable by confocal microscopy in response to pharmacologically induced ER stress.

Figure 2.. Single-particle tracking approach for detection of small oligomers.

(A) Schematic depiction of the assay. Cells expressing low levels of HaloTag-conjugated proteins are labeled with a mixture of HaloTag-conjugated dyes and imaged by oblique angle illumination. (B) Principle behind the analysis of single-particle data. Fluorescent spots are independently tracked in two channels, and correlated trajectories are identified computationally. (C) Design of the 1 x and 2 x HaloTag controls. (D) Representative frame from a movie of a cell expressing an ER-tethered 2 x tandem HaloTag and labeled with a mixture of JF549 (cyan) and JF646 (red) dyes. (E) Several frames of the boxed region in panel D, with co-localizing spots identified with arrows. (F) Percentage of correlated trajectories from cells expressing the 1 x and 2 x HaloTag controls, comparing data collected in 10 independent experimental replicates. Each data point represents a single cell, typically comprising several hundred trajectories. Error bars represent 95% confidence intervals.

Figure 2—figure supplement 1.. Orthogonal oligomerization controls.

(A) Schematic representation of the genetic dimerization strategy used as a primary control throughout the manuscript. A dual-HaloTag construct is engineered by fusing two HaloTag proteins in tandem, separated by a flexible GS linker, to the C-terminus of a single transmembrane helix targeted to the ER membrane. (B) Orthogonal dimerization control used to rule out the possibility that the internal and C-terminal HaloTags of the construct shown in panel A may have different labeling efficiencies. In this control, ER membrane-tethered HaloTag proteins assemble into constitutive dimers via an internal GST tag. (C) Tetramerization control created by combining the genetic and constitutive dimerization approaches (D) Single-particle tracking results comparing the fraction of correlated trajectories for the constructs shown in panels A, B, and C. Each data point represents a single cell. Error bars represent 95% confidence intervals.

Figure 2—figure supplement 2.. Confocal microscopy images of HaloTag controls.

Representative IRE1 KO cells transiently transfected with the 1 x HaloTag control (A), 2 x HaloTag control (B), 4 x HaloTag control (C), and untransfected (D), labeled with JF549 and JF646 dyes and imaged by spinning-disk confocal microscopy in the JF549 channel. Note the faint ‘clusters’ visible in the untransfected cell in panel D; these are common and illustrate the autofluorescent background that contributes to small apparent ‘clusters’ in cells expressing HaloTag controls of IRE1-HaloTag.

Figure 2—figure supplement 3.. Model for fraction of observed % correlated trajectories as a function of true oligomeric state.

Data points represent mean observed values for % correlated trajectories from monomeric (1 x), dimeric (2 x), and tetrameric (4 x) HaloTag control constructs. Error bars represent standard errors of the mean. The orange line is a fit of the model equation to the three data points shown (refer to “Estimation of IRE1 cluster stoichiometry” section of Materials and methods).

Figure 3.. Detection of IRE1 dimers and oligomers in live cells.

(A) Schematic depiction of the assay. IRE1-HaloTag is simultaneously labeled with HaloTag dyes of two different colors, JF549 and JF646. If the protein is purely monomeric, all single-molecule tracks are expected to be either one color or the other. If it is purely dimeric, a fraction of tracks will contain both colors. Such dual-color tracks can then be identified as correlated trajectories. (B) Single frame from a long-exposure movie (100ms per frame) of a cell in which IRE1-HaloTag is labeled with a mixture of JF549 (cyan) and JF646 (red) dyes. (C) Maximum intensity projection of the entire movie from panel B showing that single IRE1 molecules diffuse along ER tubules. (D) Single frame from a short-exposure movie (50ms per frame) of a cell in which IRE1-HaloTag is labeled with a mixture of JF549 (cyan) and JF646 (red) dyes. (E) Kymograph (time vs. position plot) along the line shown in panel D. Co-localizing diffusional IRE1 trajectory is shown with a yellow arrow. (F) Stress-induced changes in IRE1 oligomerization in response to treatment with 5 μg/ml tunicamycin (Tm), as quantified by the fraction of correlated trajectories. Green bars on the left correspond to the 1 x, 2 x, and 4 x HaloTag controls, respectively. Error bars represent 95% confidence intervals.

Figure 3—figure supplement 1.. Effect of ER stress on HaloTag controls.

(A) Single-particle tracking data showing the fraction of correlated trajectories for the 1 x and 2 x HaloTag controls, with and without a 4-hr treatment with tunicamycin. (B) Number of trajectories per movie for the four conditions shown in panel A, demonstrating that the changes in % correlated trajectories are independent of construct expression levels. Each data point represents a single cell. Error bars represent 95% confidence intervals.

Figure 3—figure supplement 2.. Effect of ER stress on the efficiency of HaloTag labeling.

Single-particle tracking data showing the fraction of correlated trajectories for the 1 x, 2 x, and 4 x HaloTag controls, as well as endogenously tagged IRE1-HaloTag cells with dye added at different times relative to the induction of ER stress. ‘Before’ indicates that cells were first treated with tunicamycin and then labeled with a mixture of JF549 and JF646 dyes, while ‘after’ indicates that cells were first labeled with the dye mixture and then stressed with tunicamycin. Each data point represents a single cell. Error bars represent 95% confidence intervals.

Figure 3—figure supplement 3.. Quantification of diffusion from single-particle trajectories.

(A) Kernel density estimates of apparent diffusion constants of HaloTag controls and IRE1-HaloTag with and without stress, obtained by conventional MSD analysis of single particle trajectories. (B) Mean posterior occupation of single-particle diffusion constants of the same four conditions as shown in panel A (see Materials and methods for details).

Figure 4.. Effects of stressors on IRE1 oligomerization.

Oligomerization of endogenously tagged IRE1-HaloTag in U-2 OS cells treated with the indicated ER stressors for the indicated amounts of time. Tunicamycin (Tm) inhibits glycosylation in the ER lumen, thapsigargin (Tg) blocks sarco/endoplasmic reticulum Ca²⁺ pumps, and dithiothreitol (DTT) triggers reduction of disulfide bonds. Green bars on the left correspond to the 1 x, 2 x, and 4 x HaloTag controls, respectively. Error bars represent 95% confidence intervals.

Figure 4—figure supplement 1.. Formation of dimers and oligomers in high- and low-expressing clones of IRE1-HaloTag cells.

Single-particle tracking data showing stress-dependent oligomerization of the high- and low-expressing IRE1-HaloTag clones. IRE1 in the lower-expressing clone remains dimeric in unstressed cells, while the shift to higher-order oligomers upon stress is less prominent than in the higher-expressing clone. Each data point represents a single cell. Error bars represent 95% confidence intervals.

Figure 4—figure supplement 2.. Trajectory density vs. percent correlation.

A plot showing the relationship between the percentage of a calculated trajectories in a given cell against the number of trajectories in the movie collected from that cell, for the four conditions indicated in the box. Solid lines represent linear fits, with the shaded regions around them showing 95% confidence intervals. Each data point represents a single cell.

Figure 5.. Effects of mutations on IRE1 oligomerization.

Oligomerization of the indicated IRE1 mutants transiently transfected into IRE1 KO U-2 OS cells and expressed under the control of the weak CMVd3 promoter. ‘IRE1-HaloTag (endogenous)’ refers to the endogenously tagged IRE1 cells that are shown in Figure 4. Error bars represent 95% confidence intervals.

Figure 5—figure supplement 1.. Trajectory counts of all mutants.

Number of trajectories per movie for all mutants shown in Figure 5, demonstrating that all constructs are expressed at comparable levels and that the changes in % correlated trajectories are independent of construct expression levels. Each data point represents a single cell. Error bars represent 95% confidence intervals.

Figure 6.. Proposed model for human IRE1 activation.

In the absence of external stress, IRE1 is pre-assembled into inactive unphosphorylated dimers via the IF1^L interface of the lumenal domain. Kinase domains within the dimer are positioned in a back-to-back (B2B) orientation, which does not allow for phosphorylation. (Gurevich and Gurevich, 2018) ER stress forces dimers to oligomerize via the IF2^L interface, placing the kinase active sites of adjacent dimers in a face-to-face (F2F) orientation that favors trans-autophosphorylation. Here and throughout the figure, the original dimer is shown in solid blue tones while the newly associated dimer is semi-transparent. (Shattil and Newman, 2004) Phosphorylation at the F2F interface results in a partially phosphorylated oligomer, wherein one protomer of each dimer is phosphorylated and one is not. The relative activity of the RNase domains in this state is unknown. (Chung, 2017) At this point, the oligomer may dissociate into partially phosphorylated dimers. (Reich et al., 1997) Another dimer associates with the partially phosphorylated dimer via the second IF2^L interface, catalyzing phosphorylation of the second protomer of the original dimer. Note that this may either occur sequentially, as shown here, or simultaneously with step 2, if multiple dimers assemble into a hexamer or larger oligomer. Phosphorylated IRE1 now has an active RNase domain and dissociates into fully active dimers (Kaufman, 1999). Eventually, dimers are dephosphorylated by phosphatases and return back into the inactive state.