ISRIB Blunts the Integrated Stress Response by Allosterically Antagonising the Inhibitory Effect of Phosphorylated eIF2 on eIF2B

Abstract

The small molecule ISRIB antagonizes the activation of the integrated stress response (ISR) by phosphorylated translation initiation factor 2, eIF2(αP). ISRIB and eIF2(αP) bind distinct sites in their common target, eIF2B, a guanine nucleotide exchange factor for eIF2. We have found that ISRIB-mediated acceleration of eIF2B's nucleotide exchange activity in vitro is observed preferentially in the presence of eIF2(αP) and is attenuated by mutations that desensitize eIF2B to the inhibitory effect of eIF2(αP). ISRIB's efficacy as an ISR inhibitor in cells also depends on presence of eIF2(αP). Cryoelectron microscopy (cryo-EM) showed that engagement of both eIF2B regulatory sites by two eIF2(αP) molecules remodels both the ISRIB-binding pocket and the pockets that would engage eIF2α during active nucleotide exchange, thereby discouraging both binding events. In vitro, eIF2(αP) and ISRIB reciprocally opposed each other's binding to eIF2B. These findings point to antagonistic allostery in ISRIB action on eIF2B, culminating in inhibition of the ISR.

Keywords: CRISPR/Cas9-homologous recombination; cell stress; drug action; eukaryotic initiation factor-2B; mRNA translation; phosphorylation; protein binding; protein biosynthesis/drug effects; protein conformation.

Graphical abstract

Figure 1

ISRIB Accelerates eIF2B Guanine Nucleotide Exchange Activity Selectively in Presence of eIF2(αP) (A) Plots of eIF2B guanine nucleotide exchange activity as reflected in time-dependent decrease in fluorescence of BODIPY-FL-GDP bound to non-phosphorylatable eIF2(α^S51A) (125 nM). Blue traces lack and green traces include ISRIB (250 nM), the gray trace lacks eIF2B. The size of the symbol reflects the concentration of eIF2B in the assay. All the data points of a representative experiment performed in duplicate are shown. The half-life of GDP binding with 95% confidence interval (CI) for each plot is indicated (observation reproduced three times). (B) As in (A) but utilizing a fixed concentration of wild-type or the indicated ISR-insensitive eIF2B mutants (40 nM), BODIPY-FL-GDP bound to non-phosphorylatable eIF2(α^S51A) (125 nM) and where indicated, eIF2(αP) (1 μM) and ISRIB (250 nM). Plotted are the mean fluorescence values ± SD of an experiment performed in technical triplicate. (C) In vivo characterization of the ISR in wild-type and mutant CHO cells of the indicated genotype. Shown are histograms of the activity of CHOP::GFP (an ISR reporter gene) in untreated cells and cells in which the eIF2α kinase GCN2 had been activated by L-histidinol in the absence or presence of ISRIB. Shown is a representative experiment reproduced three times. (D) Estimates of protein synthesis rates in CHO cells of the indicated genotype before and after activation of the eIF2α kinase PERK by thapsigargin (Tg). The lower panel is an anti-puromycin immunoblot of whole cell lysates, in which the intensity of the puromycinylated protein signal reports on rates of protein synthesis. The upper panels are immunoblots of P-eIF2α and of total eIF2α. Below is a stacked column graph of the quantified blot signals (mean ± SD, n = 3): puromycin-labeled in light green (top), P-eIF2α in orange (middle), and total eIF2α in blue (bottom).

Figure 2

eIF2(αP) and ISRIB Associate with Different Conformations of eIF2B (A) Overlay of the eIF2B apo structure (gray) and eIF2B in complex with two eIF2(αP) trimers (the α^Pγ complex; color-coded as in the adjacent cartoon). The blue spheres show the position of the C^α atoms of eIF2Bδ^E310 and δ^L314. (B) Different arrangements of the eIF2B pocket that accommodates the eIF2α-NTD: unphosphorylated in the catalytically productive conformation, and phosphorylated in complexes containing one (α^P1) or two (α^Pγ) bound eIF2(αP) trimers. Upper panels: an overlay of the productive eIF2B⋅ISRIB⋅eIF2 complex (cyan, PDB: 6O81), the eIF2B apo structure (gray), and the α^Pγ complex (magenta). For clarity, only the unphosphorylated eIF2α-NTD of the eIF2B⋅ISRIB⋅eIF2 complex is shown. Lower panels: similar alignment of the apo structure, the α^Pγ structure, and the α^P1 structure (green). Right panels: close-up views showing the displacements of helix δ-α3 between the different complexes. (C) Deformation of the ISRIB-binding pocket in eIF2B with two bound eIF2(αP) trimers (the α^Pγ complex). The eIF2B⋅ISRIB complex structure (PDB: 6CAJ) is shown in gray and the α^Pγ complex in color-coded representation (as in the adjacent cartoon). Key residues known to affect the binding or action of ISRIB are highlighted as spheres. Structures are aligned by the four C-terminal domains of the β- and δ-subunits of eIF2B for (A) and (B), and by the C^α atoms surrounding (within 10 Å) the ISRIB molecule in the eIF2B⋅ISRIB structure for (C). See also Figures S1 and S2.

Figure 3

Phosphorylated eIF2 Attenuates FAM-ISRIB Binding to eIF2B (A) Plot of fluorescence polarization signals (mean ± SD, n = 3) arising from samples of FAM-ISRIB (2.5 nM) incubated with varying concentrations of wild-type or mutant eIF2B. Where indicated, 500 nM unlabeled ISRIB was added as a competitor. K_1/2max with 95% confidence intervals (CI) is shown. (B) Plot of fluorescence polarization signals, at equilibrium, arising from FAM-ISRIB bound to wild-type eIF2B in presence of the indicated concentration of the P-eIF2α-NTD (mean ± SD, n = 3) or eIF2(αP) trimer. The data were fitted by non-linear regression analysis to a “log[inhibitor] versus response four parameter” model. IC₅₀ values with 95% CI are shown. (C) As in (B) above, plot of fluorescence polarization signals, at equilibrium, arising from FAM-ISRIB bound to wild-type or mutant eIF2B (100 nM) in presence of the indicated concentration of eIF2(αP) trimer (mean ± SD, n = 3). IC₅₀ values with 95% CI are shown.

Figure 4

Phosphorylated eIF2 Attenuates FAM-ISRIB Binding to eIF2B on a Timescale Consistent with the ISR (A) Plot of the time-dependent change in fluorescence polarization of FAM-ISRIB bound to wild-type eIF2B, following injection of 1 μM unlabeled (“cold”) ISRIB at t = 0 (green diamonds, mean ± SD, n = 3, and the fit of the first 6 min to a first order decay reaction [k_off = 0.74 min⁻¹; 95% CI, 0.68–0.9 min⁻¹; R², 0.9823], dotted green line). Control samples, unchallenged by “cold” ISRIB (blue circles, mean ± SD, n = 3) and reference samples (n = 1) from the same experiment are shown. (B) Plot of time-dependent change in fluorescence polarization of FAM-ISRIB bound to wild-type or ISR-insensitive mutant eIF2Bs (δ^E310K or δ^L314Q) (60 nM) in presence or absence of 600 nM unphosphorylated wild-type eIF2 (top panel) or non-phosphorylatable eIF2(α^S51A) (bottom panel). Where indicated, at t = 0 the eIF2α kinase PERK was introduced to promote a pool of eIF2(αP). Shown are the mean ± SD (n = 3) of the fluorescence polarization values of the PERK-injected samples. The traces were fitted to a first order decay reaction. δ^WT: t_1/2 4.7 min (95% CI 4.5–6.7, R², 0.9344); δ^E310K: t_1/2 153 min, and δ^L314Q: t_1/2 92 min (both with a poor fit to first order decay, R² <0.5). See also Figure S3.

Figure 5

ISRIB Inhibits Binding of eIF2B to the N-Terminal Domain of Phosphorylated eIF2α (A) Biolayer interferometry (BLI) traces of the association and dissociation phases of eIF2B decamers (100 nM) or eIF2B^βδγε tetramers (400 nM) in the absence (DMSO, in blue) or presence of ISRIB (1 μM, in pink) to and from the biotinylated P-eIF2α-NTD immobilized on the BLI probe. The fits to a 2-phase association and a 2-phase dissociation model are indicated by the gray dashed line. Shown are mean ± SD (n = 3) of the sample of eIF2B decamers with and without ISRIB and all the data points of the tetramer samples from a representative experiment conducted three times. (B) Left: BLI traces of consecutive association phases and a terminal dissociation phase of wild-type and the indicated eIF2B mutants in the absence (DMSO, in blue) or presence of ISRIB (1 μM, in pink) as in (A). The probe was reacted with escalating concentrations of eIF2B (9–150 nM) ± ISRIB, before dissociation in the respective buffer. Shown are all the data points of a representative experiment performed three times. Right: plots of the mean and 95% confidence interval (CI) of the plateau values of the association phases (obtained by fitting the data from the traces on the left to a 2-phase association non-linear regression model) against the concentration of eIF2B. The dotted line reports on the fit of plots to a one site specific binding with Hill slope = 1 non-linear regression model. (C) Left: time-dependent change in BLI signal in the dissociation phase of eIF2B (previously associated in absence of ISRIB) from the biotinylated P-eIF2α-NTD in the presence of escalating concentration of ISRIB. The mean and 95% CI of the fraction of the dissociation attributed to the fast phase (%Fast) was calculated by fitting the dissociation traces to a biphasic model. The fit is indicated by the gray dotted lines. The overlaying data points in light gray are shown for 0 nM (small circles on top curve) and 10 nM (large squares on bottom curve) ISRIB only. Shown are traces from a representative experiment performed three times. Right: plot of the %Fast of the dissociation reactions to the left, as a function of ISRIB concentration. The plot was fitted to an [Agonist] versus response (Hill slope = 1) non-linear regression model (dotted line) yielding an EC₅₀ of 1.8 (95% CI, 0.8–2.8) nM. See also Figure S4.

Figure 6

Attenuated ISRIB Action in Cells Lacking eIF2(αP) (A) Characterization of the ISR in Eif2S1^S51A mutant CHO cells (lacking phosphorylatable eIF2α) depleted of eIF2B subunits by CRISPR/Cas9 targeting of their encoding genes. Two different guides for either beta (β1, β2) or delta (δ1, δ2) subunit, and one guide for epsilon subunit (ε) were transfected separately. Where indicated, the cells were exposed continuously to ISRIB (1 μM), commencing at the point of transduction with the CRISPR/Cas9 encoding plasmids and continued until harvest. Shown are histograms of the CHOP::GFP ISR reporter in populations of cells 48, 72, and 96 h following eIF2B gene targeting (±ISRIB) from a representative experiment performed three times. The mean ± SD (n = 3) of the ratio of fluorescent signal of the ISR-induced population (the right peak on the histograms) to the non-induced (left peak) are plotted to the right (^∗p < 0.05 Student’s t test). (B) Immunoblot of 3xFLAG-tagged endogenous eIF2Bγ detected with anti-FLAG M2 antibodies in CHO or 3xFLAG-tagged endogenous eIF2Bβ in HeLa cell lysates that were either treated with DMSO (top panel) or ISRIB (bottom panel) and resolved on a 10%–40% glycerol density gradient in a buffer of physiological salt concentration. The position of reference proteins of the indicated molecular weight in this gradient is indicated below the image and the arrows point to the predicted position of eIF2Bβδγε tetramers and eIF2B(α)₂(βδγε)₂ decamers. (C) As in (A) above, but following CRISPR/Cas9-mediated depletion of eIF2B’s substrate eIF2, by targeting the Eif2S1 gene encoding its α-subunit. Two different guides, eIF2α-A and eIF2α-B, were transfected separately. Shown is a representative experiment performed twice. The mean ± SD (n = 2) of fluorescent signals ratio as in (A) are plotted to the right (^∗∗p < 0.05 Student’s t test).

Figure 7

A Model of the Functional Consequences of the Antagonism between eIF2(αP) and ISRIB Binding to eIF2B (A) Cartoon of the ISRIB-binding pocket in the active (ground) state of eIF2B (left), eIF2B bound by one eIF2(αP) trimer (center), and eIF2B bound by two eIF2(αP) trimers (right). (B) The binding of one eIF2(αP) trimer partially inhibits the catalytic activity by a steric block of the active site induced by the docking of the γ-subunit of eIF2(αP) onto eIF2Bγ (state I). Binding of a second eIF2(αP) trimer occludes the second active site of eIF2B by a similar steric block but also interferes with catalytic activity through allosteric inhibition that deforms both pockets for productive binding of eIF2α, strongly inhibiting the catalytic activity of eIF2B (state II). (C) The presence of ISRIB hierarchically antagonizes the binding of eIF2(αP). Weaker antagonism toward one bound eIF2(αP) trimer (states I and III), as ISRIB’s binding pocket is less deformed under these circumstances. The stronger deformation of the ISRIB-binding pocket that is coupled to the binding of a second eIF2(αP) trimer sets up a competition whereby state IV is most strongly disfavored by ISRIB (rendered opaque in the cartoon). By stabilizing the ground state and the partially inhibited state III, ISRIB pulls the equilibrium away from the strongly inhibited state II, thus antagonizing the ISR. High enough concentrations of eIF2(αP) competitively override this inhibition.