A Sephin1-insensitive tripartite holophosphatase dephosphorylates translation initiation factor 2α

The integrated stress response (ISR) is regulated by kinases that phosphorylate the α subunit of translation initiation factor 2 and phosphatases that dephosphorylate it. Genetic and biochemical observations indicate that the eIF2αP-directed holophosphatase, a therapeutic target in diseases of protein misfolding, is comprised of a regulatory subunit, PPP1R15, and a catalytic subunit, protein phosphatase 1 (PP1). In mammals, there are two isoforms of the regulatory subunit, PPP1R15A and PPP1R15B, with overlapping roles in the essential function of eIF2αP dephosphorylation. However, conflicting reports have appeared regarding the requirement for an additional co-factor, G-actin, in enabling substrate-specific dephosphorylation by PPP1R15-containing PP1 holoenzymes. An additional concern relates to the sensitivity of the holoenzyme to the [(o-chlorobenzylidene)amino]guanidines Sephin1 or guanabenz, putative small-molecule proteostasis modulators. It has been suggested that the source and method of purification of the PP1 catalytic subunit and the presence or absence of an N-terminal repeat–containing region in the PPP1R15A regulatory subunit might influence the requirement for G-actin and sensitivity of the holoenzyme to inhibitors. We found that eIF2αP dephosphorylation by PP1 was moderately stimulated by repeat-containing PPP1R15A in an unphysiological low ionic strength buffer, whereas stimulation imparted by the co-presence of PPP1R15A and G-actin was observed under a broad range of conditions, low and physiological ionic strength, regardless of whether the PPP1R15A regulatory subunit had or lacked the N-terminal repeat–containing region and whether it was paired with native PP1 purified from rabbit muscle or recombinant PP1 purified from bacteria. Furthermore, none of the PPP1R15A-containing holophosphatases tested were inhibited by Sephin1 or guanabenz.

The integrated stress response (ISR) 6 is a signal transduction pathway that couples diverse stressful conditions to the activation of a rectifying translational and transcriptional program that is implicated in biological processes ranging from memory formation to immunity and metabolism (reviewed in Ref. 1). The mammalian ISR and its yeast counterpart (the general control response) are initiated by phosphorylation of the ␣ subunit of translation initiation factor 2 (eIF2␣) on serine 51 (2,3), and its activity is terminated by eIF2␣ P dephosphorylation.
Despite genetic evidence pointing to the sufficiency of the conserved C-terminal portion of PPP1R15 in reversing the eIF2␣ P -dependent ISR in vivo (4,5,10), complexes formed in vitro between PPP1R15 regulatory subunit fragments and PP1 have not been observed to accelerate eIF2␣ P dephosphorylation. Dephosphorylation of eIF2␣ P is no faster by a complex of PPP1R15A-PP1 (or PPP1R15B-PP1) than by PP1 alone, showing that, when added as single components, PPP1R15A/B do not influence k cat or K m of PP1 toward the substrate eIF2␣ P (10). However, addition of G-actin to the binary complex of PPP1R15 and PP1 selectively accelerates eIF2␣ P dephosphory-lation. G-actin binds directly to the conserved C terminus of PPP1R15 alongside PP1 to form a ternary complex, whose affinity (K d ϳ10 Ϫ8 M) matches the EC 50 of G-actin's stimulatory effect (10,13).TheinvivorelevanceofG-actinforeIF2␣ P dephosphorylation is attested to by the finding that actin sequestration in fibers (as F-actin) enfeebles eIF2␣ P dephosphorylation, implying a role for factors that affect the actin cytoskeleton in ISR regulation (14).
The ability to dephosphorylate eIF2␣ P is an essential function in developing mammals (15). Nonetheless, inactivation of the PPP1R15A gene, which decelerates eIF2␣ P dephosphoryla-tion and prolongs the ISR, is protective in certain cellular and animal models of diseases associated with enhanced unfolded protein stress (16 -19). This has generated interest in targeting the PPP1R15A-containing holophosphatase for inhibition by small molecules (reviewed in Ref. 20), an endeavor that requires detailed knowledge of the enzymatic mode of action.
A recent report challenged the need for G-actin as a cofactor in PPP1R15A-mediated eIF2␣ P dephosphorylation (21). Instead, it suggested that a binary complex assembled from PP1␣ and a fragment of PPP1R15A (PPP1R15A 325-636 ), encompassing both the C-terminal PP1-binding region and the N-termi-  Table S1). Key residues used for truncated versions of the proteins in this study are annotated. The ER localization domain and the proline, glutamate, serine, and threonine-rich (PEST) repeats are highlighted, as are the PP1 and G-actin binding sites in the conserved C-terminal region. The MBP solubility tag is also represented in the cartoons of the constructs. B, top panel, Coomassiestained PhosTag SDS-PAGE containing resolved samples of dephosphorylation reactions (30 min at 30°C) in which 2 M eIF2␣ P was dephosphorylated by PP1 N purified from rabbit skeletal muscle in the presence or absence of PPP1R15A 325-636 -MBP (50 nM) and/or G-actin (400 nM).The position of the various protein species is indicated. eIF2␣ P and eIF2␣ 0 refer to the phosphorylated and nonphosphorylated form of the bacterially expressed N-terminal domain (residues 1-185) of eIF2␣, respectively. Note that both G-actin and PP1 N preparation gave rise to two bands: a major full-length species and minor degradation product in the case of G-actin and a PP1and tropomyosin band in the case of PP1 N (see also Fig. S1). Shown is a representative experiment of two independent repetitions performed. Center panel, plot of the rate of eIF2␣ P dephosphorylation as a function of the concentration of PP1 N from lanes 1-12 of the experiment above. Bottom panel, plot of the velocity of each enzyme relative to the mean of velocity of PP1 alone calculated from all the informative reactions in the two repeats of this experiment. Statistical significance was derived from Mann-Whitney test (ns, nonsignificant, p Ͼ 0.05; ***, p Յ 0.001). C, as in B but using bacterially expressed PP1␣ as the catalytic subunit (96, 48, 24, or 12 nM), MBP-PPP1R15A 325-636 (50 nM), and G-actin (400 nM). The assays were performed during 20 min at 30°C. Shown is a representative experiment of two independent repetitions performed.

Reconstitution of a PPP1R15A-containing holophosphatase
nal repeat-containing extension, dephosphorylates eIF2␣ P faster than PP1 alone (21). Importantly, dephosphorylation of eIF2␣ P by this active binary complex was reported to be selectively inhibited in vitro by guanabenz and Sephin1, two structurally related small molecules reputed to function in vivo as proteostasis modifiers (22,23). The new study contradicts previous observations that neither a PPP1R15A-PP1 binary complex nor a PPP1R15A-PP1-G-actin ternary complex were susceptible to inhibition by guanabenz or Sephin1 (9,13).
Here we address three important questions raised by these discrepant reports. Does the isotype of the PP1 catalytic subunit or its source (recombinant versus native) influence the requirement for G-actin by the eIF2␣ P -directed holophosphatase? What role does the N-terminal repeat-containing region of PPP1R15A play in eIF2␣ P dephosphorylation by the holophosphatase? Do these factors influence the sensitivity of eIF2␣ P dephosphorylation to guanabenz and Sephin1?

Both native PP1 and bacterially expressed PP1␣ require the presence of G-actin to promote PPP1R15A-regulated eIF2␣ P dephosphorylation
PP1 produced in Escherichia coli may differ in its enzymatic activity from PP1 purified from animal tissues, both in its substrate specificity and in its sensitivity to regulatory subunits (reviewed in Ref. 24). To determine whether the G-actin dependence of PP1-PPP1R15A-mediated eIF2␣ P dephosphorylation is a peculiarity of the bacterially expressed PP1␥ isoform used previously (10, 13), we purified the native catalytic subunit of PP1 from rabbit skeletal muscle (PP1 N ), following an established protocol (25), and compared the two PP1 preparations. Native PP1 (PP1 N ) is a mixture of PP1␣, PP1␤, and PP1␥ isoforms and gave rise to two prominent bands on SDS-PAGE (Fig.  S1A, left panel). The mass spectra of tryptic peptides derived from the PP1 N sample were analyzed by Maxquant with iBAQ (intensity-based absolute quant) to identify the major contaminating species (tropomyosin), and to estimate the relative contribution of PP1 and contaminants to the protein preparation. This enabled a comparison of the catalytic subunit content of PP1 N preparation with the bacterially expressed PP1␥, which served as a reference.
PP1 purified from rabbit muscle is a mixture of ␣, ␤, and ␥ isoforms, whereas it has been reported that the PP1␣ isoform possesses in vivo selectivity for PPP1R15A (6). Therefore, we prepared bacterially expressed PP1␣ by a method that promotes its native-like state (28). To control for effects the location of the tag might have on activity, we also generated an N-terminally MBP-tagged PPP1R15A 325-636 (MBP-PPP1R15A 325-636 ; Fig. 1A and Table S1). The holophosphatase comprised of PP1␣ and MBP-PPP1R15A 325-636 also exhibited a stringent requirement for G-actin (Fig. 1C).
A concentration-dependent stimulatory effect of PPP1R15A on eIF2␣ P dephosphorylation by the three component holoenzyme (PP1, PP1R15A, and G-actin) was observed with constructs tagged at either their N or C termini and with either native or bacterially expressed PP1 (Fig. 2, A and B). The difference in EC 50 values obtained for PPP1R15A 325-636 -MPB with PP1 N (58 nM) or MBP-PPP1R15A 325-636 with PP1␣ (6 nM) may reflect the effect of the position of the MBP tag, the contaminating tropomyosin (in PP1 N ), or both. Importantly, the data agreed with similar experiments in which PPP1R15A 325-636 and bacterially expressed PP1␥ were used, with an EC 50 of 10 nM (see Fig. 8A in Ref. 13).
G-actin also exerted a saturable concentration-dependent stimulatory effect on the activity of a three-component holophosphatase constituted with native PP1 N (Fig. 2C). The EC 50 for G-actin with PP1 N (30 nM) was similar to that observed previously using bacterially expressed PP1␥, with an EC 50 of 13 nM (see Fig. 2C in Ref. 13). Hence, despite variations in the estimated EC 50 values for PPP1R15A or G-actin, the combinations of catalytic and regulatory subunits tested showed consistent PPP1R15A and G-actin concentration-dependent enzymatic activity. These experiments, conducted in a buffer of physiological ionic strength over a physiological protein concentration range (nanomolar catalytic subunit and micromolar substrate) and over a timescale aimed to minimize the effect of substrate depletion on enzyme kinetics, indicate that neither the source of PP1 nor the position of the tag in PPP1R15A are likely to account for the reported G-actin-independent ability of PPP1R15A to stimulate eIF2␣ P dephosphorylation.

Two-fold stimulation of eIF2␣ P dephosphorylation by repeat-containing PPP1R15A in an unphysiological low ionic strength buffer
To explore the discrepant findings on the G-actin independent stimulatory activity of MBP-PPP1R15A 325-636 , we sought to reproduce the experiments reported in Ref. 21 as closely as possible. We received from the Bertolotti laboratory their expression plasmid. The encoded protein, referred to here as MBPϳPPP1R15A 325-636 (Fig. 1A), differs from the one used above (MBP-PPP1R15A 325-636 ) by the absence of three residues in the linker separating the MBP from PPP1R15A and 11 residues in the linker separating PPP1R15A from the C-terminal polyhistidine tag (Table S1). The MBPϳPPP1R15A 325-636 fusion protein was produced in E. coli and purified as described previously (21), and dephosphorylation reactions were carried out in a salt-free, low ionic strength buffer designed to mimic as closely as possible the one used in that study (50 mM Tris (pH 7.4), 1.5 mM EGTA, and 2 mM MnCl 2 , with the notable exception of 0.5 mM TCEP, added here to prevent inactivation of the catalytic subunit by oxidation).
A 2-fold stimulation of eIF2␣ P dephosphorylation by MBPϳPPP1R15A 325-636 was apparent in reactions conducted at low salt concentration (15 mM) but lost at more

Reconstitution of a PPP1R15A-containing holophosphatase
physiological concentrations (100 mM), whereas the 5-fold stimulatory effect of G-actin was observed at both low and physiological salt concentration (Fig. 3A). The stimulatory effect of MBPϳPPP1R15A 325-636 at low salt concentration depended on the N-terminal repeat-containing region of PPP1R15A (Fig.  3B), as reported previously (21), and was not observed with a nonspecific dephosphorylation substrate (Fig. S2A).
Although modest (2-fold) and confined to nonphysiological, low ionic strength conditions, this stimulatory effect was also reproducibly observed with the MBP-PPP1R15A 325-636 and PPP1R15A 325-636 -MBP proteins used in Figs  Reconstitution of a PPP1R15A-containing holophosphatase effect) and under physiological conditions, the presence of G-actin dominates the kinetics of eIF2␣ P dephosphorylation.

Lengthy incubation of the enzymatic reactions does not uncover PPP1R15A's ability to promote G-actin-independent eIF2␣ P dephosphorylation at physiological salt concentrations
Upon inhibition of the phosphorylating kinase, the eIF2␣ P signal decays with a t1 ⁄ 2 of Ͻ10 min (with no change in the total eIF2␣ content) in both cultured mouse fibroblasts (see Fig. 6 in Ref. 14) and Chinese hamster ovary cells (see Fig. 10 in Ref. 13). Despite the rapid in vivo kinetics of the dephosphorylation reaction, the experiments pointing to G-actin-independent eIF2␣ P dephosphorylation were conducted with long incubations of 16 h at 30°C (21). In the absence of other components, PP1␣ is markedly unstable at 30°C, losing about half of its activity by 1 h and all detectable activity by 3 h (Fig. S3, A and B). Thus, a stabilizing effect of a PP1 binding co-factor might have accounted for the apparent G-actin-independent stimulatory effect of MBP-PPP1R15A 325-636 on PP1␣-mediated eIF2␣ P dephosphorylation at physiological salt concentrations. However, over a range of PP1 concentrations (0.2-200 nM), the presence of MBP-PPP1R15A 325-636 failed to stimulate eIF2␣ P dephosphorylation, regardless of whether PP1 N (Fig. 4A) or PP1␣ (Fig. 4B) was used as the catalytic subunit.

Substrate recruitment by the repeat-containing PPP1R15A 325-512 region plays a secondary role in the kinetics of eIF2␣ P dephosphorylation, and its disruption is unlikely to account for sensitivity to Sephin1
PPP1R15A interacts directly with eIF2␣, both in cells (9) and in vitro (21). This interaction maps to the repeat-containing region of PPP1R15A, residues 325-512, N-terminal to PPP1R15A's PP1-binding domain (Fig. 1A) and was proposed to play an important role in the catalytic cycle of PPP1R15A-containing holoenzymes (21). However, in the presence of G-actin, PPP1R15A 325-636 -MBP and PPP1R15A 533-624 -MBP ( Fig. 1A and Table S1) stimulated eIF2␣ P dephosphorylation similarly when paired either with PP1 N (compare our Figs. 2B and 5A) or with PP1␥ (compare Figs. 8A and 2B in Ref. 13). These findings suggest that the conserved C-terminal PPP1R15 fragment that binds PP1 and G-actin simultaneously is sufficient to promote eIF2␣ P dephosphorylation and to dominate its kinetics in vitro and call into question the importance of the N-terminal repeats in PPP1R15A to the fundamentals of the holoenzyme's catalytic cycle.
We considered that an important contributory role for substrate engagement by the PPP1R15A 325-533 repeat-containing fragment to the catalytic cycle of the holophosphatase might have been masked by compensatory features that diverge between the different regulatory subunit constructs, fortuitously equalizing their activity. To address this possibility, we measured the ability of MBP-PPP1R15A 325-512 containing the repeats but lacking the C-terminal PP1 binding region ( Fig. 1A and Table S1) to compete with MBP-PPP1R15A 325-636mediated (G-actin-dependent) eIF2␣ P dephosphorylation using PP1␣ as the catalytic subunit. Minimal inhibition of the dephosphorylation reaction was observed at competitor concentrations of up to 8 M (Fig. 5B), which is a Ͼ300-fold excess over the MBP-PPP1R15A 325-636 regulatory subunit (present in the reaction at 24 nM) and a concentration of 18-fold above the reported K d of the interaction between MBP-PPP1R15A 325-512 and eIF2␣ P (21).
These data suggest that substrate recruitment by the N-terminal extension of PPP1R15A plays a secondary role in the kinetics of the dephosphorylation reaction in vitro and that the reported role of Sephin1 and guanabenz in disrupting that interaction is unlikely to make an important contribution to their pharmacological activity. Consistent with these conclusions, we found that, under physiological salt conditions where eIF2␣ P dephosphorylation depends on the concentration of PP1␣, MBP-PPP1R15A 325-636 , and G-actin, we were unable to

Reconstitution of a PPP1R15A-containing holophosphatase
Complete inhibition of PPP1R15A-mediated eIF2␣ Pdephosphorylation by Sephin 1 was reported in an assay conducted over 16 h at 30°C in a low ionic strength buffer (21). We wished to test whether the reported Sephin1 inhibition might be unmasked by this long incubation (in which the enzyme is undergoing inactivation; Fig. S3B). Using identical MBPϳPPP1R15A 325-636 and PP1␣ constructs, in an identical low ionic strength buffer and following overnight incubation at 30°C, we observed a 2-fold stimulation of eIF2␣ P dephosphor-ylation by MBPϳPPP1R15A 325-636 (similar to that noted in shorter reactions; Fig. 3). However, even under these conditions, designed to mimic as closely as possible those used in Ref. 21, the presence of 100 M Sephin1 was devoid of an inhibitory effect on substrate dephosphorylation (Fig. S4C).

Discussion
The new experiments presented here cover a range of conditions with realistic concentrations and time regimes. Incorpo- Reconstitution of a PPP1R15A-containing holophosphatase ration of multiple time points and titrations of reaction components enabled a comparison of enzyme kinetics that accounts for the effect of substrate depletion. Our observations were made with four different PPP1R15A preparations, three different PP1 preparations, and both buffer conditions used previously in our laboratory and those used in Ref. 21, all of which consistently show the requirement for G-actin as an additional co-factor in enabling PPP1R15A to stimulate eIF2␣ P dephosphorylation in vitro. Therefore, the results presented here are in keeping with previous observations that G-actin has an essential role in promoting eIF2␣ P dephosphorylation both in vitro and in vivo (10, 13, 14).
The PP1 apo-enzyme is salt-sensitive and inhibited by buffers of physiological ionic strength (29). By contrast, PP1 holoenzymes retain their regulated enzymatic activity at physiological ionic strength (30). These considerations call into question the significance of the 2-fold stimulation of eIF2␣ P dephosphorylation by PPP1R15A 325-636 observed in buffer of low ionic strength. Our experiments also cast doubt on the importance of the physical interaction between the repeat-containing region of PPP1R15A (residues 325-512) and eIF2␣ P in the substrate-specific dephosphorylation reaction carried out with physiological ionic strength. PPP1R15 regulatory subunits are found throughout the animal kingdom, but only their C-terminal ϳ70 residues are conserved (11). This C-terminal fragment contains all the information needed to promote eIF2␣ P dephosphorylation, as exemplified by its selective hijacking by herpesviruses (12) and by experimentally targeted expression in cells (see Fig. 1C in Ref. 10). In complex with G-actin, the conserved C-terminal fragment of the PPP1R15s is also able to direct PP1 to selectively dephosphorylate eIF2␣ P in vitro (Figs. 2 and 5A here and Refs. 10,13).
The prominent stimulatory role of G-actin on eIF2␣ P dephosphorylation, observed both in vivo and in vitro, should not obscure the possibility that binary complex formation with PPP1R15 might also favor eIF2␣ P dephosphorylation independently of G-actin joining the complex. Regulatory subunit binding restricts access to PP1 (24,31), favoring the phosphorylation of one class of substrates over another. Mere exclusion of some substrates from access to the catalytic subunit might accelerate eIF2␣ P dephosphorylation when levels of PPP1R15A levels are sufficiently elevated in cells, even though in vitro (and in the absence of competing substrates), the PPP1R15A-PP1 binary complex is not a faster eIF2␣ P phosphatase than PP1 alone (provided the experiments are conducted at physiological salt concentrations). As neither Sephin1 nor guanabenz affect the stability of the PPP1R15A-PP1 complex (13), it is unlikely that they achieve any measure of inhibition by weakening PPP1R15A's ability to compete with other regulatory subunits for limiting amounts of catalytic subunit. These considerations lead us to propose a dual role for PPP1R15A in cells: diverting limiting amounts of PP1 away from other substrates toward eIF2␣ P and, in conjunction with G-actin as an essential co-activator, stimulating the intrinsic rate of dephosphorylation by the holoenzyme thus formed. Actin, too, has a dual role in stimulating eIF2␣ P dephosphorylation: by stabilizing the PPP1R15-PP1 complex (14), G-actin favors the exclusion of other regulatory subunits while stimulating enzyme kinetics selectively toward eIF2␣ P (Fig. 7).
Here we present no argument against an important function for the divergent N-terminal extensions of PPP1R15 regulatory subunits. This role may play out in terms of subcellular localization (26) or protein stability (32) and might be influenced by a physical interaction with the substrate (9,21). However, our findings argue that the physical interaction noted previously between PPP1R15A residues 325-512 and eIF2␣ P (21) is unlikely to play an important role in formation of the enzymesubstrate complex required for catalysis under physiological conditions, and, hence, its reported disruption by guanabenz or Figure 6. Neither Sephin1 nor GBZ interfere with eIF2␣ P dephosphorylation. A, Coomassie-stained PhosTag SDS-PAGE containing resolved samples from dephosphorylation reactions (20 min, 30°C) in which 2 M eIF2␣ P was dephosphorylated by PP1␣ (24 nM) in the presence or absence of MBP-PPP1R15A 325-636 (60 nM) and/or G-actin (400 nM). The components were preincubated as specified with either Sephin1 (100 M), tautomycin (80 nM), or DMSO (vehicle) for 15 min at room temperature before being added to the reaction. The bottom panel shows a long exposure of the relevant section of the image above corresponding to the phosphorylated and nonphosphorylated forms of eIF2␣. Quantification of the percentage of dephosphorylation (%dP) is shown below the image. Shown is a representative experiment of three independent experiments performed. B, as in A but with guanabenz (GBZ). Shown is a representative experiment of two independent experiments performed.

Reconstitution of a PPP1R15A-containing holophosphatase
Sephin1 is unlikely to underscore an inhibitory effect on eIF2␣ P dephosphorylation.
Most importantly perhaps, the findings presented here argue that the inability of previous efforts to uncover a role for guanabenz or Sephin1 in inhibiting eIF2␣ P dephosphorylation in vitro (9,13) was unlikely to have arisen from choice of catalytic subunit, from features of the PPP1R15A regulatory subunit, or the buffer conditions used. Rather, the findings reported here, made in vitro, reinforce observations that Sephin1 and guanabenz have no measurable effect on the rate of eIF2␣ P dephosphorylation in cells (13). The recent description of PPP1R15A/ GADD34-independent cellular effects of guanabenz (33) and our observations that Sephin1-induced changes in gene expression were noted both in cells lacking PPP1R15A and in cells with nonphosphorylatable eIF2␣ (13) suggest the need to reconsider the role of these two compounds as eIF2␣ P dephosphorylation inhibitors.

Protein expression and purification
The plasmids used to express protein in E. coli and the sequence of the encoded proteins are listed in Tables S1 and S2. PPP1R15A 325-636 -MBP and PPP1R15A 533-624 -MBP were produced as described previously (13). Briefly, proteins were expressed in E. coli BL21 (New England Biolabs, catalog no. C3013) as N-terminally tagged GSH S-transferase fusion proteins and purified by tandem affinity chromatography, bound to a GSH-Sepharose 4B resin and eluted with GSH, followed by an overnight cleavage with tobacco etch virus protease (to remove the glutathione S-transferase tag), binding to amylose beads, and elution in maltose-containing buffer.

Reconstitution of a PPP1R15A-containing holophosphatase
umes of lysis buffer, and proteins were eluted with amylose elution buffer (lysis buffer ϩ 10 mM maltose). MBP-R15A 325-512 purification required an additional buffer exchange step (into lysis buffer) using Centripure P1 desalting columns (EMP Biotech, catalog no. CP-0110) to eliminate maltose (which appeared to interfere with the dephosphorylation reactions when present at high concentrations).
MBPϳPPP1R15A 325-636 (a gift from the Bertolotti laboratory) was expressed and purified as described previously (21) with minor modifications. The isopropyl ␤-D-thiogalactopyranoside-induced culture was maintained for 16 h at 18°C, and 0.5 mM TCEP was included in all buffers, throughout the purification procedure, and in the final dialysis buffer (50 mM Tris (pH 7.4), 200 mM NaCl, and 0.5 mM TCEP).

In vitro dephosphorylation reactions
Unless otherwise stated, dephosphorylation reactions were performed at a final volume of 20 l by assembling 5 l of 4ϫ solution of each component: PP1, PPP1R15A, G-actin, and eIF2␣ P (or their respective buffers). A 10ϫ assay buffer (500 mM Tris (pH 7.4), 1 M NaCl, 1 mM EDTA, 0.1% Triton X-100, and 10 mM MgCl 2 ) was diluted 1:10, supplemented with 1 mM DTT, and used to create working solutions of PP1, PPP1R15A, and eIF2␣ P at the desired concentrations. G-actin working solutions were created using G buffer (2 mM Tris-HCl (pH 8), 0.2 mM ATP, 0.5 mM DTT, and 0.1 mM CaCl 2 ). Holoenzyme components (PP1, PPP1R15A, and G-actin) were combined first, and substrate (eIF2␣ P ) was added last to initiate the reactions, which were conducted under shaking at 500 rpm and at 30°C for the specified time. Dephosphorylation reactions designed to reproduce the observations in Ref. 21 were performed in the assay buffer described therein (50 mM Tris-HCl (pH 7.4), 1.5 mM EGTA (pH 8.0), and 2 mM MnCl 2 ), with the modification that 0.5 mM TCEP was added to disfavor oxidative inactivation of the enzyme. The NaCl content of the final reaction was constrained by the contribution of the protein solutions added to each reaction. To maintain parity between reactions performed with and without PPP1R15A, an equal volume of the PPP1R15A buffer was added to reactions lacking the protein. The final salt concentration in the various reactions is noted in the figure legends.
The stability test of PP1␣ (Fig. S3) was performed by preparing a fresh 240 nM solution of PP1␣ in the assay buffer described above. Separate aliquots were preincubated either at 30°C or on ice for the specified times (30 min to 7 h, see schematic in Fig. S3A). At termination of the preincubation, 5 l of these preincubated solutions were added to 20 l of dephosphorylation reactions as described above.
Reactions were terminated by addition of 10 l of 3ϫ Laemmli buffer supplemented with 100 mM DTT and heating the samples for 5 min at 70°C. A third (10 l) of the final volume was resolved in 12.5% PhosTag SDS gels (Wako, catalog no. NARD AAL-107) at 200 V for 1 h. Gels were stained with Coomassie Instant Blue and imaged on an Odyssey imager (LI-COR, Lincoln, NE).
ImageJ was used to quantify eIF2␣ P dephosphorylation, as reflected by the intensity of the fluorescence arising from the Coomassie stain of the eIF2␣ P and eIF2␣ 0 bands resolved by the PhosTag SDS-PAGE gels and captured as a TIF file on the Odyssey imager. GraphPad Prism v8 was used to fit the plot and perform statistical analysis. Table S3 lists the number of times each experiment was performed. conceived the study, codesigned and conducted the experiments, interpreted the results, created the figures, and co-wrote the paper. Z. C. co-designed the experiments, assisted with the preparation of PP1 from rabbit muscle, interpreted the results, and edited the manuscript. M. S. C. expressed and purified PP1␣ from E. coli, interpreted the results, and edited the manuscript. W. P. oversaw the expression and purification of PP1␣ from E. coli, interpreted the results, and edited the manuscript. M. B. co-designed the experiments, oversaw the purification of PP1 from rabbit muscle, interpreted the results, and edited the manuscript. D. R. conceived the study, co-designed the experiments, interpreted the results, and co-wrote the paper.