Author response: eIF2B activator prevents neurological defects caused by a chronic integrated stress response

Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract The integrated stress response (ISR) attenuates the rate of protein synthesis while inducing expression of stress proteins in cells. Various insults activate kinases that phosphorylate the GTPase eIF2 leading to inhibition of its exchange factor eIF2B. Vanishing White Matter (VWM) is a neurological disease caused by eIF2B mutations that, like phosphorylated eIF2, reduce its activity. We show that introduction of a human VWM mutation into mice leads to persistent ISR induction in the central nervous system. ISR activation precedes myelin loss and development of motor deficits. Remarkably, long-term treatment with a small molecule eIF2B activator, 2BAct, prevents all measures of pathology and normalizes the transcriptome and proteome of VWM mice. 2BAct stimulates the remaining activity of mutant eIF2B complex in vivo, abrogating the maladaptive stress response. Thus, 2BAct-like molecules may provide a promising therapeutic approach for VWM and provide relief from chronic ISR induction in a variety of disease contexts. https://doi.org/10.7554/eLife.42940.001 eLife digest Cells must be able to respond to their changing environment in order to survive. When cells encounter particularly unfavorable conditions, they often react by activating a so-called ‘stress’ response. A group of proteins collectively known as eIF2B helps to regulate this response. In a severe neurological condition called Vanishing White Matter (VWM), the genes that produce the eIF2B proteins contain mutations that make eIF2B less active. As a result, certain cells in people with VWM are always stressed. Six years ago, researchers discovered a molecule that boosts the activity of eIF2B. In 2018, they found that it also works on various mutant forms of eIF2B found in VWM. The molecule had so far only been tested in biochemical laboratory experiments. Now, Wong et al. – including some of the researchers involved in the 2018 study – have tested whether an improved version of the molecule treats VWM in mice. The trial treatment successfully halted all signs of the disease in the mice. The molecule blunted the persistent stress response of the cells in the brain and spinal cord, primarily in a cell type that is severely affected by the human form of VWM. Cells in other parts of the body were spared. Overall, the results of the experiments suggest that an eIF2B activator may prove to be an effective treatment for VWM in humans. It could similarly be used to treat other conditions that activate this abnormal cell stress response. The molecule Wong et al. used is not suitable for use in humans, so work is continuing to find a suitable variant. https://doi.org/10.7554/eLife.42940.002 Introduction eIF2B is the guanine nucleotide exchange factor (GEF) for the GTPase and initiation factor eIF2 and modulation of its activity is central to regulation of protein synthesis rates in all eukaryotic cells. GTP-bound eIF2 associates with the initiator methionyl tRNA and this ternary complex (eIF2-GTP-Met-tRNAi) delivers the first amino acid to the ribosome. GTP is hydrolyzed and eIF2-GDP is released, requiring reactivation by eIF2B to enable a new round of protein synthesis (Hinnebusch and Lorsch, 2012). Four stress-responsive kinases (PERK, HRI, GCN2 and PKR) that detect diverse insults converge on phosphorylation of eIF2 at serine 51 of the α subunit (eIF2α). Phosphorylation converts eIF2 from a substrate of eIF2B into its competitive inhibitor, triggering the integrated stress response (ISR), which reduces translation initiation events and decreases global protein synthesis (Krishnamoorthy et al., 2001; Yang and Hinnebusch, 1996). Concomitantly, the ISR induces the translation of a small set of mRNAs with special sequences in their 5’ untranslated regions, including the transcription factor ATF4 (Harding et al., 2000; Watatani et al., 2008). ATF4 triggers a stress-induced transcriptional program that is thought to promote cell survival during mild or acute conditions but can contribute to pathological changes under severe or chronic insults (Pakos-Zebrucka et al., 2016). Vanishing White Matter (VWM; OMIM 603896) is a rare, autosomal recessive leukodystrophy that is driven by mutations in eIF2B (Leegwater et al., 2001; van der Knaap et al., 2002). The disease is characterized by myelin loss, progressive neurological symptoms such as ataxia, spasticity, cognitive deterioration and ultimately, death (Schiffmann et al., 1994; van der Knaap et al., 1997). Age of VWM onset is variable and predictive of disease progression, ranging from severe prenatal/infantile onset leading to death in months (as in the case of ‘Cree leukoencephalopathy’) to slower-progressing adult onset presentations (Fogli et al., 2002; Hamilton et al., 2018; van der Knaap et al., 2006). Because eIF2B is essential, VWM mutations are restricted to partial loss-of-function in any of the five subunits of the decameric complex (Fogli et al., 2004; Horzinski et al., 2009; Li et al., 2004; Liu et al., 2011). Nearly 200 different mutations have been catalogued to date in the Human Gene Mutation Database, which occur as homozygotes or compound heterozygotes with a different mutation in each allele of the same subunit. Reduction of eIF2B activity is analogous to its inhibition by phosphorylated eIF2α, thus it is congruent and compelling that ISR activation has been observed in VWM patient post-mortem samples (van der Voorn et al., 2005; van Kollenburg et al., 2006). We previously showed that a range of VWM mutations destabilize the eIF2B decamer, leading to compromised GEF activity in both recombinant complexes and endogenous protein from cell lysates (Wong et al., 2018). We demonstrated that the small molecule eIF2B activator ISRIB (for ISR inhibitor) stabilized the decameric form of both wild-type (WT) and VWM mutant complexes, boosting their intrinsic activity. Notably, ISRIB bridges the symmetric dimer interface of the eIF2B central core, acting as a molecular stapler (Tsai et al., 2018; Zyryanova et al., 2018). In addition, we showed that ISRIB attenuated ISR activation and restored protein synthesis in cells carrying VWM mutations (Wong et al., 2018). Although we found that ISRIB rescued the stability and activity of VWM mutant eIF2B in vitro, the ability of this class of molecules to prevent pathology in vivo remained an unanswered question. Knock-in mouse models of human VWM mutations have been characterized, and the severe mutations recapitulate key disease phenotypes such as progressive loss of white matter with concomitant manifestation of motor deficits (Dooves et al., 2016; Geva et al., 2010). Here, we generate an improved tool molecule and demonstrate that sustained eIF2B activation blocks maladaptive induction of the ISR and prevents all evaluated disease phenotypes in a VWM mouse model. Results 2BAct is a novel eIF2B activator with improved in vivo properties in rodents To interrogate efficacy in vivo, we sought a small molecule eIF2B activator with improved solubility and pharmacokinetics relative to ISRIB. To that end, we synthesized 2BAct, which has a differentiated bicyclo[1.1.1]pentyl core and, unlike ISRIB, is no longer symmetric (Figure 1A). 2BAct is a highly selective eIF2B activator and exhibited similar potency to ISRIB in a cell-based reporter assay that measures ISR attenuation (Figure 1—figure supplement 1A and Supplementary file 1A). 2BAct is able to penetrate the central nervous system (CNS) (unbound brain/plasma ratio ~0.3) and also demonstrated dose-dependent oral bioavailability using an aqueous-based vehicle (Supplementary file 1B). Additionally, 2BAct is well-suited for formulation in diet, enabling long-term dosing without effects on body weight in WT mice (Figure 1—figure supplement 1B). The molecule was well-tolerated in the animal studies described here, and did not elicit any relevant effects in a rat cardiovascular (CV) safety study; however, significant anomalies were observed in a dog CV model. This CV safety liability makes this particular molecule unsuitable for human dosing. Figure 1 with 4 supplements see all Download asset Open asset 2BAct normalized body weight gain and prevented motor deficits in male R191H mice. (A) Chemical structure of 2BAct and ATF4-luciferase reporter assay EC50. (B) Schematic of the 2BAct treatment experiment. Body weights were measured weekly for the duration of the experiment. (C) Body weight measurements of male mice along the study. Lines are linear regressions. At the 6–11 week time point when 2BAct treatment was initiated, body weights were not significantly different among the three conditions (p>0.05; two-way ANOVA with Holm-Sidak pairwise comparisons). R191H body weight was significantly lower at all subsequent time points. 2BAct-treated animals caught up to WT animals at the 8–13 week time point, and their weights were not significantly different thereafter. (D) Inverted grid test of muscle strength. Time spent hanging was measured up to a maximum of 60 s. (E–F) Beam-crossing assay to measure balance and motor coordination. Time to cross the beam was measured up to a maximum of 30 s (E), and the number of foot slips/falls was counted (F). For (C)-(F), N = 20 (WT), 19 (R191H) and 21 (R191H + 2 BAct) males were analyzed. Error bars are SD. For (D)-(F), *p<0.05; **p<0.01; ***p<10-3; nsp>0.05 by Mann-Whitney test with Bonferroni correction. https://doi.org/10.7554/eLife.42940.003 2BAct normalized body weight gain in VWM mice We generated a previously described mouse model of VWM that harbors the severe ‘Cree leukoencephalopathy’ mutation, Eif2b5R191H/R191H (hereafter referred to as R191H; Figure 1—figure supplement 2) (Dooves et al., 2016). The homologous human Eif2b5R195H mutation causes an infantile-onset, rapidly progressing form of VWM (Black et al., 1988). R191H mice recapitulate many aspects of the human disease, such as spontaneous myelin loss, progressive ataxia, motor skill deficits, and shortened lifespan (Dooves et al., 2016). We selected this severe disease allele for in vivo studies as pharmacological efficacy in this model is a stringent test for eIF2B activators and, mechanistically, should generalize to milder mutations as seen in vitro (Wong et al., 2018). We confirmed that primary fibroblast lysates from R191H embryos had lower GEF activity than WT lysates, and that 2BAct enhanced this activity threefold (EC50 = 7.3 nM; Figure 1—figure supplement 1C–D). To test efficacy, we undertook a 21-week blinded treatment study with 2BAct and measured intermediate and terminal phenotypes associated with disease progression in R191H mice (Figure 1B). 2BAct was administered orally by providing mice with the compound incorporated in rodent meal. This dosing regimen provided unbound brain exposures 15-fold above the in vitro EC50 at the end of the study, ensuring saturating coverage of the target. At the initiation of the study (6–11 weeks old), WT and R191H males had similar body weights (Figure 1C and Figure 1—figure supplement 3A). However, 1 week later, a significant difference emerged and continued to grow as R191H mice failed to gain weight (WT gain = 0.5 g/week vs. R191H gain = 0.08 g/week; p<10−4). Remarkably, the body weight of 2BAct-treated R191H males caught up to WT mice 2 weeks after beginning dosing (8–13 weeks old), at which point their rate of weight gain equalized (WT gain = 0.5 g/week vs. R191H gain = 0.55 g/week; p = 0.24; Figure 1C). Similar results were observed in female mice (Figure 1—figure supplement 4A). As lack of weight gain appeared to be the first overt phenotype, rapid normalization by 2BAct was a promising prognostic sign of efficacious target engagement. 2BAct prevented the appearance of motor deficits in VWM mice Longitudinal characterization revealed that R191H mice developed progressive, age-dependent strength and motor coordination deficits. From 8 to 19 weeks of age, R191H animals were not significantly different from WT in their performance on an inverted grid test of neuromuscular function. At 23 weeks, R191H mice showed a trend towards shorter hang times and at 26 weeks, this decrease was highly significant in both sexes (Figure 1—figure supplement 3B). In a beam-crossing test of balance and motor coordination, R191H mice were not significantly different from WT littermates at 8–19 weeks of age (Figure 1—figure supplement 3C–D). However, at 23 weeks of age, beam-crossing time was significantly increased in mutant animals, and they exhibited more foot slips and falls from the beam while crossing. The deficit in both parameters was exacerbated at 26 weeks, and some R191H animals completely failed to cross the beam within the trial cutoff of 30 s (Figure 1—figure supplement 3C). These results are consistent with the original description of the R191H model (Dooves et al., 2016). With baseline performance measured, we assessed the effect of 2BAct treatment on R191H motor skills. Placebo-treated R191H males had significantly reduced inverted grid hang times at both tested time points, as well as more coordination errors and time spent crossing the balance beam (Figure 1D–F). By contrast, 2BAct-treated R191H males were indistinguishable from WT in both assays. Full normalization was similarly observed in female animals (Figure 1—figure supplement 4B–D). Together, these data show that treatment of R191H animals with 2BAct prevented the progressive deterioration of motor function caused by VWM. 2BAct prevented myelin loss and reactive gliosis in VWM mice VWM is a leukoencephalopathy defined by progressive loss of myelin. In patients with advanced disease, an almost complete loss of cerebral white matter is observed (van der Knaap et al., 1997). Similarly, in a previous characterization of the R191H mouse model, perturbed myelination and myelin vacuolization were noted in the brain beginning at 4–5 months of age (Dooves et al., 2016; Klok et al., 2018). Given the dramatic rescue of behavioral phenotypes by 2BAct, we performed immunohistochemical analysis to examine its effects on myelin and accompanying pathologies at the end of the treatment. We focused on two heavily myelinated regions, the corpus callosum and the spinal cord. Notably, severe spinal cord pathology has recently been reported in both VWM patients and R191H mice (Leferink et al., 2018). As anticipated, R191H animals showed a clear reduction in myelin by Luxol Fast Blue staining of both regions (33% reduction in corpus callosum, p<10−4; 58% reduction in cervical/thoracic spinal cord, p<10−4; Figure 2A–D). Strikingly, 2BAct treatment maintained myelin levels at 91% and 85% of WT in the corpus callosum and spinal cord, respectively. Staining for myelin basic protein (MBP), an alternative measure of myelin content, corroborated these results in both regions (Figure 2—figure supplement 1A–D). As astrocytes have been implicated in the pathogenesis of VWM, we also stained for the astrocyte marker GFAP (Dietrich et al., 2005; Dooves et al., 2016). We found a significant increase in GFAP in both regions of placebo-treated R191H mice, which was fully normalized by 2BAct treatment (Figure 2A–B and Figure 2—figure supplement 1C–D). Figure 2 with 2 supplements see all Download asset Open asset 2BAct prevented myelin loss and reactive gliosis in the brain and spinal cord of R191H mice. (A) Representative IHC images of the corpus callosum. Scale bar, 250 μm. Inset is magnified 2X. Inset scale bar, 100 μm. (B) Quantification of staining in (A). Area of positive staining expressed as μm2. (C) Representative IHC images of the lower cervical/upper thoracic region of the spinal cord. Scale bar, 500 μm. Inset is magnified 6.8X. Inset scale bar, 50 μm. (D) Quantification of staining in (C). For (B) and (D), N = 12 mice/condition (6 males and six females combined; no significant sex differences were detected). Error bars are SD. *p<0.05; ***p<10-4; nsp>0.05 by 1-way ANOVA with Holm-Sidak pairwise comparisons. https://doi.org/10.7554/eLife.42940.008 Consistent with the literature, we noted a significant increase in Olig2 in R191H spinal cord (Figure 2—figure supplement 1C–D). Olig2 is a marker of the oligodendrocyte lineage, and its increase could indicate an attempt to compensate for the myelin loss (Bugiani et al., 2011). Additionally, we observed signs of reactive microglia in the placebo-treated R191H samples, as evidenced by a 15-fold increase in Iba-1 staining (Figure 2C–D). ATF3, an ISR target induced in the spinal cord during injury and inflammation, was also significantly increased (Dominguez et al., 2010; Hossain-Ibrahim et al., 2006). 2BAct treatment fully normalized all four of these markers (Figure 2C–D and Figure 2—figure supplement 1C–D). In an analysis of younger mice, we observed no significant differences in myelin, Iba-1 or ATF3 between R191H and WT spinal cord at the start of treatment (2 months of age; Figure 2—figure supplement 2). The time course of pathology was consistent with the emergence of motor deficits, and reflected the degenerative nature of VWM. Together, these results demonstrate that dosing with 2BAct before onset of histological signs prevents CNS pathology in a mouse model of VWM. A chronic ISR in the CNS of VWM mice is prevented by 2BAct In all eukaryotic systems, ATF4 protein expression is regulated by the level of ternary complex in cells (Harding et al., 2000; Mueller and Hinnebusch, 1986; Vattem and Wek, 2004). We postulated that the decrease in eIF2B GEF activity brought about by the Eif2b5R191H mutation would reduce levels of ternary complex, leading to upregulated ATF4 translation in R191H mice. In support of this, ISR activation has been reported in patient VWM postmortem samples (van der Voorn et al., 2005; van Kollenburg et al., 2006). To evaluate ISR activity, we measured the expression of 15 transcripts previously identified as ATF4 target genes at three different time points (2.5, 5, and 7 months) during the lifespan and development of pathology. The ISR was robustly and consistently induced at these time points in the cerebellum, forebrain, midbrain and hindbrain of R191H animals (Figure 3A and Figure 3—figure supplement 1A). Significant upregulation of all targets except Gadd34, Slc1a5 and Gadd45a was evident in cerebellum at 2.5 months; at later timepoints, these three transcripts also became significantly induced. The ISR signature was similar across all brain regions with Atf5, Eif4ebp1 and Trib3 being the most upregulated transcripts in the panel. Interestingly, we did not observe ISR induction in R191H mice at postnatal day 14 (Figure 3—figure supplement 1B). Thus, the ISR is activated sometime between 2 and 8 weeks of age through an as-yet unknown mechanism. Activation of this stress response preceded the appearance of pathology (myelin loss, gliosis and motor deficits) in VWM mice. Figure 3 with 3 supplements see all Download asset Open asset The ISR is activated in the brain of R191H mice and its induction is prevented by 2BAct. (A) mRNA expression in R191H cerebellum at 2.5 (N = 13/genotype), 5 (N = 20 WT, 19 R191H) and 7 (N = 30/genotype) months of age. (B) Western Blots of the indicated proteins from 7-month-old male cerebellum. Actin was used as a loading control. Each lane represents an individual animal. (C) Schematic of ISR activation in the context of external stressors or VWM. PP1, protein phosphatase 1. Gadd34, an ATF4-induced regulatory subunit of PP1 that targets it to eIF2. (D) mRNA expression in R191H cerebellum (N = 23 WT, 21 R191H, 24 R191H + 2 BAct) and spinal cord (N = 10/condition) at 27–32 weeks of age from the 2BAct treatment study (Figure 1B). For (A) and (D), males and females were combined as there was no significant difference between sexes. Data are shown normalized to WT transcript levels. Bars, mean ±SD. *p<0.01; **p<10-3; nsp>0.05 by Student’s t-test with Holm-Sidak correction (compared to WT). Transcripts without symbols were highly significant with p<10−4. 2BAct treatment was highly significant for all transcripts (p<0.01 vs. placebo treatment). A table of p-values from tests is available in Figure 3—source data 1. https://doi.org/10.7554/eLife.42940.011 Figure 3—source data 1 Adjusted p-values from t-tests of multiplex transcript expression quantification. https://doi.org/10.7554/eLife.42940.015 Download elife-42940-fig3-data1-v1.xlsx In addition to changes in transcript levels, we confirmed translational induction of ATF4 as well as the increase in protein levels of the negative regulator of cap-mediated mRNA translation EIF4EBP1 by Western blot analysis of 7-month-old R191H cerebellum lysates (Figure 3B). As expected for an ISR induced by eIF2B dysfunction rather than external stressors (see schematic in Figure 3C), we did not detect an increase in eIF2α phosphorylation in R191H brains. We observed greater ISR induction in the spinal cord compared to the cerebellum (Figure 3D, compare red and brown points), with the transcription factor Atf3 showing an additional 10-fold increase. The greater extent of myelin loss in the spinal cord, as well as astrocyte and microglial activation, suggests exacerbation of the phenotype due to increased ISR activation. Notably, treatment with 2BAct abolished ISR induction in both regions (Figure 3B,D). The striking attenuating effect of this molecule on the ISR is consistent with full rescue of GEF activity in vivo. To determine whether our results would extend to other VWM mutations, we examined a second mouse model of VWM bearing an Eif2b5R132H/R132H mutation, which corresponds to the disease-causing human Eif2b5R136H allele (Geva et al., 2010). This model has a normal lifespan and exhibits very mild phenotypes in comparison to R191H mice. Nevertheless, we detected significant upregulation of two ISR targets, Atf5 and Eif4ebp1, that was blocked by 4 weeks of treatment with 2BAct (Figure 3—figure supplement 2). Even though patients carry the eIF2B mutation(s) in all cell types, VWM manifests as a CNS disease with the exception of ovarian failure in late-onset female patients. In extremely rare and severe cases, renal dysplasia and hepatomegaly have also been recently reported (Hamilton et al., 2018). To evaluate the impact of the R191H mutation on other tissues, we interrogated ISR target expression in various peripheral organs. Upregulation of ATF4 targets was not detected in skeletal muscle, liver, kidney or ovaries (Figure 3—figure supplement 3), demonstrating that the CNS is particularly sensitive to a reduction in eIF2B function. 2BAct normalized the R191H brain transcriptome Because our targeted RNA panel consisted of only 15 genes, we turned to RNA-seq in order to comprehensively profile the transcriptional changes that take place in the VWM brain. We analyzed cerebellum from WT and R191H mice at 2, 5 and 7 months of age in order to assess potential changes in R191H mice as they develop pathology. We confirmed the upregulation of ISR target genes beginning as early as 2 months of age, the magnitude of which was sustained at 5 and 7 months (Figure 4—figure supplement 1A–C). By contrast, and as expected for a disease driven by dysfunction in eIF2B, we did not observe changes in expression for known downstream targets of the parallel IRE1α or ATF6-dependent branches of the unfolded protein response at any time point (Figure 4—figure supplement 1A–C). In order to identify additional classes of transcripts that distinguish R191H from WT, we performed singular value decomposition (SVD) analysis on the dataset from 2-month-old mice. We focused on genes with the largest positive and negative loadings on the first eigengene (i.e. the first singular vector from SVD [Alter et al., 2000]), which separated the samples by genotype (Figure 4A–B). The first class consisted of 473 genes with increased expression (positive loadings) in R191H mice. Unsurprisingly, GO-term enrichment analysis revealed categories that contained many known ATF4-dependent targets involved in amino acid metabolism and tRNA aminoacylation (Supplementary file 1C) (Adamson et al., 2016; Han et al., 2013). Figure 4 with 2 supplements see all Download asset Open asset R191H mice have an abnormal brain transcriptome at 2 months of age that is normalized by 2BAct treatment. (A) Scree plot showing the variance explained by each component of the SVD analysis of 2-month-old WT and R191H cerebellum. (B) Individual 2-month-old cerebellum samples plotted along the first and second components of SVD analysis. (C) Heatmap of gene expression changes in WT and R191H cerebellum at 2, 5, and 7 months of age (N = 3/genotype/time point). Shown are the 50 genes with the largest absolute loadings in the first eigengene from SVD analysis of 2 month samples. Source data for (A) -(C) are available in Figure 4—source data 1. (D) Volcano plots showing gene expression changes between R191H and WT (left) and R191H + 2 BAct and WT (right). Orange and green dots indicate transcripts that were more than 2X increased or decreased, respectively, in the R191H vs. WT plot. These dots are replicated on the R191H + 2 BAct vs. WT plot for comparison. (E) Heatmap of gene expression changes in WT and R191H cerebellum treated with placebo or 2BAct for 4 weeks. Genes are the same set plotted in Figure 4C. Colors indicate the scaled ln(TPM) from the mean abundance of the gene across all samples. For (D) and (E), N = 3/condition. Source data for (D) and (E) are available in Figure 4—source data 2. https://doi.org/10.7554/eLife.42940.016 Figure 4—source data 1 Fold-changes of transcripts identified in RNA-seq of 2-, 5- and 7-month-old cerebellum. https://doi.org/10.7554/eLife.42940.019 Download elife-42940-fig4-data1-v1.xlsx Figure 4—source data 2 Fold-changes of transcripts identified in RNA-seq in the 4-week 2BAct treatment experiment. https://doi.org/10.7554/eLife.42940.020 Download elife-42940-fig4-data2-v1.xlsx A second class comprised 600 genes with reduced expression (negative loadings). These genes were not restricted to expression in a specific cell type, but GO-term enrichment analysis revealed categories related to myelination and lipid metabolism (Supplementary file 1C), suggesting an effect on glial cells such as astrocytes and oligodendrocytes. A gene signature of perturbed myelin maintenance is detectable as early as 2 months, preceding evidence of myelin loss by histological analysis. A heatmap of the 50 genes with the largest absolute loadings on the first eigengene revealed that the expression of both classes persisted and was consistent as the animals aged (Figure 4C). Our targeted analysis revealed complete abrogation of the ISR by 2BAct treatment (Figure 3D). To test whether 2BAct could also normalize the broad downregulation of transcripts related to CNS function, we performed RNA-seq on cerebellum from 2.5-month-old WT and R191H mice treated for only 4 weeks. Remarkably, both upregulated and downregulated classes of transcripts were normalized in 2BAct-treated R191H mice (Figure 4D). Clustering of samples based on the top 50 genes from the previous analysis confirmed that 2BAct-treated R191H mice were indistinguishable from WT (Figure 4E). Thus, 2BAct normalized the defective expression of glial and myelination-related genes. Moreover, 2BAct treatment of WT mice did not significantly alter gene expression compared to placebo (Figure 4—figure supplement 2A–B). Together, these data demonstrated that 2BAct treatment normalizes the aberrant transcriptional landscape of VWM mice without eliciting spurious gene expression changes in WT mice. The ISR is activated in astrocytes and myelinating oligodendrocytes of R191H mice The robust induction of ISR targets in the brain of VWM mice raised the question of whether all CNS cell types are uniformly affected, or whether a subpopulation of cells is particularly susceptible to a decrease in eIF2B function. To address this, we performed single cell RNA-seq (scRNA-seq) on two brain regions, forebrain and cerebellum, of 2.5-month-old WT and R191H mice. Unbiased clustering of single cells from each region identified 13 and 10 clusters in the forebrain and cerebellum, respectively, that were subsequently assigned cell type identities using CNS cell type gene markers obtained from bulk RNA-seq data (Figure 5A, Figure 5—figure supplement 1 and Figure 5—figure supplement 2A) (Koirala and Corfas, 2010; Zhang et al., 2014). Cells from both WT and R191H tissues were represented in each cluster, demonstrating that transcriptionally defined cell types are not influenced by genotype at this early time point in disease progression. Next, we generated an unbiased, tissue-independent ISR target signature by using the top 50 upregulated genes from our bulk RNA-seq analysis of R191H cerebellum as input for Clustering by Inferred Co-Expression (CLIC) analysis (Li et al., 2017). CLIC identified 18 of our input genes as coherently co-expressed across 1774 diverse mouse microarray datasets, and expanded this co-expression module to a final signature comprising 95 genes (Supplementary file 1D). Figure 5 with 3 supplements see all Download asset Open asset The ISR is strongly activated in astrocytes