Dissecting the cell biology of the ISR in cognitive disorders at the single-cell resolution Authors: Bong-Kiun Kaang Published: 2026 The integrated stress response (ISR) is widely recognized as a fundamental cellular mechanism that maintains homeostasis by regulating protein synthesis in response to both intracellular and extracellular stress (1). The ISR also represents a major frontier in brain health and cognitive function. It dynamically regulates the synaptic processes that are required to form a long-term memory (1, 2). Moreover, persistent ISR activation drives neuronal dysfunction and contributes to the pathogenesis and cognitive decline across a wide range of neurodegenerative and neurodevelopmental disorders (1). Despite these advances, the specific cellular mechanisms by which sustained ISR activation impairs cognitive function remain largely unknown. The study by Torkenczy et al. (3) now demonstrates that the cellular response to chronic ISR activation varies remarkably among distinct brain cell types. This finding raises the intriguing possibility that such cell type–specific ISR responses may underlie the diverse pathological features observed across neuropsychiatric and neurodegenerative disorders. In a recent study published in Science (2026) (4), the authors generated a mouse model that enables chronic ISR activation. To create this model, they introduced the Ppp1r15bR658C mutation, a genetic variant identified in patients with cognitive impairment, into the endogenous mouse Ppp1r15b locus. PPP1R15B functions as an adaptor protein that recruits protein phosphatase 1 (PP1) to phosphorylated eIF2α, thereby promoting eIF2α dephosphorylation. The R658C mutation disrupts this process, resulting in persistently elevated levels of phosphorylated eIF2α and chronic activation of the ISR (4). This model provides a powerful system for investigating cell type–specific responses to persistent ISR activation. Using this mouse model, the authors isolated nuclei from the brain and performed single-cell RNA sequencing (scRNA-seq) together with single-cell ATAC sequencing (scATAC-seq). These complementary approaches enabled them to characterize cell type–specific changes in gene expression and chromatin accessibility and to identify the downstream factors that mediate the cellular response to sustained ISR activation (Fig. 1). Cell type–specific integrated stress responses in the forebrain. This figure summarizes the experimental strategy used to characterize cell type–specific responses to chronic ISR activation. Single-cell RNA-seq and ATAC-seq of forebrain cells from Ppp1r15bR658C ISR model mice enabled the identification of cell type–specific changes in gene expression and chromatin accessibility, as well as the TFs underlying these distinct ISR-dependent transcriptional programs. Strikingly, ATF4-dependent regulation of long-term memory was mediated primarily through GABAergic neurons rather than glutamatergic neurons. DEG, differentially expressed genes. A salient finding of the study was that transcriptional responses to chronic ISR activation were highly cell type–specific. Despite persistent ISR activation, with the exception of the downregulated Ac149090.1 gene, not a single individual gene was regulated consistently across all cell types. Instead, previously identified ISR-responsive genes exhibited distinct cell type–specific expression patterns. Twenty-seven genes, including Atf4, were commonly upregulated in both glutamatergic and GABAergic neurons, whereas 87 genes were selectively induced in glutamatergic neurons and 44 genes were specific to GABAergic neurons. Gene set enrichment analysis with KEGG database further demonstrated that chronic ISR activation elicited distinct transcriptional programs in different brain cell populations. Among canonical ISR-responsive genes, Atf4 mRNA was robustly induced in most cell types. However, because ATF4 is also regulated by signaling pathways independent of the ISR, including mTORC1-dependent translation, the authors sought to identify ISR-specific transcriptional signatures beyond Atf4. To achieve this, they applied metacell-based analytical approaches, including MiloDE and scWGCNA. These analyses identified two highly reproducible gene modules, each consisting of approximately 30 genes. The “ISR-up” module comprised genes that were consistently upregulated in glutamatergic neurons, GABAergic neurons, oligodendrocytes, and astrocytes, whereas the “ISR-down” module contained genes that were predominantly downregulated in glutamatergic and GABAergic neurons. Notably, neither module showed substantial changes in endothelial cells or somatostatin interneurons, indicating that these cell types were relatively resistant to chronic ISR-induced transcriptional remodeling. By employing SEACells analysis, the authors ultimately identified a consensus ISR signature consisting of 28 genes derived from the ISR-up module. Although their absolute expression levels varied among different cell types, these genes were consistently upregulated throughout the cerebral cortex of Ppp1r15bR658C mice, regardless of cell identity. Two-thirds of these consensus ISR signature genes have previously been identified as direct transcriptional targets of ATF4 (5). They include genes involved in amino acid metabolism and homeostasis, such as aminoacyl-tRNA synthetases (Aars, Eprs), amino acid transporters (Slc1a, Slc7a family), and enzymes involved in amino acid biosynthesis (Psph, Chac1). These findings establish a robust molecular signature of chronic ISR activation that is shared across multiple brain cell types while also highlighting the remarkable cell type–specific diversity of the ISR transcriptional response. To investigate how the ISR regulates gene expression at the single-cell level, the authors examined the relationship between chromatin accessibility and transcriptional changes during chronic ISR activation. They analyzed scRNA-seq and scATAC-seq datasets obtained from wild-type (WT) and Ppp1r15bR658C mice using chromVAR, which quantifies genome-wide chromatin accessibility at DNA regions containing specific transcription factor (TF) binding motifs. This analysis revealed striking cell type–specific differences in TF motif enrichment during ISR activation. In Ppp1r15bR658C mice, motifs recognized by AP-1 TFs, including FOS and JUN family members, were preferentially enriched in glutamatergic neurons, whereas ATF4-binding motifs were more strongly enriched in GABAergic neurons. In contrast, neither motif was significantly enriched in nonneuronal cell types. To further examine the functional significance of these chromatin changes, the authors integrated scATAC-seq and gene expression data using Signac analysis. Consistent with the motif enrichment results, genes containing ATF-binding motifs, such as Gars, exhibited higher expression in GABAergic neurons, whereas genes associated with AP-1 motifs, including Fbln2, were preferentially expressed in glutamatergic neurons. These findings indicate that cell type–specific chromatin remodeling underlies distinct transcriptional responses to chronic ISR activation. The authors next performed TF footprinting analysis for ATF4 and AP-1 components to directly assess TF–DNA interactions. This approach is based on the principle that TF binding displaces nucleosomes, thereby increasing the accessibility of the surrounding DNA to Tn5 transposase insertion. Genome-wide analysis across individual cell types demonstrated that chromatin accessibility at ATF4-binding motifs was markedly increased in GABAergic neurons, whereas motifs recognized by AP-1 components, particularly JUNB and FOSL2, exhibited greater accessibility in glutamatergic neurons. The authors next asked whether these cell type–specific differences in ATF4 and AP-1 activity contribute to learning and memory, one of the brain’s most fundamental cognitive functions. Using sophisticated mouse genetics, they first crossed Atf4flox/flox mice with Ppp1r15bR658C mice. The resulting offspring were then crossed with either Gad2-cre mice or CaMKII-cre mice to selectively delete Atf4 in inhibitory GABAergic neurons or excitatory forebrain glutamatergic neurons, respectively, in the background of chronic ISR activation. These mice were subjected to contextual fear conditioning, and long-term fear memory was assessed 24 h later. Consistent with the original Science (2026) manuscript (4), Ppp1r15bR658C mice exhibited impaired long-term memory as a consequence of persistent ISR activation. Deletion of Atf4 in forebrain glutamatergic neurons had no impact. Surprisingly, however, selective deletion of Atf4 in GABAergic neurons significantly worsened long-term memory impairment, indicating that ATF4 plays a protective role in inhibitory neurons under conditions of chronic ISR activation. These findings demonstrate that the downstream consequences of sustained ISR activation differ markedly between neuronal subtypes and further suggest that ATF4-dependent regulation of long-term memory is mediated primarily through GABAergic neurons rather than glutamatergic neurons. Finally, the authors explored the clinical relevance of their findings by analyzing publicly available single-cell transcriptomic datasets from patients with neurological disorders associated with cognitive impairment. In Down syndrome, the ISR signature genes were broadly upregulated across virtually all brain cell types. Consistent with these human data, ISR suppression in a Down syndrome mouse model reverses the cognitive and synaptic deficits (5). Similarly, increased expression of ISR-related genes was observed in Lewy body dementia and Parkinson’s disease, although the magnitude and affected cell populations varied among disorders. The authors further integrated scRNA-seq datasets with genome-wide association study data to perform single-cell disease relevance scoring (6). ISR-associated gene signatures showed significant enrichment in glutamatergic neurons in Alzheimer’s disease (AD), where ISR rescues their long-term memory deficits in AD mouse models. Interestingly, a similar association was also observed for schizophrenia, suggesting that dysregulation of the ISR in excitatory neurons may represent a common molecular feature shared by multiple neuropsychiatric and neurodegenerative disorders. “The study by Torkenczy et al. (3) now demonstrates that the cellular response to chronic ISR activation varies remarkably among distinct brain cell types.” For nearly two decades, ATF4 has been regarded primarily as a maladaptive downstream effector of the ISR in Alzheimer’s disease, and numerous studies have demonstrated that reducing ATF4 signaling improves cognition (7). The new findings challenge this prevailing view by showing that ATF4 performs a protective function specifically in GABAergic neurons during chronic ISR activation, suggesting that the impact of ATF4 depends critically on cellular context rather than simply its abundance. The authors identified a critical role for ATF4 in GABAergic interneurons in cognitive impairment caused by excessive activation of the ISR. This represents an important conceptual advance, suggesting that neurodegenerative diseases may not arise simply from random damage across all cell types, but rather from cell type–specific alterations or maladaptive responses in vulnerable populations. Furthermore, the findings suggest that the type of neurodegenerative disease may be determined, at least in part, by the specific cell populations that are most vulnerable to chronic ISR activation. This finding opens many exciting avenues for future research. For example, it would be interesting to determine which specific interneuron subtypes (PV, SOM, VIP, etc.) are involved in this process. Investigating the impact of cell type–specific expression of the R658C variant on cognitive function would also provide important insights. Furthermore, it remains an intriguing possibility that ISR-mediated cognitive dysfunction may be regulated not only in a cell type–specific manner but also through brain region–specific mechanisms. Another important question is whether further enhancement of ATF4 function specifically in interneurons could improve or restore cognitive function. This could be tested both in WT mice and in Alzheimer’s disease model mice, potentially revealing whether modulation of interneuronal ATF4 signaling represents a therapeutic strategy for cognitive decline. Finally, as single-cell technologies continue to dissect the most vulnerable cell populations in brain disorders, modulating the ISR offers promise for transforming the future of treatment for human cognitive disorders.