Emerging roles for integrated stress response signaling in homeostasis.

Integrated stress response signaling has been shown to increase with age in multiple model organisms. This increase in the homeostatic “set point” of ATF4 in relevant cell types might drive disease‐causing processes and may be explained by increased cellular burdens with age, perhaps to maintain tissue‐wide function in response to age‐related tissue degeneration. Given the myriad crucial roles of ISR in responding to stressors and restoring function, it is worth speculating whether ISR signaling mitigates or exacerbates some relevant hallmarks of aging, such as loss of proteostasis, deregulated nutrient sensing, and genome instability [ 136 ]. For example, transcriptomic analysis of different parts of the Drosophila CNS from young and aging animals found upregulation of genes that regulate DNA damage/repair mechanisms, including ISR effectors such as Xrp1 and GADD45 , along with a gene expression signature associated with loss of proteostasis [ 137 ]. Interestingly, this gene expression profile was more pronounced in the optic lobes (visual processing center) than in the central brain or ventral nerve cord [ 137 ], suggesting a putative role for ISR specifically in visual system aging (discussed below). Chronic inflammation and dysbiosis are other hallmarks of aging that ISR is likely to be a player in; these are discussed in the “ Tissue surveillance ” section below. Flies, as with humans, exhibit mild vision loss with age due to retinal degeneration. This can be exacerbated by a number of known genetic aberrations, including those that cause misfolded protein aggregation or impair ISR signaling. Loss of Drosophila ATF4 leads to acceleration of photoreceptor degeneration in moderately aged flies [ 26 ]. This crucial role of ATF4 in preserving retinal function is conserved in the mammalian visual system: Loss of ER stress response factors accelerates the onset of autosomal dominant retinitis pigmentosa (adRP) (reviewed in Ref. [ 138 ]). While the role of ATF4 activity in photoreceptor degeneration has not been directly demonstrated to be cell‐autonomous, unrestrained ATF4 expression in the fly retina accelerates retinal degeneration and ISR inhibition in aging photoreceptors delays their decline [ 139 ]. Taken together, this supports a model where an optimal “set point” of ATF4 activity governs photoreceptor function in the visual systems of both adult flies and mice. In conjunction with its role in skeletal muscle development, ISR signaling also contributes to skeletal muscle atrophy in aging mice. In WT mice, ATF4 drives GADD45A expression, which in turn drives a gene expression program in muscles that ultimately results in atrophy, including reduced protein synthesis, increased autophagy and caspase‐mediated proteolysis [ 140 ]. While this study does not employ an aging paradigm, it is possible that the ATF4/GADD45 axis of gene regulation is a strategy that muscles use to cope with age and associated stressors. This was confirmed by a more recent study which demonstrated that GADD45 expression increases with age and drives skeletal muscle atrophy [ 141 ]. Forced expression of GADD45 in mouse skeletal muscle reduced muscle mass and exercise capacity, leading to atrophy of glycolytic muscle fibers [ 141 ]. Comparative proteomics of normal and GADD45 ‐overexpression muscles found a strong GADD45‐dependent repression in metabolic networks such as TCA cycle, ATP synthesis, and pyruvate metabolism. Ultimately, this led to decreased oxidative capacity of muscles and loss of mitochondria. This suggests that a shift in energy metabolism might drive age‐related muscle atrophy downstream of ATF4/GADD45. Notably, a 2015 study examined ISR‐mediated transcriptomic changes in muscle following drug‐based interventions for age‐related skeletal muscle atrophy [ 142 ]. Treatment with two small molecules, ursolic acid and tomatidine, mitigated age‐induced skeletal muscle atrophy [ 142 ]. Transcriptomic analysis revealed that these treatments reduced transcript levels of several ATF4‐regulated genes [ 142 ]. Of note, aging skeletal muscle exhibited reduced protein synthesis in an ATF4‐dependent manner [ 142 ]. Forced ATF4 expression reduced protein synthesis in skeletal muscle, and muscle‐specific ATF4 knockout reduced age‐related atrophy [ 142 ]. Together, this supports a model whereby ISR signaling‐mediated reduced protein synthesis drives age‐related muscle atrophy in mice. Many modalities of amino acid deprivation are known to prolong lifespan, most commonly methionine restriction [ 143 , 144 , 145 , 146 , 147 ]. Given the obvious connection between amino acid deprivation and GCN2 signaling, there has been much interest over the years in determining whether ATF4 mediates lifespan extension. In yeast, C. elegans , and Drosophila models, GCN2‐ATF4 signaling has been implicated in lifespan extension mediated by amino acid restriction [ 148 , 149 ]. Though researchers speculate that amino acid‐sensing mechanisms including ISR might be implicated (reviewed in Refs [ 150 , 151 , 152 ]), a role for ATF4 in lifespan extension following dietary restriction has yet to be directly established in mammalian systems (likely due to limited availability of genetic tools for tissue‐specific deletion). Interestingly, genetic mouse models for studying lifespan, such as dwarf mice [ 153 , 154 , 155 ], implicate ATF4 activity in metabolic tissues. Genetically dwarf mice have been studied for their unique physiological features that show delayed aging and prolonged lifespan. Some of these features include robust defense mechanisms against oxidative stress, reduced insulin secretion, and signal transduction. Transcriptomic analysis of these mice compared with litter controls reveals changes in gene expression of amino acid transporters and lipases, some of which are known ATF4 targets [ 155 ]. Livers and fibroblasts of Snell dwarf mice exhibit elevated ATF4 activity compared with control mice in response to ISR triggers [ 153 , 154 ]. This could be due to a difference in the baseline levels of ATF4 protein in comparison with control animals, though such values were not reported. The specific mechanisms by which dwarf mice induce ATF4 more robustly in response to stressors such as nutrient deprivation or oxidative stress remain to be described. It was noted, however, that hepatocytes from dwarf mice exhibit elevated levels of ATF4 mRNA [ 154 ], suggesting that in addition to translational induction of ATF4 via canonical ISR activation/eIF2α phosphorylation, increased ATF4 transcription (or increased ATF4 mRNA stability) may be a contributing factor. Whether such elevated ATF4 is a necessary cellular mechanism by which dwarf mice are more stress tolerant remains to be examined. In addition to lifespan extension with dietary restriction, ATF4 also frequently features in molecular mechanisms underlying lifespan extension interventions, including litter crowding [ 156 ] and administration of rapamycin, acarbose, or hydrogen sulfide [ 157 , 158 , 159 ]. Finally, since ISR signaling is implicated in adult stem cell functions (as described earlier in this review), it is possible that changes in homeostatic ISR with age might also play a role in stem cell exhaustion, a hallmark of aging. As a corollary, mechanisms that regulate stem cell quiescence and appropriate differentiation when required could contribute to healthspan extension (reviewed in Ref. [ 160 ]) The potential of this field is underscored by studies in Drosophila that demonstrate that (a) PERK depletion specifically in ISCs improves lifespan [ 42 ], and (b) PERK hyperactivation in the intestine (while not ISC‐specific) decreases lifespan in these animals by mitigating the proliferative capacity of ISCs and propensity for dysplasia [ 161 ].

Much of this review centers on ATF4 as the primary ISR effector downstream of canonical eIF2α kinase activation (Fig.  1 ). Recent work has uncovered two noncanonical eIF2α kinases during proteotoxic and oxidative stress: FAM69C and MARK2 [ 162 , 163 ], though homeostatic roles for these kinases have not yet been described. While our current understanding of homeostatic ISR function is largely based on studies of ATF4, other ISR effectors have been described in recent years, many downstream of ATF4 [ 2 , 164 ]. Here, we discuss ATF3, another bZIP transcription factor, as an example. ATF3 is a transcriptional target of ATF4 [ 165 ] and can promote cell death or cell survival under different stress contexts in both Drosophila and mice [ 146 , 147 , 148 , 149 , 165 , 166 , 167 , 168 ]. These observations are in accordance with known roles for ISR signaling in cell death and survival in the context of malignancy (reviewed in Refs [ 10 , 169 ]). ATF3 is activated downstream of ATF4 in many contexts and is autoregulated in a negative feedback loop [ 166 ]. Developmental studies in the abdominal epithelium and the intestine suggest that ATF3 works with, or parallel to, the JNK signaling pathway transcription factor Jun to maintain homeostatic functions [ 167 , 170 ]. In the Drosophila abdominal epithelium, ATF3 levels are developmentally regulated: Highest ATF3 expression was observed at late larval and early pupal stages, followed by a decrease in mid‐pupal stages and a subsequent increase in late pupal stages [ 167 ]. This period of dynamic ATF3 expression coincides with morphogenesis of the fly abdomen, which involves the replacement of larval abdominal epithelial cells with the adult epidermis. Normally, the extrusion and apoptosis of larval epithelial cells precede the division, migration, and closure of the adult epithelial progenitors. Constant ATF3 overexpression disrupted larval epithelial cell replacement and prevented extrusion and closure of the adult abdominal epithelial layer [ 167 ]. Clonal ATF3 overexpression caused perdurance of larval epidermal cells in the adult abdomen, suggesting that a reduction in ATF3 expression might be required for clearance of the larval abdominal epithelium during metamorphosis [ 167 ] In both cases, ATF3 promotes cell survival by cooperating with the JNK signaling pathway effector, Jun. In a different context—intestinal regeneration following infection—ATF3 negatively regulates JNK signaling [ 170 ]. Direct regulation of cell turnover/survival mechanisms by ATF3 may broadly underlie homeostatic maintenance: This idea is corroborated by recently identified roles for ATF3 in the maintenance of mouse pulmonary endothelial cells [ 171 ] and suppression of tumorigenesis [ 172 ]. Taken together, these observations indicate that ATF3 performs diverse cellular functions depending on the cell types and/or contexts. Like ATF4, homeostatic and adaptive functions may differ greatly within the same tissues. Much more work is needed to further elucidate the links between dynamic ATF3 expression and cell survival and removal during morphogenesis and adult tissue homeostasis and whether these processes are directly downstream of ATF4.

SN, LG, and DV contributed to conceptualization, review and editing; SN and LG contributed to writing—original draft; DV contributed to supervision.

Comparison of transcriptomic analyses across the literature suggest that the ATF4 targetome during adaptive responses is context‐specific; this appears true as well for homeostatic ISR functions. The studies we have summarized to this point describe homeostatic roles for ATF4 during developmental and adult stages. These functions fit into broader cell biological themes that we have identified as “emerging functions” of homeostatic ISR signaling. These groupings are shown, with relevant examples discussed in this review, in Table  2 . Each of these themes represents potential avenues to continue building on our understanding of how ISR signaling functions in physiological homeostasis. Below, we highlight three broad cellular functions of ISR that appear in multiple cellular contexts: secretory cell function, biological oscillators, and tissue surveillance. Given the critical role of the ER in processing proteins destined for secretion, it is perhaps expected that cell types with a perpetually high secretory burden might be among those that employ ISR signaling for their homeostatic function. This was proposed by Harding et al . in 2001 [ 173 ] following their discovery that loss of PERK impairs secretory functions in the pancreas, which led to defects in insulin secretion and contributed to the onset of diabetes mellitus. Several studies, including theirs, have implicated PERK (via the Unfolded Protein Response, or UPR) in managing ER stress in highly secretory cells (reviewed in Ref. [ 174 ]). ATF4 and ISR have since been implicated as well in secretory cell functions, as was described earlier in this review. Global gene expression analysis showed that loss of ATF4 or PERK in mouse fibroblasts led to reduced expression of several genes involved in secretory pathways [ 66 ]. These data agree with our recent discovery that Drosophila ATF4 in homeostatic fat tissue regulates synthesis of vitellogenins (i.e., yolk) that are secreted and trafficked to oocytes [ 125 ] Vitellogenins, or yolk proteins, are produced in adipocytes (in Drosophila ) or hepatocytes (in vertebrates) and for uptake by maturing oocytes [ 175 , 176 , 177 ]. We also discovered that fat body ATF4 signaling mediates neuropeptide secretion to promote ovulation [ 125 ]. Given that ISR signaling is required for secretory cell function in numerous contexts—those highlighted in this review include osteoblasts, lens secondary fiber cells, adipocytes, and airway epithelial cells—future work might uncover even more themes regarding the role of ATF4 in nonautonomous physiological functions, beyond the function of secretory cells themselves. In reviewing the literature, we found that ISR signaling has roles in various cyclical processes, including circadian rhythms, the cell cycle, and female reproductive hormone cycles, on different time scales. This section will explore the potential role of ISR signaling in regulating or maintaining various biological oscillators. Several recent works have implicated a role for ISR activation in regulating the circadian rhythm. One in particular demonstrated that GCN2 acts upstream of ATF4 to modulate the duration and rhythmicity of the circadian clock in mammals [ 178 ]. This study, performed in GCN2 null mice, showed that mutants have disrupted circadian rhythms. Cyclical phosphorylation of eIF2α‐induced cyclical translation of ATF4, which in turn directly activated the transcription of the key circadian gene PER2 , thereby regulating the clock. Another recent work furthers our understanding of ISR activation in the circadian clock using computational modeling [ 179 ]. Simulations in this study revealed a potential role for ATF4 in the main timekeeping hub of the mammalian circadian clock, the suprachiasmatic nucleus (SCN). While their analyses showed that the ISR‐mediated regulation of clock genes is important for maintaining the clock, simulations of “jetlag,” where mice were temporarily exposed to phase‐shifted light/dark, showed that ISR signaling alters wake/sleep entrainment. This is supported by empirical evidence that GCN2 null mice can be re‐entrained when the correct light/dark conditions are altered [ 178 ], arguing that the role of ISR in the SCN is to reinforce the robustness of the clock oscillator, but not necessarily to maintain it. These conclusions are also consistent with other known roles for ATF4 and the ISR pathway in learning and memory formation as well as synaptic plasticity [ 110 , 180 , 181 , 182 ]. In addition to the key role of ISR in the central clock (SCN), recent work has also uncovered an important function of PERK in the peripheral clock of the liver. Similar to GCN2 disruption in the SCN, liver‐specific PERK mutation results in a disrupted clock, and animals who are refractory to “phase shift” or “jetlag” [ 183 ]. PERK‐dependent activation of ATF4 directly regulates the transcript abundance of the two master regulators of the circadian clock, BMAL1 and CLK. Previous work has shown that PERK/ATF4 activates the expression of a microRNA, miR‐211 , which antagonizes CHOP to promote cell survival [ 184 ]. PERK/ATF4 activation of miR‐211 in the liver also represses BMAL1 and CLK , affecting the oscillation of the liver clock. Together, this underscores the potentially critical roles of ISR activation in different tissues, and the tissue specificity of upstream kinases. Notably, livers from such jetlagged animals accumulate ATF4 and CHOP proteins and show acute and rapid hepatic necrosis when subjected to ER stress. This suggests a crucial role for ISR in coordinating detoxifying functions of the liver with the circadian rhythm. Indeed, ATF4 has been shown, in a clock‐dependent manner, to also regulate the expression of xenobiotic efflux transporter, ABCG2 in the intestine [ 185 ]. With the importance of circadian regulation of xenobiotic efflux becoming more appreciated in tissues other than the liver and intestine, such as in the blood brain barrier of Drosophila, it is likely that ISR‐mediated circadian clock regulation of xenobiotic efflux [ 186 , 187 ] may represent an important oscillating mechanism to maintain homeostasis daily. Photoperiod‐dependent function for ISR signaling may extend beyond the animal kingdom; plants only have one conserved eIF2α kinase, GCN2, which responds to abiotic stresses [ 188 ]. Among its many roles, GCN2 activation has a known role in abscisic acid (ABA) homeostasis [ 189 ], which is one of the five major plant hormones that is important for regulating leaf abscission—the process by which leaves are shed seasonally in response to light and temperature changes. Further, recent work has shown that plant GCN2 is activated by light, which is induced by oxidative stress in chloroplasts—suggesting that the role of GCN2 in circadian rhythms may be shared across plant and animal kingdoms. Additionally, while no ATF4 ortholog has been identified in plants, several bZip transcription factors have been shown to regulate leaf senescence as well as the circadian clock [ 190 ]. This leads us to speculate whether this may be either a convergent or a conserved evolutionary process whereby similar strategies are used to sense and respond to environmental factors, such as changing temperature and light (abiotic) cues to promote leaf dropping and dormancy in deciduous angiosperms. The canonical cell cycle oscillator contains four phases: two gap phases (G1 and G2) separating DNA synthesis (S) and mitosis (M), followed by cytokinesis [ 191 ]. Expectedly, uncontrolled cell proliferation is most frequently associated with cancers and underlies at least two key hallmarks of cancer—sustained proliferation and enabling replicative immortality [ 192 ]. ISR dysregulation, however, is implicated in other hallmarks of cancer: resisting cell death, insensitivity to antigrowth factors, and metabolic reprogramming [ 192 ]. Since the impact of ISR activation in cancers has been extensively reported and reviewed elsewhere [ 193 ], this section will instead focus on the putative role of ATF4 during the canonical cell cycle under nonpathological conditions. Two phases of the cell cycle are prone to stress, particularly genotoxic and proteotoxic stress: DNA synthesis and mitosis [ 194 ]. ATF4 levels have been shown to oscillate during the cell cycle, with the highest expression during the S phase [ 195 ]. Since ATF4 is known to activate key DNA damage response pathways (discussed above in “ISR signaling as a driver of aging,” and reviewed in Ref. [ 164 ]), it may be playing a surveillance role during S phase, a stage which is susceptible to DNA damage due to replication fork stalling and replication‐transcription machinery collisions [ 194 , 196 ]. Whenever DNA damage is sensed at S and M phases of the cell cycle, checkpoints are activated to ensure fidelity, integrity, and robustness of the cell cycle. Checkpoint activation occurs through a kinase cascade that prevents cell cycle progression until damage is repaired. Conversely, prolonged activation of checkpoints signals lack of resolution of damage in the cell cycle, resulting in activation of apoptosis pathways [ 194 , 196 ]. In yeast, CK1 kinase has been demonstrated to play an important role in the mitotic checkpoint [ 197 ]. CK1 also potentially plays a role in targeting ATF4 for degradation by successive rounds of phosphorylation [ 195 ]. Stabilizing ATF4 using non‐degradable mutants results in arrested cell cycle progression, suggesting that ATF4 levels might need to be low at key phases of the cell cycle and that this may be regulated by kinases like CK1. One important caveat of this study was that it was performed using mitosis‐blocking drugs to synchronize cells, which can inherently evoke a stress response [ 198 ]. Future studies using modern fluorescent cell cycle reporters such as FUCCI [ 199 ] will be required to examine this relationship more closely throughout different phases of the cell cycle and in other physiological contexts. Such an imaging‐based approach was recently adopted by a study that found ATF4 activation in a subpopulation of very slow‐cycling quiescent cells in a prolonged G0 state. Intriguingly, this study also demonstrated that nutrient deprivation stress can influence the “depth” and duration of quiescence in vitro , potentially implicating GCN2 in this process [ 200 ]. These highlighted roles for stage‐specific ATF4 activation go beyond previous studies, which have shown that acute ISR activation inhibits cell proliferation and promotes cell death in multiple cell types [ 10 , 193 , 201 , 202 , 203 ]. In addition to activating DNA damage response, ATF4 also has known roles in nucleotide metabolism, particularly purine biosynthesis [ 204 ]. This function may also underlie ATF4 upregulation specifically during the S phase of the cell cycle, when the levels of nucleotides dictate DNA synthesis in S‐phase progression. Recent work has also shown that p53, an important tumor suppressor and transcription factor regulating genome maintenance, has a shared transcriptional target network with ATF4 [ 205 ]. Transcriptomic analysis and chromatin‐binding assays identify 26 shared targets, some of which (such as GADD34 and SESN2) could act to stop proliferation. Among the other targets identified in this study are key genes in the MAPK pathway, which act in response to several extrinsic mitogenic stimuli. These observations suggest that shared targets of ATF4 and p53 play a role in protein synthesis, which is consistent with emerging metabolic roles of p53 and other cell cycle regulators such as DP [ 206 , 207 , 208 ]. The authors speculate that the ATF4/p53 shared network of targets represents a switch from anabolic to catabolic state when these pathways are disrupted, leading to cell cycle disruptions as well. This enticing speculation will be well‐served by context‐dependent analysis with more temporal, stage‐specific disruptions to reveal how the interplay between genome and stress surveillance during S‐phase impacts the cell cycle oscillator. Intriguingly, ATF4 activation is noted in S‐phase but not mitosis, despite the fact that (a) mitosis is the most stress‐vulnerable phase of the cell cycle, and (b) selective translation of key mitotic proteins is regulated by factors that also regulate ATF4 translation (such as DENR) [ 209 , 210 ]. Uncoupling of translation regulation factors and ATF4 levels during mitosis makes temporal sense as mitosis lasts 40–60 min [ 211 ], which is insufficient time for ATF4 to evoke a transcriptional program. Additionally, high levels of ATF4 at inappropriate times during the cell cycle could lead to arrest, if not cell death [ 2 ]. Dysregulated ATF4 activation during periods of low translation during the cell cycle might not only stall cell cycle progression (which would activate checkpoints) but also result in aberrant ISR activation, which could prove catastrophic. In summary, ATF4 oscillation in phase with DNA synthesis during the cell cycle represents an elegant way for cells to integrate genome surveillance and metabolic needs during a prolonged, sensitive phase of the cell cycle. The reproductive hormone oscillator in most mammalian females—estrous cycle in mice, menstrual cycle in humans—encompasses physiological changes that prepare them for mating, conception, and pregnancy. These cycles regulate multiple cell types, largely in the ovary and uterine lining (endometrium), that orchestrate phases of egg development, ovulation, mate receptivity, and endometrial changes to prepare for egg implantation. ATF4 protein level changes have been shown to oscillate during the estrous cycle in the mouse endometrium [ 212 ]. ATF4 protein levels were significantly higher in the endometrium during estrus (the mate‐receptive phase immediately preceding ovulation), while ATF4 mRNA levels did not change appreciably except for a very small increase during estrus. Given the known role of ATF4 in regulating angiogenesis [ 87 , 88 ], the function of ATF4 during the estrus phase may be to promote blood vessel formation in the endometrium, potentially by positively regulating VEGF expression. ATF4 and phospho‐eIF2α levels also appear to cycle in the corpus luteum [ 213 ], which is responsible for progesterone secretions that govern the luteal phase (metestrus and diestrus in mice). Studies performed in human ovary‐derived cells in culture suggest that ATF4 might drive ovulation, as ovarian cells from patients with decreased ovulation (polycystic ovary syndrome) expressed lower levels of ATF4 [ 214 ]. In agreement with this, our recent work in Drosophila females demonstrated that ATF4 activity promotes ovulation during homeostasis [ 125 ], suggesting that ISR modulation of reproductive status is a highly conserved process. ATF4 levels change dynamically in the endometrium during early pregnancy in mice [ 215 ]. ATF4 levels increase around implantation (Day 1–4 post fertilization) but decrease by Day 7 (post implantation) during pregnancy. Injecting an anti‐ATF4 antibody resulted in reduced embryo implantation, demonstrating the physiological requirement of ATF4 expression in this tissue. These findings allow for a speculative role for ATF4 in building up initial vasculature that aids in implantation and embryo viability. Together, these studies suggest that cyclical expression of ATF4 during reproductive cycles may underlie the specification of endometrium function where the tissue has to grow and diminish in a cyclical fashion. More generally, oscillating ISR activation could be a functional way to integrate endocrine cycles with biological outputs such as rapid tissue growth and shedding, and conversely, aberrant activation of ISR at certain stages may underlie pathologies such as endometriosis, polycystic ovary syndrome (PCOS), and pregnancy complications [ 214 , 216 ]. An emerging facet of ISR signaling is its demonstrable role in critical stages of various biological oscillating systems. We posit that these may represent a common role for ISR activation in contributing to the robustness of biological switches. Progression of oscillators such as the cell cycle and circadian clock are characterized by a series of irreversible state transitions [ 217 , 218 , 219 ]. This irreversibility is often conferred by molecular signaling networks known as bistable switches. One factor that ensures effective activation of such molecular switches is the reduction of ‘noise’ in the system [ 220 ]. ISR activation and the resulting global repression of translation might play a role in contributing to the robustness of biological switches at critical transition points by reducing such “noise.” Future systems‐level studies could explore this potential homeostatic role for ISR in such processes. Beyond oscillators, this potential role of ISR can even be extrapolated to other switches/irreversible state transitions such as morphogenetic changes, apoptosis, and developmental stage transitions [ 219 ]. Long‐term homeostasis requires constant surveilling and elimination of agents that might disrupt it. This can include invading pathogens as well as individual (or small groups of) unhealthy cells within a tissue. A number of studies have implicated ISR factors in two key surveilling processes: at the systemic level, immune defense, and at the tissue level, cell competition. The literature consensus is that ISR plays a critical role in the immune response across phyla from Drosophila [ 221 ] to humans (reviewed in Ref. [ 222 ]). These critical roles for ATF4 in multiple immune cell types across multiple contexts have been extensively reviewed previously [ 223 ]. Such roles for ISR signaling in the immune system can also be extrapolated to cells like microglia, often referred to as immune surveillants of the central nervous system in Drosophila and also preserve homeostasis therein [ 224 ]. Correspondingly, recent work suggests this might implicate ISR signaling, as ATF4 signaling was found to promote production of proinflammatory cytokines within microglia [ 225 ]. Together, these findings tantalizingly position ISR toward a homeostatic “surveillance” role in the immune system, ready to respond to general, ubiquitous threats. The more recently discovered evidence for ISR signaling in surveillance of cellular fitness is discussed in greater detail below. Multicellular organisms have evolved a robust tissue surveilling mechanism that prioritizes the health of a tissue at the expense of individual, lesser “fit” cells. This process, termed cell competition, has been well studied over the last couple of decades for its ability to maintain long‐term tissue homeostasis by eliminating cells that, via genetic and/or physiological changes, have become less fit for tissue occupancy (“loser” cells) relative to their wild‐type neighbors (reviewed in Ref. [ 226 , 227 , 228 ]). A cell competition environment can also arise when a subset of cells become supercompetitors (“winner” cells), outcompeting their wild‐type neighboring cells for tissue occupancy via growth/proliferation [ 229 , 230 , 231 ]. In both scenarios, cells with higher relative fitness induce apoptotic cell death in the adjacent lesser‐fit cells. Cell competition has been observed in homeostatic regulation of the Drosophila imaginal wing disk epithelium [ 232 ], mouse embryo [ 233 ], stem cell niches (either by outcompeting neighboring stem cells for niche occupancy or by inducing cell death) [ 234 , 235 ], and is implicated in tumorigenesis [ 236 , 237 ]. Much of our molecular understanding of “loser” cells is drawn from studies of ‘ Minute ’ mutants, a collection of mutations in ribosomal protein‐encoding genes [ 238 , 239 , 240 ], suggesting that reduced translation primes cells to become loser cells. A series of recent studies have implicated PERK signaling in cell competition [ 241 , 242 , 243 ]. Surprisingly though, these studies in the Drosophila wing disk have determined that Minute loser cells are driven toward competition‐mediated apoptosis due to proteotoxic stress. Minute mutant cells exhibit proteasome dysfunction, and the resulting accumulation of protein aggregates drives loser cell status [ 241 ]. These protein aggregates trigger induction of the ISR effector Xrp1 autonomously in loser cells, which further promotes protein aggregation in a feed‐forward loop. Either expression of Xrp1 or loss of GADD34, a regulatory subunit of a protein phosphatase that contributes to dephosphorylation of eIF2α (Fig.  1 ), are sufficient to confer loser status in wing imaginal disk cells [ 244 ]. Consistently, a recent study using a mouse epiblast cell competition model found that the loser cell transcriptome exhibited an ISR signature, including higher abundance of ATF4 transcript [ 233 ]. Thus, ATF4 expression might serve as a rheostat for cellular health as organisms evaluate quality control to maintain tissue homeostasis.

In this review, we have curated the myriad roles of ISR signaling across metazoa under homeostatic conditions. We covered the conserved, crucial roles of ISR signaling in development, tissue homeostasis, physiology, and aging. Due to the expansive role of this signaling pathway and nomenclature diversity, there are likely many papers on ISR and ATF4/CREB2/crc that we may not have read/covered; this was unintentional. Our goal was to underscore how often these so‐called stress response factors serve important cellular functions outside the context of exogenous stress. To conclude, below we highlight several outstanding questions in the field; future inquiries addressing these gaps will significantly further our understanding of ISR in homeostasis and how its dysregulation may impact or drive disease states. Several instances of ISR activity regulating homeostatic metabolic function in tissues such as the liver and adipose have been discussed in this review. Emerging evidence suggests that this may extend beyond the liver and adipose to other cell types that experience high metabolic burden, such as pancreatic β‐cells or cholesterol‐rich macrophages [ 28 , 245 ]. Other such examples include the critical roles of ISR in regulating glucose metabolism in osteoblasts [ 89 ], in regulating the expression of xenobiotic efflux pumps in the intestine cells [ 185 ], and in nonautonomous neuronal regulation of metabolism [ 130 ]. Together, this growing body of literature points to a function for ISR in serving a function beyond metabolic adaptation [ 121 ] as a potential master regulator of metabolism. As with other cell‐type‐specific ISR signaling paradigms, the metabolic outcomes may be context‐dependent and coregulated by cell‐type‐specific transcription factors. With large‐scale metabolome‐wide assays becoming more affordable, and ever‐increasing numbers of publicly available data sets, this area of research is primed for meta‐analysis‐driven systems biology approaches to understand how ISR regulates metabolism. Recent work has intriguingly uncovered a “massed‐spaced effect” in non‐neuronal cells that relies on CREB activity [ 246 ]. The “massed‐spaced effect” refers to the presentation of a stimulus to cells at regular intervals to promote better memory consolidation at the tissue level. The interplay between CREB and ATF4/CREB2 has several known roles in neuronal function and memory consolidation (discussed earlier in this review under “ Neurons ”) and in other contexts [ 110 , 111 , 180 , 247 ]. When compounded with ideas presented above on ISR set point and oscillatory patterns, a plausible notion emerges that ISR signaling could serve as a reinforcer of tissue level “memory” in non‐neuronal tissues. An exciting potential research avenue would be to examine whether complex, systems‐level behaviors are driven by (a) variation in the extent, duration, or periodicity of ISR activation, or (b) diversity of dimerization partners and transcriptional targets across tissues in multicellular organisms. While ISR signaling directs a myriad of developmental and homeostatic functions, as detailed thoroughly in this review, it is important to remember that ISR signaling can be further activated upon stress induction in each of these tissues. Intuitively, the homeostatic and adaptive functions of ATF4 within any one tissue or cell type are necessarily different, presenting a paradox. We offer three (nonexhaustive) hypotheses for how a cell and/or tissue might achieve this mechanistic feat despite the presence of ISR signaling under homeostatic and stress‐adaptive conditions. First, it is possible that ATF4 acts in concert with context‐specific cofactors to regulate the expression of distinct gene cohorts during homeostasis and during a stress response. Indeed, many cofactors have been identified for ATF4 in various contexts—almost all in the context of adaptive response—including p300 [ 248 ], PCAF [ 249 ], CHOP [ 250 ], Jun [ 251 ], and others (Fig.  1 ). CHOP and CREB are also induced downstream of ATF4 during adaptive responses, supporting the idea that upon stress induction, ATF4 might alter the cofactor landscape to favor induction of stress response versus homeostatic target genes. Second, a shift in the ATF4 targetome between homeostasis and stress could be influenced by other parallel and/or cooperative signaling pathways that are induced with stress, including IRE1/XBP1, mTOR signaling, and JNK signaling. There are several reported instances of genetic and pharmacological interactions between these pathways (reviewed in Ref. [ 164 ]), though the precise molecular mechanisms underlying this crosstalk are not reported. Third, ISR‐mediated selective mRNA translation via phosphorylation of eIF2α and 4E‐BP induction (reviewed in Ref. [ 9 ]) might selectively repress homeostatic ATF4 effectors while enabling translation of critical stress response genes (Fig.  1 ). This possibility is inferred from our study showing that enteric bacterial infection activates ISR signaling, resulting in selective translation of antimicrobial peptide‐encoding mRNAs, but suppression of other non‐adaptive mRNAs [ 221 ]. These models are not mutually exclusive: for example, activation of mTOR or IRE1 pathways could lead to increased expression of an ATF4 cofactor that enables a shift away from its homeostatic functions. Additionally, cohorts of ATF4 cofactors might be subject to selective translation under certain stress conditions. A great deal of future research will be required to test any one or a combination of these hypotheses. Chronic ISR activation in stress contexts is known to trigger maladaptive responses that underly various disease states such as neurodegenerative disorders and metabolic disorders (reviewed in Ref. [ 35 ]). These observations raise an important question: does persistent physiological activation of ISR during homeostasis in certain tissues eventually result in pathology? The answer might be difficult to mechanistically distinguish from age‐dependent effects of ISR signaling, since age is a known factor that contributes to increased stress burden (reviewed in Ref. [ 252 ]). Even so, we offer two considerations regarding potential consequences of persistent ISR activation and how they may relate to age‐related decline in tissue function. First, some cell types might be intrinsically more tolerant to persistent ISR activation. Indeed, different cell types incur vastly different outcomes of ATF4 activity based on the context. For example, adipocytes and photoreceptors rely on ATF4 activity for their cellular function as discussed above, whereas epidermal cells die due to increased ISR signaling during development [ 253 ]. Second, cells that rely on constitutive ISR signaling for their function may be more prone to age‐related decline in tissue function due to such chronic ISR activation. This idea is supported by studies in the aging mouse retina where ISR inhibition in young animals leads to vision defects but such inhibition in older animals actually mitigates age‐dependent retinal degeneration [ 139 ]. Other cell types discussed in this review (such as adipocytes or keratinocytes) might similarly exhibit ATF4‐driven, age‐dependent pathologies or function decline. The molecular components that define the distinctions between these various ISR signaling paradigms—constitutive requirement in young tissues, chronic activation in older tissues, etc—remain to be determined. Integrated stress response signaling and ATF4 activity have been shown to regulate several cellular mechanical processes, including head eversion in Drosophila pupae [ 40 ], wound healing‐associated cytoskeletal rearrangements in mammalian keratinocytes [ 84 ], and migration of differentiating mouse fetal HSCs and Drosophila ovarian border cells [ 73 , 254 ]. While the phenotypic consequences of ATF4 depletion in migratory cells have been documented, in many cases the specific transcriptional targets that orchestrate cellular mechanical changes are understudied. In some contexts, ATF4 has been implicated in the regulation of Rho GTPases and Cdc42 [ 255 , 256 ], both critical mediators of cytoskeletal dynamics and cellular mechanics. Whether these or other specific targets are consistently activated across diverse cell types remains unknown. Comparative transcriptomic analyses on homeostatic Drosophila and mammalian tissues offer a promising avenue for identifying conserved, overlapping ISR targets that contribute to biomechanical modulation. Such analyses could reveal whether a core set of ATF4 effector genes mediate cellular mechanics. ATF4 function in biomechanical cellular properties appears to employ both cell‐autonomous and non‐autonomous effects of ISR activation. While ATF4 can directly influence a cell's intrinsic mechanical properties by promoting the expression of cytoskeletal remodeling factors such as those described above, its impact extends beyond individual cells. Specifically, ATF4 can regulate the secretion of extracellular matrix (ECM) remodeling factors, such as matrix metalloproteinases (MMPs) [ 257 ] thereby influencing collective cell migration and tissue‐scale changes. Better understanding of homeostatic regulation of cellular biomechanics by ATF4 may also help resolve several seemingly confounding observations in the field of cancer biology, where ATF4 has been demonstrated to promote or suppress tumor invasiveness and cancer cell migration under different conditions [ 257 , 258 , 259 , 260 ]. By bridging the gap between our understanding of the homeostatic and pathological roles of ISR in regulating cellular mechanics, we can gain valuable insights into the mechanisms driving cancer metastasis and other pathologies associated with aberrant cellular mechanics. While many of the homeostatic contexts we have described in this review employ ATF4 constantly (like secretory cells whose function persists over the lifetime of the cell), other cell types might also need ways to negatively regulate homeostatic ISR activity (like ISCs that exhibit a downregulation of ISR signaling upon differentiation, or oscillators that show dynamic ATF4 expression). By extension, mechanisms that enable a shift to the “off” state for ISR activity might also be critical for coordinating cellular functions in homeostasis. Prime examples of such an “off” state being physiologically relevant are revealed by phenotypes for loss of GADD34 in tissues where no homeostatic ISR is detected (e.g., Drosophila wing disks [ 54 ]). Notably, ATF4 is an obligate heterodimer and is regulated heavily by the proteasome [ 2 , 164 ]. It will be interesting to learn more about how cellular context, including cofactor or ubiquitin ligase expression, influences ATF4 transcriptional output (e.g., targetome, active versus repressive state) in the various “homeostatic” contexts described above. In sum, our hope is that this review provides demonstrated and speculative frameworks for current and future researchers to interrogate the homeostatic role of ISR in the aforementioned tissues (and others yet to be characterized) in the absence of an obvious stressor.

In 1926, the physiologist Walter Cannon coined the term “homeostasis,” derived from the Greek word roots homeo (meaning “similar to”) and stasis (meaning “standing still”). He aimed to extend the principle of equilibrium, which deals in “relatively simple physiochemical states in closed systems,” to describe the complex cooperations between cells and tissues that comprise physiology [ 1 ]. Cannon notes in his writing: “The word [homeostasis] does not imply something set and immobile, a stagnation. It means a condition—a condition which may vary, but which is relatively constant.” In other words, “homeostasis” refers to not a static physiological state but an adaptive process by which a system (group of cells, organ, individual) can return to a stable state following insults or “stressors,” whether small or large. Cells and organisms have evolved elaborate molecular mechanisms that mediate such a return to homeostasis. A well‐studied example of such a mechanism is the integrated stress response (ISR), often referred to as a “homeostatic program.” There is an abundance of literature describing the function of ISR in restoring homeostasis following an obvious insult—either cell‐extrinsic, as with nutrient deprivation or immune challenge, or cell‐intrinsic, as with ER stress or oxidative stress [ 2 ]. These and other stressors activate one of four classical ISR kinases in vertebrates (Fig.  1 ): PKR‐like endoplasmic reticulum kinase (PERK), general control nonderepressible 2 (GCN2), protein kinase R (PKR), and heme‐regulated inhibitor (HRI). Phosphorylation of the translation initiation factor eIF2α by any of these kinases leads to reduced initiator methionine availability, and thus a reduction in the rate of global translation [ 3 ]. These suppressive conditions intriguingly induce the translation of a select suite of mRNAs that bear unique structural features in their 5′ leaders [ 4 , 5 ]; the best studied of these encodes the basic leucine zipper (bZIP) protein—activating transcription factor 4 (ATF4) [ 5 , 6 , 7 , 8 , 9 ]. There are also several other transcription factors that are implicated downstream of ISR kinase activation, some of which are known to dimerize with ATF4 in many pathological contexts [ 2 , 10 ]. Schematic of integrated stress response (ISR) signaling showing the core components. Schematic showing ISR components, protein kinase R like endoplasmic reticulum kinase (PERK), general control nonrepressible 2 (GCN2), protein kinase R (PKR), heme‐regulated inhibitor (HRI), and activating transcription factor 4 (ATF4) discussed herein. Also depicted is a well‐studied negative regulator of the ISR activation—growth arrest and DNA damage inducible protein 34 (GADD34), interaction partners of ATF4 (p300, Jun, CCAAT/enhancer‐binding protein (C/EBP)), and transcriptional targets eukaryotic initiation factor 4E‐binding protein (4E‐BP) and C/EBP homologous protein (CHOP). See Table  1 for nomenclature across commonly used model organism species. ATF4 is highly conserved across evolution [ 11 , 12 , 13 ], while the number of ISR kinases varies across species: yeast encode for GCN2 and HRI‐related orthologs [ 14 , 15 ], Caenorhabditis elegans and Drosophila for PERK [ 16 , 17 ] and GCN2 [ 18 , 19 ], and most vertebrates encode for all four (PERK, GCN2, PKR, and HRI) [ 20 , 21 ]. Table  1 lists nomenclature for these core ISR signaling components across popular model organisms. Most instances of elevated ATF4 protein expression are attributed to changes in translation initiation dynamics as a result of ISR kinase activation by obvious exogenous insults, although a handful of studies also report transcriptional upregulation of ATF4 in response to stress [ 22 , 23 ]. ATF4 mRNA is detected near‐ubiquitously in both Drosophila melanogaster and mice [ 24 , 25 ], though the relative transcript abundance varies across tissues. In mice, the highest transcript abundance was found in calvaria, brain, thymus, liver, and lung [ 25 ]. ATF4 protein has been robustly detected in the eye, osteoblasts, adipose tissue, and liver; its role in these and other tissues will be discussed later in this review. In Drosophila , homeostatic ATF4 protein has only been reported in adult adipocytes [ 26 ], though ATF4 translation has been detected in photoreceptors [ 27 ] and ATF4 transcriptional activity has been observed in larval adipocytes, midgut enterocytes, and salivary glands [ 27 ]. Summary of core ISR pathway genes and their nomenclature across species. The constitutive expression of ATF4 in the above tissues inspires a perplexing question: What does homeostasis mean for a cell or tissue that is continuously faced with a high stress burden? For example, cell types with high secretory capacity, such as β‐islet cells (of the pancreas), adipocytes, and photoreceptors, are constantly experiencing ER stress [ 26 , 28 , 29 , 30 ]. Likewise, detoxifying tissues such as the liver or kidney experience elevated oxidative stress since they encounter noxious chemicals that cause increased generation of reactive oxygen species (ROS) [ 31 , 32 , 33 ]. In these cases, “homeostasis” is no longer just a series of stress‐adaptation cycles to return to a stable state, as Cannon first postulated. Rather, one can imagine that maintaining tissue homeostasis sometimes requires equipping the composite cells to tolerate a chronic or persistent cellular stress burden. As previously mentioned, ISR signaling has been studied extensively for its role in adaptive stress responses [ 2 ] and in driving numerous pathologies [ 34 , 35 ]; this review will highlight the less‐often discussed roles of ISR signaling in maintaining organismal and tissue homeostasis under favorable conditions. Here, we describe known roles for ISR signaling components in homeostatic functioning of several tissues across both D. melanogaster and mammals, including adipocytes, hemocytes, skeletal muscle, photoreceptors, and others (see Graphical Abstract). Table  2 provides a summary of the cell types and homeostatic processes that we discuss. We also speculate on other examples of “burdens of function” where ISR signaling may be required but has not yet been interrogated. Much of what is currently known about the function of ISR factors in non‐adaptive homeostasis comes from loss‐of‐function phenotypes in model organisms. Some of the conclusions reported in this review are explicitly stated in the original publications, and some are inferred here based on published data. Using these, we aim to reframe ISR signaling as a workhorse for maintaining cell and tissue homeostasis, in addition to its widely appreciated role as an adaptive mechanism that restores homeostasis following acute stress. Summarizing known roles of ISR under homeostatic conditions across various organisms.

The authors declare no conflict of interest.

As in Drosophila , multiple ATF4 loss‐of‐function models have also been established in mice [ 62 , 63 , 64 ]. Careful viability analysis in the mouse model found a similar (~ 30%) rate of lethality in ATF4 mutants ( CREB‐2 −/− in mice) as that observed for Drosophila crc hypomorphs [ 62 ]. As with the Drosophila mutants, ATF4‐ mutant mice bear several phenotypes, including skeletal defects [ 25 , 63 ], runted development [ 63 ], fetal anemia [ 63 ], and microphthalmia [ 62 , 64 ]. These and other defects in ATF4 knockout mice, and the implications for a homeostatic role of ATF4, are discussed below. Perhaps the best characterized role for ATF4 in mammalian tissue development is in the differentiation and physiological function of osteoblasts, which form from mesenchymal stem cells and are responsible for bone synthesis. ATF4 was found to be a relevant substrate for the serine/threonine kinase RSK2, which is required for osteoblast differentiation [ 65 ]. Loss‐of‐function mutations in RSK2 underlie Coffin–Lowry Syndrome, which is characterized by skeletal and cognitive defects. Mice deficient for RSK2 exhibit decreased osteoblast differentiation and lower bone mass, which was phenocopied by loss of ATF4 [ 65 ]. RSK2 and ATF4 were found to be required for synthesis of collagen, a primary bone component, in cultured osteoblasts [ 65 ]. The defect in collagen synthesis was corrected by the addition of nonessential amino acids to growth media, leading to the postulation that the bone mass defects in RSK2 and ATF4 mutants were secondary to a decrease in amino acid transport, which is a known function of ATF4 [ 66 ]. This postulation was corroborated by a later study showing a striking rescue of the low bone mass defects in ATF4 and RSK2 mutant mice by rearing animals on a high‐protein diet [ 67 ]. There are additional mechanisms that may explain how ATF4 regulates osteoblast differentiation. ATF4 is thought to directly regulate the expression of osteocalcin, which is required for terminal osteoblast differentiation [ 25 , 67 ]. An ATF4‐responsive element was identified in the osteocalcin (OG2) locus. Osteocalcin is expressed in bone progenitor cells and is required for osteoblast differentiation [ 25 ], suggesting that the role of ATF4 and ISR signaling in osteoblast differentiation is cell‐autonomous. Additionally, ATF4 expression is induced by parathyroid hormone (PTH), a key regulator of bone development, specifically promoting anabolic activity in osteoblasts [ 68 ]. Treatment of cultured osteoblasts with PTH led to elevated ATF4 transcript abundance [ 69 ]. Attenuated bone growth upon loss of PTH/ATF4 has been attributed to diminished anabolic activity [ 68 ]. PTH was also found to stimulate ER stress in osteoblasts in a PERK‐dependent manner, and loss of ATF4 led to decreased expression of osteoblast differentiation genes [ 70 ]. Consequently, PERK knockout mice also display skeletal system defects [ 70 ]. Interestingly, to the best of our knowledge, this seems to be the only recorded example wherein a developmental process explicitly requires induction of a stressor (such as dysregulation of ER homeostasis). Loss of ATF4 results in fetal anemia, suggesting a role for ATF4 in the development of hematopoietic stem cells (HSCs) in mice [ 63 , 71 ]. In the mouse fetal liver, a significant pool of HSCs is supported by a stem cell niche, which comprises both endothelial and stromal cells. During gestation, HSCs emerge from the mesoderm and migrate to seed the fetal liver around 8–9 days post coitum [ 72 ]. ATF4‐deficient HSCs exhibited decreased migration as well as decreased ability to self‐renew and differentiate [ 73 ]. Analysis of livers from ATF4 −/− fetuses also found that their HSCs were less competent to repopulate the blood lineage following irradiation [ 73 ]. HSC transplantation experiments demonstrated that, in the absence of ATF4, stromal cells of the niche were less adept at supporting wild‐type HSCs [ 73 ]. RNA‐seq and chromatin immunoprecipitation revealed angiopoietin‐like protein 3 (Angptl3) to be a direct ATF4 transcriptional target and a critical effector of ATF4 in HSCs: Rescue of ATF4 ‐knockout HSCs with Angptl3 enabled these stem cells to more efficiently reconstitute the HSC pool in irradiated mice [ 73 ]. ATF4 has separately been shown to be required for erythropoiesis using ex vivo models of erythroid differentiation [ 71 ]. This activity of ATF4 in erythroblasts was shown to be downstream of HRI [ 71 ]; loss of HRI also causes severe anemia [ 74 , 75 ]. It remained unknown, until recently, whether ATF4 regulates fetal HSCs and adult HSCs erythropoiesis through similar mechanisms. A recent study using conditional knockout mice showed that ATF4 directly regulates transcription of the small subunit ribosome protein, Rps19BP1, leading to impaired ribosome biogenesis in erythroid progenitor cells [ 76 ]. Since erythroid lineage specification is characterized by increased ribosome biogenesis and protein translation [ 77 ], the discovery of ATF4 as a regulator of Rps19BP1 has interesting mechanistic implications in ribosomopathies such as Diamond–Blackfan anemia, which are also caused by defective ribosome biogenesis [ 78 ]. Eye lens development includes sequential steps of primary and secondary lens fiber cell differentiation. Primary cells migrate during differentiation, and secondary cells undergo elongation as part of their differentiation program [ 79 ]. The first reported loss‐of‐function phenotype for mammalian ATF4 is microphthalmia, the lack of an external lens structure [ 62 , 64 ]. Histological analysis revealed that the differentiation and migration of primary lens fiber cells were unaffected by the loss of ATF4 [ 62 , 64 ]. These early studies noted that in embryonic stages, ATF4 depletion did not result in defects in the retina, cornea, or iris, but did result in disrupted elongation and differentiation of secondary fiber cells, ultimately leading to the degeneration of this structure in late embryonic stages. Restoring ATF4 specifically in the lens was sufficient to rescue lens development defects in ATF4 knockout mice, suggesting ATF4 is necessary and sufficient for lens development [ 64 ]. Interestingly, the lens development defects were accompanied by hyperplasia of the secondary lens fibers, where the number of cells was increased, but all the cells appeared to have normal morphology [ 64 ]. The differentiation of lens secondary fiber cells is marked by an increase in the secretory function of these cells; lens fiber cells secrete proteins called crystallins that are responsible for ensuring the refractive index of the lens [ 79 ]. ATF4 itself is not required for the secretion of crystallin proteins in the lens [ 64 ]; rather, the role of ATF4 may be to protect these highly secretory cells from degeneration in late embryogenesis. This is supported by results from another study wherein defective lens development caused by ATF4 deletion was rescued by concomitant deletion of the pro‐apoptotic factor p53 [ 62 ]. This is not the only context where ATF4 and P53 have been shown to act synergistically. ATF4‐p53 interactions are discussed in the “ ISR signaling in oscillators ” section later in this review. These results together argue that ATF4 is important not only for the development of secondary fiber cells in the lens, but also for ensuring the survival of these cells throughout mammalian ocular development. Additionally, independent works in vertebrate models (mice and chicken) have determined that BMP signaling is necessary for eye secondary lens fiber development [ 80 , 81 ], crystallin secretion, and cell elongation, the latter of which is also dependent on ATF4. This suggests that, similar to the wing and skeletal systems, development of the eye lens may also involve interplay between ATF4 and BMP signaling pathways. It is unknown which ISR kinase is upstream of ATF4 in the mammalian lens. However, the ISR kinase, PERK has been implicated in being required for proper visual system function in Drosophila where electrical recordings from PERK mutants show aberrant electroretinograms [ 46 ], which could be due to the absence of protective ATF4 signaling and subsequent cell death in Drosophila retinas, which contain highly secretory photoreceptors. The known and potential roles of ISR in visual system homeostasis and as a potential driver of aging in the visual system are discussed in detail in the following sections. The pleiotropic phenotypes of ATF4 mutant mice and Drosophila leading to neonatal lethality likely mask many cell‐type‐specific developmental defects. Fortunately, cell culture‐based models have served as an excellent model system to study the role of ISR in some contexts in vitro . An example is the role of ATF4 in keratinocyte differentiation [ 82 ]. Differentiation and maturation of keratinocytes involves a series of transcriptional changes that drive morphological changes and can be achieved by the addition of calcium to the growth medium in culture [ 83 ]. Using polysome profiling in this, one study found that the differentiation of keratinocytes in vitro is marked by a global repression of translation, a direct consequence of ISR activation indicated by upregulated phosphorylation of eIF2α and increased ATF4 levels. These results were also consistent with results from human skin samples, which showed higher phospho‐eIF2α in the suprabasal epidermis where differentiated keratinocytes reside, in comparison with basal undifferentiated keratinocytes. Interestingly, authors determined that, in vitro , GCN2 depletion results in improper differentiation of keratinocytes, indicating that the GCN2 axis of ISR activation is necessary for keratinocyte differentiation. However, inducing phosphorylation of eIF2α alone was not sufficient to drive differentiation [ 82 ], suggesting that GCN2‐ATF4 signaling acts cooperatively with other pathways required for keratinocyte differentiation. Additionally, ISR signaling via GCN2 activation has recently been implicated in the crucial barrier function of keratinocytes [ 84 ]. GCN2 deletion in cultured keratinocytes as well as in wounded mice resulted in impaired wound healing [ 84 ]. Conversely, GCN2 activation in wounded cells results in actin remodeling, allowing cells to modulate cytoskeletal changes to close the wound [ 84 ]. ROS activation is required at the leading edge of keratinocytes for appropriate cytoskeletal rearrangements; interestingly, this also requires GCN2 activity [ 84 ], suggesting that GCN2‐mediated activation of ISR is required for proper functioning of keratinocytes in their crucial barrier maintenance role. Whether a specific stressor upstream of GCN2 is required for this function remains to be uncovered. Recent work in primary cell culture found that early expression of ATF4 activity, downstream of PERK, is important for airway cell differentiation and airway epithelial development. The mammalian airway epithelium has multiple cell types: basal stem cells, multiciliated cells, and secretory cells including goblet cells and club cells. Knockdown experiments in a mouse in vitro airway epithelial cell culture model revealed that ATF4 expression is required for secretory cell differentiation [ 85 ]. ATF4 depletion or pharmacological PERK inhibition resulted in defective secretory cell differentiation from basal stem cells. ATF4 depletion also resulted in assembly defects of multiciliated cells but not their function [ 85 ], implying that ATF4 is important specifically for development of the airway epithelium. PKR inhibition did not result in impaired differentiation, suggesting ISR kinase specificity in the context of airway epithelium development. (Roles for GCN2 and HRI have not yet been tested in this model.) The authors of this study proposed that PERK activation is specifically responsible for ATF4 induction in airway cells, which activates transcriptional programs that specify secretory cell lineages in the mammalian airway epithelium. ChIP‐seq analysis of ATF4 in mouse tracheal epithelial cells (MTEC) demonstrated that ATF4 can directly bind regulatory elements of secretory cell markers: SMAD5 and BMPR1B [ 85 ]. Taken together, the airway secretory epithelium, like osteoblasts, frames ATF4 activation as a precursor for differentiation of secretory lineages by direct transcriptional regulation of TGF‐β/BMP [ 85 ]. Another theme emerges from this recent work: translation changes associated with ISR signaling are required for differentiation of multiple cell types (club and goblet cells in the airway epithelium, as well as keratinocytes and HSCs described earlier in this review).