A beneficial adaptive role for CHOP in driving cell fate selection during ER stress

CHOP supports EdU incorporation during ER stress in primary mouse embryonic fibroblasts

C/EBP-family transcription factors generally stimulate differentiation and inhibit proliferation (Johnson, 2005; Nerlov, 2007), and CHOP likely acts as a dominant-negative member of this family (Ron and Habener, 1992). Thus, despite CHOP’s characterization as growth arrest-inducible and its putative role in cell cycle arrest (Harris et al, 2001; Hendricks-Taylor and Darlington, 1995; Mihailidou et al, 2010), we speculated that CHOP might have an unappreciated role in proliferation that might ordinarily be obscured by the acutely cytotoxic doses of chemical agents typically used to elicit ER stress and probe CHOP’s function. To test this idea, we compared the incorporation of the thymidine analog 5-ethynyl-2’-deoxyuridine (EdU) in synchronized wild-type (w.t.) and Chop knockout (Chop^−/−) primary mouse embryonic fibroblasts (MEFs) that were treated with either 2.5 or 50 nM thapsigargin (TG), a stressor that causes ER stress by blocking ER calcium reuptake. The doses used were one- to two orders of magnitude lower than typically used to elicit ER stress, yet are still sufficient to activate the UPR. Moreover, as we have shown, MEF cultures can still undergo net proliferation during treatment with 2.5–10 nM TG despite the perturbation (Rutkowski et al, 2006).

The cells used for this experiment were isolated by timed intercross of animals heterozygous for a previously described constitutive Chop null allele (Zinszner et al, 1998). To ensure that any results were robust and attributable to the presence or absence of CHOP and not to unrelated line-to-line differences, experiments were performed in at least two separate wild-type (w.t.) and Chop^−/− lines. Under these conditions, there was no apparent difference in EdU incorporation between w.t. and Chop^−/− cells in the absence of stress, indicating that knocking out CHOP does not affect basal proliferation (Figs. 1A and EV1A). In this experiment, treatment with 2.5 nM TG trended toward slightly diminished EdU incorporation in Chop^−/− cells but not in wild-type cells—though this trend did not reach statistical significance—while 50 nM TG almost entirely prevented EdU incorporation in Chop^−/− cells yet allowed for modest but significant incorporation in w.t. cells (Fig. 1B). Comparable results were observed in both synchronized (Fig. 1B) and non-synchronized cells (Fig. EV1B). A similar phenotype was observed when MEFs were treated with the mechanistically unrelated ER stressor tunicamycin (TM, Figs. 1C and EV1C), suggesting that the CHOP-mediated EdU incorporation difference is not stressor-specific. Because EdU incorporation reads out on S-phase progression, these data suggest that CHOP promotes cell proliferation. We found that deletion of CHOP increased the percentage of cells in G2 at the expense of G1 cells (Fig. EV1D). There was also decreased mRNA expression of the cell cycle regulators Pcna, Cyclin A2, and Cyclin E1, and increased expression of the negative regulator of PCNA p21, although Ki67 and Cdk6 expression were also elevated (Fig. EV1E). These data are consistent with CHOP facilitating cell proliferation but not acting as a direct regulator of cell cycle progression. The difference in proliferation was also observed in the liver cell line TIB-73 when CHOP was deleted using CRISPR/Cas9 targeting (Figs. 1D and EV1F); therefore, the phenotype is not cell-type specific. The effect was also observed in MEFs after immortalization by large T-antigen (Fig. EV1H), but not in the immortal MEF line 3T3 (Fig. EV1I), indicating that there are at least some scenarios where the difference is lost. A potential role for CHOP in promoting proliferation was first tentatively observed but otherwise unexplored in the original characterization of Chop^−/− cells and animals (Zinszner et al, 1998).

CHOP is necessary and sufficient to support EdU incorporation during ER stress

The difference in EdU incorporation between w.t. and Chop^−/− cells might reflect a direct role for CHOP in proliferation, or a compensatory mechanism that arises in embryos in which CHOP has been deleted. Thus, we created an allele in which CHOP could be controlled with temporal precision. In this allele, termed “FLuL” (Frt-Luciferase-LoxP), a gene trap cassette expressing a splice acceptor, nanoLuciferase (nLuc), and a transcriptional terminator, together flanked by Flp sites, blocks expression of the CHOP open reading frame, which resides in Chop exons 3 and 4 (Fig. 2A). Deletion of the cassette by the FLPo recombinase creates a floxed allele (“fl”) that restores CHOP expression (Fig. 2B), which can then be again eliminated by the action of CRE (“KO”) (Fig. 2C). Consistent with previous findings with global CHOP knockout mice (Zinszner et al, 1998), Chop^FLuL/FLuL mice were viable, fertile and grossly indistinguishable from wild-type (w.t.) mice. Primary Chop^FLuL/FLuL MEFs expressed nLuc, and its expression was greatly enhanced by TM (Fig. 2D), which was expected since nLuc should be under the same transcriptional and translational control as CHOP itself. At the same time, CHOP expression was completely lost in Chop^FLuL/FLuL MEFs but restored in Chop^fl/fl MEFs in which the gene trap had been deleted by FLPo (Fig. 2E). The FLuL allele could be manipulated both in vivo and in vitro—in the latter case by adding recombinant adenovirus expressing FLPo to restore CHOP in the FLuL allele, or expressing CRE to delete CHOP in the floxed allele. In either case, Ad-GFP is used as a control. Demonstrating the feasibility of this approach, treatment of Chop^FLuL/FLuL cells with Ad-FLPo eliminated transcripts expressing nLuc and restored transcripts expressing Chop exons 3–4; as expected, all transcripts from the Chop allele were stress-dependent in their expression (Fig. 2F). Stress-mediated induction of Chop exons 1–2 (which are represented in both the Chop transcript and the nLuc transcript) was actually higher in cells treated with FLPo (Fig. 2F, top), which implies either that the nLuc transcript is degraded more rapidly than is the Chop transcript, or that CHOP contributes to its own regulation such that restoration of CHOP protein expression by FLPo enhances transcription from the locus (or both). This in vitro deletion approach has the advantage of allowing the effects of CHOP to be examined in cells of the same origin, which is not the case when using MEF lines derived from separate embryos isolated from heterozygote intercrosses.

Further validating these lines, w.t. and Chop^FLuL/FLuL MEFs were treated with different concentrations of TG for either 24 or 36 h. PARP cleavage, an indicator of apoptotic cell death, was more prominent in w.t. MEFs than in Chop^FLuL/FLuL MEFs, particularly at 50 nM TG (Figs. 2G and EV1G). Similarly, FLPo-infected (and hence wild-type) MEFs from the Chop^FLuL/FLuL background released more lactate dehydrogenase (LDH) than GFP-infected MEFs after TG treatment, suggesting that FLPo-directed CHOP re-expression in Chop^FLuL/FLuL MEFs enhanced cell death under ER stress (Fig. 2H). Taken together, these data validate in this new model the well-described role of CHOP in promoting cell death upon ER stress.

The CHOP-dependent effect on EdU incorporation observed in Fig. 1 was also seen in MEFs harboring this new Chop allele. Wild-type or Chop^FLuL/FLuL MEFs were treated with 50 nM TG or 500 ng/mL TM for 24 h, and EdU incorporation was evaluated by flow cytometry. As in Fig. 1, no consistent difference in basal EdU incorporation was seen between w.t. and Chop^FLuL/FLuL MEFs. However, there was more EdU incorporation in w.t. MEFs compared to Chop^FLuL/FLuL MEFs during both TG and TM treatment (Fig. 3A,B). Induction of CHOP expression in Chop^FLuL/FLuL MEFs by provision of Ad-FLPo was sufficient to increase EdU incorporation upon TG treatment (Figs. 3C and EV2A), while deletion of CHOP by provision of Ad-CRE to Chop^fl/fl MEFs decreased EdU incorporation (Figs. 3D and EV2B). Taken together, these results robustly establish that CHOP is both necessary and sufficient to support EdU incorporation under ER stress.

CHOP supports proliferation during ER stress through restoration of protein synthesis

Given the well-established role of CHOP in promoting cell death, we examined the relationships among EdU incorporation, cell cycle progression, and cell death. First, double staining for EdU and the cell death marker cleaved PARP showed the EdU-positive and dying populations of cells to be essentially mutually exclusive (Fig. 4A). In addition, the pan-caspase inhibitor Q-VD-OPH did not alter the propensity of wild-type cells to incorporate EdU over Chop^FLuL/FLuL MEFs (Fig. 4B). Therefore, EdU-positive cells do not represent a population of dying cells, nor are they affected by the ability of other cells to die. Rather, EdU-positive cells progress through the cell cycle. To demonstrate this, we utilized the reversible ER stressor- dithiothreitol (DTT). Cells were pulsed with 5 mM DTT for 2 h, EdU was incorporated after DTT washout by pulse labeling, and the cell cycle progression of the EdU-positive cells over time was assessed by propidium iodide (PI) staining of DNA content. As we have shown, cells can survive and proliferate despite repeated pulses with high concentrations of DTT (Wu et al, 2007). EdU-positive cells progressed from S-phase through G2/M such that, by 12 h after the EdU pulse, almost all of the EdU-positive cells had returned to the 2 N DNA content characteristic of G1 (Figs. 4C and EV3A–D). This was true of both w.t. and Chop^FLuL/FLuL EdU-positive cells, though as expected there were far fewer of the latter. We also considered the possibility that dying cells release some mitogenic factor that would stimulate proliferation of non-dying cells independent of the receiving cells’ Chop genotype (Willy et al, 2015). To test this idea, we carried out a medium switch experiment where the culture medium of TG-treated w.t. or Chop^FLuL/FLuL MEFs was collected and applied to non-treated w.t. MEFs, in which EdU incorporation was then assessed. Under these conditions, whether the media came from w.t. or Chop^FLuL/FLuL MEFs had no effect on EdU incorporation in the receiving cells (Fig. EV3E). Taken together, these data show that CHOP promotes cell cycle progression and argue that this effect does not likely arise as an indirect consequence of CHOP-mediated cell death.

CHOP has been proposed to transcriptionally regulate apoptosis (Han et al, 2013; Hu et al, 2018; Krokowski et al, 2013; Zinszner et al, 1998), inflammation (DeZwaan-McCabe et al, 2013; Scaiewicz et al, 2013; Willy et al, 2015), and lipid metabolism (Chikka et al, 2013) among other processes. While the functions of CHOP might depend on the cell type in which it is expressed and the complement of other transcription factors with which it might interact, in MEFs at least, its essential impact on gene expression appears to be the restoration of protein synthesis, which occurs both through CHOP-dependent upregulation of the eIF2α phosphatase GADD34 (Marciniak et al, 2004) and the broader induction of tRNA synthetases and other genes that augment the protein biosynthetic capacity (Han et al, 2013; Krokowski et al, 2013). CHOP-dependent enhancement of protein synthesis was shown to contribute to cell death under stress by enhancing the burden of nascent proteins on an already-stressed ER (Marciniak et al, 2004). However, this effect was tested under stress conditions severe enough that cells would have had little likelihood of ever restoring ER homeostasis. Thus, we considered that augmenting protein synthesis might benefit cells that are able to restore ER homeostasis with a proliferative advantage. To test this idea, we took advantage of ISRIB (Integrated Stress Response Inhibitor), a chemical that counteracts the suppression of protein synthesis under stress by allosterically modulating the GTP exchange factor eIF2B in such a way as to prevent its engagement by eIF2α (Sekine et al, 2015; Sidrauski et al, 2013; Zyryanova et al, 2021). Under these conditions, ISRIB dampened eIF2α-dependent signaling including expression of at least ATF4 and possibly CHOP as well (Fig. 4D). And, as expected, ISRIB largely or completely prevented a short treatment (4 h) of TG from inhibiting protein synthesis, as assessed by ³⁵S incorporation, and this effect was seen in both CHOP-expressing cells (Chop^fl/fl treated with Ad-GFP) and CHOP-null cells (Chop^fl/fl cells treated with Ad-CRE) (Fig. EV3F, G). ISRIB increased EdU incorporation upon TG treatment in CHOP-null cells, whereas it had no such effect in wild-type cells (Fig. 4E). Conversely, treatment of cells with the GADD34 inhibitor salubrinal, which perpetuates the stress-dependent inhibition of protein synthesis (Boyce et al, 2005), diminished EdU incorporation (Fig. 4E), as did siRNA-dependent knockdown of Gadd34 (Fig. 4F,G). From these data, it is likely that the proliferative function of CHOP is carried out at least in part by its previously described role in reversing the attenuation of protein synthesis caused by eIF2α phosphorylation.

CHOP confers a competitive advantage on cells under mild ER stress

The observation that CHOP promotes proliferation raises the possibility that, during stresses of mild intensity (which presumably better reflect the sorts of stresses encountered in normal physiological scenarios), CHOP might confer a functional benefit that has not been appreciated before. To test this prediction in the most sensitive and rigorous way, we subjected cells of identical origin either expressing or lacking CHOP to a growth competition assay. The experiment was performed in both directions: either Chop^FLuL/FLuL cells were treated with Ad-GFP or Ad-FLPo; or Chop^fl/fl cells were treated with Ad-GFP or Ad-CRE. We used quantitative PCR to identify each allele (Fig. 5A) and confirmed that each primer pair was both specific (Fig. 5B) and efficient (Fig. 5C) in detecting the allele against which it was designed. We then mixed each pair of cells in independent replicate plates at a 1:1 ratio and cultured them for up to 3 passages, in the presence of either vehicle or 2.5 nM TG, with the media and stressor refreshed every 48 h. The cells were passaged upon reaching confluence, with the replicates being kept separate from each other, and with an aliquot kept from each plate for qPCR. As we have previously shown, 2.5 nM TG does not appreciably diminish the proliferative capacity of MEFs, even though that dose is sufficient to activate the UPR (Rutkowski et al, 2006). When cells were cultured in stressor-free media, there was no significant change in the ratio of either the FLuL-to-Flox alleles or the Flox to KO alleles (Fig. 5D,E)—unsurprising since no detectable CHOP is expressed in the absence of stress (e.g., Fig. 4E). However, CHOP-expressing cells of both origins significantly outcompeted CHOP-null cells when challenged with 2.5 nM TG (Fig. 5D,E). These data suggest that CHOP augments the functional recovery of cells from or adaptation to mild ER stress.

Single cell analysis confirms a proliferative role for CHOP

The fact that CHOP promotes both cell death and proliferation, in mutually exclusive populations, suggests that the response of a population of cells to a stressor is not uniform. Thus, we wanted to better understand how CHOP expression corresponded to cell fate in individual cells. To accomplish this, we used flow cytometry to detect endogenous CHOP expression in cells subjected to ER stress. For this purpose, we used a monoclonal antibody whose specificity for CHOP we have previously demonstrated (DeZwaan-McCabe et al, 2013). We treated cells with a dose of TG (5 nM) that we have previously shown permits net proliferation in MEFs (Rutkowski et al, 2006). Over a time course of that mild TG treatment, the expression of CHOP increased uniformly in the cell population, reaching its maximum by 8 h, at which point all cells expressed CHOP. However, at subsequent time points, the cells began to separate into two populations, one of which retained maximal CHOP expression and the other of which became CHOP-negative, with relatively few cells in an intermediate state (Fig. 6A). In contrast, cells treated with 100 nM TG rapidly became CHOP-positive, to the same extent as cells treated with 5 nM TG, but remained so through the time course (Fig. 6B). This behavior was mirrored in vivo, in animals challenged by an IP injection of TM. Eight hours after challenge, most if not all hepatocytes expressed CHOP, as seen by immunostained nuclei, whereas at a later time point a few cells remained strongly CHOP-positive while CHOP was undetectable in the remainder (Fig. 6C). Under conditions of severe stress (100 nM), there was little or no effect of CHOP on the overall ER stress burden, as assessed by splicing of the IRE1α target Xbp1; however, during mild stress (5 nM), CHOP enhanced UPR activation at the same time points (particularly 16 h) at which the population of cells had split (Fig. 6D, E). Thus, CHOP can promote at least two mutually exclusive cell fates (death and proliferation) and is expressed in at least two distinct temporal patterns (transient and persistent).

These findings raise the question of whether the transience of CHOP expression has any bearing on whether or not cells proliferate. Unfortunately, because CHOP is a nuclear antigen, flow cytometry requires fixation and permeabilization, thus preventing meaningful downstream analysis of how these two populations of cells differ. Moreover, CHOP expression is extensively regulated transcriptionally, post-transcriptionally, and translationally, and probably by its degradation (Ma et al, 2002; Palam Baird and Wek, 2011; Rutkowski et al, 2006; Ubeda et al, 1999; Zinszner et al, 1998), meaning exogenous fluorescent reporters are unlikely to faithfully reflect the dynamics of endogenous CHOP. Indeed, even our nLuc gene trap, though it is knocked into the Chop locus and (presumably) subject to the same transcriptional and translational control as is endogenous CHOP, shows a considerable basal expression that CHOP itself does not (Fig. 2D,E).

Thus, as an alternate approach, we used single-cell RNA-seq (scRNA-seq) to better characterize how Chop expression relates to cell fate, while recognizing that mRNA expression is an imperfect readout for such a complex process. We mixed Chop^FLuL/FLuL cells treated with either Ad-GFP (CHOP-null) or Ad-FLPo (CHOP-expressing) and challenged them with 10 nM TG for 16 h (Fig. EV4A–D). We chose this dose because it is the lowest at which we can reliably observe a CHOP-dependent difference in EdU incorporation (Figs. 3C,D and EV4A). The fact that the nLuc cassette is present only in the Chop^FLuL/FLuL cells but not the Chop^fl/fl cells (Fig. 2A,B) and that nLuc is expressed even under non-stressed conditions (Fig. 2D), allowed us to discriminate cells of each genotype within the population by their expression of nLuc or lack thereof. Indeed, while there was a small percentage of cells that express neither Chop nor nLuc (cluster 6), the expression profiles of Chop and nLuc are otherwise essentially mutually exclusive (Fig. 7A). UMAP (Uniform Manifold Approximation and Projection) analysis separated cells into two major groups of cells, and the Leiden algorithm identified 15 potentially distinct clusters within the population (Fig. 7A; Dataset EV1). Pathway analysis revealed that clusters 0 and 8 comprised proliferating cells, as the “Chromosome Segregation” pathway was dramatically enriched in both groups, and to a much lesser extent, if at all, in any other group (Figs. 7B and EV4E). “DNA-dependent DNA replication” was also enriched in clusters 0 and 8, and also in clusters 6 and 7. Notably, the “Response to ER stress” pathway was not enriched in clusters 0 or 8. The conclusion that clusters 0 and 8 represent proliferating cells was supported by expression of the proliferation markers Mki67 (Fig. 7C) and Pcna (Dataset EV1). Notably, while there are some nLuc-expressing cells (all in group 0), the majority of cells in the two groups are wild-type with respect to Chop expression (Fig. 7A,B). Yet, among the groups with substantially expressed Chop (1, 2, 0, 8, and 13), Chop was lowest in groups 0 and 8 (Fig. 7D). Our pathway results also showed that groups 1, 2, 3, 5, 9, 10, 13, and 14—essentially, the upper right portion of the UMAP plot plus the small group 14—represent cells with an activated UPR at that time point (Fig. 7B), with the UPR (or “response to ER stress”) being the most highly upregulated pathway in clusters 2 and 9 (Fig. EV4E). These results support the model that CHOP promotes proliferation, and also suggest that this proliferation occurs in cells with attenuated UPR signaling.

Supporting the Xbp1 splicing data in Fig. 6, the scRNA-seq data illustrates that CHOP promotes the maintenance of UPR signaling. We arrive at this conclusion based on groups 4 and 7. These two closely related groups comprise predominantly nLuc-expressing (and therefore CHOP-negative) cells (Fig. 7A). UPR activation is lower in these cells than in the clusters in the upper right quadrant of the UMAP graphs, as seen both in lower expression of nLuc (Fig. 7A), the ER cochaperone Dnajc3, and other UPR target genes (Fig. 7C and Dataset EV1). (We chose Dnajc3 to display because, unlike many other genes encoding ER chaperones, it is expressed across a fairly wide dynamic range, making differences in its expression more obvious). A broader analysis of UPR targets genes also supports that groups 4 and 7 are not characterized by UPR activation (Fig. EV4F). Yet these two groups also showed elevated expression of the Integrated Stress Response (ISR) marker genes Rars, Chac1, Atf4, and Trib3 (Figs. 7C and EV4F). The ISR refers to the gene regulatory pathway that is induced downstream of eIF2α kinases, including PERK but also PKR, HRI, and GCN2 which are sensitive to other types of stress (Harding et al, 2003). Therefore, while the ISR is one component of the UPR, it can be functionally dissociated from the UPR, as appears to be the case in clusters 4 and 7. From this result, we conclude that loss of CHOP increases the likelihood that a cell will have a gene expression profile consistent with suppression of the UPR but hyperactivation of the ISR. This finding is also consistent with the previously published observation that CHOP restrains ISR activity (Kaspar et al, 2021). We also found when we profiled the expression of a larger range of UPR target genes that a majority of them (12/20) were more highly expressed in the wild-type UPR-activated clusters (1, 2, 13, 14) than in the knockout UPR-activated clusters (4, 7, 9, 10, 11; 7 genes out of 20) (Fig. EV4F). While several of the genes that are more highly expressed in CHOP-positive cells have been identified as direct CHOP targets (Ero1l, Wars, Bip, Ppp1r15a) (Han et al, 2013), most have not.

To gain more insight into the functional consequences of CHOP activity, we compared the extent to which individual genes in the dataset correlated with expression of Chop versus with nLuc, under the assumption that the genes most strongly dependent on CHOP would show the largest difference in correlation between CHOP and nLuc in a way that removed ER stress itself as a confounding variable (since cells with high expression of either CHOP or nLuc would be those cells with strong ISR or UPR activation). Consistent with the conclusions from global RNA-seq data (Han et al, 2013), there was little difference in the expression of a sampling of apoptosis-related genes previously implicated as CHOP targets between cells of the two genotypes, suggesting that they are not strongly affected, if at all, by CHOP under these conditions (Fig. 7E). In contrast, among the 20 UPR target genes shown in Fig. EV4F, 16 of them correlated more strongly with CHOP than with nLuc—some of them substantially so, including previously-identified CHOP targets (Ppp1r15a/Gadd34, Ero1l, Trib3, Wars) but also other UPR target genes not previously linked to CHOP or PERK but instead tied to the ATF6 or IRE1 pathways (Pdia4, Pdia6, Dnajc3, Edem) (Adamson et al, 2016; Lee Iwakoshi and Glimcher, 2003). This finding that most UPR genes are more strongly upregulated when CHOP is present—including genes that are probably not direct targets of CHOP—is consistent with the idea that CHOP can aggravate ER stress and thus lead to increased expression of some UPR targets that are not themselves directly regulated by CHOP. Interestingly, when we performed pathway analysis of the genes with the strongest correlation difference between CHOP and nLuc, we found that pathways of DNA replication and cell cycle control were enriched (Fig. 7F; Dataset EV2), which provides further evidence that a major consequence of CHOP action during moderate ER stress conditions is cell cycle progression and proliferation.

Generation of the Chop FLuL and Chop Flox allele

Chop^FLuL/FLuL mice were generated by CRISPR/Cas9-mediated targeting, carried out by the University of Iowa Genome Editing Facility, using techniques based on (Miura et al, 2018). C57BL/6J mice were purchased from Jackson Labs (000664; Bar Harbor, ME). Male mice older than 8 weeks were used to breed with 3–5-week-old super-ovulated females to produce zygotes for pronuclear injection. Female ICR (Envigo; Hsc:ICR(CD-1)) mice were used as recipients for embryo transfer. All animals were maintained in a climate-controlled environment at 25 °C and a 12/12 light/dark cycle.

Chemically modified CRISPR-Cas9 crRNAs and CRISPR-Cas9 tracrRNA were purchased from IDT (Alt-R® CRISPR-Cas9 crRNA; Alt-R® CRISPR-Cas9 tracrRNA (Cat# 1072532)). The crRNAs and tracrRNA were suspended in T10E0.1 and combined to 1 μg/μl (~29.5 μM) final concentration in a 1:2 (μg:μg) ratio. The RNAs were heated at 98 °C for 2 min and allowed to cool slowly to 20 °C in a thermal cycler. The annealed cr:tracrRNAs were aliquoted to single-use tubes and stored at −80 °C.

Cas9 nuclease was also purchased from IDT (Alt-R® S.p. HiFi Cas9 Nuclease). Cr:tracr:Cas9 ribonucleoprotein complexes were made by combining Cas9 protein and cr:tracrRNA in T10E0.1 (final concentrations: 300 ng/μl (~1.9 μM) Cas9 protein and 200 ng/μl (~5.9 μM) cr:tracrRNA). The Cas9 protein and annealed RNAs were incubated at 37 °C for 10 min. The RNP complexes were combined with single-stranded repair template and incubated an additional 5 min at 37 °C. The concentrations in the injection mix were 100 ng/μl (~0.6 μM) Cas9 protein and 20 ng/μl (~0.6 μM) each cr:tracrRNA and 40 ng/μl single-stranded repair template.

Pronuclear-stage embryos were collected in KSOM media (Millipore; MR101D) and washed 3 times to remove cumulous cells. Cas9 RNPs and double-stranded repair template were injected into the pronuclei of the collected zygotes and incubated in KSOM with amino acids at 37 °C under 5% CO₂ until all zygotes were injected. Fifteen to 25 embryos were immediately implanted into the oviducts of pseudo-pregnant ICR females. Guide RNA sequences were as follows:

Ddit3_5PACTAATGATGGTGTGTCGGGA

Ddit3_3PCCTGCACCAAGCATGAACAGT

Correct targeting was verified by sequencing through the allele in founder mice. Chop^fl/fl mice were generated by breeding Chop^FLuL/+ mice with the FLP delete strain Pgk1-flpo in the C57BL/6J background (Jackson labs strain 011065; (Wu et al, 2009)) and then breeding the allele to homozygosity. All animal usage was approved by and in accordance with Institutional Animal Care and Use Committee procedures Protocol #3021076.

Primers used for genotyping were:

Primer A: TGGATCTGGCAGGGTCAAAG

Primer B: CCCAACCCCTTCCTCCTAC

Primer C: TGGAAAGGACATACATTCCA

With these primers, the w.t. product is 270 bp, the FLuL product is 194 bp, and the Flox product is 382 bp.

Cell Culture and drug treatments

Primary MEFs of different genotypes were isolated by timed intercrosses of Chop^+/− (Zinszner et al, 1998), Chop^FLuL/+, or Chop^fl/+ animals. MEFs of the various genotype combinations from the FLuL allele are available upon request. All MEF experiments were repeated in at least two independently-derived cell lines to confirm robustness of data. Female mice were euthanized at 13.5 days post-coitus and MEFs were isolated following the procedure described by Marian et al, 2013. Genotypes of isolated MEFs were determined by PCR. For cell culture, MEFs were maintained in DMEM containing 4.5 g/L glucose (Invitrogen, USA) supplemented with 10% FBS, 1% L-glutamine, penn/strep, amnio acids, and non-essential amino acids (Invitrogen, USA), at 37 °C in a 5% CO₂ incubator. Cell cycle synchronization was carried out by culture for 48 h in serum-free medium prior to stress treatments in complete medium. For chemical treatment, 1000X stocks of TG (EMD/Millipore), TM (EMD/Millipore), ISRIB (EMD/Millipore), Q-DV-OPh (Cayman), ISRIB (EMD/Millipore), and Salubrinal (EMD/Millipore) were made in DMSO and 200X stocks of DTT were made in 1x PBS, and single-use aliquots were frozen. For non-treated cells, DMSO or PBS was added to the same concentration as for treated cells. The plating densities of cells were 2.0–2.5 × 10⁵ cells/well in 12-well plates (for pulse labeling assay), 3.0–4.0 × 10⁵ cells/well in six-well plates (for protein and RNA analysis), 7.0–8.0 × 10⁵ cells/well in 60 mm dishes (for flow experiments), and cells were allowed to rest overnight before stressor treatment. For growth competition experiments, the media was removed and replaced with fresh media containing stressor every 48 h. Both treated and non-treated groups were passaged when non-treated cells reached ~90% confluency. During passaging, both groups were plated at the same density, rested overnight, and then treated again with DMSO or stressor-containing media the next day. For Q-DV-OPh, the stock concentration of the drug was 10 mM (in DMSO). Cell cultures were periodically confirmed to be mycoplasma-negative. In vivo challenge with TM was performed intraperitoneally with TM or DMSO diluted in PBS.

Other cell culture

To delete CHOP in TIB-73 cells (RRID:CVCL_4383), gRNAs targeting the N-terminus of CHOP were cloned into the pX458 plasmid backbone that also expresses Cas9 and GFP separated by a P2A cleavage site (Ran et al, 2013). TIB-73 cells were transfected with either empty vector or gRNA-containing vector using Lipofectamine 3000 (Thermofisher, USA). Forty-eight hours after transfection, GFP positive cells (from both groups) were sorted by flow cytometry and plated on 96-well plates (1 cell per well). After obtaining single cell-derived colonies, cells were trypsinized and expanded, and screened for CHOP expression by treatment with TM (5 μg/ml for 24 h) followed by western blot. For GADD34 knockdown, MEFs were seeded on 60 mm cell culture dish at 0.7 × 10⁶ cells/dish. Cells were allowed to attach overnight before siRNA transfection. siRNA transfection was performed following manufacture’s protocol (MEF Avalanche Transfection Reagent, EZ biosystems, EZT-MEFS-1) using either scrambled negative control siRNA (IDT, 51-01-14-04) or validated GADD34 siRNA (IDT, mm.Ri.Ppp1r15a.13.2). Cells were treated with TG 24 h after transfection. For LDH assay, cells were treated with vehicle or TG for 24 h and cytotoxicity was assessed using the CyQuant LDH Assay (ThermoFisher) according to the manufacturer’s protocol.

Flow cytometry and immunostaining

EdU (final conc 10 μM) was added directly to the culture medium 4 h before cell harvest unless specified elsewhere. Upon harvest, cells were trypsinized and resuspended in 1X PBS for EdU staining. EdU staining was performed using a Click-iT EdU Alexa Fluor 488/594/647 Flow Cytometry Kit (Thermofisher, USA). After staining, cells were resuspended in 1X PBS and analyzed using an LSRII flow cytometer (Becton Dickinson). For PARP staining, after EdU staining cells were resuspended in incubation buffer (0.5% BSA in 1x PBS) in the dark at RT for 30 min. After incubation, cells were then resuspended in antibody staining solution (cleaved-PARP antibody (1:100 dilution, Cell Signaling Technology, 94885) in incubation buffer) and incubated at RT for 1.5 h. Cells were then washed three times with 1x PBS and incubated in secondary antibody staining solution (1:500 dilution, goat anti-rabbit Alexa488 (Thermofisher A-11008)) at RT for 1 h. Cells were then washed four times in 1x PBS and subjected to flow cytometry. For cell cycle analysis, wild-type and Chop^FLuL/FLuL primary MEFs were cultured in 5 mM DTT-containing media for 2 h to induce ER stress. The DTT- media was then changed to fresh, stressor-free media after washing with 1x PBS. Fifteen hours later, EdU was directly added to culture medium at a final concentration of 10 μM. After a further 30 min, the EdU-containing media was changed to normal culture media after washing in 1x PBS. The cell cycle status of EdU-labeled cells was evaluated at 0, 6, and 12 h post-EdU pulse. EdU staining was as above. After EdU staining, cells were resuspended in PI staining solution (100 μg/RNase, 20 μg/mL PI in 1x PBS) and incubated in the dark at 37 °C for 30 min. Samples were analyzed on a Becton Dickinson LSRII immediately after PI staining. Data was analyzed using FlowJoV10. For flow experiments, SSC-FSC voltage was set based on pilot experiments on the same cell type. We first gated on the FSC-A-SSC channel to gate for the main cell population, and then doublets were excluded by an additional gate on the FSC-W channel. Proper voltage for fluorescent quantification (e.g. EdU, cleaved-PARP, CHOP) were set based on the results from negative controls (secondary antibody-only). For cell cycle analysis, PI gating was set to only include cells that contain 2N and 4N, excluding cells with polyploidy.

For CHOP immunostaining in MEFs, cells were fixed in 4% formaldehyde for 10 min at RT and permeabilized with 0.1% Triton X-100 (in 1x PBS) for 20 min at RT. After permeabilization, cells were washed twice with 1x PBS, resuspended in 245 μL incubation buffer (10% BSA in 1x PBS), and incubated at RT for 30 min. After blocking, CHOP antibody (Santa Cruz, sc-7351) was added at a 1:50 dilution, and cells were then incubated at RT for 1 h for primary incubation. After primary incubation, cells were washed twice with incubation buffer and twice with 1x PBS and then resuspended with Alexa-488-conjugated anti-mouse antibody (Invitrogen, A21202) diluted 1/500 in incubation buffer at RT in the dark for 1 h. Cells were then washed three times with 1x PBS and analyzed by flow cytometry. Detection of CHOP in fixed liver tissue was as described (DeZwaan-McCabe et al, 2013), following IP injection of 1 mg/kg TM or vehicle diluted in PBS.

Isolation of paired MEFs lines

At passage 2, Chop^FLuL/FLuL or Chop^fl/fl MEF (at least two lines of each genotype) were infected with either Ad-GFP or Ad-GFP-FLPo (for Chop^FLuL/FLuL MEFs) or Ad-GFP-CRE (for Chop^fl/fl MEFs) at an M.O.I. of 250. Adenoviruses were prepared by the University of Iowa Viral Vector Core using the RAPAd system (Anderson et al, 2000). Seventy-two hours after infection, MEFs were trypsinized and resuspended in 1x PBS. Viable GFP-positive MEFs were sorted by FACS (Becton Dickinson Aria II) after Hoechst 334342 staining and FACS. 5 × 10⁵ cells of each pair were sorted out from each group, cultured on cell culture dishes and expanded for future use. All experiments were performed using cells below passage 10. Matched cell lines are available upon request.

RNA and protein analysis

Protein lysates were processed for immunoblots as described (Rutkowski et al, 2006). Primary antibodies include: CHOP (Santa Cruz sc-7351 RRID:AB_627411 for flow cytometry or Proteintech 15204-1-AP RRID:AB_2292610 for western blot), PARP (Cell Signaling Technology 9542 RRID:AB_2160739), ATF4 (Cell Signaling Technology 11815 RRID:AB_2616025), PeIF2α (Thermo 44-728G), total eIF2α (Cell Signaling Technology 9722 RRID:AB_2230924), and Calnexin (loading control; Enzo ADI-SPA-865 RRID:AB_10618434). Nanoluciferase activity was assessed using the Nano-Glo In-Gel Detection kit (Promega, USA) following the manufacturer’s protocol. qRT-PCR, including primer validation by standard curve and melt curve analysis, was also as described (Rutkowski et al, 2006). Briefly, RNA was isolated using Trizol (Thermofisher, USA) following the manufacturer’s protocol, and RNA concentrations were evaluated using the Qubit RNA Broad Range kit (Invitrogen, USA). 400 ng of RNA was used for reverse transcription using PrimeScript RT Master Mix (Takara, USA). PCR reactions were performed using TB Green Premix Ex Taq (Takara, USA). Gene expression was normalized against the average of two loading controls (Btf3 and Ppia).

qRT-PCR Primers

Btf3 forward: CCAGTTACAAGAAAGGCTGCT reverse: CTTCAACAGCTTGTCCGCT

Ppia forward: AGCACTGGAGAGAAAGGATT reverse: ATTATGGCGTGTAAAGTCACCA

FLuL/Chop exon 1-2 forward: TTGAAGATGAGCGGGTGGCA reverse: CTTTCAGGTGTGGTGGTGTA

FLuL exon 2-nLuc forward: GTGTTCCAGAAGGAAGTGCA reverse: CTTTGGATCGGAGTTACGGA

Chop exon 3-4 forward: GGAAGCCTGGTATGAGGAT reverse: CCACTCTGTTTCCGTTTCCT

Gadd34 forward: GAGATTCCTCTAAAAGCTCGG reverse: CAGGGACCTCGACGGCAGC

³⁵S pulse labeling

For protein synthesis experiments, cells were labeled using an ³⁵S Met/Cys labeling mixture at 200 µCi/ml (EasyTag Express³⁵S; Perkin Elmer) for 30 min in media that contained drug treatments and that was 20 percent MEF culture medium as above and 80 percent DMEM lacking Met/Cys and with dialyzed FBS. After labeling, cells were lysed in lysate buffer (1% SDS, 100 mM Tris (pH 8.9)) followed by vigorous boiling. Samples were separated by electrophoresis, the gels were dried using a vacuum dryer, and then exposed to film at −80 °C (BioMax MS with a Transcreen LE intensifying screen).

Media switch assay

Wild-type and Chop^FLuL/FLuL primary MEFs (two lines of each genotype) were treated with vehicle or 50 nM TG for 20 h (original cells). Media switch was performed by collecting media from each group of these original cells and applying the collected media to a separate set of non-treated w.t. primary MEFs (media-receiving cells). The original cells were washed three times with 1X PBS and maintained in fresh, stressor-free media. After media switch, EdU was added directly to both original cells and media-receiving cells, and cells were harvested 4 h after the addition of EdU. The data shown is aggregated from 2 lines (duplicates in each line).

Growth competition assay

Genomic DNA was isolated using the Qiagen Puregene Cell Kit (158043). An extra RNase treatment step was added to avoid the potential contamination of RNA. DNA concentrations from different groups were then quantified using Qubit dsDNA Broad Range kit (Invitrogen, USA) and diluted to the same concentration (10 ng/µL) for PCR. Extension time in the PCR program was 15 s to preclude amplification of long products. The sequences of different primers are listed below. Allele enrichment was normalized to the Hprt locus.

Hprt forward: CTGCCTCTGCCTCCTAAATG reverse: TGTCGTCTCCCAGAGGATTC

FLuL forward: CAACCTGGACCAAGTCCTT reverse: ATTTGGTCGCCGCTCAGAC

Flox forward: CATGTGATCATCTGGACAAC reverse: GATGGTGTGTCGGCCACAC

KO forward: AGAGTTGGATCTGGCAGGGT reverse: GAGAGACAGACAGGAGGTGA

Single-cell RNA sequencing

FACS sorted Ad-GFP- and Ad-FLPo-treated Chop^FLuL/FLuL MEFs were allowed to grow for at least two passages and the Ad-FLPo-treated cells verified for restoration of CHOP expression and loss of nLuc activity before use. After quality control, GFP and FLPo cells were trypsinized and the cell concentration was determined using a Countess Automated Cell Counter (Thermofisher, USA). The final concentration was the average of three reads from 3 different aliquots of cell suspension. Cells of different genotypes were then mixed at a 1:1 ratio (total number) and plated on a 10 cm plate and allowed to rest overnight before treatment with either 10 nM TG or vehicle. Upon harvest, cells were trypsinized and centrifuged for 3 min at RT. The pellet was then resuspended in HBSS using large orifice tips. The cell suspension was filtered using a 70 µM cell strainer to eliminate aggregates. Cell viability was determined by Countess after Trypan Blue staining. Only samples with over 90% viability proceeded to the next step. Single-cell RNA-sequencing was performed by the Iowa Institute of Human Genetics using the 10x Genomics Chromium Single-Cell System and a Chromium Next GEM Single Cell 3’ GEM Library&Gel Bead Kit (10xGenomics, USA), and a total of 5000 cells were sequenced.

For data analysis, nanoluciferase sequence was added to the Mus musculus 10 (mm10) to build a new reference genome. scRNA-seq data were then aligned to reference genome by Cellranger (v. 4.0.0) count. The filtered expression matrix with cell barcodes and gene names was further loaded with the ‘Read10X’ function of the Seurat (v.4.0.0) R package. For quality control, cells with the number of detected genes (nFeature_RNA) <= 250 and mitochondrial content >5% were removed. Next, the Seurat pipeline was used for data normalization, selection of highly variable feature genes, and clustering with default parameters in most cases. Specially, the UMAP algorithm was conducted for dimensionality reduction with top 30 principal components and a resolution of 0.8 was used for ‘FindClusters’ function. We also used the python Scanpy package to perform clustering analysis and obtained similar results with Seurat. To identify differentially expressed genes for each cluster, ‘FindMarkers’ function (only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25) was used. We further used the ‘enrichGO’ function in the ClusterProfiler package to conduct GO analysis. Bonferroni correction was used for the adjustment of p value for enrichment analysis. In addition, the significance level for gene expression between different groups was determined by Wilcoxon test. The correlation coefficient of gene expression was calculated by the Spearman Correlation method.

Statistical analysis

Continuous variables were reported as the mean ± standard deviation. For comparisons of multiple groups, ANOVA was used with correction (Tukey or Šídák) for multiple hypothesis testing, using GraphPad Prism. For Figs. 1B, 4G, a Mann–Whitney test or a Welch’s ANOVA, respectively, were used because of non-normality of residuals. A post-correction alpha of 0.05 was used to determine statistical significance. Data points represent biological replicates, usually from one experiment (independently treated and analyzed wells) where the experiment was also repeated independently at least once, and sometimes, where noted, from multiple experiments conducted similarly. For most experiments, 3–4 samples were used. Blinding was not performed. No data points/samples were excluded.