CHOP promotes the transition to chronic integrated stress response signaling with suppression of hepatocyte identity

Reviewer #1 (Public review):

Summary:

The predominant view on CHOP's functions during ER stress is that it promotes cell death. This is in contrast to a handful of reports in the literature that claim that CHOP is a positive regulator of protein synthesis during chronic ER stress, and therefore is part of the adaptation program to ER stress. These previous studies were performed in tissue culture cells. Velarde and co-authors have used a mouse model of induction of mild ER stress to study the function of CHOP in hepatocytes.

Major strengths and weaknesses of the methods and results:

The authors use state-of-the-art mice to manipulate (i) CHOP and (ii) ATF6, a protective factor of ER proteostasis, and address the hepatocyte responses to mild ER stress in vivo and in cultures. Validated gene expression programs are well correlated to liver pathology in the mouse models. This is a very well-done study.

The authors clearly show that CHOP transitions hepatocytes under mild ER stress to a chronic ISR state, which is phenocopied by ATF6-depleted hepatocytes. So the conclusion that CHOP exacerbates ER stress in hepatocytes during mild ER stress is correct. It is also clear that CHOP targets negatively the transcription of hepatocyte identity genes, which opens a new direction of studies on the function of CHOP in secretory cells in general.

Conclusion:

This is a significant study that will benefit different research fields, and specifically studies on proteostasis, as was recently highlighted in Nat. Str. Mol. Biol. by experts in the field.

To this reviewer, the importance of the study is that it links the function of a transcription factor (CHOP) to stress intensity (mild versus severe) in a physiological experimental model (hepatocyte function and pathology).

Reviewer #2 (Public review):

The Unfolded protein response (UPR) and related integrated stress response (ISR) are critical signaling systems for cell survival in response to acute stresses. While the UPR directs critical adaptive gene expression, certain chronic stresses switch this pathway towards cell death and disease. An important question concerns the mechanisms by which the UPR switches from being adaptive to maladaptive. Prevailing models focus on the transcription factor CHOP (DDIT3 or GADD153), whose levels are enhanced via the UPR, and extended/amplified amounts of CHOP are suggested to boost death-related gene expression. However, the literature and this manuscript point out a number of observations that do not neatly fit with this model, suggesting that there are still unresolved processes by which CHOP adjusts cell outcomes via the UPR.

This manuscript features a nice hepatocyte-targeted knockout of CHOP to discern the contribution of CHOP in the transition between adaptive and maladaptive outcomes. The key ideas presented in this study are that CHOP-directed gene expression is focused on protein synthesis, metabolism, and hepatocyte identity. In the progression of the UPR, CHOP expression can lead to resumption of protein synthesis, which can assist in the translation of the UPR-directed transcriptome, which includes ATF6/XBP1-directed genes that aid the processing capacity of the endoplasmic reticulum (ER). However, enhanced nascent protein can further stress the ER. CHOP directs gene expression in both the first phase- acute and second phase-chronic in the UPR, and the pivotal decision lies in the transition between the phases.

Overall, the manuscript includes some new ideas as well as refinements of earlier ones for CHOP-determination of UPR-directed cell fate. The CHOP-hepatocyte knockout mouse model helps to delineate the different tissue functions of CHOP, which has been a problem for some earlier studies. The manuscript progression of experiments is solid, and experimental design and documentation are rigorous. The manuscript text is largely clear, but there are portions that would benefit from fuller explanations of ideas.

There are three points of concern. First, the manuscript model (Figure 7) lays out a timeline for the progression of the UPR between two phases. The study is not always clear about the times assayed, and there appears to be a single time point for measurements. Second, there is emphasis on protein synthesis changes in the model. It is true that the literature argues that resumption of protein synthesis concurrent with stress damage (i.e., GADD34-directed gene expression) is a key reason for the potentially debilitating effects of CHOP (e.g., Marciniak et al 2004, Han et al 2013). However, the manuscript does not feature protein synthesis measurements. Inclusion of bulk protein synthesis measurements in the context of this model system would strengthen the study and support for the model. Finally, for this reviewer, some of the most interesting ideas center on CHOP-directed transcription of genes that regulate hepatocyte identity. There is solid evidence for direct CHOP regulation of these genes, but the manuscript does not really develop and test the ramifications of these networks on cell fate during ER stress.

Reviewer Concerns:

(1) The abstract packs in a lot of information. The ideas would not be clear to a general reader. Furthermore, the UPR and ISR are referred to in the second-to-last sentence, but not defined earlier in the abstract.

(2) There are some typos/grammar concerns.

(3) ATF4 diminished with CHOP-depletion (Figure S2A). What is the mechanism here? Does this complicate the analysis of CHOP-directed gene expression? How does this fit with Figure 6J? The timelines for TM treatment are critical. The authors should more fully explain the time courses in the experiments.

(4) Figures 2 and 3: There is a discussion on enhanced protein synthesis with loss of CHOP (reduced GADD34 expression). What is the time point - 8 hours TM? Emphasize, explain, and justify time points of experiments here and in later panels. It would strengthen the model with direct measurements of protein synthesis. The authors could include GADD34 protein measurements in these panels. Figure 3 - panel D - some abbreviations are not standard.

(5) Figure 4: One of the most interesting in the manuscript is the transcription factors downstream of CHOP that are linked with hepatocyte differentiation and metabolism. The manuscript would be bolstered by developing some of these target genes into the Figure 7 transition model.

(6) Figure 6: The comparison of CHOP and ATF6 target genes is a highlight of the manuscript. The literature on this topic is complex, and there are some suggestions that CHOP can be downstream of ATF6. Furthermore, there were some earlier models by Walter and others about extended induction of Perk (death) vs induction of other UPR sensors (survival) (e.g. PMID: 17991856). It would be helpful in the Discussion to delineate between these models and their critical differences.

Reviewer #3 (Public review):

In this manuscript, the authors aim to understand the function of the transcription factor CHOP, which is known to promote cell death during severe stress in the ER. The authors note that CHOP is induced during less severe stress, but its functional output is not well understood in these cases. Here, they study the effects of conditional knockouts of CHOP in hepatocytes of mice challenged with chemical inducers of ER stress.

Tunicamycin (an ER stress inducer) injection leads to the upregulation of CHOP and lipid accumulation in the liver, but no significant cell death in the experiments outlined here. Conditional knockout of CHOP results in a number of differences in the way hepatocytes respond to stress, notably resulting in lower steatosis.

There are two main findings supported by the data presented here. First, the authors show that CHOP suppresses the expression of ONECUT, a master regulator of hepatocyte differentiation and metabolism, during ER stress. They show by ChIP-seq that CHOP binds to the promoter region of this gene, and by RNA-seq that ONECUT expression is suppressed by ER stress in a CHOP-dependent manner. Many predicted targets of ONECUT1 were also suppressed by ER stress in a CHOP-dependent manner, though they were not bound directly by CHOP. The data support a model where CHOP down-regulates hepatocyte metabolism and identity via regulation of ONECUT1. This is a new and interesting finding, perhaps explaining the steatosis phenotype of livers that accompanies ER stress, although this was not tested directly.

The second main finding of this paper is that CHOP deletion leads to an interesting assortment of effects on genes related to the ER stress response and integrated stress response (ISR). As expected, based on prior work, CHOP deletion led to more phosphorylation of eIF2alpha (CHOP is known to upregulate the phosphatase for this translation factor). However, unexpectedly, this did not cause increased expression of ATF4 (a transcription factor whose upregulation during stress is dependent on eIF2alpha phosphorylation) and its downstream targets; in fact, CHOP deletion had the opposite effect on these. In other words, CHOP seems to both turn off the initiating signal for the ISR (namely, eIF2alpha phosphorylation) and also promote the downstream signaling events that rely on this initiating signal. It makes sense that cells would do this, as restoring translation would be important for realizing the effects of the massive changes in gene expression initiated by ER stress, and yet this would exacerbate stress in the short term, so it would be counterproductive to also turn off the entire stress-regulated program. Having a factor (perhaps CHOP) that coordinates these two events makes sense. It will be interesting in future work to understand the mechanisms behind this regulation.

Finally, CHOP deletion led to less activity of other aspects of the ER stress response, notably IRE1 (determined through measurement of XBP1 splicing and RIDD of Bloc1s1). This is explained by the continued phosphorylation of eIF2alpha in these knockouts, as the continued attenuation of translation would lessen the burden of misfolded proteins in the ER. Somewhat confusingly, the same pattern is not seen in downstream targets of XBP1. Less splicing, coupled with perhaps less translation of the spliced mRNA, should result in less active transcription factor and lower expression of its target genes in the CHOP KO. This is not observed in Figure 2, although the more global gene expression analysis suggests that all stress-dependent gene expression changes were weaker in the CHOP KO livers.

The authors characterize the effects of CHOP, promoting restoration of protein synthesis and the accompanying exacerbation of stress while preserving the signaling that should relieve ER stress, as a switch from an acute to chronic phase of ER stress. This is mirrored in their analysis of ATF6 in a similar series of experiments. Although this is an interesting framework for thinking about the stress response, whether CHOP is the key factor or a supporting actor in regulating this transition will require a better understanding of the mechanisms involved.

Author response:

We thank the reviewers for their assessment of our work and their comments. We are grateful for their evaluation of our findings as fundamental and convincingly supported, and for their appreciation of the relative scope of this manuscript and of future work. The most direct requests for new experimental data are from reviewer #2, who asks for direct assessment of the effects of CHOP deletion on expression of GADD34 and on protein synthesis. We agree that these are important experiments to conduct for the revision.

The reviewers requested more clarity on the experimental logic of the paper and on the place of our findings in the broader context of ER stress signaling, which we will be happy to provide in a revised manuscript. These revisions will include a more explicit consideration of how the regulation of metabolic genes by CHOP contributes to its effects in the liver independently of its role in regulating eIF2a dephosphorylation.

In particular, there were concerns about the logic of the time points chosen that we feel are important to also address here. For analysis of ChopHKO animals, all experiments were carried out 8 hours after ER stress challenge. This is because, as we show in Fig. 1B and also in our previous paper on CHOP (1), this is the time point at which CHOP expression is at its maximum. Thereafter, hepatocytes become heterogeneous with respect to whether they do or do not express CHOP. This is an interesting finding because it suggests that CHOP is part of a cellular switch, and potentially even an effector of that switch—a point currently raised in the Discussion but worth further highlighting in a revision. At the practical level, it means that discerning the contribution of CHOP to ER stress signaling and adaptation at subsequent time points will require sophisticated single cell analyses that can discriminate cells that express CHOP from cells that do not, which are an important future direction.

In contrast, for Atf6aHKO animals, all experiments were carried out 48 hours after ER stress challenge. As we have previously shown (2), at short time points after a stress challenge, such as 8 hours, there is very little difference in ER stress signaling between wild-type animals and those lacking ATF6a. The reason for this lack of distinction is that the major targets of ATF6a are ER chaperones and the like. Because adaptation to ER stress in the early phases of the response depends more on non-transcriptional mechanisms such as inhibition of protein synthesis and IRE1-dependent mRNA decay (RIDD), the failure to fully upregulate ATF6a targets is initially of little consequence. It is only at later time points when wild-type animals restore ER homeostasis and largely silence ER stress signaling. In contrast, at these same later points, animals lacking ATF6a show evidence of persistent ER stress, most notably in the form of persistent Xbp1 mRNA splicing and profound suppression of metabolic genes. Although the 8 hour time point for experiments in ChopHKO animals differs from the 48 hour time point for Atf6aHKO animals, the two lines of experimentation are united by the persistence of ongoing ER stress and of ISR signaling despite diminished eIF2a phosphorylation at the points when the presence of CHOP or the absence of ATF6a are of the greatest impact. A revised manuscript will present this logic more clearly.

References

(1)  Liu K, et al., EMBO Reports 25, 228 (2024)

(2)  Rutkowski DT, et al., Dev. Cell 15, 829 (2008)