The protease DDI2 regulates NRF1 activation in response to cadmium toxicity

Summary DNA-damage inducible 1 homolog 2 (DDI2) is a protease that activates the transcription factor NRF1. Cellular models have shown that this pathway contributes to cell-stress adaptation, for example, on proteasome inhibition. However, DDI2 physiological function is unknown. Ddi2 Knock-out (KO) mice were embryonic lethal. Therefore, we generated liver-specific Ddi2-KO animals and used comprehensive genetic analysis to identify the molecular pathways regulated by DDI2. Here, we demonstrate that DDI2 contributes to metallothionein (MT) expression in mouse and human hepatocytes at basal and upon cadmium (Cd) exposure. This transcriptional program is dependent on DDI2-mediated NRF1 proteolytic maturation. In contrast, NRF1 homolog NRF2 does not contribute to MT production. Mechanistically, we observed that Cd exposure inhibits proteasome activity, resulting in DDI2-mediated NRF1 proteolytic maturation. In line with these findings, DDI2 deficiency sensitizes cells to Cd toxicity. This study identifies a function for DDI2 that links proteasome homeostasis to heavy metal mediated toxicity.


INTRODUCTION
The molecular mechanisms of cell stress response play a central role in cell homeostasis as well as in pathogenesis in different diseases such as cancer and neurodegeneration (Cubillos-Ruiz et al., 2017). The DNA Damage Inducible 1 Homolog 2 (DDI2) is an aspartic protease containing a retroviral protease-like (RVP) domain, highly conserved in eukaryotes (Perteguer et al., 2013;Siva et al., 2016). DDI2 is ubiquitously expressed and implicated in cell stress response. However, DDI2 function, regulation, and downstream molecular mechanisms are largely unknown. Recent data from Caenorhabditis elegans, yeast, and human cells demonstrate the involvement of DDI2 in protein homeostasis by regulating proteasome activity (Koizumi et al., 2016;Lehrbach and Ruvkun, 2016), as well as in cell stress response during DNA replication (Kottemann et al., 2018;Serbyn et al., 2020). Koizumi and collaborators reported that DDI2 modulates proteasome subunits through activation of the transcription factor Nuclear factor erythroid-2-related factor 1 (NRF1, encoded by the NFE2L1 gene) (Koizumi et al., 2016;Rousseau and Bertolotti, 2018). At basal, NRF1 is glycosylated in the ER and translocated to the cytoplasm, where the proteasome continuously degrades it. However, NRF1 accumulates in the cytoplasm on proteasome inhibition and is proteolytically cleaved by DDI2. This process releases an active form of NRF1 that translocates to the nucleus and induces a transcriptional program including the expression of proteasomal (PSM) genes. This adaptation mechanism is known as the proteasome bounce-back response (Koizumi et al., 2016;Radhakrishnan et al., 2010Radhakrishnan et al., , 2014Zhang et al., 2007). Consistent with NRF1 contribution to proteasome homeostasis, homozygous deletion of Nfe2l1 is lethal in mice (Chan et al., 1998) and late-stage deletion of Nfe2l1 in neuronal cells leads to neurodegeneration (Lee et al., 2011). Although NRF1 target genes are poorly understood, it is described that NRF1 binds the antioxidant response elements (ARE), found in the promoter region of PSM and metallothionein (Mt) genes. In addition, the depletion of Nfe2l1 in the mouse liver impairs the expression of Mt1 and Mt2 genes (Ohtsuji et al., 2008). MT1/2 are small cysteine-rich and metal-binding proteins, involved in metal-ions chelation and detoxification of Cadmium and Arsenic. (Habeebu et al., 2000;Klaassen et al., 2009;Ohtsuji et al., 2008;Wang et al., 2020;Zhao et al., 2011). Cd is a toxic heavy metal widely present in the environment and exhibits cytotoxic and carcinogenic effects in both mice and humans. It is known that Cd induces the accumulation of ubiquitinated proteins, affects the Ubiquitin Proteasome System (UPS), and causes cytotoxicity and apoptosis in both mouse and human cells (Figueiredo-Pereira et al., 1997;Yu et al., 2011). Moreover, the UPS has been shown to mediate the degradation of abnormal proteins and cell resistance to Cd exposure (Jungmann et al., 1993). Previous studies suggested

DDI2 modulates steady-state metallothionein expression in vivo
The aspartic protease DDI2 is highly conserved between species and ubiquitously expressed; however, its function is poorly understood. DDI2 was recently described to mediate the activation of NRF1 during proteasome inhibition (Koizumi et al., 2016). To further elucidate DDI2 function, we generated DDI2 deficient mice. Full body depletion of DDI2, using gene-trap technology, was embryonic lethal at middle gestational stage with decreased viability observed at E9.5 ( Figures 1A and S1), confirming previous observations (Siva et al., 2020). However, a DDI2 liver-specific depletion, using conditional deletion of DDI2-floxed under albumin control was viable and the toxicology results were similar to WT animals at steady state, except for catalase activity, that was reduced ( Figure S2A). At non-stressed conditions, DDI2-KO mice presented similar liver tissue structure and morphology compared to the WT animals. To identify DDI2 downstream target genes, we performed mRNA sequencing using liver samples from DDI2-KO mice and littermate controls ( Figure 1B). Gene ontology (GO) analysis of molecular function showed that many genes affected by DDI2 deficiency belong to the category of DNA binding and transcriptional regulators ( Figure S2B). Several genes involved in the metabolic process (Atp4a, Slc25), cellular response to ions and transcription regulation (Mt1, Mt2) are disturbed in DDI2-deficient livers ( Figure 1B and S2B). The top downregulated hits identified in the genetic screening were the metallothionein genes, Mt1 and Mt2 ( Figure 1B). MT genes were previously described to be under the control of the transcription factor NRF1 (Ohtsuji et al., 2008), iScience Article suggesting that DDI2 may contribute to NRF1 activation at basal in the mouse liver. These findings were confirmed by real-time PCR mRNA expression analysis in the liver of deficient animals ( Figure 2A). To analyze stress-induced Mt genes expression, we challenged mice with heavy metals. Intraperitoneal injection of acute doses of cadmium (Cd) induced a robust Mt gene expression in the liver of WT animals. In contrast, DDI2-KO mice presented a reduced Mt1 and Mt2 genes induction after 16 h of Cd treatment (Figure 2A). Importantly, we found that Cd exposure triggered proteolytic activation of NRF1 in the mouse liver iScience Article ( Figure 2B) as well as in human HepG2 cell line ( Figure S3A). We observed that NRF1 protein levels increase following Cd treatment because of an impairment in NRF1 protein degradation whereas NRF1 mRNA levels remained unchanged ( Figure S3B). Moreover, similar to previous studies in human cell lines treated with proteasome inhibitors (Koizumi et al., 2016), we found that DDI2 depletion in mouse hepatocytes prevented Cd-mediated NRF1 cleavage and activation in the liver ( Figure 2B). In addition, we confirmed that, similarly to proteasome inhibitors, treatment with Cd can promote NRF1 cleavage in human HepG2 cells ( Figure S3C).
To further analyze the contribution of DDI2 to Cd exposure, we immunohistochemically stained MT proteins in liver tissue sections. We confirmed that MT is increased following 16 h of Cd injection ( Figures 2B and2C). The WT mouse showed a high MT protein expression with a heterogeneous pattern throughout the liver ( Figure 2C, mid-panel). When we stained a transversal cut of the blood vessel, it was possible to identify the high levels of secreted MT protein ( Figure 2C, top right panel). These results confirmed the previous reports suggesting that MT proteins chelate Cd molecules in the liver, the MT-Cd complexes are transported to the bloodstream to be later filtered in the kidneys and eliminated from the body through urine (Sabolic et al., 2010). In contrast, after Cd exposure, DDI2-KO animals expressed a diffused pattern of MT in the liver, and showed less secreted MT into the blood vessels than WT animals ( Figure 2C, bottom panel).
In addition to NRF1 activation, we also observed an accumulation of NRF2 protein following Cd treatment, independent of DDI2 ( Figure 2B). Our results align with previous reports where NRF2 was described to mediate an adaptive cellular response to Cd-induced oxidative stress (Chen et al., 2021;Chen and Shaikh, 2009). To determine if NRF2 could be involved in MT induction in the absence of DDI2, we generated a double KO mouse (d-KO) with complete deficiency of functional NRF2 and liver-specific DDI2 deletion.
Our results indicated that NRF2 is not involved in MT induction following Cd treatment. In the NRF2-KO mouse, we observed MT genes induction following Cd treatment, similar to the WT mouse. In contrast, the double DDI2 and NRF2-KO mice presented a reduced MT response following Cd treatment, identical to that observed in the DDI2-KO mice ( Figure 2D). These findings demonstrate the partial but specific contribution of the DDI2-NRF1 pathway to the Cd-mediated MT-gene expression.
Cadmium induces DDI2-mediated cleavage and nuclear translocation of NRF1 in human HepG2 cells To further confirm the MT gene induction following Cd treatment, we used an NRF1-responsive luciferase reporter assay (8xARE-Luc construct characterized in (Ohtsuji et al., 2008)). We transfected the reporter in the human hepatocyte cell line (HepG2). Consistently with MT expression observed in the mouse model, the MT-promoter was activated following Cd treatment ( Figure 3A). In addition, in line with previous reports (Chen et al., 2014), Arsenic or Zinc treatments could also increase the MT-promoter activity. In contrast, Bortezomib (BTZ) or Carfilzomib (CFZ), two proteasome inhibitors that promote DDI2-mediated NRF1 activation, could not induce MT-promoter activation per se, indicating that Cd, in addition to NRF1, relies on additional factors to trigger MT gene expression. In agreement with our in-vivo data where we demonstrate that MT activation is independent of NRF2 activity, the chemical activation of NRF2 with tert-butylhydroquinone (tBHQ) treatment did not induce the MT-promoter activation in HepG2 cells ( Figure 3A).
Cd-induced NRF1 accumulation and cleavage is rapid and occurs within hours in a DDI2-dependent manner ( Figure 3B). In addition, cell fractionation studies confirmed that proteolytically cleaved NRF1 localized in the nucleus upon Cd treatment whereas uncleaved NRF1 accumulated in the cytoplasm ( Figure 3C).
To further confirm the contribution of DDI2 and NRF1 to Cd-mediated MT gene expression, we engineered HepG2 with DDI2 and NRF1 deficiency. In agreement with the observations made in DDI2-KO mice, we found that DDI2 or NRF1 deficiency comparably affected Cd-induced MT-promoter activity ( Figure 3D). Moreover, NRF1-deficient cells presented undetected levels of MT protein at both basal and Cd-treated conditions ( Figure S3A).

Cadmium inhibits proteasome activity
Although we analyzed NRF1 activation, we observed increased detection of ubiquitinated proteins following Cd treatment ( Figure 3C), suggesting that Cd may affect proteasome activity. Because ll OPEN ACCESS iScience 25, 105227, October 21, 2022 iScience Article proteasome inhibition can trigger the proteolytic activation of NRF1 (Koizumi et al., 2016;Radhakrishnan et al., 2014), we hypothesized that proteasome impairment could be the upstream mechanism of NRF1 activation following Cd exposure. Surprisingly, we found that Cd treatment in HepG2 cells impaired the proteasome activity, similarly to the proteasome inhibitor, CFZ. The treatment with Cd decreased most proteolytic activities of the proteasome including, chymotrypsin, trypsin, and caspase-like activities (Figure 4A). To determine if Cd directly binds and impairs the proteasome function, we evaluated the proteasome activity using cell-free purified proteasome complexes in the presence of Cd molecules. As expected, CFZ directly interacted with the purified proteasome and blocked the substrate degradation.
In contrast, Cd did not inhibit the proteasome in the cell-free assay ( Figure 4B), indicating that its mechanism of action is indirect and may affect upstream events required for proteasome function.
Consistent with proteasome inhibition, we observed in mice, that DDI2 deficiency in livers increased K48linked ubiquitinated proteins ( Figure S3D). However, in these conditions we could not observe a robust induction of proteasome genes ( Figure S4A). In contrast, treatment with cadmium decreased PSMA4 mRNA levels in wildtype and NRF2 deficient animals. Because DDI2 deficiency impaired expression of this gene at iScience Article basal no additional defect was observed on treatment with Cd ( Figure S4A). In contrast, in HepG2 cells, cadmium increased PSMA4 gene expression following Cd treatment. As expected, both DDI2-or NRF1-KO cells failed to modulate PSMA4 induction ( Figure S4B). These observations indicate that the proteasome bounce back response can be activated but is not always engaged as observed in mice.

DDI2 contributes to increased cell survival on cadmium-induced toxicity
Metallothioneins contribute to the detoxification of heavy metals. We, therefore, tested whether DDI2mediated MT expression could decrease Cd-mediated toxicity. We analyzed MT genes expressions in HepG2 cells harboring DDI2 deficiency. In line with the promoter-reporter assays, we observed decreased induction of the human homologs of MT genes on treatment with Cd compared with the control population ( Figure 5A). We also found that the expression of previously described NRF1 target genes, like NQO1, were slightly reduced in Cd-treated HepG2 cells deficient for DDI2, compared to the WT cells ( Figure 5A). Moreover, we showed that cellular survival is decreased in DDI2-or NRF1 deficient cells exposed to Cd (Figure 5B), further suggesting that depleting DDI2, consequently preventing NRF1 activation, sensitizes cells to Cd toxicity. Then, we tested if mouse lacking DDI2 in the liver are more sensitive to Cd toxicity. Surprisingly, we did not verify differences in mouse survival, when comparing WT and DDI2-KO, following Cd exposure ( Figure S5). However, this observation may be due to the fact that other organs, compensate for MT proteins expression as reported previously (Sabolic et al., 2010), accounting for Cd detoxification and mouse survival. Recently, it has been suggested that MT promotes chemoresistance during metal-based cancer chemotherapies (Borchert et al., 2020;Rodrigo et al., 2021). Cisplatin is a platinum-based compound, commonly used in chemotherapy treatments alone or in combination with other drugs. To test if the DDI2-NRF1-MT axis promotes cell survival and resistance to cancer chemotherapies, we treated HepG2 cell with cisplatin and evaluated the cell survival. Similar to what we observed in cells treated with Cd ( Figure 5B), HepG2 cells lacking DDI2 or NRF1 are more sensitive to cell death following cisplatin treatments compared to WT cells ( Figure 5C), further indicating that DDI2 and NRF1 can modulate responses to heavy metal exposure.

DISCUSSION
-This study reports a mouse model to investigate DDI2-associated molecular mechanisms and physiological functions. We used this model to describe the involvement of DDI2 during cell stress response against Cd toxicity. We demonstrated that on Cd exposure in mice, DDI2mediated cleavage and activation of NRF1 and regulated downstream metallothionein (MT) expression in both human and mouse models. Using liver-specific DDI2-KO animals, we found that very few genes were significantly affected by DDI2 deficiency in untreated mice. Unexpectedly, no defects in well-known NRF1 targets such as the proteasome subunits (Koizumi et al., 2016;Lehrbach and Ruvkun, 2016) were identified, except for Mt1 and Mt2 expression. These results are consistent with previous reports showing that liver-specific NRF1-KO mice are viable and display reduced expression levels of Mt1 and Mt2 (Ohtsuji et al., 2008). These data strongly suggest that DDI2 is upstream of NRF1 signaling pathway in the liver and contributes to Mt expression. Although it has been proposed that both NRF1 and NRF2 share a common target DNA motif, here, we demonstrate that NRF1 controls Mt gene expression independently of NRF2. Our observations agree with previous reports showing that both WT and NRF2-KO mice present similar Mt expression levels in both Cd-treated and untreated conditions (Chen et al., 2021). However, the activation of MT is not exclusively controlled by the DDI2-NRF1 pathway. Mt genes are activated by different response elements, present in the gene promoter, in response to stress stimuli. Previous reports suggest that Cd can displace zinc (Zn) and copper (Cu) from MT conjugates. Consequently, free Zn ions are able to bind and induce the activation of MTF-1 transcriptional factor. Active MTF-1 binds to MRE in the DNA and eventually stimulates the synthesis of Mt mRNA (Sabolic et al., 2010). Interestingly, Mt null mice are highly sensitive to Cd toxicity (Masters et al., 1994). In comparison, no lethal toxicity was reported in MTF-1-KO mice (Wimmer et al., 2005), and in this study, we did not observe increased lethality in DDI2-KO when treated with 8 mg/kg of Cd. Altogether, Mt genes are regulated by distinct mechanisms, including DDI2-NRF1 axis, and are essential to cope with metal toxicity.
The proteasome is a key machinery in eukaryotic cells responsible to cope with cellular toxicity and maintain homeostasis. Importantly, cancer treatments, including proteasome inhibitors, are based on the activation of cell stress to induce apoptosis of targeted tumor cells (Manasanch and Orlowski, 2017). Understanding the molecular mechanisms of cell stress response is crucial to design new therapeutic approaches and to bypass the cancer resistance to treatments. Here, we reported that cellular exposure to metals, as Cd, blocks the proteasome function, increases ubiquitinated proteins, and induces the DDI2-NRF1 stress response pathway to promote the cell survival in hepatocytes. Several studies suggested that metal complexes may regulate proteasome activity (Verani, 2012). However, the molecular mechanisms are largely unknown, these may include the blockade of the UPS by inhibiting de-ubiquitinating enzymes (DUBs) or proteasome assembly chaperones (Chen et al., 2018;Padmanabhan et al., 2016;Xu et al., 2021).
In addition to Cd detoxification, MTs can bind to metal-based drugs, like cisplatin, promoting chemo-resistance in cancer treatments (Borchert et al., 2020;Rodrigo et al., 2021;Si and Lang, 2018). MT expression in tumors is associated with cancer treatment resistance (Si and Lang, 2018) and targeting MT proteins has been suggested to be a powerful tool to boost response rates to metal-based therapies (Borchert et al., 2020). Our data suggest that inhibiting the protease activity of DDI2 could be a suitable therapeutic target to dampen the resistance to metal-based treatments. Nelfinavir is a clinically approved HIV-inhibitor, also described to inhibit DDI2 (Gu et al., 2020). Previous reports demonstrated a beneficial effect of targeting DDI2 using nelfinavir in cancer treatments (Chow et al., 2009) including in Multiple Myeloma (Besse et al., 2018;Driessen et al., 2016;Fassmannova et al., 2020;Gu et al., 2020;Hitz et al., 2019). However, it remains to be evaluated if nelfinavir affects results from inhibition of DDI2-downstream activation of NRF1-MT cellular response.
Altogether, our findings identify heavy metals as activators of the DDI2-NRF1 pathway and downstream MT-chelating proteins. Our results contribute to the understanding of the metal-compounds secretion from the body, as well as the chemotherapy resistance to metal-based treatments.

LIMITATION OF THE STUDY
The contributions of DDI2 during development are still unknown. DDI2-deficient mice die at the mid-gestational stage. Additional studies are necessary to understand the cause of death. Comparing these defects with those observed in NRF1 deficient animals could hint at additional functions for DDI2, possibly independent of NRF1. Although we extrapolate that the phenotypes observed in DDI2 deficiency are mainly a consequence of lack of NRF1 activation, we cannot exclude the contribution of other DDI2 substrates yet to be characterized. For example, NRF3, a close homolog of NRF1, could contribute to DDI2 functions in tissues where it is expressed.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We thank to Masayuki Yamamoto, Laurie H. Glimcher, Nicolas Fasel, Pascal Schneider and Margot Thome for providing essential material. We thank to the Mouse Facility, Mouse Pathology Facility, Metabolomics Plat-form and Genomics facility of the University of Lausanne for technical assistance. We thank to the members of the Department of Immunobiology for the discussion and helpful input. This work was supported by a grant from the Swiss National Science Foundation (310030_173152) and Human Frontier Science Program CDA-00059.

DECLARATION OF INTERESTS
The authors declare no competing interests.

Luciferase assay
For each condition, one million cells were transfected with 0.9 mg MT-Firefly-luciferase reporter plasmid (Ohtsuji et al., 2008) provided by Masayuki Yamamoto (Tohoku University Graduate School of Medicine); 0.1 mg Renilla plasmid was used as the internal control. Dual-Luciferase Assay System (Promega, Madison, MI, USA) was performed 48 h after transfection, according to the manufacturer's instructions. Specific Firefly luciferase activity was calculated from light intensity measurements and normalized against Renilla luciferase activity as internal control.

QUANTIFICATION AND STATISTICAL ANALYSIS
Representative results from at least three independent experiments were shown. Statistical comparisons were made as described in the Figure legends with GraphPad Prism. The significance of the difference was set at p values <0.05. * denotes p<0.05; **p<0.01; ***p<0.001. n indicates the number of independent experiment, except in Figures 2A, 2D, and S5 where it indicates number of animals per group and Figure S1, where N indicates total number of animals.