20S immunoproteasomes remove formaldehyde-damaged cytoplasmic proteins suppressing caspase-independent cell death

Immunoproteasomes are known for their involvement in antigen presentation. However, their broad tissue presence and other evidence are indicative of nonimmune functions. We examined a role for immunoproteasomes in cellular responses to the endogenous and environmental carcinogen formaldehyde (FA) that binds to cytosolic and nuclear proteins producing proteotoxic stress and genotoxic DNA-histone crosslinks. We found that immunoproteasomes were important for suppression of a caspase-independent cell death and the long-term survival of FA-treated cells. All major genotoxic responses to FA, including replication inhibition and activation of the transcription factor p53 and the apical ATM and ATR kinases, were unaffected by immunoproteasome inactivity. Immunoproteasome inhibition enhanced activation of the cytosolic protein damage sensor HSF1, elevated levels of K48-polyubiquitinated cytoplasmic proteins and increased depletion of unconjugated ubiquitin. We further found that FA induced the disassembly of 26S immunoproteasomes, but not standard 26S proteasomes, releasing the 20S catalytic immunoproteasome. FA-treated cells also had higher amounts of small activators PA28αβ and PA28γ bound to 20S particles. Our findings highlight the significance of nonnuclear damage in FA injury and reveal a major role for immunoproteasomes in elimination of FA-damaged cytoplasmic proteins through ubiquitin-independent proteolysis.

toxicant with many sources of exposure, such as tobacco smoking 20 , off-gassing from consumer goods and emissions by combustion processes 15 . Inhalation FA exposures are linked with higher risks for respiratory 21 and other cancers 22,23 . FA carcinogenicity is commonly associated with the formation of genotoxic DNA-protein crosslinks (DPCs) involving histones 15,21 due to conjugation of FA with the abundant Lys ε-amino groups within these proteins. Protein damage by FA is quite extensive, as evidenced by a rapid heat shock response and extensive protein polyubiquination 24 . Vulnerability of the nervous system to toxic effects of FA 18,19,25 is consistent with its protein-damaging properties.
Here we examined whether i-proteasomes are involved in responses to FA damage, considering that they display higher activities toward basic proteins 14 . We found that FA triggered the disassembly of 26S i-proteasomes promoting ubiquitin-independent removal of damaged cytoplasmic proteins and suppressing long-term cytotoxic effects. Thus, one of the nonimmune functions of i-proteasomes is protection against proteotoxicity by ubiquitous FA and possibly by other aldehydes. Our findings are also important for the mechanistic understanding of FA toxicities, demonstrating that protein damage outside the nucleus contributes to the development of adverse effects.

Results
Gene expression of i-proteasome subunits. We selected human lung H460 and IMR90 cells as our biological models, which we have used in the past for the characterization of genotoxic signaling and proteotoxicity by FA 24,26,27 . Based on gene expression by qRT-PCR, i-proteasomal components constituted 28.7% of all catalytic subunits in H460 cells and 16% in IMR90 cells (Fig. 1A). In both cell lines, LMP7 accounted for approximately 15-20% of β5 subunits with chymotrypsin-like activity, which is the main proteolytic activity of proteasomes. Acute and chronic FA treatments of normal IMR90 cells produced no significant changes in gene expression of i-proteasomal components (Fig. 1B). These negative results were not caused by technical factors, as IMR90 cells showed a strong upregulation of all three i-proteasomal subunits by interferon-γ (4 ng/ml, 24 h: 3.1 ± 0.4, 49 ± 11.4 and 3.7 ± 0.3-fold for LMP7, LMP2 and LMP10 respectively, n = 3).

FA cytotoxicity.
Mutations in i-proteasome proteins result in specific human disorders, however, knockouts of individual i-proteasome genes in mice did not cause any overt pathophysiological changes 4,5 , suggesting that the absence of i-proteasome proteins leads to compensatory changes. In agreement with pathological phenotypes of mutated i-proteasomes in humans, pharmacological inhibition of i-proteasomes in mice produced clear physiological effects. A selective inhibitor of LMP7 activity, ONX-0914 28 , is a commonly used tool for interrogation of i-proteasome functions in vitro and in animal models of human diseases. Treatments of H460 lung cells with 0.3 or 0.6 μM ONX-0914 (LMP7-i) resulted in the appearance of a slower moving LMP7 band, which was equivalent to a gain of 0.5-1 kDa ( Fig. 2A). The magnitude of the observed shift corresponds to a covalent attachment of a single LMP7-i (581 Da molecular weight). LMP2 showed a larger shift (2.5-3 kDa), corresponding to its unprocessed form, which probably results from the mutually dependent incorporation and activation of i-proteasome components 29, 30 . In agreement with qRT-PCR data (Fig. 1B), FA did not alter protein levels of LMP2 or LMP7 ( Fig. 2A). To test LMP7-i effects on the constitutive proteasomes, we measured levels of unstable transcription factors HIF1α and p53, which both undergo proteasome-dependent degradation in unstressed cells. In contrast to the proteasome inhibitor MG132, LMP7-i had no effect on the stability of HIF1α and p53 (Fig. 2B). LMP7-i also had no effect on Ser326 phosphorylation of the proteotoxic stress-sensitive transcription factor HSF1, which showed a massive upregulation by MG132. Importantly, the selected dose of our positive control MG132 caused only a partial proteasome inhibition as evidenced by the lack of free ubiquitin depletion (Fig. 2B). The addition of LMP7-i for 6 h produced no changes in cell cycle and DNA replication (Fig. 2C). Longer 24 h incubations with ≥0.5 μM LMP7-i led to modest (15-20%) decreases in the colony formation (Fig. 2D), indicating that i-proteasomes play some role in the normal physiology of H460 cells. The impact of i-proteasomes on FA cytotoxicity was first examined by the clonogenic assay, which integrates all forms of cell death. The presence of LMP7-i during 3 h FA exposures and the subsequent 24 h recovery significantly diminished clonogenic viability of cells (Fig. 2E). Since FA is a potent replication stressor 26,27 , we also tested cytotoxicity of two other replication stressors hydroxyurea and camptothecin. Hydroxyurea causes stalling of replication forks by depleting cells of dNTPs whereas camptothecin produces DNA-topoisomerase I crosslinks. Unlike FA, cotreatments with LMP7-i and hydroxyurea or camptothecin did not elicit significant changes in clonogenic viability (Fig. 2F,G). Thus, i-proteasomes do not have a general role in cell recovery from replication stress.
Genotoxic signaling by FA. FA conjugates to lysine-rich basic proteins such as histones and i-proteasomes have higher activity on these types of proteins 14 . Thus, i-proteasomes could be potentially involved in recovery from FA-histone damage that includes DNA-histone crosslinks and FA-histone modification. FA-induced DPCs and chromatin damage triggered activation of the apical kinases ATR 26 and ATM 27 , respectively. Inhibition of i-proteasomes did not alter phosphorylation of either ATM (CHK2, KAP1) or ATR (CHK1, p53) substrates by FA ( Fig. 3A-C). Consistent with the normal activation of p53 (Fig. 3A,C), upregulation of its target, the CDK inhibitor p21, was also unaltered by LMP7-i (Fig. 3C). The presence of LMP7-i during short FA treatments also produced no changes in ATM-or ATR-dependent phosphorylation (Fig. 3D). Ser139-phosphorylated histone H2AX (known as γ-H2AX) is a well-established marker of genotoxic stress 31 . In FA-treated cells, γ-H2AX was found exclusively in S-phase cells and its levels were elevated by inhibition of standard proteasomes 32 . We confirmed that γ-H2AX was present only in the S-phase, as indicated by identical levels of γ-H2AX-positive and γ-H2AX/ EdU double-positive cells (Fig. 3E,F). I-proteasome inactivation did not alter the overall or S-phase-specific formation of γ-H2AX.
Scientific RepoRts | 7: 654 | DOI:10.1038/s41598-017-00757-w Caspase activation and cell death. FA-treated H460 cells undergo apoptosis, which is in part mediated by the transcription factor p53 26 . We found that the FA-induced production of a caspase-generated PARP fragment and cleaved (active) executioner caspase 7 were not affected by i-proteasome inhibition (Fig. 4A,B). LMP7-i activity was not lost during prolonged incubations, as evidenced by the continuing presence of the slower migrating LMP7 form (Fig. 3B). We next investigated LMP7-i effects on survival of cells with blocked caspase activation. We included a previously validated dose of the pancaspase inhibitor Q-VD-Oph 33 during FA treatments and the subsequent 72 h recovery and examined the colony formation. As expected based on the detection of activated caspases, Q-VD-Oph significantly decreased clonogenic toxicity of FA (Fig. 4C). We also found that i-proteasome inhibition enhanced clonogenic toxicity of FA even in the presence of Q-VD-Oph (Fig. 4D). These findings together with the normal activation of the proapoptotic transcription factor p53 ( Fig. 3A,C) indicate that i-proteasomes are protective against a caspase-independent cell death.   LMP7 band and the higher dose also produced a slower moving LMP2 band (Fig. 5A). In further experiments, we used 0.3 μM LMP7-i in IMR90 cells. Similar to H460, i-proteasome inhibition in IMR90 cells did not alter early genotoxic responses to FA, such as p53 phosphorylation, γ-H2AX production or DNA synthesis ( Fig. 5A-C). However, i-proteasome inactivation impaired a long-term recovery of IMR90 cells from FA damage, resulting in the depletion of S-phase (Fig. 5D) and accumulation of G1 cells (Fig. 5E). Thus, i-proteasomes also play a protective role against FA toxicity in primary human cells.

HSF1 activation.
We have recently identified FA as a potent inducer of the heat shock-responsive transcription factor HSF1 24 . HSF1 normally resides in the cytosol but accumulation of misfolded proteins induces its phosphorylation, nuclear translocation and binding to chromatin 34 . We first examined HSF1-Ser326 phosphorylation, which has shown a robust dose-dependent response to FA 24 . I-proteasome inhibition strongly enhanced Ser326 phosphorylation by FA in IMR90 cells whereas HSF1 protein levels remained unchanged (Fig. 6A). A slower mobility of HSF1 in FA samples is typical for proteotoxic conditions and reflects its phosphorylation at multiple sites 34 . FA-treated IMR90 and H460 cells with inactive i-proteasomes also showed higher amounts of nuclear Ser326-phosphorylated (Fig. 6B,C) and total HSF1 (Fig. 6D). Overall, these results indicate elevated levels of FA-induced proteotoxic stress in cells with disabled i-proteasomes.
Native gel analysis of proteasomes. A native gel electrophoresis allows separation of 20S and 26S particles that are subsequently identified by western blotting for their specific components 35 . We found that a majority of standard core 20S particles (~65%), detected by the presence of the chymotrypsin-like protease PSMB5, was uncapped in control H460 and IMR90 cells (Fig. 7A). These values are consistent with the results in other mammalian cells 3 . In response to FA, incorporation of regular 20S particles into capped 26S proteasomes slightly decreased in H460 (from 36.6 ± 3.5% to 30.1 ± 1.5%) but remained the same in IMR90 cells (36.1 ± 2.2% for control versus 35.5 ± 0.4% for FA). Similar to constitutive 20S proteasomes, a majority of 20S i-proteasomes identified by immunoblotting for LMP7 that replaces PSMB5 in i-proteasomes was also uncapped in control cells (Fig. 7B). Unexpectedly, we found that FA caused severe losses of 26S i-proteasomes in both H460 and IMR90 cells. The overall LMP7 protein amounts were not altered by FA treatments (Fig. 7C). All 19S particles detected by its base component Rpt2 were incorporated into 26S proteasomes irrespective of FA treatments in H460 cells (Fig. 7D). In primary IMR90 cells, 19S proteasomes were present in both free and 26S forms and their distribution was not noticeably altered by FA. Thus, 26S i-proteasomes but not standard 26S proteasomes were sensitive to FA-induced disassembly. Proteolytic activity of 20S core particles is promoted by their association with small PA28 (11S) activators: cytoplasmic PA28αβ and nuclear PA28γ 3 . PA28αβ is particularly important for stimulation of 20S i-proteasomes. We found that FA increased by approximately 2-fold a fraction of PA28αβ that was bound to 20S particles (74.8 ± 2.8% from 36.2 ± 2.8 in controls, n = 2, p = 0.027) (Fig. 7E, left panel). A similar change was found when 20S-bound PA28αβ was normalized to the amount of the 20S component PSMB5 (1.9-fold increase by FA). FA also induced a higher association of PA28γ with 20S particles (Fig. 7E, right panel) although the fraction of 20S-bound PA28γ was lower than that for PA28αβ. Ubiquitin reserves and protein polyubiquitination. The disassembly of 26S i-proteasomes and the increased association of PA28 activators with 20S particles indicate a shift towards ubiquitin-independent proteolysis. This raises the question whether cells with inactivated i-proteasomes experience a higher stress on the ubiquitin system due to the need for ubiquitination of additional proteins for destruction by standard 26S proteasomes. One measure of ubiquitin usage is the amount of free ubiquitin 36 . Inhibition of i-proteasomes during 1 h FA treatments produced either no impact (IMR90 cells) or only a modest depletion of free ubiquitin (H460 cells) (Fig. 8A). The inactivity of i-proteasomes during longer FA incubations caused >2-fold depletion of free ubiquitin, which showed no changes in cells with functional i-proteasomes (Fig. 8B,C). Consistent with the dynamics of free ubiquitin, cells with disabled i-proteasomes contained higher levels of polyubiquitinated cytoplasmic proteins as detected with two antibodies (Fig. 8D).

Discussion
In this work we obtained evidence for the importance of i-proteasome activity in recovery of human cells from injury by carcinogenic FA. Previous studies have found altered FA cytotoxicity in cells with certain DNA repair deficiencies, demonstrating the toxicological significance of DNA damage [37][38][39] . DPCs are generally considered as the main form of FA-induced DNA damage 15,21 . DPC repair involves a proteolytic removal of crosslinked proteins, which can be performed by a DNA-dependent protease SPRTN in conjunction with DNA replication 40,41 .  Inhibition of the constitutive proteasomes has also impaired removal of FA-produced DPCs in human cells 42 and altered genotoxic signaling responses in a manner that was consistent with a diminished repair of DPCs 32 . Histones are the main nuclear proteins modified by FA due to its high reactivity with the side-chain amino group of lysine. Chromatin damage by FA without the involvement of DNA also triggers activation of the genotoxic stress-sensitive kinase ATM 27 . Although i-proteasomes exhibit a higher activity towards basic proteins such as histones 14 , the protective function of i-proteasomes against FA did not result from their involvement in repair of chromatin or DNA. This conclusion is supported by several experimental observations. FA-induced DPCs are potent blockers of DNA replication 26,27 due to the inability of the ring-shaped replicative helicase complexes to progress over the steric block imposed by the bulkiness of DPCs 43 . Our results on the normal recovery of DNA synthesis and the normal formation and decay of the genotoxic stress marker γ-H2AX in cells with suppressed i-proteasome activity indicate that the removal of replication-blocking FA-DNA lesions was not affected. The presence of unrepaired DPCs leads to accumulation of cells in the G2 phase, which was not observed in FA-treated cells with inactivated i-proteasomes in contrast to inhibition of standard proteasomes 32 . FA-triggered DNA damage signaling responses include DPC-linked phosphorylation of CHK1 and p53 by ATR 26 and chromatin damage-induced CHK2 and KAP1 phosphorylation by ATM 27 . None of these signaling responses were altered by i-proteasome inhibition, further strengthening the argument that the prosurvival role of i-proteasomes involved cellular recovery from nongenotoxic injury by FA.
We have recently found that FA induces proteolytic polyubiquitination of proteins throughout the cell, especially in the cytoplasm, and causes a rapid activation of the cytosolic protein damage sensor HSF1 24 . Elevated levels of the activating Ser326 phosphorylation and chromatin binding of HSF1 in cells with inhibited LMP7 indicate that i-proteasome activity was probably most important for removal of FA-damaged cytosolic proteins. This suggestion is further supported by higher amounts of polyubiquitinated cytoplasmic proteins in cells with disabled i-proteasomes. FA caused a near complete release of the catalytic 20S i-proteasome from 26S i-proteasomes but little or no dissociation of standard 26S proteasomes, indicating that i-proteasome activity was shifted from ubiquitin-dependent to ubiquitin-independent proteolysis. Suppression of i-proteasome activity showed a greater usage of ubiquitin in FA-treated cells, pointing to a larger burden of proteins that required ubiquitination. The switch to ubiquitin-independent proteolysis of FA-damaged proteins benefits stressed cells by saving ubiquitin for tagging and proteolysis of other proteins and preserving ATP due to energy independence of 20S-mediated protein degradation. There are no other reports on a stress-induced disassembly of 26S i-proteasomes although the dissociation of standard 26S proteasomes is a known protective response against oxidized proteins 12,44 . Oxidation of protein cysteines 45 and the protein chaperone HSP70 46 have been implicated in the destabilization of standard 26S proteasomes by oxidative stress. It is possible that similar processes are involved in the disassembly of 26S i-proteasomes by FA, which reacts with NH 2 /SH-groups resulting in the formation of damaged/misfolded proteins that can titrate HSP70 away from 26S i-proteasome promoting dissociation of the regulatory 19S particle. The typical substrates for ubiquitin-independent proteolysis in unstressed cells are intrinsically unstable or unstructured proteins 3,12 , suggesting that these proteins could be particularly vulnerable to misfolding in response to chemical modifications of Lys and Cys by FA. The conformational flexibility of unstructured polypeptides can bring reactive Cys/Lys in a close proximity, allowing FA-induced crosslinking of two amino acids and thereby fixing distorted structures. When not promptly removed, severely misfolded proteins form aggregates that are resistant to proteasome-mediated proteolysis and can exert chronic toxic effects 1 . This pathological mechanism can explain delayed cell cycle abnormalities caused by i-proteasome inhibition in FA-damaged cells.

Methods
Chemicals. ONX-0914 (A4011) and Q-VD-Oph (A1901) were from ApexBio. Formaldehyde (F8775), camptothecin (C9911) and hydroxyurea (H8627) were from Sigma. Interferon-γ was purchased from Thermo Scientific (RIFNG50). Cells and treatments. H460 and IMR90 cells were purchased from the American Type Culture Collection and cultured as previously described 27 . Cells were treated with FA and other stressors in complete growth media. ONX-0914 (LMP7-i) was added to cells 1 h before FA and present during FA treatments. For induction of i-proteasomes, cells were treated with 4 ng/ml interferon-γ for 24 h.
Cell cycle analysis. A recently described procedure was followed 32 . DNA synthesis was measured by EdU labeling (10 µM, 1 h). DNA ploidy was determined by propidium iodide staining (40 µg/ml, 30 min at room temperature). Flow cytometry data were acquired on FACSCalibur (BD Biosciences) and analyzed by the CellQuest Pro software.
Clonogenic survival. H460 cells were seeded onto 60-mm dishes (400 cells/dish) and treated with chemicals on the next day. After 6-8 days of growth, colonies were fixed with methanol and stained with the Giemsa solution.
qRT-PCR. Total cellular RNA was purified with TRIzol Reagent (Ambion) and used for reverse transcription reaction (RT First Strand Kit, Qiagen). Real-Time PCR reactions were prepared using the RT SYBR Green ROX qPCR Mastermix and primers from Qiagen and run on the ViiA7 Real-Time PCR System (Applied Biosystems). Expression of i-proteasome subunits after FA treatments was determined by the ΔΔCt method using B2M, GAPDH and TBP mRNAs for normalization. Calculations of the relative gene expression for catalytic subunits within the same type of activity (β1, β2 or β5) included the following steps: 1) determination of the ratio (Ri) of the i-proteasome subunit expression to the standard subunit expression by subtracting the corresponding Ct values and using the resulting number as the power in the binary logarithm and 2) determination of the percentage of the i-proteasome subunit using the following equation: (Ri/1 + Ri) × 100%. The overall percentage of i-proteasomes was calculated by combining individual percentages for β1i, β2i and β5i subunits and dividing by 3.
Statistics. Two-tailed, unpaired t-test was used for the evaluation of statistical differences between the groups.