A naturally-occurring dominant-negative competitor of Keap1 against its inhibition of Nrf2

Transcription factor Nrf2 is a master regulator of antioxidant and/or electrophile response elements (AREs/EpREs)driven genes involved in homeostasis, detoxification and adaptation to various stresses. The cytoprotective activity of Nrf2, though being oppositely involved in both cancer prevention and progression, is critically controlled by Keap1 (Kelch-like ECH-associated protein 1) as an adaptor subunit of Cullin 3-based E3 ubiquitin ligase, that is a key sensor for oxidative and electrophilic stresses. Now, we first report a novel naturally-occurring mutant of Keap1, designated Keap1ΔC, which lacks most of its C-terminal Nrf2-interacting domain essential for inhibition of the CNC-bZIP factor. This mutant Keap1ΔC is yielded by translation from an alternatively mRNA-spliced variant lacking the fourth and fifth exons, but their coding sequences are retained in the wild-type Keap1 locus (with no genomic deletions). Although this variant was found primarily in the human highly-metastatic hepatoma (MHCC97H) cells, it was widely expressed at very lower levels in all other cell lines examined. No matter whether Keap1ΔC retains less or no ability to inhibit Nrf2, it functions as a dominant-negative competitor of Keap1 against its inhibition of Nrf2-target genes. This is due to its antagonist effect on Keap1-mediated turnover of Nrf2 protein.

the ETGE and DLG motifs within the Neh2 domain of Nrf2 [15], which acts as a degron targeting the CNC-bZIP protein to ubiquitin-mediated proteasomal degradation pathways.
Whilst global knockout of Nrf2 in the mouse does not alter normal physiological development and growth with no spontaneous pathological phenotypes, all Nrf2 -/animals are more susceptible than wild-type mice to chemical cytotoxicity [16]. By contrast, global knockout of its inhibitor Keap1 causes a constructive enhancement in both protein stability and transactivation activity of Nrf2, but leading to death of all Keap1 -/--deficient mice within three weeks after birth [17,18]. This lethality is principally attributed to malnutrition caused by severe hyperkeratosis within the upper digestive tract (i.e. esophagus and forestomach). Further analysis of Keap1:Nrf2 double knockout mice has confirmed that the recognizable phenotype of Keap1 -/--deficient mice is attributable to the constitutive activation of Nrf2. To avoid such death of murine pups before weaning, conditional hepatocyte-specific Keap1 -/mice were generated, as an expected result of viable animals with no apparent abnormalities, but a nuclear moderate accumulation of Nrf2 is elevated in order to activate the CNC-bZIP factor and confer its potent resistance against acute drug toxicity [19]. Despite cytoprotective function of Nrf2 supported by graded expression of Keap1, a marked decrease of its levels to less than 50% results in increased mortality in a study of 2-year-old mice [20]. Overall, these facts demonstrate that the benefits of modest cytoprotective activation of Nrf2 against acute toxicity are hormetic, but its constitutive activation beyond a certain threshold is rather disadvantageous to long-term survival. The notion is supported by permanently hyperactive Nrf2 allowing for its transfiguration to act as an unrecognized mediator of oncogenesis and promote cancer cell survival in tumorigenesis [21,22]. Hence, such distinct dual roles of Nrf2 (to become a good friend or dangerous foe) in tumor prevention and progression have led us to take careful account of its opposing potentials to implicate in cancer treatment.
In this present study, we first report a discovery of a naturally-occurring mutant of Keap1 (designated Keap1 ΔC ), which is expressed primarily in human highly-metastatic hepatoma MHCC97H cells, albeit it was widely expressed at much lower levels in all other cell lines examined. Keap1 ΔC is identified to arise from translation of an alternatively mRNA-spliced variant lacking the fourth and fifth exons of wild-type Keap1 (but with no genomic deletion mutants). Subsequent sequencing of the cDNA products from MHCC97H cells revealed that overlapped double peaks emerged from the 1326th nucleotide of the Keap1-coding region (Fig. 1C). To clarify the overlapped double peaks of the above-described Keap1 cDNA sequences, they were subcloned into the pcDNA3 expression vector, before 14 clones were randomly selected and identified by PCR with a pair of Keap1 primers. The resulting electrophoresis of PCR products showed two types of bands with distinct sizes, though both were migrated closely to the standard 2000-bp DNA marker, on 1.0% Agarose gels (Fig. 1D). These two bands of Keap1 cDNAs were also determined by further sequencing of all selected clones. The result demonstrates that the relatively larger bands represent the wild-type full-length cDNA sequence of Keap1 without any mutation, whilst the shorter bands are owing to a loss of the entire fourth and fifth exons when compared with the intact wild-type (Fig.   1E). Together with not any of the genomic deletion mutants of the Keap1 gene locus in these cells examined, these findings have led us to postulate there exists a novel alternatively-spliced variant of Keap1 transcript in MHCC97H cells. Further bioinformatic analysis of the spliced variant revealed a loss of the C-terminal 180-aa residues of Keap1 (called Keap1 ΔC , Fig. 1F). This results in a constructive deletion mutant of most of the DGR domain and adjacent CTR region of Keap1 essential for its interaction with Nrf2 [23,26]. Lastly, the nucleotide sequence of the variant Keap1 ΔC along with its amino acid sequence had been submitted to the GenBank developed by NCBI, with the accession No.

Differential expression of Keap1 ΔC in distinct cell lines.
To determine whether Keap1 ΔC is also expressed in other cell lines rather than MHCC97H cells, two distinct pairs of primers targeting specifically for this variant or wild-type Keap1 ( Fig. 2A) were, according to their sequence similarity and difference, designed with the same upstream primers coupled with distinct downstream primers retaining the identical tetranucleotide (5'-CCTC-3') at their 3'-ends, for a precision measure of real-time qPCR. Subsequent results illustrated that different products for Keap1 and Keap1 ΔC exhibited the same compatible melt peaks (Fig. 2B, left) Further examinations revealed that endogenous Keap1 and Keap1 ΔC mRNAs were widely but diversely expressed at varying extents in distinct types of 15 human somatic cancerous and non-cancerous cell lines (Fig. 2E), including 7 hepatoma-derived cell lines HepG2, MHCC97H, MHCC97L, Huh7, Hep3B, SMMC7721 and Hepa1-6, one lung cancer cell line A549, and other 4 female cancer cell lines MCF7 (from breast cancer), HeLa (from cervical carcinoma), SKOV3 and A2780 (the latter two from ovarian cancer), together with 3 non-cancerous cell lines of HL7702 (liver), Hacat (skin) and HEK293 (kidney). Overall, expression of Keap1 ΔC mRNA was at relatively higher levels determined primarily in MHCC97H, Huh7 and A2780 cells than other cell lines, but its bona fide abundances were also presented at considerable lower levels than the value of wild-type Keap1 mRNA in all examined cell lines.

An interference of Keap1 ΔC with Keap1 for its inhibition of Nrf2-target genes
To investigate whether inhibition of Nrf2 by Keap1 is interfered by distinct expression of Keap1 ΔC in four different cell lines, thus we examined basal protein levels of Keap1 and Nrf2, along with its downstream target genes HO-1, GCLM and NQO1. Western blotting showed a significant high abundance of Nrf2 only in MHCC97H cells, rather than other three cell lines determined (Fig. 3A1). This result is too coincidental to be just a coincidence, along with almost none of Keap1 detected in MHCC97H cells (Fig. 3A2), but as accompanied by relative higher levels of Keap1 ΔC (Fig. 2E). This finding indicates that the increase of Nrf2 should be attributable to remarked expression of Keap1 ΔC . This notion is also supported by additional opposing fact that no increased, but considerable lower, levels of Nrf2 ( (Fig. 2E), led to marked decreases of Nrf2 to similar lower levels that measured from HL7702 cells ( Fig. 3A1). Together, the non-cancerous HL7702 cells are allowed for maintaining proper constructive thresholds of Keap1 and Nrf2 to be expressed at controllable lower levels. In turn, it is inferable to exist a potential relevance of malignant hepatoma development to the constructive increases in basal Nrf2 or Keap1 proteins, which should also be affected by differential expression of Keap1 ΔC . Further examinations of these cancerous and non-cancerous cells revealed a coupled positive and negative correlation of Nrf2 and Keap1, respectively, with basal protein expression levels of Nrf2-target genes HO-1 and GLCM, but not NQO1 (Fig. 3A3 to A5). Such an exception of NQO1 that was highly expressed in HL7702 and HepG2 cell lines (Fig. 3A5), though accompanied by lower levels of Nrf2, indicates that it may also be regulated by other CNC-bZIP family members (e.g. Nrf1) and other transcription factors (e.g. AP-1) [27,28].
To determine whether inhibition of Nrf2-mediated gene transcription by Keap1 is interfered by Keap1 ΔC , both MHCC97H and HepG2 cell lines were co-transfected with a Keap1 or Keap1 ΔC expression construct together with an ARE-driven luciferase reporter. The results showed that over-expression of Keap1 caused a significant decrease in ARE-Luc reporter activity to ~30% of control levels (Fig. 3B, that were measured from transfection of MHCC97H cells with an empty pcDNA3 vector and the reporter). However, such a decreased transactivation activity of ARE-Luc reporter gene mediated by endogenous Nrf2 appeared to be, at least in part, disinhibited by endogenous Keap1 ΔC to a certain extent, albeit over-expression of Keap1 ΔC only resulted in slight repression of the reporter gene activity in the same MHCC97H cells (Fig. 3B). By contrast, Nrf2-transacting ARE-Luc reporter activity was almost unaffected by ectopic Keap1 or Keap1 ΔC in HepG2 cells (Fig. 3C). This is postulated to be owing to a considerable higher background of endogenous Keap1, so that Nrf2-targeted gene reporter may be not over-regulated by further over-expression of ectopic Keap1 or Keap1 ΔC in HepG2 cells.
To further determine the interfering role of Keap1 ΔC in the negative regulation of Nrf2 by Keap1, HepG2 and MHCC97H cell lines were transfected with a Keap1 or Keap1 ΔC expression construct alone or plus Nrf2 plasmids. As expected, Western blotting showed that endogenous protein levels of Nrf2 were obviously reduced ( Fig. 3D1 & E1, lanes 2 vs 1) in both cell lines that had been allowed for further over-expression of ectopic Keap1 (Fig. 3D5, D6, E5 and E6). However, forced expression of ectopic Keap1 ΔC led to a significant augment in abundances of endogenous Nrf2 protein (lanes 3 vs 1). Further examinations also revealed that ectopic Keap1, but not Keap1 ΔC , led to greater or less extents of distinct decreases in total levels of both ectopic and endogenous Nrf2 proteins ( Fig. 3D1 and E1, cf.   lanes 5 &6 with 4). Moreover, variations in protein expression levels of different ARE-driven genes HO-1 (Fig. 3D2& E2), but not GCLM (Fig. 3D3 & E3), were correlated positively with changed abundances of Nrf2 per se in the presence or absence of ectopic Keap1 or Keap1 ΔC , implying that HO-1 is an optimal marker of Nrf2-target genes.
Intriguingly, real-time qPCR showed that both endogenous and ectopic mRNA levels of Nrf2 were almost not influenced by over-expression of either Keap1 or Keap1 ΔC in HepG2 cells, but the latter two regulators were required for striking suppression of endogenous Nrf2, but were also involved in significant promotion of its ectopic expression in MHCC97H (Fig. 3F), suggesting that Nrf2 mRNA expression and/or its stabilization might be monitored by a proper status of Keap1 or Keap1 ΔC . Importantly, further examinations revealed that both basal and Nrf2-mediated levels of HO-1 mRNA were markedly repressed by ectopic Keap1, rather than Keap1 ΔC , in HepG2 cells (Fig. 3G). By contrast, MHCC97H cells gave rise to an increased expression background of endogenous HO-1 mRNA, but it was not further enhanced by Nrf2 over-expression, albeit its expression was obviously reduced by Keap1, but with almost none or less of inhibition by Keap1 ΔC (Fig. 3G). This implies that transcription of HO-1 is regulated by Nrf2 in a steady-state system. In addition, basal GCLM and NQO1 l mRNAs appeared to be unaffected by either Keap1 or Keap1 ΔC (Fig. 3H & I), but a marginal increase in Nrf2-induced GCLM rather than NQO1 expression was diminished by Keap1. This was roughly not or less decreased by Keap1 ΔC in MHCC97H and HepG2 cells.  $p<0.01, n=33) and/or significant decreases(*p<0.01, n=33) were determined with error bars (S.E.).

Keap1 ΔC acts as a dominant-negative competitor against intact Keap1
To provide a better understanding of the axiomatic reason why Keap1 ΔC is enabled for its interference with Keap1, we conducted molecular modeling of two possible dimeric complexes of Keap1 ΔC (only retaining two Ketch motifs) with intact Keap1 (with all six Ketch motifs and adjacent C-terminal region, which are essential for direct association with Nrf2) (Fig. 4A). These models were based on a known crystal structure of Keap1 in complex with the N-terminal Neh2 region of Nrf2 (i.e. 3WN7 deposited in PDB) [15,29,30]. From formation of an invalid dimer of Keap1 ΔC :Keap1 and Keap1 ΔC :Keap1 ΔC (Fig. 4A, right panels), it is therefore deduced that they do not only enable intact Keap1 to be simply consumed, but act as a potential dominant-negative competitor of Keap1 against its functional interaction with Nrf2.
To verify whether Keap1 ΔC has a marginal negativity to regulate Nrf2 through its residual DGR region, we created a series of expression constructs. Of note, the Keap1 N321 mutant lacks the entire DGR domain, but retains its N-terminal 321-aa region, while Keap1 ΔN is an N-terminal deletion mutant, which comprises only 303 aa covering the entire DGR domain of Keap1 to its C-terminal end (Fig. 4B). Subsequently, ARE-driven luciferase reporter assays of HepG2 (Fig. 4C) and MHCC97H (Fig. 4D) cell lines demonstrated that Nrf2-mediated transactivation activity was N-terminally V5-tagged polypeptide of ~8-kDa was, by coincidence, yielded from the putative N-terminal proteolytic processing of Keap1 or Keap1 ΔC (Fig. 4I, lanes 14 & 15 in upper two panels). Moreover, immunoblotting of Keap1 ΔN unraveled that additional post-synthetic processing of Keap1 might also occur within a region closer to its C-terminal end, such that the majority of the C-terminally-tagged V5 ectope of this mutant was truncated off the protein, albeit the remaining portion of Keap1 ΔN was still recognized by Keap1-specific antibody (Fig. 4G, lane 6 vs Fig. 4H, lane 12).

An antagonist effect of Keap1 ΔC on Keap1-mediated turnover of Nrf2
To further explore the potential mechanism by which Keap1 ΔC de-represses the negative regulation of Nrf2 by Keap1, distinct settings of pulse-chase experiments were conducted in different cell lines. The COS-1 cells that had been transfected with an Nrf2-expression construct alone or in combination with an additional construct for Keap1 or Keap1 ΔC , before being treated with cycloheximide (CHX, which inhibits biosynthesis of nascent proteins), were subjected to determination of whether Keap1 ΔC exerts an antagonist effect on Keap1-mediated turnover of Nrf2. As anticipated, over-expression of Keap1 caused a strikingly abrupt decrease in the abundance of Nrf2 to 16% of the basal expression levels in the non-Keap1-transfected cells (Fig. 5A). Conversely, a substantial increase to 1.67-fold amounts of Nrf2 resulted from co-expression of Keap1 ΔC . Further examination revealed that over-expression of Keap1 ΔC rendered the half-life of Nrf2 turnover to be prolonged from 0.19 h (=11.4 min) to 0.28 h (=16.8 min) following treatment of cells with CHX (Fig. 5B). Even after 1-h treatment of cells with CHX, 20% of Nrf2 abundance was retained by Keap1 ΔC , before being gradually decreased to its putative minimum of ~7% by 2 h of the chemical treatment (Fig. 5A). However, just a 7% minimum of Nrf2 was also exhibited at time points of 0.31 h (=18.6 min) or 1.07 h (=64.2 min) respectively, following CHX treatment of cells that had been transfected with Keap1 or not (Fig.   5B).
The above-described data demonstrate that Keap1 ΔC is not a negative regulator of Nrf2, but conversely may act as an antagonist against endogenous Keap1, leading to disinhibition of keap1-mediated Nrf2 turnover. To address this hypothesis, COS-1 cells that had been co-transfected with expression constructs for Keap1, Keap1 ΔC or both together with Nrf2 plasmids were subjected to further pulse-chase experiments. As the time of CHX treatment was extended to 2 h, a fairly smooth tendency of decreases in the Nrf2 abundance from its maximum (1.67-fold) to its minimum (0.12-fold) was observed in cells co-expressing Keap1 ΔC and the CNC-bZIP protein (Figs. 5, C and D). By sharp contrast, the former minimum of Nrf2 was closer and even almost equal to additional starting level (0.15-fold) of this protein determined immediately before Keap1/Nrf2-co-expressing cells were treated with CHX (Fig. 5C, cf. the first lane with the last one). However, the Keap1-led decrease of Nrf2 appeared to be partially mitigated, with a half-life slightly prolonged from 0.23 h to 0.41 h, by co-transfection of Keap1 ΔC (Figs. 5, C and D). This finding indicates that Keap1 ΔC exerts an antagonist effect on keap1 in monitoring the turnover of Nrf2 protein.
Further time-course analysis of CHX-treated MHCC97H cells revealed that endogenous Nrf2 protein levels were modestly reduced by over-expression of Keap1, rather than Keap1 ΔC (Fig. 5E), but its half-life was obviously extended by Keap1 ΔC from 0.39 h (=23.4 min) to 0.62 h (=37.2 min) following CHX treatment (Fig. 5F). Taken together with the data shown in Fig. 2E, these suggest that endogenous Keap1 ΔC may compete Keap1, leading to de-repression of Keap1-mediated turnover of Nrf2 protein in MHCC97H cells. However, a marked increase in the endogenous CNC-bZIP protein was observed following treatment of HepG2 cells with proteasomal inhibitor MG132 (at 10 mol/L) (Fig. 5G). Notably, over-expression of Keap1 rather than Keap1 ΔC caused significant decreases in baseline abundance and MG132-stimulated accumulation of endogenous Nrf2 protein in HepG2 cells. Taken together, these results demonstrate that Keap1, but not Keap1 ΔC , is required for targeting Nrf2 to the proteasomal-mediated degradation pathway (Fig. 5H). acting as a dominant-negative competitor of Keap1. This is due to the fact that Keap1 ΔC can occupy the place in the formation of an invalid dimer with Keap1 or itself, no matter whether it only retains less or no ability to inhibit Nrf2. It is important to note that this mutant Keap1 ΔC has an antagonist effect on Keap1-mediated turnover of Nrf2 by proteasomal degradation pathway. In addition, upon dissociation of Nrf2 from Keap1, the CNC-bZIP factor will be allowed for spatio-temporal translocation into the nucleus before transactivating ARE-driven cytoprotective genes against oxidative stress or other biological stimuli.

CONCLUSION
As the most important inhibitory regulator of Nrf2, Keap1 plays a key role in monitoring the protein degradation of this CNC-bZIP factor and its subcellular locations, before regulating ARE-driven target genes [32,33]. Here, we report a novel discovery of Keap1 ΔC , as a naturally-occurring mutant of Keap1 with a constructive deletion of its C-terminal 180-aa residues essential for its directly binding to the Neh2 domain of Nrf2. The Keap1 ΔC mutant is generated from translation of the alternatively Keap1 mRNA-spliced variant lacking both the fourth and fifth exons, but their coding sequences are retained in wild-type Keap1 gene locus (i.e. with no genomic deletions). The variant Keap1 ΔC was determined primarily in the human highly-metastatic hepatoma MHCC97H cells, albeit is also widely expressed at very lower levels in all other cell lines examined. Further determination of Keap1 ΔC reveals that this mutant protein has a potent at forming an invalid dimer with Keap1 or itself, no matter whether it only retains less or no ability to inhibit Nrf2. As such being the case, the mutant Keap1 ΔC is also identified to act as a dominant-negative competitor

Expression constructs and transfection
The human full-length Nrf2, Keap1 and Keap1 ΔC cDNA sequences were subcloned into a pcDNA3 vector, respectively.
Deletion mutants of Keap1 were created by inserting appropriate PCR-amplified cDNA fragments into the above-described vector. Related primers were listed in Table 1. These mutants included Keap1 ΔN (with a deletion of the N-terminal 321-aa from Keap1), Keap1 N321 (only retaining the N-terminal 321-aa region). In addition, they were N-terminally or C-terminally tagged by the V5 ectope to yield distinct strategic expression constructs. Following all these constructs were verified by DNA sequencing, they were transfected into experimental cells (2.5×10 5 ), which had been allowed for growth to a confluence of 80% in 35-mm culture dishes. The transfection was carried out for 15-min in a mixture of indicated constructs (2 g/ml) with Lipofectamine 3000 (Invitrogen, USA), and then allowed for a recovery from transfection before being experimented.

Luciferase reporter assay
A pARE-Luc reporter plasmid [34], which was used for a measure of the transactivation activity of ARE-driven gene mediated by Nrf2, and pRL-TK, which is an internal control of Renilla luciferase reporter, together with indicated expression constructs were co-transfected into HepG2 or MHCC97H cells. At about 24 h after transfection, the luciferase reporter activity was measured by using dual luciferase reporter assay (Promega, USA).

Real-time quantitative PCR
Experimental cells that had been transfected or not been transfected with the indicated plasmids were subjected to isolation of total RNAs by using the RNAsimple Kit (Tiangen Biotech CO., Beijing). Subsequently, 500 ng of total RNAs was added in a reverse-transcriptase reaction to generate the first strand of cDNA (with Revert Aid First Strand Synthesis Kit from Thermo). The synthesized cDNA was served as the template for qPCR, in the GoTaq® qPCR Master Mix (from Promega), before being deactivated at 95℃ for 10 min, and amplified by 40 reaction cycles of 15 s at 95℃ and 30 s at 60℃. The final melting curve was validated to examine the amplification quality, whereas expression of mRNA for β-actin was viewed as an internal standard control.