β-TrCP-mediated ubiquitination and degradation of liver-enriched transcription factor CREB-H

CREB-H is an endoplasmic reticulum-resident bZIP transcription factor which critically regulates lipid homeostasis and gluconeogenesis in the liver. CREB-H is proteolytically activated by regulated intramembrane proteolysis to generate a C-terminally truncated form known as CREB-H-ΔTC, which translocates to the nucleus to activate target gene expression. CREB-H-ΔTC is a fast turnover protein but the mechanism governing its destruction was not well understood. In this study, we report on β-TrCP-dependent ubiquitination and proteasomal degradation of CREB-H-ΔTC. The degradation of CREB-H-ΔTC was mediated by lysine 48-linked polyubiquitination and could be inhibited by proteasome inhibitor. CREB-H-ΔTC physically interacted with β-TrCP, a substrate recognition subunit of the SCFβ-TrCP E3 ubiquitin ligase. Forced expression of β-TrCP increased the polyubiquitination and decreased the stability of CREB-H-ΔTC, whereas knockdown of β-TrCP had the opposite effect. An evolutionarily conserved sequence, SDSGIS, was identified in CREB-H-ΔTC, which functioned as the β-TrCP-binding motif. CREB-H-ΔTC lacking this motif was stabilized and resistant to β-TrCP-induced polyubiquitination. This motif was a phosphodegron and its phosphorylation was required for β-TrCP recognition. Furthermore, two inhibitory phosphorylation sites close to the phosphodegron were identified. Taken together, our work revealed a new intracellular signaling pathway that controls ubiquitination and degradation of the active form of CREB-H transcription factor.

CREB-H is identical or very similar to other β -TrCP degrons found in erythropoietin receptor, rotavirus nonstructural protein NSP1, transcription factor Nrf2, β -catenin and transcriptional co-activator YAP1 (Fig. 2B). Moreover, multiple alignment of CREB-H-ΔTC among different species revealed two additional clusters of evolutionarily conserved serine and threonine residues flanking the DSGIS motif: SxxxSxxxSxxxS located at residues 73-85 and SxxxxS/T located at residues 95-100 ( Fig. 2A). Because phosphorylation of degron-proximal serine and threonine residues critically affects β -TrCP recognition in many substrates 29,30 , it would not be surprising that these residues in CREB-H-ΔTC might also have a regulatory function in β -TrCP-dependent degradation. In addition, identical or very similar β -TrCP recognizing motifs were also found in other members of the CREB3 subfamily (Fig. 2B).
Finally, we compared the stability of CREB-H-ΔTC in HepG2 cells with or without β -TrCP overexpression by cycloheximide chase assay. Diminution of CREB-H-ΔTC was more pronounced in the presence of β -TrCP than in its absence (Fig. 4A, lanes 6-8). Taken together, our results supported the notion that CREB-H-ΔTC is a novel substrate of SCF β-TrCP E3 ubiquitin ligase.   found to have minimal influence on polyubiquitination of CREB-H-ΔTC-Δ81-90 (Fig. 4E, lane 5). In the control group with overexpression of β -TrCP, polyubiquitination of CREB-H-ΔTC was much more robust and CREB-H-ΔTC protein was marginally detected in both the input and the precipitate (Fig. 4E, lane 3). Hence, our results consistently demonstrated the essentiality of the DSGIS degron in β -TrCP binding, polyubiquitination and degradation of CREB-H-ΔTC.

Characterization of β-TrCP
For most β -TrCP substrates, phosphorylation of the β -TrCP degron is usually required for their binding to β -TrCP 29,30 . With this in mind we constructed and characterized different phosphorylation mutants of CREB-H-ΔTC (Fig. 5A) to determine whether the interaction between β -TrCP and CREB-H-ΔTC is phosphorylation-dependent.
Second, further mutational analysis was carried out to identify the serine residues, the phosphorylation of which would be critical for β -TrCP recognition. Interestingly, whereas CREB-H-ΔTC-A3 with non-phosphorylatable S85A, S87A and S90A did not interact with β -TrCP, CREB-H-ΔTC-A2 carrying non-phosphorylatable S87A and S90A was still capable of binding to β -TrCP (Fig. 5C, lanes 3 and 4). Finally, we explored how phosphorylation within the SDSGIS degron might affect transcriptional activity of CREB-H-ΔTC. S87 and S90 within the SDSGIS motif were replaced one by one by alanine and aspartate (Fig. 5A). Luciferase reporter assay was performed on the CRE promoter which is known to be highly responsive to CREB-H-ΔTC. CREB-H-ΔTC-S90A displayed increased transcriptional activity but the CREB-H-ΔTC-S90D mutation had inhibitory effect (Fig. 5E, groups 6 and 7). CREB-H-ΔTC-S87A was also transcriptionally more active than CREB-H-ΔTC-S87D (Fig. 5E, groups 4 and 5). The activity difference between non-phosphorylatable and phosphomimetic mutants of CREB-H-ΔTC suggested that phosphorylation at S87 and S90 might indeed be critical in governing β -TrCP recognition.

S73 and S77 are degron-proximal inhibitory phosphorylation sites of CREB-H-ΔTC.
The presence of evolutionarily conserved serine residues in the flanking regions of the β -TrCP phosphodegron in CREB-H-ΔTC ( Fig. 2A) prompted us to ask whether their phosphorylation might have regulatory roles in β -TrCP-dependent ubiquitination and degradation of CREB-H-ΔTC. In this regard, we noted that phosphoserines and phosphothreonines adjacent to the degron in other β -TrCP substrates such as Snail, YAP and Nrf2 are also known to be influential in substrate recognition and degradation [39][40][41] .
To shed light on the roles of serine phosphorylation in the region of amino acid residues 73-77 proximal to the degron, we constructed non-phosphorylatable mutant CREB-H-ΔTC-MA, in which S73, S76 and S77 had been replaced by alanine. Opposite to this, a CREB-H-ΔTC-MD mutant with serine-to-aspartate substitution at the same three positions was also created (Fig. 6A). Notably, CREB-H-ΔTC-MA migrated faster on SDS-PAGE than CREB-H-ΔTC, whereas no noticeable change in the electrophoretic mobility rate of CREB-H-ΔTC-MD was found (Fig. 6B, lanes 2 and 3). The electrophoretic mobility shift indicated that one or more sites among S73, S76 and S77 might be post-translationally modified. Constitutive phosphorylation was most probable since the phosphomimetic form migrated as fast as the wild-type protein. Luciferase reporter assay confirmed the rise and fall of the transcriptional activity of CREB-H-ΔTC-MA and CREB-H-ΔTC-MD, respectively (Fig. 6C, groups 3 and 4), indicating that phosphorylation of this region affects protein function.
We generated another mutant named CREB-H-ΔTC-A7, which contains both A4 and MA mutations (Fig. 6A). This mutant was not only as stable as CREB-H-ΔTC-A4 but also migrated as fast as CREB-H-ΔTC-MA (Fig. 6B, lanes 5-7). In addition, CREB-H-ΔTC-A7 had much higher transcriptional activity on the CRE-Luc reporter than CREB-H-ΔTC-A4 (Fig. 6D, groups 4 and 5), indicating the enhancing effect of the MA mutation on CREB-H-ΔTC-dependent transcription. When we used anti-V5 to pull down CREB-H-ΔTC and its mutants from cells, multiple protein species were found in the precipitates (Fig. 6E). Treatment with calf intestine phosphatase eliminated the multiple protein bands in the CREB-H-ΔTC and CREB-H-ΔTC-MA immunoprecipitates (Fig. 6F, lanes 2 and 4), providing crucial support to the phosphorylation of CREB-H-ΔTC.
To identify the exact phosphorylation sites among S73, S76 and S77, additional mutants were generated and tested. Because single mutants CREB-H-ΔTC-S73A, CREB-H-ΔTC-S76A and CREB-H-ΔTC-S77A only had weak phenotypes (data not shown), we introduced them in the background of CREB-H-ΔTC-A4 (Fig. 6A). Fast-migrating species of both CREB-H-ΔTC-S73A-A4 and CREB-H-ΔTC-S77A-A4 were observed in SDS-PAGE (Fig. 6G, lanes 2 and 4). In contrast, fast migration was not observed for CREB-H-ΔTC-S76A-A4, CREB-H-ΔTC-A4 or CREB-H-ΔTC (Fig. 6G, lanes 1, 3 and 5). Consistent with the electrophoretic mobility patterns, transcriptional activities of CREB-H-ΔTC-S73A-A4 and CREB-H-ΔTC-S77A-A4 on the CRE-Luc reporter Scientific RepoRts | 6:23938 | DOI: 10.1038/srep23938 were higher than that of CREB-H-ΔTC-A4 and close to that of CREB-H-ΔTC-A7 (Fig. 6H, groups 4, 6, 7 and 8). Taken together, our data suggested that both S73 and S77 are inhibitory phosphorylation sites of CREB-H-ΔTC.  reporter (Fig. 6D, group 4). In this part of our study we made use of the CREB-H-ΔTC-A4 mutant to investigate the impact of CREB-H-ΔTC stabilization on target gene transcription in cultured cells. We and others have previously identified phosphoenolpyruvate carboxykinase (PEPCK) gene as a target of CREB-H in the regulation of gluconeogenesis 4,7 . In addition, three genes critically involved in lipid metabolism, namely FGF21, FSP27β and APOA4, have recently been found to be CREB-H target genes in the liver 12,18,20 . On the other hand, CREB-H-ΔTC has also been shown to induce cell secretion and activate MMP13 gene transcription in HEK293 cells 14 . Hence, we chose these five genes for analysis of the transcriptional activity of CREB-H-ΔTC-A4. CREB-H-ΔTC was capable of activating the transcription of PEPCK, FGF21, FSP27β and APOA4 in HepG2 or Hep3B hepatoma cells as well as that of MMP13 in HEK293 cells, but the stimulating effect of CREB-H-ΔTC-A4 mutant was more pronounced on all five genes (Fig. 7A,C,E,G,I, groups 2 and 3). Consistent with the RT-qPCR results, the transcriptional activity of CREB-H-ΔTC-A4 as reflected in the expression of luciferase reporter driven by all five promoters was also higher than that of CREB-H-ΔTC (Fig. 7B,D,F,H,J, groups 2 and 3). Plausibly, stabilization of CREB-H-ΔTC led to enhancement of transcriptional activation of target genes both in magnitude and in duration.
To further verify the role of β -TrCP in the regulation of PEPCK gene expression, the two β -TrCP-targeting siRNAs were used to deplete endogenous β -TrCP expression in HepG2 cells. The knockdown effect was highly specific to β -TrCP and the siRNAs had no influence on the expression of CREB-H mRNA (Fig. 8A,B). By contrast, the expression of PEPCK transcript induced by CREB-H-ΔTC was significantly boosted when endogenous β -TrCP was depleted (Fig. 8C, groups 3 and 4). Thus, the activation of PEPCK gene expression by CREB-H was regulated by β -TrCP.

Discussion
In the present study, we uncovered a new mechanism by which SCF β-TrCP ubiquitin ligase mediates the ubiquitination and degradation of CREB-H-ΔTC, the active form of liver-enriched transcription factor CREB-H. An evolutionarily conserved β -TrCP phosphodegron with the SDSGIS motif was identified and characterized. In addition, the degron-proximal S73 and S77 were identified as the inhibitory phosphorylation sites for β -TrCP-dependent ubiquitination and degradation. Based on our findings, we propose a model for regulation of CREB-H-ΔTC activity (Fig. 9). In response to as yet uncharacterized physiological stimuli, CREB-H-ΔTC is phosphorylated sequentially by two kinases, the second of which might be casein kinase II (CKII) as demonstrated in another study 38 . Phosphorylation by CKII at the degron is required for β -TrCP binding and recognition. According to our results, phosphorylation at three serine residues S85, S87 and S90 might be minimally required. In addition, inhibitory phosphorylation at S73 and S77 within the degron-proximal region, plausibly by glycogen synthase kinase 3 (GSK-3) distinct from CKII as suggested in the other study 38 , also has regulatory function. CREB-H-ΔTC phosphorylated at S73 and S77 has low activity and is probably sensitized to subsequent degron phosphorylation by CKII leading to K48-linked polyubiquitination and proteasome-mediated degradation. As a result, transcription of CREB-H target genes, such as PEPCK, FGF21, FSP27β , APOA4 and MMP13, is down-regulated.
At the completion of our study, another work documenting SCF β-TrCP -dependent degradation of CREB-H-ΔTC was published 38 . The two studies came to the same conclusion. The results presented were largely consistent and complementary. However, most of the mutants constructed and analyzed, some of the results obtained, as well as the emphases of the two papers were different. Whereas two kinases CKII and GSK-3 were identified in the other study, K48-linked polyubiquitination of CREB-H-ΔTC was analyzed in more detail in our work. On the other hand, the S87A S90A mutant was largely deficient for binding with β -TrCP in the other study 38 . In contrast, the β -TrCP-binding activity was unaffected in the same CREB-H-ΔTC-A2 mutant in our work, but a triple mutant CREB-H-ΔTC-A3 (i.e. S85A, S87A and S90A) lost the capability to bind with β -TrCP (Fig. 5C). Our results therefore suggested that S85, S87 and S90 in CREB-H-ΔTC might be minimally required for β -TrCP recognition. We are currently performing animal studies to clarify whether the CREB-H-ΔTC-A3 and CREB-H-ΔTC-A4 mutants are constitutively active and β -TrCP-dependent CREB-H-ΔTC degradation is physiologically relevant in vivo.
Although the two studies corroborated with each other to support the new model for SCF β-TrCP -dependent ubiquitination and degradation of CREB-H-ΔTC, several questions remain unanswered. First, direct evidence for S85 phosphorylation is required although we have demonstrated the essential role of this residue in β -TrCP binding. Additional phosphorylation sites including S81, S95 and T100 also merit further analysis. Second, in addition to CKII and GSK-3 reported in the other study 38 , other kinases acting on the SDSGIS motif and its flanking regions remain to be identified and characterized. Based on bioinformatic analysis, we speculate that S73 and S77 might be phosphorylated by proline-dependent kinases in addition to GSK-3, whereas S87 and S90 might be targeted by acidophilic serine kinases in addition to CKII 42 . Finally, the mechanism by which dephosphorylation of S73 and S77 augments CREB-H-ΔTC activity remains unclear. In this regard, we have compared the subcellular localization of CREB-H-ΔTC and CREB-H-ΔTC-MA but both were found in the nucleus (data not shown), excluding the possibility of cytoplasmic sequestration of CREB-H-ΔTC-MA. Other mechanisms should be explored in future study.
The relationship between the β -TrCP phosphodegron and its proximal region warrants further investigations. One possibility is that phosphorylation of the degron-proximal region "primes" or sensitizes degron phosphorylation. Such "priming phosphorylation" has been found in many other β -TrCP substrates 39,40,43 . One example is Snail with a DS (0) GxxS (+4) xxxS (+8) xxS (+11) motif, in which the β -TrCP degron is bolded. In this case, priming phosphorylation at the serine residue at + 11 position is required for subsequent phosphorylation at serine residues at +8, +4 and 0 positions 39,44 (panels A,C,E,G,I). Relative mRNA expression levels were normalized to the levels of β -tubulin transcript. Fold activation values represent the means ± standard deviations of three independent measurements. Concurrently, luciferase reporter assays were performed with the indicated reporters in HepG2, Hep3B and HEK293 cells as in Fig. 5E (panels B,D,F,H,J). Results are representative of triplicate experiments and error bars indicate the standard deviations. Statistical analysis was performed with two tailed Student's t test. *p < 0.05. **p < 0.01.***p < 0.001.  The phosphodegron of CREB-H-ΔTC, DS (0) Gφ S (+3) , is very similar to but does not exactly match the canonical β -TrCP-binding motif DS (0) Gφ xS (+4) . However, several substrates without the x residue, such as RV-NSP1, YAP and Nrf2, have been identified, indicating that the x residue is not essential 40,41,46 . Interestingly, residue at −2 position of DS (0) Gφ S (+3) always prefers phosphorylatable residues (S, T or Y). Experimentally, we also demonstrated the requirement of S85 for the binding of CREB-H-ΔTC to β -TrCP. Thus, we propose the phosphorylatable residue at −2 position should be important for the binding of the SDSGφ S motif to β -TrCP. In other words, phosphorylation at −2 position might be "priming" degron phosphorylation at other sites. Alternatively, phosphorylation at −2 position could provide a negative charge to increase the binding of the SDSGφ S motif to β -TrCP. Previous studies have shown that negatively charged residue is most critical to β -TrCP binding 47 .
We have found that CREB3L4 contains the same β -TrCP recognizing sequence (SDSGIS) as CREB-H. In addition, the SDSDGS motif in CREB3L1 and the SDSEGS motif in CREB3L2 are very similar to the motif found in CREB-H. Plausibly, β -TrCP-mediated degradation is a common mechanism shared by most of these proteins. CREB3L1 and CREB3L2 are the substrates of Fbxw7. Both Fbxw7 and β -TrCP are F-box proteins of the SCF E3 ubiquitin ligase and they share many similar properties. Fbxw7 recognizes an S/TxxxS/T motif, in which S/T are phosphorylated 48 . It will be of interest to see whether Fbxw7 might also target CREB-H and CREB3L4. Interestingly, within CREB3L2 and CREB3L1 sequence, the reported Fbxw7-binding motif (TPPSS) is immediately next to the putative β -TrCP-binding motif (Fig. 2B). Fbxw7 and β -TrCP can cooperatively degrade substrates 49 . Thus, it will be intriguing to determine whether Fbxw7 and β -TrCP might cooperatively mediate the degradation of some CREB3 subfamily proteins.
The expression and activity of CREB-H are regulated at multiple levels including proteolytic activation, phosphorylation, ubiquitination and degradation. Most of the results are obtained with CREB-H-ΔTC in our study and in the other work 38 . CREB-H-ΔTC is the physiologically active form of CREB-H. The full-length CREB-H tethered to the ER membrane is an inactive protein with no transcriptional activity. However, since the phosphodegron and flanking regions are present in both CREB-H and CREB-H-ΔTC, it is not surprising if CREB-H degradation is regulated through the same mechanism. Indeed, we showed that CREB-H was also subjected to K48-linked polyubiquitination. Thus, CREB-H might also be targeted by β -TrCP in addition to other ER-bound E3 enzymes involved in ERAD. β -TrCP is an E3 ligase that plays an important role in the regulation of various cell signaling pathways governing cell division, metabolism and oncogenesis 29,31 . Thus, our demonstration of β -TrCP-dependent ubiquitination and degradation of CREB-H-ΔTC provides new opportunities to explore how CREB-H degradation and activity are linked to other signal transduction pathways converging on β -TrCP. β -TrCP has been suggested as an important new target for anti-cancer therapy 27 . By the same reasoning, our findings also reveal a new strategy of controlling CREB-H activity by enhancing or decreasing the expression and activity of β -TrCP. For example, small-molecule activators and inhibitors of the two kinases that phosphorylate CREB-H-ΔTC could be harnessed for therapeutic modulation of the expression of CREB-H target genes.  Table 1. Plasmid pSV-RLuc was purchased from Promega. Plasmids pCRE-Luc and pPEPCK-Luc have been described previously 4 .

Methods
Plasmid FLAG-β -TrCP-ΔFbox and plasmids expressing various CREB-H-ΔTC mutants were generated by site-directed mutagenesis using reagents supplied by Agilent. Mutagenic primers were designed using the web-based QuikChange Primer Design Program (www.agilent.com/genomics/qcpd; Agilent).
Cell culture and transfection. Human hepatoma cell line Hep3B and human embryonic kidney cell lines HEK293 and HEK293T were cultured in Dulbecco's Modified Eagle's Medium containing 10% fetal bovine serum (Life Technologies). Human hepatoma cell line HepG2 was grown in Eagle's Minimum Essential Medium (Life Technologies) supplemented with 10% fetal bovine serum. Transfection of cells was performed as previously described 56,57 .
For proteasome inhibition, transfected cells were treated with 10 μM MG132 (Calbiochem) for 6 hours before harvest. For protein turnover assay, 200 μM cycloheximide (Sigma-Aldrich) was added into transfected cells 5 hours before harvest.

Co-immunoprecipitation, phosphatase treatment and in vivo polyubiquitination.
Co-immunoprecipitation was carried out as described 58 . Cells were lysed with NP40 buffer supplemented with protease inhibitor cocktails. Antibodies were recovered by incubating with recombinant protein G agarose (Invitrogen) for 2 hours, followed by overnight incubation with the remaining cell lysate at 4 °C. The protein G agarose was collected and washed at least three times with wash buffer (50 mM Tris-Cl, pH7.4, 800 mM NaCl, 1 mM EDTA, 1% NP-40 and 0.2% Triton X-100) after incubation. The immunoprecipitates were separated by SDS-PAGE and analyzed by immunoblotting.
Phosphatase assay was performed as described 59 . HEK293T cells were cultured in 60 mm dishes and transiently transfected with V5-CREB-H-ΔTC or its mutant forms. The protein lysates were incubated with mouse anti-V5 antibody and recombinant protein G agarose overnight. Next day the beads were rinsed with cold PBS and treated 10 units of calf intestine alkaline phosphatase (Fermentas) for 1 hour at 37 °C. After treatment, the beads were pelleted, washed twice with wash buffer and analyzed by immunoblotting.
For in vivo polyubiquitination, cells were co-transfected with the MYC-ubiquitin plasmid in combination with the wide-type or mutant V5-CREB-H-ΔTC plasmids. Forty-eight hours after transfection, total protein was collected, immunoprecipitated with anti-V5 antibody, resolved with SDS-PAGE and immunoblotted with anti-MYC antibody.

RNA interference.
RNA knockdown experiments were performed as described 57 . HepG2 and HEK293T cells were transfected with 100 nM siRNA using Lipofectamine 2000 (Invitrogen). The siRNA sequences are as follows: 5′ -AAGUGGAAUU UGUGGAACAU CdTdT-3′ for siβ -TrCP#1 43 ; 5′ -GAGAGAGAAG ACUGUAAUAdT dT-3′ for siβ -TrCP#2 and 5′ -GCUACCUGUU CCAUGGCCAdT dT-3′ for siGFP 60,61 . Real-time PCR analysis. Total RNA was extracted from cells using the RNAiso Plus reagent (Takara) according to the manufacturer's instructions and then incubated with DNase I (Ambion) to remove remaining genomic DNA. Reverse transcription was performed using the Transcriptor First Strand cDNA Synthesis reagents (Roche). Real-time quantitative PCR was conducted in the presence of SYBR Premix Ex Taq (Takara) with the StepOnePlus real-time PCR system (Applied Biosystems). The mRNA level of β -tubulin was determined as an internal control and the results were analyzed using StepOne Software v2.3 (Applied Biosystems). Primer sequences used are listed in Supplementary Table 1.