CREB-H belongs to a new subfamily of bZIP transcription factors
In search of proteins closely related to LZIP, we identified several mouse expressed sequence tag clones that might encode novel bZIP proteins homologous to LZIP. DNA sequencing of one clone (IMAGE 1887290) confirmed that this cDNA codes for a protein of 479 amino acids harboring a characteristic bZIP domain (GenBank accession no. AF392874). The coding sequence is reasoned to be complete in light of the existence of an in-frame termination codon immediately upstream of the putative initiating ATG. More than 75% of the residues in this novel protein are similar to those in LZIP.
At the time of initial identification, this mouse LZIP-like protein had no existing counterparts in the databases, except for a paralogous protein termed OASIS/CREB3L1 (36
). Subsequently, a highly homologous human protein was identified in a search for liver-specific transcription factor through random sequencing of cDNAs in libraries derived from human liver. This protein designated CREB-H is specifically expressed in hepatocytes (20
). In light of the high homology between the two proteins (67% identity and 73% similarity), the LZIP-related protein we identified likely represents the mouse ortholog of human CREB-H. Notably, the official name for CREB-H as approved by Mouse Genome Informatics is cAMP responsive element binding protein 3-like 3 protein (CREB3L3; http://www.informatics.jax.org
). For simplicity and consistency, hereafter we will call it mouse CREB-H. The mouse CREB-H locus was found to locate in chromosome 10. The gene is 8.5 kb in length and contains 12 exons.
Interestingly, several additional mammalian bZIP proteins significantly homologous to LZIP and CREB-H have been identified more recently. These include AIBZIP/Atce1/CREB4/CREB3L4 (37
) and human BBF2H7/CREB3L2 (39
). These proteins contain a short stretch of hydrophobic amino acids, which was predicted to be a TM domain by two programs HMMTOP (Hungarian Academy of Sciences; http://www.enzim.hu/hmmtop
) and PredictProtein (Columbia University; http://cubic.bioc.columbia.edu/predictprotein
). The C-terminal TM domain immediately adjacent to the bZIP region is highly conserved in this unique group of bZIP proteins. Indeed, phylogenetic analysis based on protein sequences segregates these bZIP proteins into a new subfamily distinct from CREB and ATF6.
CREB-H preferentially recognizes CRE, ATF6 and box B elements
Human CREB-H binds to CRE and box B (20
). However, mouse Atce1 closely related to CREB-H has been shown to recognize κB motif instead of CRE (38
). To clarify this issue, recombinant polyhistidine-tagged mouse CREB-HZ protein, which contains the bZIP region (amino acids 206–318), was expressed and purified from E.coli
. This protein was then tested for its ability to bind various 32
P-labeled DNA elements using the EMSA assay. In addition to CRE, box B and κB site, canonical ATF6-binding element (26
) was also examined because ATF6 has a TM domain (40
) and shares some similarity to CREB-H subfamily proteins. ERSE-I and ERSE-II are two different ATF6-responsive elements found in the promoter of GRP78 (32
) and Herp (33
) genes, respectively. CREB-H was found to bind strongly to canonical CRE, ATF6 and box B elements (, lanes 1, 5 and 6), but not to AP-1 or κB motif (lanes 2 and 4). CREB-H also bound less potently to CRE in the rat PEPCK gene, ERSE-I and ERSE-II (lanes 8–10), but it hardly recognized Sp1 and C/EBPα motifs (lanes 3 and 7). Results from the competition EMSA assay using 10- and 50-fold cold oligonucleotides verified the specific interaction of CREB-H with CRE, ATF6 and box B elements ().
Figure 1 CREB-H preferentially recognizes CRE, ATF6 and box B elements. (A) EMSA assay. EMSA was performed using 0.2 μg of purified CREB-HZ protein (amino acids 206–318), 3 pmol of 32P-labeled oligonucleotides representing the indicated DNA binding (more ...)
Proteolytic processing of CREB-H in cultured cells
We raised two polyclonal antisera (α-CH8 and α-CH9) in rabbits against the N-terminal 84 amino acids of CREB-H. We then performed western blotting with these antibodies to probe CREB-H protein in Hepa1-6 hepatoma cells that were either mock-transfected (, lanes 1 and 5), transfected with an untagged CREB-H-expressing plasmid (, lanes 2–4) or transfected with a FLAG-tagged CREB-H expressing plasmid (, lanes 6–8). Consistent with our results from northern blot analysis (see below in ), the amount of endogenous CREB-H protein in Hepa1-6 cells was very low (, lanes 1 and 5). In contrast, several additional α-CH8/9-reactive protein bands were observed in CREB-H-overexpressing Hepa1-6 cells (, lanes 2 and 6).
Figure 2 Proteolytic processing of CREB-H in hepatoma cells. (A) Schematic representation of the structure of CREB-H and CREB-HΔTC. (B) Western blot analysis. Empty vector (mock; lanes 1 and 5), pFLAG-CREB-H (lanes 2–4) and pcDNA-CREB-H (lanes (more ...)
Figure 6 CREB-H is underexpressed in human HCC tissues. Surgically resected HCC (T) and adjacent non-cancerous liver tissues (NT) from 26 patients were subjected to RNA extraction and subsequent RT–PCR using primers that amplify 286 bp cDNA fragment of (more ...)
To verify the identity of these reactive bands, we tested the pre-immune sera (, lanes 4 and 8) and performed protein blocking assay by pre-incubating α-CH8/9 with an excess amount of recombinant CREB-H protein (, lane 3 and 7). Because neither pre-immune nor depleted α-CH8/9 recognized the bands of ~80 and 50 kDa in size (, lanes 3, 4, 7 and 9), these bands should represent specific CREB-H-derived species. While the ~80 kDa proteins (, lanes 2 and 6, bands with an arrow) plausibly represent the full-length CREB-H and FLAG-CREB-H, the 50 kDa species (bands with an arrowhead) might be a cleaved product (, compare lane 2 to 3, and lane 6 to 7). Both forms of CREB-H might have extensive post-translational modifications because their sizes are larger than predicted.
Since α-CH8/9 recognizes the N-terminus of CREB-H, the appearance of 50 kDa α-CH-reactive species suggests site-specific proteolysis of CREB-H at the C-terminus. To verify the specificity of α-CH8/9 and to further characterize the proteolytic processing of CREB-H, we expressed two versions of CREB-H proteins (, full-length CREB-H and truncated CREB-HΔTC) fused to GFP. In CREB-HΔTC, a C-terminal part including the TM domain was removed (). Comparison of this artificially truncated protein with the proteolytic fragments of CREB-H might shed light on the cleavage site in CREB-H. To this end, Hepa1-6 cells were transfected with GFP-CREB-H and GFP-CREB-HΔTC plasmids and the cell extracts were examined by western blotting ().
With increasing amounts of protein loaded, α-CH9 was able to detect a band of ~110 kDa in size that corresponds to GFP-CREB-H (, upper panel, lanes 2–4). The blot was re-probed with a monoclonal anti-GFP antibody (α-GFP; , lower panel, lanes 3 and 4). The reaction of the 110 kDa band with both α-CH9 and α-GFP verified the identity of GFP-CREB-H and it also lent further support to the specificity of α-CH9.
Likewise, both α-CH9 and α-GFP recognized the GFP-CREB-HΔTC protein of 75 kDa (, lane 5). Interestingly, in cells that had been transfected with a GFP-CREB-H plasmid but not a GFP-CREB-HΔTC construct, an α-CH9- and α-GFP-reactive band of about the same size was also observed (, compare lane 4 with lane 5, bands with arrows). This species plausibly represents a processed form of CREB-H that had been cleaved at a specific site. Because the size of this cleavage product is almost identical to that of GFP-CREB-HΔTC, the actual cleavage site was predicted to be very close to the junction between bZIP and TM regions (). It is noteworthy that this cleaved species of 77 kDa shown in corresponds to the 50 kDa band in . The difference in molecular mass (~27 kDa) was accounted for by GFP. In addition, when we performed immunoprecipitation with α-CH8 and extracts of cells transfected with either FLAG-CREB-H or GFP-CREB-H plasmid, two major protein species corresponding to the full-length and cleaved forms of CREB-H, respectively, were found in the precipitates ().
Next, we asked whether proteolysis also occurs in HepG2 cells stably transduced with a lentivirus expressing CREB-H. In keeping with results obtained from transfected cells, a strong ~80 kDa band and a weaker ~50 kDa species reactive to α-CH9 were observed in extracts of lentivirus-transduced cells (cf. , lane 2 with , lane 5). These two protein bands were not seen when α-CH9 was depleted with recombinant CREB-H (, lane 4). As a size reference for the ~50 kDa band observed in lentivirus-transduced cells, protein extract of HepG2 cells expressing CREB-HΔTC, which did not react to depleted α-CH9 (, lane 5), was also examined (cf. lane 3 with lane 2).
CREB-H activates CRE- and ATF6 element-dependent transcription
Human CREB-H has been shown to activate box B-dependent transcription (20
). Box B is a fat-body-specific enhancer in Drosophila
and box B-like sequences in mammalian cells have not been characterized (18
). In light of the strong binding of CREB-H to CRE and ATF6 element (), we next asked whether CREB-H was able to drive transcription from CRE or ATF6 element. To address this, we co-transfected a CREB-H expressing plasmid and either pCRE-Luc or pATF6-Luc reporter plasmid into HepG2 hepatoma cells and assayed for luciferase activity. For comparison, we also tested the activities of ATF4, ATF6, CREB and C/EBPα on CRE and ATF6 enhancers (). While the full-length CREB-H weakly activated CRE-dependent transcription (, 1.6-fold activation), it was capable of stimulating the ATF6 enhancer to 4.9-fold (). This potency is comparable with that of ATF4, CREB and C/EBPα (11.3-, 3.9- and 6.9-fold activation, respectively).
Figure 3 Transcriptional activity of CREB-H and potentiation by cAMP. Empty vector (VEC) and plasmids (0.3 μg of each) expressing CREB-H, CREB-HΔTC (CHΔTC), ATF4, ATF6, CREB and C/EBPα were individually co-transfected with pCRE-Luc (more ...)
It is generally accepted that removal of the TM domain from ATF6 and LZIP leads to the activation of these transcription factors (40
). The proteolytic cleavage of ATF6 can be induced in vivo
in response to ER stress (40
), but the physiological stimuli for the activation of LZIP remain elusive. CREB-H also contains a TM domain, which is probably removed through site-specific proteolytic cleavage () as exemplified in the cases of ATF6 (42
) and SREBPs (43
). To test the hypothesis that CREB-H could also be activated through regulated proteolysis, we asked whether the truncated CREB-HΔTC devoid of the TM and C-terminal domains () might have a higher transcriptional activity. Indeed, CREB-HΔTC acted as a transcriptional activator on both CRE and ATF6 element, resulting in ~100- and 30-fold increase of luciferase activity (). In this setting, CREB-HΔTC is more potent than CREB, ATF4, ATF6 and C/EBPα. This raised the possibility that the truncated CREB-HΔTC might be functionally equivalent to the physiologically active form of CREB-H generated through site-specific proteolysis in vivo
CREB-H is activated in response to cAMP stimulation
cAMP is a ubiquitous second messenger and regulator of gene transcription (45
). In one well-characterized pathway, cAMP activates PKA, which phosphorylates and activates CREB (17
). In addition, C/EBPs are also responsive to cAMP stimulation (6
). This prompted us to investigate the regulation of CREB-H by cAMP. We measured the activities of CREB-H and CREB-HΔTC on the pCRE-Luc and pATF6-Luc reporters with or without the addition of Fsk, which stimulates adenylate cyclase leading to increased cAMP levels (). Fsk did not activate ATF6 element-dependent transcription significantly with the exception of CREB-H, which was stimulated mildly by Fsk (, ~80% increase in luciferase activity). As to CRE-dependent transcription, both CREB-H and CREB-HΔTC were modestly responsive to Fsk stimulation, resulting in up to 100% increase in reporter activity (, 3.7- and 227.8-fold activation, respectively). However, in the control experiment in which the empty vector alone was co-transfected, Fsk also stimulated CRE-dependent transcription (, 3.7-fold activation). Thus, while it remains unclear as to whether Fsk specifically stimulates CREB-H, Fsk activation of endogenous transcription factors alone unlikely accounted for the increase of CRE-dependent luciferase activity observed in the presence of CREB-HΔTC and Fsk. Thus, cAMP might be directly or indirectly involved in regulating CREB-H and CREB-HΔTC.
CREB-H activates PEPCK promoter
While the DNA-binding and gene-activating abilities of LZIP, CREB-H and other transcription factors in the same subfamily have been well documented (20
), their physiological targets remain to be identified. As a first step toward identifying the gene targets of CREB-H (20
), we asked whether it might bind and activate CRE-dependent transcription in liver. In this regard, the gene encoding gluconeogenic enzyme PEKCK has been extensively used as a model for studying the regulation of CREB- and C/EBP-dependent transcription in liver (49
). Based on this reasoning, we have demonstrated that CREB-H binds modestly to the CRE in rat PEPCK promoter (, lane 8). Because CREB-H binds moderately to the rat PEPCK CRE, here we set out to investigate whether it might activate the PEPCK promoter. Indeed, CREB-HΔTC but not the full-length CREB-H stimulated the PEPCK promoter activity when overexpressed in hepatoma cell line Hep3B (, lane 3). PEPCK is abundantly expressed in liver and is an important metabolic enzyme that controls gluconeogenesis in both liver and adipose tissues. It is well known that cAMP/PKA can activate the PEPCK gene transcription through the cAMP response unit in the PEPCK promoter (6
). As such, when the cAMP pathway is stimulated by either treatment of Fsk or the expression of PKA catalytic subunit in Hep3B cells, the PEPCK promoter is stimulated (, lanes 4 and 7) (52
). Consistent with our finding that Fsk was able to further enhance CREB-HΔTC activation of CRE (), we observed that the expression of CREB-HΔTC cooperated with either PKA or Fsk in the activation of PEPCK promoter (, lanes 6 and 8).
Because the reporter assays were performed in cells transfected with PEPCK-Luc plasmid, we next investigated the binding of CREB-H to genomic PEPCK promoter in CREB-H-expressing cells using ChIP assay (). We observed that α-CH8 but not pre-immune IgG precipitated a CREB-H–DNA complex containing endogenous PEPCK promoter (, panels 1 and 2, lane 3 compared with lane 2). In a control experiment, α-CH8 did not pull down an irrelevant genomic sequence in human MAD1 promoter (, panel 3, lane 3). In further support of the specificity of the interaction between CREB-H and the PEPCK promoter, ChIP assays performed in non-CREB-H-expressing Hepa1-6 cells (see ) yielded no CREB-H–DNA complex (, panel 4), whereas the PEPCK promoter was amplified from the CREB-H-containing immunoprecipitate prepared from CREB-H-expressing Hepa1-6 cells. Thus, our results consistently revealed that CREB-H specifically associates with PEPCK promoter in vivo.
Figure 5 CREB-H is exclusively expressed in the liver of adult mouse and is differentially expressed in mouse cancer cell lines. Northern blot analysis was performed on mouse tissues (A) and cancer cell lines (B). CREB-H transcripts are arrowed. Western blotting (more ...)
Subcellular localization of CREB-H
We expressed CREB-H and CREB-HΔTC proteins in MCF-7 cells and determined their subcellular localization. CREB-H localized predominantly in the cytoplasm (, panel 1), whereas CREB-HΔTC was exclusively in the nucleus (panel 5). Because ATF6 with a TM domain localizes to ER membrane and is activated by intramembrane proteolysis (40
), we investigated the association of CREB-H with ER by co-staining cells for CREB-H and calnexin, an ER marker. As shown in , full-length CREB-H co-localized with calnexin (panels 3 and 4 compared with panel 7). These results suggest that CREB-H, similar to ATF6, localizes to the ER.
Figure 4 Subcellular localizations of CREB-H and CREB-HΔTC. (A) pcDNA-CREB-H (panels 1–4) and pcDNA-CREB-HΔTC (panels 5–7) plasmids were separately transfected into MCF-7 cells. Cells were fixed and co-stained with α-CH8 (more ...)
Next, we examined the localization of CREB-H protein in HepG2 cells homogenously expressing V5-tagged CREB-H from a lentiviral vector. Again, a typical cytoplasmic localization pattern was observed when we stained the lentivirus-transduced cells with anti-V5 (, panels 1 and 5). We noticed a significant co-localization of CREB-H and calnexin (, panels 1–4). While the staining patterns of CREB-H and the Golgi marker GM130 were distinct, a fraction of CREB-H was also found in the Golgi apparatus (, panels 5–7). Thus, the full-length but inactive CREB-H () is retained predominantly in the ER. On the contrary, CREB-HΔTC acts as a potent transactivator plausibly owing to constitutive nuclear localization.
CREB-H expression is abundant in adult liver but reduced in hepatoma cells
We performed northern blot analysis of CREB-H mRNA in mouse tissues and cells. Consistent with previous findings in human (20
), CREB-H mRNA was exclusively and abundantly detected in adult mouse liver (, lane 5). On the other hand, the CREB-H signal was not observed in a mouse hepatoma cell line Hepa1-6 (, lane 7). In addition, we also found that human CREB-H transcript is underexpressed in most hepatoma cell lines tested (data not shown). It is, however, noteworthy that mouse CREB-H transcript is highly expressed in 3T3 fibroblasts immortalized by mammary sarcoma virus (M-MSV-BALB/3T3), cancerous subcutaneous adipocytes (L-M) and NB41A3 cancer cells derived from neuroblastoma (, lanes 4, 5 and 12).
Consistent with results from northern blot analysis, CREB-H protein was detected in mouse liver, but not in brain or heart (, lanes 1–3). The identity of the endogenous CREB-H protein in mouse liver was verified by comparing with untagged recombinant CREB-H (, lane 4) and by antibody depletion experiments (, lanes 5–8). Notably, the truncated CREBΔTC form was undetectable in the blot, implicating that CREB-H is not constitutively active in mouse liver.
Because CREB-H is abundantly expressed in normal adult liver but underexpressed in liver cancer cell lines, we speculated that CREB-H might be critically involved in hepatocarcinogenesis. With this hypothesis in mind, we performed RT–PCR screening of CREB-H transcripts in human HCC tissues. The CREB-H mRNA was significantly underexpressed in HCC samples when compared with corresponding non-tumorous liver (, P < 0.001 by Wilcoxon test). A more than 2-fold reduction in steady-state amounts of CREB-H mRNA was observed in 11 of the 26 HCC tissues (see for representative RT–PCR results and for a scatterplot).
Growth suppressive activity of CREB-H in hepatoma cells
The underexpression of CREB-H in HCC cells and tissues ( and ) suggests that it might play a role in hepatocarcinogenesis. Considered together with the tumor suppressive activities of other bZIP proteins, such as C/EBPα (11
), C/EBPβ (53
) and LZIP (22
), we queried whether CREB-H might also serve a growth suppressive function in the liver. We carried out a growth assay in HepG2 cells using transient co-transfection of pSV-βgal, a β-gal expression vector driven by SV40 promoter, plus excess amount (10×) of CREB-H expressing plasmids. This ensures that the blue cells are the CREB-H transfected cells when staining with X-gal. This method has previously been used to study the growth suppressive activity of various proteins, including C/EBPα and C/EBPβ (9
On day 1 after co-transfection, slight differences were seen in the number of blue cell clusters among the groups that had received vector alone, full-length CREB-H and CREB-HΔTC (, panels 1–3). On average, four cell clusters (of ≥4 cells) were seen in vector-transfected cells, whereas two and one clusters were evident in CREB-H- and CREB-HΔTC-overexpressing cells, respectively (, columns 1–3). However, on day 5 after transfection, significant differences were observed among vectors, CREB-H and CREB-HΔTC groups (, panels 4–9; , columns 4–6). On average, four cell clusters increased to seven cell clusters in vector-transfected cells, whereas only one or less than one cell cluster was found in CREB-H- and CREB-HΔTC-overexpressing cells (, columns 4–6). Western blot analysis confirmed that the protein expression levels of CREB-H and CREB-HΔTC in HepG2 cells on day 1 and day 5 post-transfection were similar (data not shown). Thus, CREB-H and CREB-HΔTC inhibited the proliferation of HepG2 cells.
Figure 7 Overexpression of CREB-H suppresses growth in cultured human cells. (A) pcDNA3.1/V5 vector, pcDNA3.1/V5-CREB-H and pcDNA3.1/V5-CREB-HΔTC were co-transfected with pSV-βgal at a ratio of 10:1 into HepG2 cells. Cells were harvested and stained (more ...)
To further characterize the growth inhibitory activity of CREB-H, we measured 5-bromo-2′-deoxyuridine (BrdU) incorporation in CREB-H- and CREB-HΔTC-expressing HepG2 cells using confocal microscopy. BrdU incorporation indicates the cellular activity of DNA synthesis that dictates competency of cell proliferation. This method has been widely used to study cell growth and cell cycle progression. In particular, it has been adopted to demonstrate the growth suppressive role of transiently expressed C/EBPα in cultured cells (12
). In our control experiment, the expression of β-gal did not affect BrdU incorporation (, panels 1–3). In stark contrast, BrdU staining was undetectable in either CREB-H- or CREB-HΔTC-expressing HepG2 cells (, panels 4–9; compare transfected cells with arrowheads to surrounding untransfected cells with arrows). Quantitative analysis by cell counting indicates that none of the CREB-H- or CREB-H-positive cells was incorporating BrdU, whereas ~50% (47.6 ± 5.0%) of control cells expressing β-gal were positive for BrdU. In another word, BrdU incorporation was significantly reduced in CREB-H- or CREB-HΔTC-expressing cells compared with control cells (P
< 0.0001 by t
-test). Thus, the activation of CREB-H can inhibit S-phase entry and cell proliferation.
Figure 8 Overexpression of CREB-H inhibits S-phase entry. (A) pcDNA3.1/V5-β-gal, pcDNA3.1-CREB-H and pcDNA3.1-CREB-HΔTC were transfected into HepG2 cells for 16 H. Cells were then treated with 10 μM BrdU for 6 h, and fixed. Cells were subsequently (more ...)