| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Research Resource |
Department of Molecular Medicine and Surgery (O.M.V., R.M., E.R.-B., G.N., A.F.-M.), Karolinska Institutet, 17176 Stockholm, Sweden; Molecular Endocrinology Group (L.F.-P.), Department of Clinical Sciences, University of Las Palmas de Gran Canaria, 35011 Las Palmas de Gran Canaria, Spain; Department of Biochemistry and Molecular Biology (D.J.C., J.W., M.O.), Oregon Health & Science University, Portland, Oregon 97239-3098; and Computational Biology Unit (B.L.), Bergen Centre for Computational Science, University of Bergen, N-5008 Bergen, Norway
Address all correspondence and requests for reprints to: Amilcar Flores-Morales, Center of Molecular Medicine (CMM), L8:01, Karolinska Hospital, 17176 Stockholm, Sweden. E-mail: Amilcar.Flores{at}ki.se.
ABSTRACT
The GH-activated signal transducer and activator of transcription 5b (STAT5b) is an essential regulator of somatic growth. The transcriptional response to STAT5b in liver is poorly understood. We have combined microarray-based expression profiling and phylogenetic analysis of gene regulatory regions to study the interplay between STAT5b and GH in the regulation of hepatic gene expression. The acute transcriptional response to GH in vivo after a single pulse of GH was studied in the liver of hypophysectomized rats in the presence of either constitutively active or a dominant-negative STAT5b delivered by adenoviral gene transfer. Genes showing differential expression in these two situations were analyzed for the presence of STAT5b binding sites in promoter and intronic regions that are phylogenetically conserved between rats and humans. Using this approach, we showed that most rapid transcriptional effects of GH in the liver are not results of direct actions of STAT5b. In addition, we identified novel STAT5b cis regulatory elements in genes such as Frizzled-4, epithelial membrane protein-1, and the suppressor of cytokine signaling 2 (SOCS2). Detailed analysis of SOCS2 promoter demonstrated its direct transcriptional regulation by STAT5b upon GH stimulation. A novel response element was identified within the first intron of the human SOCS2 gene composed of an E-box followed by tandem STAT5b binding sites, both of which are required for full GH responsiveness. In summary, we demonstrate the power of combining transcript profiling with phylogenetic sequence analysis to define novel regulatory paradigms.
GH IS THE MAIN physiological regulator of somatic growth (1). STAT5b is a transcription factor that associates with the intracellular domain of the GH receptor (GHR) and becomes phosphorylated by Janus kinase (JAK) 2 upon GH stimulation. Phosphorylated STAT5b dimerizes and translocates to the nucleus where it binds to 
-interferon activated sequences (GAS)-like elements in the DNA and promotes the transcription of some GH-regulated genes. The importance of STAT5b as a mediator of GH action has been clearly demonstrated both in vitro and in vivo (2, 3, 4). In humans, at least two different STAT5b mutations have been described resulting in severe growth retardation in children (5, 6). Interestingly, the growth curves of these individuals are practically indistinguishable from those of children with inactivating mutations on the GHR (Laron syndrome) or IGF-I, a key mediator of GH action. These findings constitute a clear demonstration of the unique role of STAT5b as a transducer of GH signals leading to somatic growth. Accordingly, STAT5b has been shown to bind to intronic regions of the IGF-I gene (7) and to regulate its hepatic transcription in response to GH (7). The key role of STAT5b in GH action has also been demonstrated in animal models. Mice expressing truncated GHR variants that can activate JAK2 but fail to bind or efficiently activate STAT5b show severe growth retardation (8).
Although current data support a role for STAT5b in the effects of GH on somatic growth, the full range of functions of STAT5b needs further investigation using models in which the protein can be specifically eliminated from the liver or other tissues. One observation pertinent to this idea comes from analysis of mice with hepatic inactivation of the glucocorticoid receptor (GR). GR appears to interact with STAT5 and act as a coactivator. The GR-STAT5b interaction is independent of the DNA binding activity of the GR and appears to result in increased transcriptional activity of STAT5b-regulated genes (9). Accordingly, the liver-specific GR knockout mouse shows a dramatic reduction in body size and reduced hepatic expression of STAT5b-dependent genes, such as IGF-I and acid-labile subunit (ALS) (10).
STAT5b response elements have been characterized in a small number of GH-regulated genes, including IGF-I (7), cytokine-induced SH2-containing protein (CIS) (11), and serine protease inhibitor 2.1 (12). Induction by GH of IGF-I, IGF binding protein 3, ALS, SOCS3, and SOCS2 is impaired in livers from STAT5b-deficient mice, but it is unclear whether each gene is a direct target of STAT5b or whether regulation occurs through other mechanisms (13). Indeed, although SOCS3 gene expression is rapidly stimulated by GH, no STAT5 response elements have been found in its proximal promoter (14). Study of the GH-STAT5b pathway is complicated by the fact that several other transcription factors are also regulated by GH, including STATs 1 and 3, the 
, ß, and 
isoforms of CCAAT/enhancer-binding protein (15) (16, 17), hepatic nuclear factor (HNF)-1 (18), HNF-6 (19), interferon regulatory factor-1 (20), mitochondrial transcription factor 1 (21), c-jun, c-fos (22, 23), forkhead box m1b (24), runt-related transcription factor 2 (25), homeobox A1 (26) and V-erbA-related protein-2 (ear-2) among others. In addition, STAT5b can regulate the function of other transcription factors through both direct and indirect interactions that are independent of its DNA binding activity. This is exemplified by its effects on peroxisome proliferator-activated receptor , where direct association results in the mutual inhibition of the transactivation activity of both proteins (27).
In this study, we have used microarray analysis to evaluate the contribution of STAT5b to the acute transcriptional response to GH in vivo. We have analyzed the effects of a single pulse of GH on gene expression in the liver of hypophysectomized rats in the presence of either constitutively active (CA) or a dominant-negative (DN) STAT5b delivered by adenoviral gene transfer. Using this approach, we have identified novel hepatic targets of STAT5b. Detailed analysis of one of these target genes, SOCS2, defines its direct transcriptional regulation by STAT5b upon GH stimulation and demonstrates how a combination of functional experiments and the application of new bioinformatics tools can lead to novel mechanistic insights.
RESULTS
Identification of STAT5b-Regulated Genes in the Liver
We previously developed a series of recombinant adenoviruses encoding NH2-terminally Flag-tagged variants of rat STAT5b, including a DN protein, containing a substitution of phenylalanine for tyrosine 699 (STAT5bDN-Y699F), and a CA version, with alteration of asparagine 642 to histidine (STAT5bCA-N642H). Using these reagents, we showed that STAT5b played a central role in the stimulation of IGF-I gene transcription by GH in vivo (7) and also was involved in the regulation of several other GH-activated genes in the liver (7). We have now used a similar approach to characterize other STAT5b-responsive genes among GH-induced mRNAs in the liver.
To identify genes that might be direct targets of STAT5b, we first evaluated differences in gene expression in livers from hypophysectomized male rats infected with adenovirus (Ad)-enhanced green fluorescent protein (EGFP), Ad-STAT5bDN, or Ad-STAT5bCA, and treated with GH or vehicle for 30, 60, or 120 min (Fig. 1
). Variations in mRNA abundance were expressed as fold changes relative to expression levels in hypophysectomized rats infected with Ad-EGFP. As assessed using SAM statistics and a FDR of 5%, we identified 387 genes differentially regulated in Ad-STAT5bCA infected livers when compared with Ad-EGFP infected controls (see Fig. 1 and supplemental Table 1, which is published as supplemental data on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org). We found that 110 of those genes increased in abundance by more than 60%, whereas 250 decreased to the same extent (supplemental Table 1). Fifty-four of these genes appeared to require active STAT5b, as their expression was diminished in livers infected with Ad-STAT5bDN and was not increased by GH (Table 1). Of this group, only 19 genes showed a rapid increase in expression after GH treatment in rats infected with the control virus, Ad-EGFP (Table 1). We also identified 179 transcripts that were rapidly regulated by GH in the livers of Ad-EGFP infected hypophysectomized rats (supplemental Table 2), with 52 mRNAs being increased in abundance within 2 h of a single GH pulse, and 97 being decreased. Because only 20% of the GH-stimulated genes in livers infected with Ad-EGFP or Ad-STAT5bCA showed a decline in abundance in livers expressing STAT5bDN, our results suggest that other signaling pathways in addition to STAT5b are important for the rapid transcriptional response to GH.
|
|
, in bold), were direct transcriptional targets of STAT5b. To answer this question, we first used a computational approach to examine each of these genes for the presence of STAT5b binding sites within putative regulatory regions located within phylogenetically conserved areas between the rat and human orthologues. If this cohort of genes is a direct target of STAT5b, then there should be an increased frequency of phylogenetically conserved STAT5b binding sites within their promoters and/or regulatory regions compared with controls whose expression was not induced by GH. A further increase in the frequency of conserved STAT5b binding sites would be expected in the subset of Ad-STAT5bCA-stimulated genes whose mRNA levels were reduced by Ad-STAT5bDN and were also increased by GH in livers expressing Ad-EGFP.
To identify putative STAT5b binding sites, we adapted the methodology developed by Lenhard et al. (28) for parallel analysis using GRID computation. In essence, this method focuses the search for DNA elements to regions that are conserved between different species, thereby reducing the noise and increasing the probability of finding physiologically relevant transcription factor binding sites. For each gene, genomic sequences containing all the introns plus 7000-bp upstream of the transcription initiation site and 3000 bp downstream of the last exon were extracted from the University of California, Santa Cruz (UCSC) rat genomic database and aligned with the orthologous human sequences using LAGAN, a system for rapid global alignment of two homologous DNAs (29). The most conserved regions were identified and the presence of STAT5b binding sites (consensus 5'TCCNNNGAA3', where N is any nucleotide) was assessed using a weight matrix obtained from JASPAR (30). The frequency of STAT5b sites was calculated for each subgroup of genes (see Table 2 for detailed information on the location of each site) and compared with the background level, calculated from a set of 320 randomly selected genes that displayed no change in mRNA expression in our experiments. As seen in Fig. 2A, we found a total of 6860 different STAT5b binding sites in the 110 genes whose expression was enhanced in livers infected with Ad-STAT5bCA. Of these potential sites, only 238 were located within phylogenetically conserved regions. Thus, 47% of the genes whose mRNA levels are elevated in livers expressing STAT5bCA were found to have conserved STAT5b binding sites, a number that was not statistically different from the 37% of control genes that appeared to contain conserved STAT5b binding sequences despite no effect of GH (Fig. 2, C and D). When additional criteria were added to the response to STAT5bCA, first, of reduced expression in the presence of STAT5bDN, and second, of increased expression by GH in control livers, then the proportion of genes having predicted conserved STAT5b binding sites rose to 57% and 84%, respectively (Fig. 2D). For this latter group, the correlation between GH and STAT5b-dependent gene activation and the presence of phylogenetically conserved putative STAT5b binding sites reached a high level of statistical significance (P < 0.01, 
2 test; Fig. 2D).
|
|
). We began by analyzing whether oligonucleotides corresponding to predicted phylogenetically conserved STAT5b binding sites in SOCS2, Frizzled 4, and emp-1 genes could bind STAT5b using EMSAs. These genes were selected based on their physiological relevance and evidence of GH regulation in independent experiments (33). Five newly predicted binding sites were tested; tandem sites in SOCS2 and Frizzled 4 and a single site in the emp-1 gene. Figure 3 shows results using IR-labeled double stranded oligonucleotide probes for each element and nuclear protein extracts from Cos-7 cells transfected with expression plasmids for the mouse GHR and either wild-type or CA rat STAT5b, and treated with GH or vehicle for 60 min. As depicted in the figure, GH induced binding of wild-type STAT5b to all the probes (Table 3). Both frizzled 4 oligonucleotides and the emp-1 probe rendered a single protein-DNA complex, whereas the SOCS2 probe, which contains tandem STAT5b site, formed two discrete protein-DNA binding complexes. CA active STAT5b was able to bind the oligonucleotides in the absence of hormone. Competition experiments using homologous DNA or the same elements mutated in the putative STAT5b binding sites confirmed specific binding to each protein-DNA complex, whereas antibody supershift experiments showed that STAT5b interacts with all the oligonucleotides. These results show that the predicted phylogenetically conserved elements could bind STAT5b in vitro and potentially could contribute to the regulation of these genes by GH.
|
|
are results of time course experiments using hepatic nuclear protein extracts from GH-deficient rats acutely treated with recombinant rat GH. Inducible binding to the SOCS-2 elements was observed beginning by 30 min after systemic GH injection, and was maximal at this time point. A similar kinetic pattern of inducible protein-DNA complexes was observed for a high-affinity STAT5b site, 5' distal RE-1, that we recently characterized in the rat IGF-I locus (34). As a control, EMSAs were performed with a double-stranded probe that recognizes members of the Sp1 family of transcription factors, which are not regulated by GH (2). As expected, protein-DNA binding was constant over the same time course, demonstrating that both the quantity and quality of nuclear protein extracts were comparable before and after hormone treatment.
|
, lower panel), could displace the interaction of STAT5b with a 32P-labeled DNA probe containing the well-characterized STAT5 binding site (SPIGLE1) from the serine protease inhibitor 2.1 gene promoter (3). As seen in the lower panel of Fig. 4, nuclear proteins from hepatoma BRL-4 cells stimulated by GH were able to bind the canonical STAT5b binding elements found in the serine protease inhibitor 2.1 (SPIGLE1) and the protein-DNA complex was supershifted by an antibody against STAT5b (12). Increasing amounts of the SOCS2 double-stranded oligonucleotide containing the tandem STAT5b sites successfully inhibited this interaction in a dose-dependent manner, whereas mutant oligonucleotides in which both sites were disrupted were ineffective. As assessed using competitor DNAs containing mutations in either STAT5b site, the presence of a single wild-type sequence could still interfere with SPIGLE1 for STAT5b binding, indicating that each of the tandem elements can bind STAT5b in vitro with high affinity.
GH Rapidly Stimulates Binding of STAT5b to Chromatin in the SOCS2 Gene and Induces SOCS2 Gene Transcription
We next used chromatin immunoprecipitation (ChIP) assays to determine whether the predicted tandem STAT5b binding sites in the first intron of the rat SOCS2 gene are capable of binding STAT5b in vivo in a GH-dependent way. As seen in Fig. 5A, little STAT5b could be detected in hepatic chromatin from GH-deficient rats in association with this region of the SOCS2 gene in the absence of hormone. Within 1 h of systemic GH injection, STAT5b was found at this site, and binding persisted for up to 4 h. Similar kinetics of induction were observed on chromatin at the IGF-I HS7 region, which has been shown previously to bind STAT5b after GH stimulation (7), but not within the non-GH-regulated ß-actin gene, or at intron 3 of the rat IGF-I gene.
|
, transcription directed by the SOCS2 promoter was stimulated within 1 h of systemic GH treatment and remained enhanced for up to 6 h. Inducible IGF-I gene transcription also was detected, whereas ß-actin gene expression remained constant. Thus, in vivo GH treatment rapidly stimulates binding of STAT5b to part of the first intron of the rat SOCS2 locus, coincident with potent stimulation of SOCS2 gene transcription.
The SOCS2 Intron Mediates GH-Activated and STAT5b-Dependent Gene Transcription
We subsequently studied the transactivating properties of genomic fragments corresponding to the predicted human SOCS2 promoter and intron 1, when placed in front of a luciferase (LUC) reporter gene (P4, P5, and P7; see Fig. 6A). Recombinant plasmids were transfected into 293T cells, and the effect of GH treatment on the reporter activity was analyzed. As shown in Fig. 6B, GH had no effect on SOCS2 promoter 1 (P1)-LUC fusion genes, but stimulated the activity of reporter plasmids containing segments of intron 1 that included the tandem STAT5b sites up to 5-fold. Similar results were observed with LUC reporter genes that contained the minimal thymidine kinase (TK) promoter and SOCS2 intron fragments P5 and P7. The importance of the two adjacent intronic STAT5b sites for hormonal regulation was next evaluated after engineering point mutations into each sequence in the context of the P5 DNA fragment. As illustrated in Fig. 7, a 4-fold increase in LUC activity was detected after GH stimulation when both sites were present. When the most 5' site was inactivated by a double mutation, the basal level of LUC activity increased slightly, but the fold regulation by GH was reduced. Mutations in the 3' site or in both sites dramatically reduced basal LUC activity and blocked the effects of GH. The tandem STAT5b elements reside within a highly conserved DNA segment that also includes an E-box motif (5'-CACGTG-3') in close proximity (Fig. 7). The introduction of a mutation into this latter site also reduced basal transcriptional activity and abrogated the response to GH, indicating that additional conserved elements within the first intron of the SOCS2 gene contribute to hormonal regulation.
|
|
Viral gene transfer makes it possible to manipulate the levels of transcription factors in vivo. In this study, transcript profiling was used to identify changes in hepatic gene expression after adenoviral gene transfer of DN and CA variants of STAT5b. Among all STAT5b-induced genes identified with this approach, we had the objective to define genes that directly bind STAT5b as a consequence of GH stimulation. We identified novel functional sites through the analysis of phylogenetically conserved STAT5b binding sites in the promoter and intronic regions of several GH-regulated genes. Among them, a tandem STAT5b binding site was found in the first predicted intron of the SOCS2 human gene. GH treatment in vivo combined with ChIP assays demonstrated that STAT5b binds to the tandem region upon hormone stimulation and promotes de novo SOCS2 gene transcription. Reporter gene assays demonstrated that 250 bp of the first intron of the human SOCS2 gene mediated a GH response that could be eliminated by mutating one of the STAT5b binding sites.
Despite the importance of STAT5b in regulating the immune response and the actions of GH, very few STAT5b target genes have been identified. Among known hepatic GH-regulated genes, hormone response elements within IGF-I, ALS, and CIS have been shown to bind STAT5b upon GH treatment in vivo (34). In immune cells, bcl-xL (35), cyclin D1 (36), IL-2 receptor (37), and Ig heavy chain genes (38) are also direct targets of STAT5 transcriptional actions. In a recent survey in BaF3 cells, 17 unique STAT5b binding sites were identified, the majority of which were located within the proximal promoter and adjacent introns (39). Interestingly, little overlap is evident between the known effects of STAT5b in the liver and in B lymphocytes. This highlights the importance of tissue-specific aspects of chromatin accessibility and structure for transcriptional regulation by STAT5b. Consequently, biologically meaningful predictions of functional STAT5b binding sites in gene promoters are challenging and the presence of evolutionary conserved STAT5 binding sites is not sufficient by itself. Added information on actual gene regulation is therefore needed. In the group of genes up-regulated in STAT5bCA expressing livers, we only find a modest over representation of phylogenetically conserved STAT5 binding sites. The additional information of gene regulation by DN STAT5b and after GH induction was essential to define a group of genes where STAT5 binding sites were over represented. In this study we predict that 16 genes are GH-dependent STAT5b targets, and this prediction critically depends on the experimental design, the microarrays used, and the biological functions of the STAT5 mutants used. It is furthermore essential to experimentally verify in silico predictions— previous functional studies support our prediction of STAT5b gene targets in three cases (IGF-I, CIS, and CYP2C12), and the present data add SOCS2 as another direct STAT5b gene target. The remaining genes are, in the absence of additional functional data, to be regarded as putatively regulated by STAT5.
Our analysis failed to reveal a higher frequency of phylogenetically conserved STAT5b binding sites (TTCN3GAA) in noncoding regions of hepatic genes rapidly induced by GH in vivo (see supplemental Table 2). Accordingly, most of the effects of GH observed in control livers were not reversed by overexpression of STAT5bDN-Y699F. This is somewhat surprising given the key physiological role of STAT5b in the growth-promoting actions of GH, and suggests that, in contrast to a model where STAT5b controls many of the transcriptional actions of GH, only a small number of GH-responsive, STAT5b target genes exist in the liver. These include IGF-I, ALS, and SOCS2 (here identified), which are all key regulators of somatic growth (33, 40). Because several signaling networks are rapidly activated by GH in addition to STAT5b, our results indicate that these pathways may be responsible for controlling the majority of GH-regulated genes in the liver.
We have shown previously that in livers expressing STAT5bDN, there is increased activity of STAT1 STAT3, MAPK, and AKT (31, 41). This may explain the enhanced GH responsiveness of several genes including c-fos, c-myc, growth response protein CL-6, and nerve growth factor I-B, among others, compared with controls (see supplemental Tables 2 and 3). Another explanation for the limited inhibitory effects of STAT5bDN on GH-mediated gene activation is that some of the transcriptional actions of STAT5b may not require its direct binding to DNA. For example, STAT5b has been shown to positively cooperate with several transcription factors including CCAAT/enhancer-binding protein ß (42), HNF4 (43), HNF6 (19), high mobility group HMG-I(Y) (44) and octamer-binding transcription factor-1 (38). This possibility seems less likely because these interactions should occur in the nucleus and STAT5bDN is known to reduce both the dimerization and nuclear translocation of endogenous STAT5b. On the other hand, collaborative or synergistic interactions with other nuclear factors cannot be discarded because the inhibitory effects of STAT5bDN may be incomplete. Indirect actions of STAT5b could also explain the fact that we detected almost twice as many GH-repressed as induced genes in both GH-treated livers and in livers infected with Ad-STAT5bCA. STAT5b is known to inhibit the transcriptional activity of several nuclear receptors, including on peroxisome proliferator-activated receptor , and the GR, mineralocorticoid, estrogen, and thyroid hormone receptors (45), which are key regulators of hepatic gene expression. Another possibility is that transcriptional inhibition can be promoted by STAT5b upon binding to DNA within selected genes. This remains to be demonstrated but is plausible given the observation that both histone acetyl transferase and histone deacetylase activities can be recruited to different promoters by STAT5 (46, 47, 48). Indirect actions also may explain the apparently contradictory results regarding the involvement of STAT5b in the regulation of SOCS3 gene transcription by GH. In livers from STAT5b-deficient mice, induction of SOCS3 by GH is impaired (13). In contrast, rapid transcriptional induction of SOCS3 by GH is not blocked by STAT5bDN (34), and STAT5b is not recruited in a GH-dependent way to the GAS elements found in the proximal promoter (14, 34).
We have now demonstrated that STAT5b binds to the first intron of the SOCS2 gene and promotes its transcription in response to GH. The STAT5b binding site identified in this study is within a region containing a tandem repeat of two GAS sequences (5'TTCNNNGAA3') separated by two nucleotides. In addition, we have identified a nearby E-box that is essential for both basal and GH-dependent reporter gene activation. The arrangement of an E-box followed by a tandem repeat of GAS sequences is not only conserved in rodents, humans, and dogs but is also found in the SOCS2 gene promoter of more distant species such as zebra fish. Although EMSAs demonstrated that the tandem GAS sequences are both bona fide STAT5b binding sites, they are unlikely to bind STAT5b simultaneously. Previous studies have shown that, unlike STAT5a, STAT5b is unable to form tetramers when bound to adjacent binding sites that are separated by either three or eight nucleotides, as observed in the CIS gene promoter (32). Indeed, attempts to select STAT5b and STAT5a binding sites from a pool of random double-stranded oligonucleotides indicates that STAT5b favors the formation of dimers on GAS tandem sequences, whereas STAT5a can form dimers or tetramers (49). Our mutational analysis of the two STAT5b binding sites in the first intron of the SOCS2 gene indicates that the more 3' site is required for GH-dependent reporter gene expression. In contrast, mutation of the 5' site appears to increase basal reporter gene expression in the absence of GH, suggesting that it may have an inhibitory role. Whether this latter site binds STAT5b or interacts with other transcription factors remains to be determined. In this context, it is interesting to notice that at least three different protein-DNA complexes are detected in EMSAs using liver nuclear extracts and the oligonucleotide containing the predicted tandem STAT5b binding site. We have not yet identified the transcription factor that binds the phylogenetically conserved E-box in the SOCS2 gene (50). However, it is possible that such a factor is activated by GH because mutation of the E-box eliminates GH-stimulated reporter gene expression. Two potential factors of relevance in this context are c-myc and sterol regulatory element binding protein (SREBP) 1a. Both proteins are known to bind to E-boxes in the promoters of several mitogenic (c-myc) (51) or lipogenic (SREBP1a) (52) genes and are able to stimulate target gene transcription. It is also known that their expression and activities are stimulated by GH (53, 54) (see supplemental Table 2). Interestingly, we find that the expression of several lipogenic genes, such as SCD-2 and FABP5, that are known targets of SREBP1a, is increased in livers infected with Ad-STAT5bCA. Future studies will elucidate whether SREBP1a and STAT5b collaborate in controlling the transcription of these genes.
The existence of a potent negative feedback loop in the regulation of GHR signaling has been long suspected due to the transient nature of its intracellular actions (55, 56). The finding that inhibition of protein synthesis is enough to prolong activation of JAK2 and STAT5 in response to GH demonstrated that this loop requires GH-stimulated de novo synthesized proteins (56). Several hepatic signaling molecules are known to negatively regulate GH action, including the tyrosine phosphatases SHP1 and PTP1b, and three members of the SOCS family, SOCS2, SOCS3, and CIS (57). The latter group has been shown to be transcriptionally induced by GH (58). Analysis of genetically modified mice has demonstrated that inactivation of SOCS2 induces gigantism (59), whereas deficiency of CIS and SOCS3 yield no obvious growth phenotype (60). On the other hand, CIS overexpression reduces the growth rate of mice (60). It thus appears likely that CIS and SOCS2 are responsible for the negative feedback loop acting on the GHR, although their actions are not strictly redundant. Interestingly, the transcription of both SOCS2 and CIS genes is regulated by GH through binding of STAT5b to cis-acting hormone response elements. These findings provide a direct mechanistic link between GHR action and transcriptional induction of negative regulatory molecules. As a consequence, hepatic STAT5b has a dual role in control of somatic growth. It exerts positive effects through the transcriptional induction of genes such as IGF-I and ALS, and also has negative actions through the stimulation of negative regulators such CIS and SOCS2. Although they share a common inductive mechanisms, the overall regulation of CIS and SOCS2 gene expression in response to GH is different, as judged by the different steady-state levels of their respective mRNAs (61). Whereas CIS mRNA induction is rapid and levels are transient, SOCS2 gene expression increases steadily with time (58, 61). These differences probably cannot be explained by STAT5b activation alone, but rather depends in part on differences in mRNA turnover or the actions of additional transcription factors.
In summary, we have demonstrated how analysis of phylogenetically conserved STAT5b binding sites can lead to the identification of new hormone response regions in GH-activated genes. Through this approach we have shown that STAT5b is a key transcription factor for SOCS2 gene expression in the liver, thereby demonstrating the power of the combination of phylogenetic sequence analysis and functional studies to define novel regulatory paradigms.
MATERIALS AND METHODS
Materials
The STAT5b (G-2) and STAT5 (C17X) antibodies were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). The anti-flag (M2) antibody was from Sigma (St. Louis, MO). Recombinant rat GH was from the National Hormone and Pituitary Program, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health (NIH). A collection of 7000 70-oligomer oligonucleotides used for microarray fabrication was obtained from QIAGEN. Oligonucleotides for gel mobility shift assays were HPLC purified and obtained from Thermoelectron (Waltham, MA). Deoxyribonuclease I was from Promega (Madison, WI). All other chemicals were reagent grade and were purchased from commercial suppliers.
Recombinant Plasmids and Viruses
Adenoviruses for EGFP and encoding modified versions of rat STAT5b have been described (7), including STAT5bDN (Y699F) and STAT5bCA (N642H). The GH-responsive plasmid SPIGLE-LUC contains six copies of the GH response element found in the promoter of the Sp2.1 gene preceding the minimal TK promoter (62). The GHR expression vector was a generous gift from Dr. Nils Billestrup and has been previously described (33).
Animal Studies
Animal experiments have been outlined previously (7). Male 7-wk-old hypophysectomized Sprague-Dawley rats were obtained from Harlan Sprague-Dawley (Indianapolis, IN) and were housed at the Oregon Health & Science University (OHSU) Animal Care Facility on a 12-h light, 12-h dark schedule with free access to food and water, according to NIH guidelines. All procedures were preapproved by the OHSU Animal Care and Use Committee. Glucocorticoids (cortisol phosphate, 400 µg/kg·d) and T4 (10 µg/kg·d) were replaced by daily sc injections, and GH deficiency was confirmed by failure to grow during a 2-wk observation period. After this interval, rats were injected via tail vein with 0.2 ml of sterile PBS containing 2 x 1010 plaque-forming units of Ad-EGFP, Ad-STAT5bDN, or Ad-STAT5bCA plus 2 x 109 plaque-forming units of helper virus Ad-tTA. For studies of GH action, rats were injected ip with either vehicle (saline) or 1.5 µg/g recombinant rat GH 48 h after infection. After an incubation period of 30–120 min, the animals were anesthetized with pentobarbital (50 mg/kg ip) and killed. In other experiments, 9-wk-old male hypophysectomized rats were injected ip with either vehicle (saline) or 1.5 µg/g of recombinant rat GH. At 0.5-, 1-, 2-, 4-, or 6-h intervals rats were anesthetized with pentobarbital and killed. Liver nuclear proteins, chromatin, and RNA were isolated as outlined below.
RNA Isolation
Total liver and hepatic nuclear RNA was isolated as described previously (7). RNA concentration was determined spectrophotometrically at 260 nm, and its quality was assessed by agarose gel electrophoresis.
Expression Profiling Using Microarrays
Microarrays containing 70-oligomer oligonucleotide probes for 7000 rat protein-coding genes were fabricated essentially as previously described (63). Oligonucleotides were dissolved in printing solution (Corning, Corning, NY) and spotted twice on a chemically modified GAPS glass slide (Corning) using a Microgrid microarray fabrication robot (Biorobotics, UK). Each of the microarrays used was quality checked by measuring DNA autofluorescence to confirm that all probes were present in the array. Total hepatic RNA (5 µg) was treated with deoxyribonuclease I (Promega) according to the manufacturer’s instructions and was labeled by reverse transcription in the presence of Cy5-CTP using the Pronto labeling kit (Corning). An equal amount of the Universal rat RNA reference (Stratagene, La Jolla, CA) was labeled with Cy3-CTP, mixed with the Cy5-labeled target, purified and mixed with 40 µl of the hybridization buffer supplied by the Pronto System (Corning). The labeled probe mix was added to the array and placed in a sealed hybridization chamber (Corning, NY) for 15–18 h at 65 C, after which the array was washed and dried. The arrays were scanned immediately using a GenePix scanner (Axon Instruments, Union City, CA). Image analysis was performed using the GenePix Pro 6.0 software (Axon Instruments). Automatic flagging was used to localize absent or very weak spots, which were excluded from analysis. Fluorescence (Cy5/Cy3) ratios were normalized as described previously (33) using the LOWESS (Locally Weighted Scatter Plot Smoother) method in the SMA (Statistics for Microarray) package (www.bioconductor.org).
Microarray Data Analysis
Because of the use of a universal RNA reference in the hybridizations, the experimentally obtained and normalized Cy5/Cy3 fluorescent ratios are presented as the expression ratio in relation to the gene expression level in Ad-EGFP-infected control untreated livers. Identification of differentially expressed genes was performed using the SAM statistical technique (64). Given the short time span used in the experiment and the fact that STAT5b DNA binding activity do not differ between different time points, no distinctions were made between the different time points, which were treated as replicates. A q value was assigned for each of the detectable genes in the array. This value is similar to a P value, measuring the lowest false discovery rate at which differential expression of a gene is considered significant. In this study, genes with a false discovery rate of less than 5% were identified as differentially expressed. The raw data and completed list of regulated genes is available as supplemental Table 3. An additional selection requirement was added to this statistically based criterion based on the absolute changes in the gene expression ratios. A value of 1.6 (60%) was chosen to denote differences in expression levels, even though smaller changes in gene expression may have important biological consequences.
Prediction of STAT5b Binding Sites
To analyze the presence of STAT5b binding sites in the promoter regions of the genes identified by the microarray analysis, we created the Prometheus software application. Prometheus was designed to predict aspects of promoter architecture for genes that may be regulated by a common mechanism. It uses weight matrices that describe transcription factor (TF) binding sites (TFBS) to find putative cis elements within genomic regions that are phylogenetically conserved between human and mouse or rat genomes. The sequence alignments were performed using the pairwise aligner LAGAN, which is based on the CHAOS alignment tool (29). Because tests where different types of coregulated genes are analyzed taken together with the many combinatorial aspects of transcription factor binding sites are likely to put a strain on computational power, we applied and tested a parallel approach. Alignment and TFBS searching were processed using the GRID paradigm. Two runtime environments have been set up on NorduGrid to execute the alignment and TFBS finding algorithms: /APPS/BIO/LAGAN-1.2 and /APPS/BIO/TFBS-0.5.0. Both algorithms run in parallel (for each accession number) using NorduGrid as virtual organization. Prometheus has a web interface developed in Perl, but the most demanding tasks are implemented as C extensions to Perl or C++. To obtain the genomic coordinates and sequences, Prometheus uses the GeneLynx Database and the GeneLynx application programming interface (65). The most recent genomic sequences for human, mouse, and rat were updated from the UCSC Genome Informatics data repository and used in the AT platform (65) to produce cross-species alignments of genomic sequences using multiple alignment algorithms. Once a set of lists is processed and TFBS are found, a statistical analysis, based on the 2 test, is provided. This analysis compares the frequency of appearance of specific TFBS in the two sets of genes.
RT-PCR
Total or nuclear RNA (5 µg) was reverse-transcribed in a final volume of 20 µl using the RT-PCR kit (Invitrogen, Carlsbad, CA) with either oligo-deoxythymidine primers (for total RNA) or random hexamers (for nuclear RNA). Each PCR contained 0.5 µl of cDNA. Primer sequences are listed in Table 4. The linear range of product amplification was established in pilot studies for each primer pair, and the cycle number that reflected the approximate midpoint was used in final experiments. This varied from 18–25 cycles for total RNA and from 25–30 cycles for nuclear RNA. Results were quantified by densitometry after electrophoresis through 1.5% agarose gels.
|
(0.2–1 pmol) were 5'-labeled with IRDye 700. Proteins and DNA were incubated for 60 min on ice and the products were separated by electrophoresis through nondenaturing 5% polyacrylamide gels in 1x Tris borate/EDTA [90 mM Tris, 90 mM boric acid, and 2 mM EDTA (pH 8.3)] at 170 V for 25 min at 20 C. Results were analyzed using an Odyssey infrared imaging system and version 1.2 analysis software (LiCoR Biosciences, Lincoln, NM). Antibody supershift and competition studies were as described (2, 35), with either the unlabeled competitive wild-type SOCS-2 and mutant SOCS-2 oligonucleotides listed in Table 3. The double-stranded oligonucleotide SPIGLE1 (0.2–1 pmol) (top strand, 5'-TGTTCTGAGAAATA-3') was labeled with 32P and used as probe in EMSA as previously described (12). In Fig. 4B, competitor DNAs included double-stranded oligonucleotides containing the STAT5 binding sites found within the first SOCS2 intron, with or without specific mutations as indicated. Results were visualized using PhosphorImager scanning (Fuji, Stamford, CT). All experiments were performed in at least three different occasions with comparable results.
ChIP Assays
Initial steps were modified from published protocols, as described previously (7). For each time point, 200 mg of rat liver was minced and incubated at 20 C in 10 ml of DMEM plus 1% formaldehyde on a rotating platform for 15 min, followed by addition of 1.5 ml of 1 M glycine and incubation for an additional 5 min. After centrifugation at 200 x g for 5 min at 20 C, the pellet was suspended in 1 ml of PBS, transferred to a 2 ml Dounce, and homogenized with 10 strokes of a tight fitting pestle (Jencons, West Sussex, UK). After brief centrifugation in a microcentrifuge, the pellet was suspended in 400 µl of lysis buffer [50 mM Tris-HCl, 5 mM EDTA, 1% SDS (pH 8.1) plus protease inhibitors], and incubated for 15 min at 4 C. Each sample was sonicated at 4 C using a total of five pulses for 15 sec each of a Branson microtip sonicator at setting no. 5 interspersed with 30 sec incubations on ice. After centrifugation at 14,000 x g for 10 min at 4 C in a microcentrifuge, the supernatant was diluted to 4 ml with buffer [20 mM Tris-HCl, 150 mM NaCl, 2 mM EDTA (pH 8.1) plus protease inhibitors], and 1-ml aliquots were used for immunoprecipitation, with 3 µl of the STAT5 C17X antibody and 45 µl of a 50% slurry of protein A agarose beads. DNA was extracted using the Qiaquick PCR DNA purification kit, following a protocol from the supplier, and was suspended in 50 µl of 10 mM Tris-HCl, 1 mM EDTA (pH 8.0). PCRs were performed with 1 µl of DNA using the primer pairs listed in Table 4. The linear range of product amplification was established for each primer pair in pilot studies, and the cycle number that reflected the approximate midpoint was used in final experiments. This varied from 28–30 cycles. Results were visualized after electrophoresis through 1.5% agarose gels. All experiments were performed on at least three separate occasions with comparable results.
Cloning of the SOCS2 Promoter Region
Four different DNA fragments were prepared from the human SOCS2 gene by PCR of genomic DNA using specific primers derived from the UCSC Genome Bioinformatics database (http://genome.ucsc.edu/) database. Primers for each DNA segment area as follows: SOCS2 promoter (P1, 3048 bp)—top strand: 5'-CCAACCTCAGCTCTGCTCTC-3', bottom strand: 5'-CTGCTCGAAAGCTTGAAAGG-3'; intron 1 (P4, 2500 bp)—top strand: 5'-GTCTACCGGAGACGTTGGAG-3', bottom strand: 5'-RGTCACCCACTGATCGCCTAC-3'; intron 1 (P5, 696 bp)—top strand: 5'-CCGGGATGTTTAGAGGAACC-3', bottom strand: 5'-CCTCCCCGGTAACTCTCC-3'; intron 1 (P7, 250 bp)—top strand: 5'-GCTCTCACCCGGTCTTCC-3'; bottom strand: 5'-GACGAGACTTGGCAAGAGTT-3'. The DNA sequence of each segment was verified. Specific restriction sites for NheI (New England Biolabs, Beverly, MA) were added to the ends of each DNA fragment, and the products were cloned into both pGL2-Basic Vector (Promega) which lacks a promoter, and pTK-LUC control vector, which contains the thymidine kinase promoter (62).
The SOCS2 P5 DNA fragment was modified by site-directed mutagenesis using the QuikChange kit (Stratagene) to engineer mutations into the E-box element, and one or both STAT5b binding sites (see Fig. 7). All mutations were verified by DNA sequencing.
Cell Culture, Transient Transfections, and Reporter Gene Assays
Cos-7 cells (ATCC CRL-1651) were incubated in antibiotic-free DMEM plus 10% fetal bovine serum, and were cotransfected with expression plasmids for the mouse GHR and wild-type or CA rat STAT5b, using Transit-LT1 and a protocol from the supplier. After incubation for 24 h, serum-free media were added containing 1% BSA for 4 h, followed by addition of recombinant rat GH (40 nM final concentration) or vehicle for 60 min. Cells were then harvested, and nuclear proteins isolated (7). Reporter assays were performed into 293T cells (3 x 105) on 24-well tissue culture dishes with lipofectamine 2000 (Invitrogen), according to the provider’s guidelines. Cells were cotransfected with an expression plasmid for the porcine GHR (100 ng) and promoter-reporter plasmids for ß-galactosidase (100 ng), and either SPIGLE-LUC (100 ng), or the different SOCS2 gene constructs indicated in Figs. 6 and 7 (400 ng). Cells were incubated for 24 h, followed by washing with PBS, and incubation in serum-free DMEM for 2–4 h before human GH (40 nM final concentration) was added. Sixteen hours later, cells were harvested and cell lysates used for LUC assays. All results were normalized to ß-galactosidase activity.
|
This work has been supported by grants to A.F.-M. from the Swedish Research Council, Wallenberg Foundation and the Swedish Institutet. L.F. was supported by Grants from Ministerio de Sanidad y Consumo (FIS 1/1000 to L.F.) and Ministerio de Ciencia y Tecnología (PETRI1995-0711 and SAF2003-02117). P.R. was supported by National Institutes of Health (NIH) Research Grant 1RO1 DK069703, and D.J.C. was supported by NIH Training Grant F32 DK070447.
Disclosure Statement: The authors have nothing to disclose.
First Published Online September 28, 2006
Abbreviations: Ad, Adenovirus; ALS, acid-labile subunit; CA, constitutively active; ChIP, chromatin immunoprecipitation; CIS, cytokine-induced SH2 protein; DN, dominant-negative; EGFP, enhanced green fluorescent protein; emp-1, epithelial membrane protein-1; GAS, -interferon activated sequences; GHR, GH receptor; GR, glucocorticoid receptor; HNF, hepatic nuclear factor; JAK, Janus kinase; LUC, luciferase; SOCS2, suppressor of cytokine signaling 2; SREBP, sterol regulatory element binding protein; STAT5b, signal transducer and activator of transcription 5b; TF, transcription factor; TFBS, TF binding sites; TK, thymidine kinase.
Received for publication February 27, 2006. Accepted for publication September 19, 2006.
REFERENCES
is a component of the growth hormone-regulated network of liver transcription factors. Endocrinology 141:1686–1692 contribute to growth hormone-regulated transcription of c-fos. J Biol Chem 274:31597–31604. FEBS Lett 444:155–159[CrossRef][Medline] (PPAR) signaling pathways. Growth hormone inhibition of pparalpha transcriptional activity mediated by stat5b. J Biol Chem 274:2672–2681. Mol Endocrinol 20:647–660 chain gene regulation: importance of two widely separated IL-2 response elements. Immunity 15:159–172[CrossRef][Medline]-helical region of the STAT5 transactivation domain. J Biol Chem 278:45340–45351
