ABSTRACT
The ERK5 mitogen-activated protein kinase (MAPK) differs from other MAPKs in possessing a potent transcriptional activation domain. ERK5 −/− embryos die from angiogenic defects, but the precise physiological role of ERK5 remains poorly understood. To elucidate molecular functions of ERK5 in the development of vasculature and other tissues, we performed gene profile analyses of erk5 −/− mouse embryos and erk5 −/− fibroblast cells reconstituted with ERK5 or ERK5(1-740), which lacks the transactivation domain. These experiments revealed several potential ERK5 target genes, including a proapoptotic gene bnip3, known angiogenic genes flt1 and lklf (lung Krüppel-like factor), and genes that regulate cardiovascular development. Among these, LKLF, known for its roles in angiogenesis, T-cell quiescence, and survival, was found to be absolutely dependent on ERK5 for expression in endothelial and T cells. We show that ERK5 drives lklf transcription by activating MEF2 transcription factors. Expression of erk5 short hairpin or a dominant-negative form of the ERK5 upstream activator, MEK5, in T cells led to downregulation of LKLF, increased cell size and upregulation of activation markers. Thus, through its kinase and transcriptional activation domains, ERK5 regulates transcriptional responses of cell survival and quiescence critical for angiogenesis and T-cell function.
Cellular differentiation programs are regulated through molecular mechanisms that guide signals generated in response to external stimuli, to ultimately induce changes at the gene level. The members of the mitogen-activated protein kinase (MAPK) family regulate a wide range of cellular responses. The MAPK pathways are induced via sequential phosphorylation and activation of a canonical three-kinase cascade, MAPK kinase kinase (MAPKKK)-MAPK kinase (MAPKK)-MAPK (28, 57). Conventionally, activated MAPKs catalyze phosphorylation of substrates at serine or threonine residues N terminally adjacent to a proline (Pro at +1 position), a change that induces a variety of responses including gene transcription, apoptosis, and proliferation. We previously described an alternative mode of activation utilized by the ERK5 MAPK. ERK5 possesses a unique transcriptional coactivator domain, which mediates protein-protein interactions with the myocyte enhancer factor 2 (MEF2) transcription factors and provides a potent coactivator function toward MEF2-driven transcription (22). In immature T lymphocytes, activation of ERK5 induces immediate-early transcription of the nur77 orphan steroid receptor gene via MEF2 proteins, an event that ultimately triggers apoptosis (6, 53, 54). Interestingly, ectopic expression of the C-terminal coactivator domain of ERK5 is sufficient on its own to induce MEF2-dependent transcription (22), demonstrating that this domain represents a functionally independent module. Other studies have shown that ERK5 also activates MEF2 proteins by the conventional mechanism of phosphorylation (23, 24, 33, 56), indicating that ERK5 may utilize diverse mechanisms to affect its downstream targets. Thus, the ability of ERK5 to both phosphorylate substrates and directly interact with and coactivate transcription factors point to unique functions of ERK5 in transcriptional regulation.
Members of the MEF2 family of transcription factors (MEF2A, -B, -C, and -D in vertebrates) regulate the expression of multiple genes critical for myogenic differentiation (34). Targeted disruption of the mef2c gene results in embryonic lethality with defects in early cardiomorphogenesis and angiogenesis (1a, 29, 30), while deficiency of mef2a gene leads to sudden heart failure after birth (37, 51). The MAPKs ERK5 and p38 increase transcriptional activity of the MEF2 proteins through phosphorylation (34) and, uniquely in the case of ERK5 as discussed above, also through tethering itself as a powerful transcriptional coactivator. In addition, activation of the MEF2 proteins requires Ca2+- and calmodulin-dependent kinase (CaMK) to disrupt their association with repressors such as class II histone deacetylases (HDACs; HDAC-4, -5, -7, and -9) (34) and to enhance interactions with histone acetyltransferase (HAT) proteins, p300, and PCAF (34). Heterodimerization with other transcription factors also guides the activities of the MEF2 proteins to specific target genes and enhances their transactivational capacity. For example, MEF2 heterodimerizes with members of the MyoD family of transcription factors to bind and activate cis-regulatory regions found in many muscle differentiation genes (2).
Mechanisms of blood vessel stabilization include recruitment and differentiation of perivascular cells such as vascular smooth muscle cells and fibroblasts, enhancement of cell-cell adhesion to strengthen structural integrity of the endothelium, and optimization of endothelial cell responses to prolong their survival (7, 35). We and others previously reported that disruption of the gene encoding ERK5 led to angiogenic defects and embryonic lethality around day 10 or 11 of gestation in mice (14, 41, 45, 55). Deficiency of ERK5 appears to affect the integrity of the endothelium and correlates with decreased proliferation and viability of endothelial cells (14). However, the requirements for ERK5 are not likely to be limited to endothelial cell responses. We showed that erk5 −/− embryos expressed increased levels of vascular endothelial growth factor (VEGF) (45), which is produced by multiple cell types and is known to play important roles in angiogenesis (11). Overexpression of ERK5 inhibited the activity of the vegf promoter, suggesting that ERK5 acts as a transcriptional repressor toward the vegf gene (45). Although the molecular mechanism of transcriptional repression by ERK5 has not been resolved, these data indicate that ERK5 may regulate transcription both positively (as with nur77) and negatively (vegf).
To further elucidate how ERK5 regulates angiogenesis during development and to identify target genes for ERK5-mediated responses, we performed microarray analyses to compare gene profiles between erk5 +/− and erk5 −/− mouse embryos and the corresponding mouse embryonic fibroblast (MEF) cells. We analyzed gene expression profiles in cells that were cultured under normoxic or hypoxic conditions to assess contributions of ERK5 in cellular responses to hypoxic stress, an important stimulus for angiogenesis. In addition, we utilized ERK5 reconstitution in MEF cells to identify genes that are regulated in an ERK5-dependent manner. Our study indicates that ERK5 is involved in the regulation of several genes, including those that have previously been implicated in angiogenesis, apoptosis, and cell survival. We show that a gene known to play an indispensable role in blood vessel stabilization and T-cell quiescence, the lung Krüppel-like factor (LKLF or KLF2), is a primary and direct target of ERK5 and MEF2. We also provide evidence that ERK5-dependent regulation of LKLF has functional importance in T-cell activation. Our study thus indicates that ERK5, a broadly expressed MAPK, plays a critical role in the regulation of tissue-restricted developmental genes, supporting its physiologic requirement in multiple differentiation programs, including angiogenesis and hematopoiesis.
MATERIALS AND METHODS
Microarray analyses. (i) Sample preparation.Individual embryos were genotyped first and pooled. Total RNA was isolated from whole embryos at embryonic day 9.5 (E9.5) by LiCl method. Approximately 5 μg of total RNA was amplified in vitro, first converting to cDNA using an oligo(dT)24 primer containing the T7 recognition site and SuperScript II reverse transcriptase (Invitrogen), followed by in vitro transcription with T7 RNA polymerase using a kit (Ambion). Then, 5 μg of amplified RNA was labeled and hybridized as previously described (19). Analyses with MEFs were performed similarly, with the difference that total RNA was isolated by using TRIzol (Invitrogen) according to manufacturer's recommended protocol. These RNA samples were not amplified in vitro but were directly labeled and used for hybridization.
(ii) Statistical analyses.Statistical analyses were performed as described previously (19). Each group of comparisons was performed in three independent experiments, and the array contained duplicate spots. The quality of individual spots was evaluated through the SPOT program to exclude data that failed to meet the criteria predetermined by the program. Subsequently, the data were analyzed by using the statistical analysis of microarrays (SAM) (48), and a list of significantly altered genes was generated based on the average fold change (mean log2 ratio for individual genes). A gene was scored as being significantly altered if the d value and the average fold change were ≥2. Genes that were scored as significant in any group were compiled into a list and analyzed through a hierarchical clustering algorithm (cluster analyses) using CLUSTER 2.12 (8) and visualized with TREEVIEW to establish patterns of genes that are similarly regulated.
Quantitative reverse transcription-PCR (RT-PCR).Total RNA was isolated as described previously (17). All RNA samples were treated with DNase I (20 min in 10 mM MgCl2, 1 mM dithiothreitol, 0.5× Tris-EDTA; 0.01 U/μl), phenol-chloroform extracted, and ethanol precipitated. One microgram of total RNA was reverse transcribed with oligo(dT)15 primers (Promega) using SuperScript reverse transcriptase (Invitrogen). SYBR green PCR was performed with 20 ng of cDNA template using commercial kits (Applied Biosystems and Bio-Rad) and GeneAmp 5700 system (Applied Biosystems). Each sample was analyzed in triplicates, and the amounts of templates were estimated by linear regression against the known standard and normalized to internal control. The primer sequences for the PCR were as follows: LKLF forward, (5′-CGC CAC TAC CGA AAG CAC) and reverse (5′-CGC ACA AGT GGC ACT GAA AG), BNIP3 forward (5′-AGG ATT CTC GCC TTG CTG TC) and reverse (5′-GCA ACA AAA CTG ACC ACC CAA), IGF-BP3 forward (5′-CAA GTT CCA TCC ATC CAT G) and reverse (5′-CCG TGG CCT TTT TTG ATG AC), GATA-4 forward (5′-TGT GTA GCA GGC AGA AAG CAA) and reverse (5′-GAT CAC CCA CCG GCT AAA GA), RORα forward (5′-AGG CTC GCT AGA GGT GGT GTT) and reverse (5′-TGA GAG TCA AAG GCA CGG C), and γ-actin forward (5′-GCA CCT AGC ACG ATG AAG ATT AAG) and reverse (5′-GCC ACC GAT CCA GAC TGA GT).
In situ hybridization.Whole-mount in situ hybridization of mouse embryos was performed as described previously (15). Briefly, embryos fixed in 4% paraformaldehyde were permeabilized in proteinase K solution (10 μg/ml), hybridized with digoxigenin (DIG)-labeled BNIP or LKLF riboprobes, generated by using the cDNA clones (a 700-bp BNIP3 and a 500-bp LKLF cDNA fragments amplified by PCR) from the microarray library. After being washed, the samples were incubated with alkaline phosphatase-conjugated anti-DIG antibody (Roche), washed again, and visualized by reaction with BM Purple substrate (Roche). The photographs were taken by using Nikon camera attached to Zeiss microscope. For section in situ hybridization, E9.5 embryos were harvested and fixed in fixation buffer (ethanol-acetic acid-paraformaldehyde), dehydrated, embedded in paraffin blocks, and sectioned at 5-μm thickness onto Superfrost plus slides. These slides were rehydrated as per standard protocol and hybridized with DIG-labeled lklf riboprobe overnight at 68°C in hybridization mix (1× salt, 50% fomamide, 10% dextran sulfate, yeast RNA at 1 mg/ml, 1× Denhardt). After washing, a tyramide signal amplification process was applied by using a commercial kit (Perkin-Elmer), followed by chromogenic horseradish peroxidase substrate reaction with diaminobenzidene, counterstained with Hematoxylin QS (Vector Laboratories), and mounted in 80% glycerol-phosphate-buffered saline.
Cloning, transient transfection, and reporter assay.The wild-type and mutant LKLF reporters were constructed with oligonucleotides representing the 39-bp sequences derived from the region from −141 to −103 of the lklf promoter. 5′ overhang sequences representing XhoI and SalI restriction sites were added for insertion of the annealed oligonucleotides into the SalI site in the Δ56luc vector (54). The mutations represent those that were described previously (44). The sequences for these oligonucleotides are as follows: wild type (WT), 5′-(TCGA) CCG CCA GGC TTA TAT ACC GCG GCT AAA TTT AGG CTG AGC-3′; A mutant, 5′-(TCGA) CCG CCA GGT CCG TTA ACC GCG GCT AAA TTT AGG CTG AGC-3′; and B mutant, 5′-(TCGA) CCG CCA GGC TTA TAT ACC GCG GCT GGG CTT AGG CTG AGC-3′.
Each construct was verified by DNA sequencing, and those containing one copy of the response element (annealed oligonucleotides) were chosen for further analyses.
The ERK5 full-length and 1-740 retrovirus constructs were made by shuttling the NheI(blunted)-NotI cDNA fragments from the previously described pCI plasmid constructs (22) into pBS-IRES and then cloning the NotI fragments into the NotI site of the MSCV-puro vector. The Mek5(A) and Mek5(D) retroviral constructs were generated by PCR amplification and cloned into the pCMV internal ribosome entry site green fluorescent protein (GFP) retrovirus packaging vector and verified by DNA microsequencing. The erk5 shRNA retroviral construct was generated by inserting a PCR-amplified U6 promoter-erk5 (sense-linker-antisense) PCR product into the HindIII site in the Banshee retrovirus vector, as previously described (16). The erk5 target sequence was 5′-ATTGTGGCTGAAATTGAGGACTT-3′ and was kindly provided by Michael Cooke at Genomics Institute of the Novartis Research Foundation.
The MEF cells were transiently transfected with the maximum of 7 μg of plasmid DNA using LT1 transfection reagent (Mirus) according to the manufacturer's recommendations. The cells were harvested 24 h posttransfection, and reporter activities were measured by using the Dual-Luciferase Reporter Assay Systems kit (Promega) in Monolight luminometer. Where indicated, transfected cells were incubated in a modular chamber flushed with hypoxic gas mixture containing 1% O2 for additional 16 h before harvesting. For normalization, a Renilla luciferase construct was cotransfected.
EMSA, ChIP, and flow cytometric assay.Electrophoretic mobility shift assay (EMSA) was performed essentially as described previously (21, 43). Anti-TEF1 antibodies were purchased from BD Pharmingen, and anti-pan MEF2 antibodies were from Santa Cruz Biotechnology, Inc. Chromatin immunoprecipitation (ChIP) assays were performed as previously described with minor modifications (10). Five million erk5 +/− MEF cells were used per immunoprecipitation with 3 μg of anti-HA (mock) or anti-MEF2D antibody. The primers used were lklf −188F (5′-CTT GAG GAG CGC AGT CCG GGC), lklf −56R (5′-CTA GGA GGC GTC GAC GGA AAC), aprt F (5′-TGC TAG ACC AAC CCG CAC CCA GAA G), and aprt R (5′-TCG TGA CCG CAC CTG AAC AGC AC). Flow cytometric analysis was performed as described previously (18). Phycoerythrin-conjugated anti-CD62L and anti-CD25 antibodies were purchased from BD Pharmingen.
Retroviral transduction.The retrovirus plasmids were individually transfected into HEK293T cells by using the LT1 transfection reagent, together with the helper virus construct, pCL-ECO (36). At 48 to 72 h posttransfection, the viral supernatants were harvested and used to infect DO11.10 hybridoma cells or previously activated primary lymph node T cells. DO11.10 cells were infected by culturing the cells in viral supernatant containing Polybrene (8 μg/ml) for 24 h. Primary lymph node cells were infected according to the protocol described previously (50). Briefly, purified lymph node T cells (R&D T-cell enrichment columns) from C56BL/6 mice were activated in vitro by culturing in the presence of plate-bound anti-CD3 and anti-CD28 antibodies for 48 h. The media were replaced with retroviral supernatant cocktail containing Polybrene (8 μg/ml) and HEPES (1 mM) and spin infected at room temperature for 90 min. After 2 days in culture, live cells were purified by using Lympholyte-M (Cedarlane) and analyzed by flow cytometry or cultured further.
T-cell stimulation and immunoblotting.Primary lymph node T cells from C57BL/6 mice were preactivated with anti-CD3 and anti-CD28 as described above. After Lympholyte-M enrichment, cells were stimulated with mouse recombinant interleukin-7 (IL-7; R&D) at 20 ng/ml or plate-bound anti-CD3 and anti-CD28 and cultured at 37°C. At indicated time points, cells were harvested and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis as previously described (45). The anti-phosphoERK5 antibody was purchased from Biosource.
RESULTS
Microarray analyses and identification of transcriptional targets of ERK5.The angiogenic phenotype associated with ERK5 deficiency in mouse embryos is first manifested at day 9.5 to 10 of gestation and is characterized by the failure of proper blood vessel remodeling and stabilization (41, 45, 55). Previously, we showed that the levels of VEGF and its receptor Flt1 are elevated in ERK5 mutant embryos (45), providing a potential mechanism for the defects in vascular organization and maturation.
However, it is likely that ERK5 regulates transcription of other genes involved in angiogenic and/or hypoxic responses. We thus sought to identify additional transcriptional targets of ERK5. To this end, we performed microarray analyses using our UCB/UCSF custom cDNA arrays, which represent approximately 2,700 verified cDNA clones of mostly known genes (19, 40). By comparing gene expression profiles of ERK5-proficient and ERK5-deficient embryos at E9.5, we identified a number of genes that were altered at the earliest time point where erk5 −/− embryos showed developmental defects. In addition, we compared the gene profiles of hypoxia-treated erk5 +/− and erk5 −/− MEF cells with those grown under normoxic conditions, to define ERK5-regulated genes induced by hypoxia. Use of the cell culture system also allowed us to address whether ERK5 directly regulates gene expression in response to hypoxia versus induction of hypoxia-responsive genes in erk5 −/− embryos as a secondary effect of other developmental defects. Hybridizations were performed at least three times for each set, and significantly induced or repressed genes were analyzed by SAM statistical analyses software and then by hierarchical clustering algorithm (19). As shown in Fig. 1, we observed that many of the genes elevated in erk5 −/− embryos (in red) were also induced by hypoxia treatment in erk5 +/− MEF cells. However, the inducibility of these hypoxia-responsive genes was greater in erk5 −/− MEF cells compared to erk5 +/− cells. In contrast, none of the genes that were reduced in erk5 −/− embryos (in green) was hypoxia-responsive genes. Thus, these genes may be negatively regulated by ERK5.
Microarray analysis of gene expression profiles in erk5+ and erk5− mouse embryos and MEF cells. Gene expression profiles were compared between erk5+ (erk5 +/+ and +/−) and erk5 −/ − embryos at E9.5 or between erk5 +/ − and erk5 −/ − MEF cells that were cultured under normoxic (20%) or hypoxic (1% oxygen) condition for 8 h. Hybridization scheme was designed to identify genes that were differentially regulated in erk5 −/ − compared to erk5 +/ − cells under hypoxic or normoxic conditions (erk5 +/ −; reference, erk5 −/ −; test), erk5 −/ − compared to erk5 +/ − E9.5 embryos (erk5 +/ −; reference, erk5 −/ −; test), and hypoxia-inducible response of erk5 −/ − or erk5 +/ − cells (normoxia; reference, hypoxia; test). Genes determined to be significantly altered in any hybridization pair were clustered by using CLUSTER 2.12. Rows correspond to individual genes and columns to hybridization pairs. The tree diagram on the left indicates groups of genes that are similarly regulated. Several groups of genes potentially regulated in an ERK5-dependent manner are also indicated. These include hypoxia-inducible genes, genes that are elevated in erk5 −/ − embryos, genes that are derepressed in response to hypoxia in erk5 −/ − MEF cells, and genes that are repressed in erk5 −/ − embryos, as shown on the right.
To identify genes that are directly regulated by ERK5, we generated cell lines from erk5 −/− MEF cells where ERK5 expression was restored by retroviral transduction of wild-type or ERK5(1-740) construct. The latter lacks the C-terminal transactivation domain and would allow us to identify genes directly regulated by the ERK5 coactivator domain. Immunoblot analyses indicated that the relative levels of ERK5 and ERK5(1-740) expressed in these cells were comparable, although both expressed ERK5 proteins at slightly reduced levels compared to the endogenous ERK5 expression in erk5 +/− cells (Fig. 2A). To further exclude genes whose expression was affected by the retrovirus, cells transduced with empty virus were used as control in these analyses. Cluster analyses defined groups of genes that were repressed (in green) or induced (in red) in empty virus-infected erk5 −/− cells, compared to cells reconstituted to express ERK5 (Fig. 2B, middle two columns). These genes are thus induced or repressed, respectively, in an ERK5-dependent manner. Most genes that respond to restored ERK5 expression do so without correlation to relative oxygen tension. Significantly, many of the genes that were induced or repressed upon ERK5 reconstitution (Fig. 2B, middle two columns) did not score as significantly induced or repressed in erk5 +/− compared to erk5 −/−, MEF cells (Fig. 2B, first two columns). Therefore, expression of only a few genes strictly correlated with ERK5 expression (labeled as ERK5-dependent induction I and II and ERK5-dependent repression I and II), and these genes are likely to represent the true transcriptional targets of the ERK5-dependent regulation. These data underscore the general importance of profile comparison between multiple conditions and suggest that wild-type MEF cells, which arise independently of mutant MEF cells, may not be the best control for the mutant cells. Use of a reconstitution system is thus necessary to properly correlate a mutation with its phenotypic consequences.
Microarray analysis of MEF cells reconstituted with ERK5. (A) Immunoblotting analysis showing ERK5 expression in reconstituted MEF cells. erk5 −/ − MEF cells were transduced with empty virus (lane 1), virus encoding full-length ERK5 (lane 2) or truncated ERK5 [ERK5(1-740), lane 3]. erk5 +/ − MEF cells serve as control of endogenous ERK5 expression (lane 4) and erk5 −/− MEF cells as negative control (lane 5). The lower nonspecific (NS) band serves as the loading control. (B) To identify genes that strictly correlate with ERK5 expression, gene expression profiles were compared between erk5 −/− MEF cells (empty) and erk5 −/− MEF cells reconstituted with full-length ERK5, cultured under normoxic or hypoxic condition as described in Fig. 1. As control, erk5 +/− and erk5 −/− MEF cells were included. Genes that are regulated in a similar manner in both comparisons are marked as strictly dependent on ERK5. Enhanced expression in erk5 −/− MEF cells or vector-reconstituted cells indicates ERK5-dependent repression, and conversely, reduced expression indicates ERK5-dependent induction (I and II). To determine the requirement for the C-terminal transactivation domain, gene profiles of cells expressing ERK5 (1-740) truncation mutant and those expressing full-length ERK5 were also compared.
The transcriptional coactivator function of ERK5 is modulated through its C-terminal domain. To assess the requirement for ERK5 C-terminal transactivation domain for transcription, we compared the gene profiles of cells expressing the full-length ERK5 and those expressing ERK5(1-740) truncation mutant. As shown in Fig. 2B (last two columns), ERK5(1-740) evoked a similar gene profile as the full-length ERK5. This pattern indicates that the transactivation domain is not required for hypoxia-inducible responses in MEF cells. Importantly, we did not observe any genes that were induced by ERK5(1-740) compared to full-length ERK5, suggesting that the ERK5 C terminus acts primarily as a transcriptional activator and not a repressor domain.
Altered gene expression associated with ERK5 deficiency in mouse embryos.Based on the microarray analysis above, only a few genes correlate with ERK5 expression in both embryos and fibroblast cells. LKLF (lung-Krüppel-like factor) stood out as one of the most significantly repressed genes in the absence of ERK5. We evaluated expression of LKLF and several other genes that were identified through our microarray analyses of embryos. These include bnip3 (a BH3 only proapoptotic member of the Bcl-2 family), flt1, rorα, and gata4 (genes implicated in cardiovascular development) and igf-bp3 (insulin-like growth factor-binding protein 3). As shown in Fig. 3, quantitative RT-PCR analysis confirmed that erk5 −/− embryos expressed significantly increased levels of bnip3 (>6-fold), rorα (3-fold), and igf-bp3 (2-fold), and significantly decreased level of lklf (>5-fold) compared to control. No significant changes were detected for GATA-4. Increased flt1 (vegfr-1) expression in erk5 −/− embryos was also seen (data not shown) and is consistent with our previously published findings (45). Vegf-a, which we previously demonstrated to be elevated in expression in ERK5 mutant embryos, was not represented in our custom array.
Quantitative RT-PCR analyses. Relative expression of select genes in erk5 + and erk5 − embryos at E9.5 was analyzed. Total RNA isolated from whole embryos was converted to cDNA and analyzed by SYBR green RT-PCR. The graph represents relative levels of transcripts for individual genes in erk5 − samples expressed as fold over erk5 + control value, shown arbitrarily as 1 (indicated with the dark line). For each gene, the experiments were performed with at least three independent sets of samples, and the average values were used to calculate the final fold over control values.
ERK5-mediated repression of bnip3, a proapoptotic gene.The genes that were most significantly affected by loss of ERK5, bnip3 and lklf, were further analyzed to define potential requirements for ERK5 in their regulation. Consistent with the microarray data, the levels of bnip3 transcripts were indeed elevated dramatically in erk5 −/− embryos compared to control at E9.5 (Fig. 4A). However, at E9.0 and earlier, bnip3 expression was normal. This pattern of expression suggests that BNIP3 is not directly responsible for the angiogenic defects in erk5 −/− embryos, since changes in the expression of BNIP3 coincide with the emergence of the angiogenic phenotype at around E9.5. Whole-mount in situ hybridization analyses further indicated that bnip3 expression was detectable only at low levels in control embryos at E9.5 (Fig. 4B), with expression being restricted mostly to the myocardium (in situ hybridization of embryo sections [data not shown]). However, in erk5 −/− embryos the expression of bnip3 was dramatically increased and, moreover, the pattern of expression was less restricted (Fig. 4B), suggesting that loss of ERK5 increases bnip3 expression systemically.
ERK5 negatively regulates the bnip3 gene. (A) Quantitative RT-PCR analysis of erk5 +/− and erk5 −/− embryos at E9.5 and E9.0 for bnip3 expression. Total RNA derived from whole embryos was converted to cDNA and analyzed by SYBR green RT-PCR with bnip3-specific primers. Relative values for bnip3 transcript levels were normalized against γ-actin and expressed as the average of triplicate samples. The error bars represent standard deviations. The graph shown is representative of three independent sets of samples for E9.5 and two sets for E9.0. (B) Whole-mount in situ hybridization analyses of erk5 +/− and erk5 −/− embryos at E9.5. A bnip3-specific riboprobe was used to detect bnip3 transcripts in E9.5 embryos. Bnip3 expression levels are elevated in erk5 −/− embryos compared to erk5 +/− embryo. No difference was observed between erk5 +/− and erk5 +/+ embryos (not shown). (C) Bnip3 reporter assay. The bnip3 promoter activity was measured by transiently transfecting erk5 +/− or erk5 −/− MEF cells with the bnip3 reporter construct and measuring the responses in cells cultured under normoxic or hypoxic condition for 16 h posttransfection. Where indicated, plasmids expressing the subunits of the HIF1 transcription complex (HIF1), HIF1α and ARNT, were also cotransfected.
BNIP3 is a BH3 domain-containing proapoptotic molecule, whose expression is inducible by hypoxia (3, 12, 46). Thus, its upregulation in the absence of ERK5 could be secondary to hypoxic stresses brought on by the angiogenic defects in ERK5 mutant embryos. Alternatively, ERK5 could directly repress transcription of the bnip3 gene. To differentiate between these two models, we measured the response of the bnip3 promoter in MEF cells using reporter assays. The bnip3 promoter was responsive to hypoxia in the absence of ERK5, and this response was further enhanced with enforced expression of HIF-1α and ARNT, the components of the HIF1 transcription complex that has been shown to activate bnip3 transcription (3). Strikingly, the hypoxia-inducible responses of the bnip3 reporter in erk5 −/− MEFs were greater than those in erk5 +/− control cells. This was most clearly indicated by the increased fold induction after hypoxia treatment, although the basal response under normoxic conditions was also somewhat elevated in erk5 −/− compared to erk5 +/− cells (Fig. 4C). Together, these data support the notion that ERK5 negatively regulates hypoxia-inducible transcription of the bnip3 gene.
LKLF is a primary target of ERK5.Since angiogenic defects are apparent by E9.5 in the absence of ERK5, we hypothesized that the targets of ERK5 that contribute to angiogenesis would be affected at developmental stages preceding E9.5 before the defects arise initially. The lung Krüppel-like factor (LKLF and KLF2) is one gene that fits these criteria. Quantitative RT-PCR analyses showed that the levels of lklf transcripts were dramatically reduced in erk5 −/− embryos at E9.5, E9.0, and as early as E8.5 (Fig. 5A). These results were also confirmed by in situ hybridization. At E9.5 in control embryos, expression of lklf was observed in vascular structures, most prominently the endocardium and intersomitic vessels (Fig. 5B). In contrast, the erk5 −/− mutants completely lacked vasculature-associated expression of lklf, although some residual staining was detected in the head. At E9.0 lklf expression was the highest in the endocardium and the umbilical veins in the erk5 +/− controls but entirely absent from the erk5 −/− embryos. To further confirm the disappearance of LKLF, paraffin-embedded sections of E9.5 ERK5-deficient and control embryos were analyzed by in situ hybridization. Expression of LKLF, most strongly associated with endothelial cells lining the blood vessels (Fig. 5Ca and b) and the endocardium (Fig.5Cc) in the control embryos, were undetectable in ERK5-deficient embryos (Fig. 5Cd to f). These data strongly imply that ERK5 is absolutely required for the expression of LKLF and that the lklf gene serves as a primary target for ERK5-dependent regulation in cardiovascular tissues. Indeed, disruption of the lklf gene has been shown to result in angiogenic defects and embryonic lethality at E12.5, supporting its role in blood vessel stabilization during embryonic development (26, 52). Thus, transcriptional regulation of LKLF by ERK5 provides one plausible link between ERK5 and angiogenesis, although other genes (such as the previously identified target VEGF) likely play a role as well. Importantly, our analysis demonstrated that lklf is not hypoxia-inducible, since its expression level did not change after the hypoxia treatment in erk5 +/− MEF cells (Fig. 1).
ERK5-deficient embryos lack expression of LKLF. (A) Quantitative RT-PCR was performed to compare relative abundance of lklf transcripts in erk5 +/− and erk5 −/− embryos at E9.5, E9.0, and E8.5. The values represent the average of triplicate samples, normalized to γ-actin. The error bars indicate standard deviation. The data shown is representative of five (E9.5) and three (E9.0 and E8.5) independent experiments. (B) In situ hybridization was performed to assess expression of lklf in E9.5 (top panels) and E9.0 (bottom panels) erk5 +/− (left) and erk5 −/− (right) embryos. At E9.5, lklf expression associated with the endocardium (arrow) and intersomitic vessels (arrowheads) in the control embryo is absent from the mutant embryo. At E9.0, LKLF expression is detectable in the endocardium and umbilical veins (arrowhead) in the control embryos but is lacking in the mutant embryos. (C) Expression of lklf in transverse sections of E9.5 control (a to c) and erk5 −/− (d to f) embryos was analyzed by in situ hybridization. LKLF is detected in endothelial cells (arrows) of major blood vessels (a and d; magnification, ×40), in smaller developing vessels (b and e; magnification, ×20), and in the endocardium of the embryonic heart (c and f; magnification, ×20). En, endocardium.
To define ERK5-dependent regulation of the lklf gene, we analyzed the DNA sequences of the lklf promoter region. Alignment of ∼2-kb DNA sequences of the mouse and the human lklf promoter revealed a region of near identity between −135 and −70 in the mouse sequence (Fig. 6A). No additional sequence of similarity could be detected further upstream. A previous study of the mouse lklf promoter had also indicated that transcription from the lklf promoter required DNA sequences between −243 and −72 (44). This activity appeared to be regulated through two AT-rich sequence motifs contained between −141 and −103, although proteins that bind to these regions were not identified. When we analyzed this sequence, we identified one perfect consensus motif for MEF2-binding site at −119 to −110 of the mouse lklf promoter (B element), which was conserved in the human sequence (Fig. 6A). The adjacent 5′ AT-rich motif (A element) bears homology to binding sites for a variety of transcription factors, including TEF-1, a protein that has been shown to interact with MEF2 and regulates gene expression in muscle cells (20, 32) or possibly even MEF2 (25).
Regulation of the lklf gene by ERK5 and MEF2. (A) Comparison of the DNA sequences for the mouse and human lklf promoter region. DNA sequence alignment indicates the mouse (top) and human (bottom) lklf promoter sequences share one region of homology. Sequence identity is marked by “.” in the human sequence. The oligonucleotide sequences (−141/−103) used to generate the reporter construct is marked, along with the consensus motifs (A and B) representing potential transcription factor binding sites. The boxes indicate nucleotide residues that are mutated in the mutant reporter constructs. In the A mutant CTTA was changed to TCCG, and in the B mutant AAAT was changed to GGGC (44). The TATA box and the translation start site are also indicated. The nucleotide residue numbers correspond to the mouse sequence. (B) plklf(−141/−103) response to ERK5 and p38 MAPKs. The plkf(−141/−103) reporter construct containing one copy of the response element was transfected into erk5 −/− MEF cells with ERK5 or p38 alone or together with upstream activators or inhibitors. Relative responses are shown as the normalized luciferase activity. The experiment shown is representative of four independent experiments. Δ56luc is the backbone construct containing only the c-fos minimal promoter and a luciferase cassette. (C) Response of mutant reporter constructs. The wild-type or mutant versions of the plklf(−141/−103) reporter, wherein the two transcription factor-binding motifs were individually altered (indicated as “A” and “B” mutants), was transfected into erk5 −/− MEF cells with ERK5 or ERK5(1-740) without or with MEK5(D). Shown are representative results of three independent experiments. (D) Response of plklf(−141/−103) to mutant forms of ERK5. Wild-type plklf(−141/−103) reporter was cotransfected into erk5 −/− MEF cells with a full-length wild-type ERK5 (FL), ERK5 (1-740) (ΔTA), ERK5 kinase domain alone [ERK5 (1-400), kinase], or ERK5 lacking the kinase domain [ERK5 (400-806), TA] without or with MEK5(D). (E) Induction of the endogenous lklf gene by ERK5. erk5 −/− MEF cells transfected transiently with plasmids encoding MKK6, MKK6(E), MEK5(D) and ERK5, or MEK5(A) and ERK5 were analyzed by quantitative RT-PCR at 36 h posttransfection. Shown are relative lklf transcript levels normalized against γ-actin and averaged over triplicate samples.
To determine whether ERK5 activates transcription of the lklf gene through MEF2, we generated reporter constructs in which wild-type or mutant versions of the −141/−103 sequence element was inserted upstream of the c-fos minimal promoter to drive expression of the luciferase reporter gene. These constructs, each containing a single copy of the putative TEF-1/MEF2 response element, were transiently transfected into ERK5-deficient MEF cells, together with ERK5 and MEK5, the upstream activator for ERK5 (Fig. 6B). The p38 MAPK has previously been shown to phosphorylate and activate MEF2 transcription factors (13, 58) and was implicated in TRAF2-mediated LKLF gene regulation (31). Thus, we also tested whether the lklf reporter responded to expression of p38 and MKK6(E), an activated form of MKK6. We found that transfection of either ERK5 or MEK5(D), an activated mutant form of MEK5, failed to induce any reporter activity, whereas cotransfection of ERK5 and MEK5(D) dramatically increased the reporter expression (Fig. 6B). MEK5(D) also activated the reporter in erk5 +/− MEF cells (data not shown), indicating that endogenous ERK5 can drive reporter gene expression. In contrast, MKK6(E) failed to induce reporter activity via endogenous p38 or even when a plasmid encoding p38 was coexpressed (Fig. 6B). These data demonstrate that ERK5 induces responses through the −141/−103 promoter region and plays a nonredundant role in such activation that is distinct from the effects of the p38 pathway.
To further determine the requirements for the putative MEF2 site (“B”) and the adjacent AT-rich element (“A”) in ERK5-dependent activation of the lklf promoter, we tested reporters bearing mutations within the A or B motifs. Alteration of either A or B site completely abrogated the reporter response to activated ERK5, indicating that both elements were required for the activity (Fig. 6C). In addition, a mutant form of ERK5 lacking the C-terminal transactivation domain [ERK5(1-740)] was significantly less effective at inducing the reporter activity, confirming the importance of this domain in transcriptional activation. To further dissect the requirements for various domains of ERK5, we tested mutant versions of ERK5 for their ability to activate the lklf reporter (Fig. 6D). Again, ERK5(1-740) mounted reduced levels of activity compared to full-length ERK5. Importantly, the ERK5 kinase domain (amino acids 1 to 400) alone was completely ineffective at inducing the lklf promoter, indicating the absolute requirement for the ERK5 C-terminal domain toward activating transcription. In contrast, ERK5(400-806), encoding only the C-terminal domain, constitutively induced transcription and did not require activation by MEK5. Expression of wild-type and mutant forms of ERK5 was confirmed by anti-ERK5 immunoblotting (22; data not shown). The fact that the C-terminal domain alone enhanced lklf promoter activity suggests that the catalytic domain normally plays a regulatory role toward controlling the activity and/or accessibility of the transactivation domain. Consequently, the removal of the catalytic domain and the regulatory constraints associated with it leads to constitutively enhanced responses.
Next we tested whether ERK5 regulates the endogenous lklf promoter in a manner reflected by the −141/−103 reporter by monitoring the endogenous lklf transcript levels in transiently transfected erk5 −/− MEF cells using quantitative RT-PCR. Figure 6E clearly demonstrates that activation of ERK5 [achieved by coexpression of ERK5 and MEK5(D)] significantly increased lklf transcript abundance compared to cells transfected with vector plasmid or with ERK5 and MEK5(A), a catalytically inactive, dominant-negative form of MEK5. In contrast, expression of the p38 activator, MKK6(E), failed to induce LKLF expression even when p38 was cotransfected and overexpressed (Fig. 6E and data not shown). These results indicate that the response of the minimal region represented by the −141/−103 reporter accurately mimics the ERK5-dependent regulation of the endogenous lklf promoter and activation of ERK5 is sufficient to induce expression of lklf.
To investigate whether the lklf(−141/−103) element binds to MEF2 transcription factors, we performed EMSAs. We used the lklf(−141/−103) element as the labeled probe and incubated with nuclear extracts from MEFs. For competition, we used unlabeled lklf(−141/−103), as well as the A mutant, B mutant, and the A+B double-mutant oligonucleotides. As shown in Fig. 7A, we detected one major protein-DNA complex that forms with the lklf(−141/−103) probe. Addition of anti-MEF2 (MEF2D-specific or pan-MEF2) antibodies supershifted migration of this complex (lanes 3 and 4), confirming that MEF2 proteins bind to this region. Antibodies that recognize TEF1 inhibited, rather than supershifted, formation of the complex (lane 2), suggesting TEF1 may also bind to this element. As expected, lklf(−141/−103) competed against itself (lanes 5 and 6). The A mutant competed as well as the wild-type (lanes 7 and 8), whereas the B mutant failed to compete (lanes 9 and 10). The double mutant did not compete at all (lanes 11 and 12). Since the A mutant is functionally unresponsive to activated ERK5 (Fig. 6C), these results indicate that binding of MEF2 to a single element (B element) is not sufficient for activation of transcription. This is consistent with the finding that MEF2 activities are influenced by interactions with other transcription factors (2). In this instance, the lklf promoter is regulated by MEF2 and the protein that binds to the A element, which most likely is TEF-1.
lklf(−141/−103) binds to MEF2. (A) An EMSA was performed with the radiolabeled lklf(−141/−103)probe in the presence of various antibodies or competing oligonucleotides. For supershift analysis, preimmune serum (lane 1) or 1 μg each of the antibodies to TEF1 (lane 2), MEF2D (lane 3), or pan-MEF2 (lane 4) was added to the binding reaction. For oligonucleotide competition, 25 or 50 ng each of the wild-type (lanes 5 and 6), A mutant (lanes 7 and 8), B mutant (lanes 9 and 10), or A+B double mutant (lanes 11 and 12), lklf(−141/−103) oligonucleotides was included to test for its ability to compete for protein binding. (B) lklf(−141/−103) competition with chicken MLC2v probe. Labeled MLC2v probe was incubated with nuclear extracts in the presence of bovine serum albumin (BSA, lane 1), preimmune serum (lane 2), anti-TEF-1 antibody (lane 3), anti-MEF2D antibody (lane 4), pan-MEF2 antibody (lane 5), or cold oligonucleotides representing wild-type (lanes 6 and 7), “A” mutant (lanes 8 and 9), “B” mutant (lanes 10 and 11), or “A+B” mutant (lanes 12 and 13) versions of the lklf(−141/−103) at 25 and 50 ng. MEF2-containing upper complexes, TEF-1-containing lower complexes, as well as nonspecific DNA-protein complex, are indicated with arrows. (C) ChIP assays were performed to demonstrate in vivo binding of MEF2 to the lklf promoter. PCR amplification with primers specific for the lklf promoter (top) or APRT (bottom) demonstrates specific enrichment of the lklf promoter pulled down with the MEF2D antibody.
To further test whether TEF1 binds to the lklf(−141/−103) region, we utilized the chicken myosin light chain (MLC2v) probe previously described to contain a TEF1 and a MEF2 binding sites (32). As shown before, anti-TEF1 antibody was not able to supershift the lower-migrating protein-DNA complex (Fig. 7B), although this complex was previously described as containing the TEF1-like proteins (32). In contrast, anti-MEF2D or pan-MEF2 antibodies supershifted the upper migrating complex, confirming that MEF2 indeed binds to this probe. The addition of the lklf(−141/−103) oligonucleotides abrogated formation of both the MEF2-containing and the TEF1-containing complexes, indicating that the MLC2v and the lklf(−141/−103) oligonucleotides competed to bind to the same proteins. Interestingly, the A mutant lklf oligonucleotides competed, as well as the intact lklf(−141/−103), to abrogate formation of both the MEF2- and TEF1-containing complexes, whereas the B mutant oligonucleotides failed to compete at all. This strengthens the possibility that MEF2 recruits the TEF1 complex. A similar result has previously been reported using mutated TEF-1 MLC2v oligonucleotides as competitors (32). Together with our reporter experiments, the EMSA studies showed that presence of an intact MEF2 site is by itself insufficient and that additional modification(s) mediated through the adjacent AT-rich motif is absolutely required for lklf gene expression in response to activated ERK5. Binding of MEF2 to lklf(−141/−103) in vivo was evaluated by the ChIP assay. Using an antibody specific for MEF2D, we were able to pull down the lklf promoter from nuclear extracts of MEF cells as determined by PCR amplification with primers that flank the lklf(−141/−103) region (Fig. 7C). Specific interaction of MEF2 with this region is verified by the failure of amplification with primers that recognize unrelated sequences (APRT or adenine phosphoribosyltransferase promoter).
ERK5 regulates activation responses in T cells.In addition to its requirement in endothelial cells, LKLF has been shown to regulate the activation state and survival of mature T cells. The ability of ERK5 to modulate the expression of lklf and to regulate activation of T cells was assessed by generating DO11.10 T-cell hybridoma cells that were retrovirally transduced to express a dominant-negative Mek5 [Mek5(A)] or constitutively activated Mek5 [Mek5(D)] protein. To determine the requirement for ERK5, we also generated retrovirus that expresses erk5 small hairpin RNA (shRNA). This retrovirus produces erk5-specific 22-nucleotide small interfering RNA hairpins under the control of the U6 RNA polymerase III promoter and coexpresses GFP. The efficacy of this shRNA was tested in mouse tumor cells that were transduced and sorted individually based on GFP expression. The erk5 shRNA suppresses accumulation of the endogenous ERK5 proteins, as shown by anti-ERK5 immunoblot analyses (Fig. 8A). We also tested the specificity of the erk5 shRNA by analyzing for expression of ERK1/2, p38, and JNK MAPKs in these cells and determined that erk5 shRNA did not affect other members of the MAPK family (data not shown).
ERK5 regulates LKLF expression and modulates activation state in T cells. (A) In the left top panel, an anti-ERK5 immunoblot shows that erk5 shRNA successfully knocks down ERK5 protein levels in B16 mouse melanoma cells infected with the erk5 shRNA retrovirus and sorted for GFP expression. In the left lower panel, DO11.10 hybridoma cells were transduced with retrovirus encoding Mek5(A), Mek5(D), or erk5 shRNA. The expression of lklf in retrovirally transduced DO11.10 cells was assessed by quantitative RT-PCR shown as values normalized against γ-actin and averaged over triplicate samples. These analyses were performed on unsorted populations, which typically contained >70% GFP+ cells except for erk5 shRNA-expressing cells, which represented ca. 30% of the population. In the right panel, retrovirus-infected DO11.10 cells were analyzed for expression of CD62L by flow cytometry to determine T-cell activation status. Shown are the overlays of CD62L profiles on GFP+ gated populations. The profiles of Mek5(A) (solid line)-, Mek5(D) (dotted line)-, and empty virus (gray shaded)-expressing cells are shown on the top, and the profiles for erk5 shRNA (solid line)- and empty virus (gray shaded)-expressing cells are shown below. (B) The same retroviruses were used to infect activated primary lymph node T cells. Expression of lklf was quantified by SYBR green RT-PCR on unsorted populations, typically containing 30% GFP+ cells. Relative abundance is shown as fold over control values (top, indicated arbitrarily as 1 and marked by the arrow). Activation status of retrovirus-infected cells was assessed. An overlay of forward-scatter and CD62L profiles for empty virus (gray shade)- and erk5 shRNA (solid line)-infected cells is shown (bottom). (C) Activation of ERK5 in response to IL-7 or TCR cross-linking was assessed by anti-phospho ERK5 immunoblotting. Previously activated purified lymph node T cells were cultured in the presence of recombinant IL-7 (20 ng/ml) or plate-bound anti-CD3 (500A2, 1 μg/ml) and anti-CD28 (1:1,000 dilution of ascites fluid) antibodies for various lengths of time, as indicated. Total lysates were resolved by SDS-PAGE and blotted with antisera specific for phosphorylated ERK5 (top) or total ERK5 (bottom).
The effects of ERK5 inhibition or activation on LKLF expression in T-cell hybridomas were assessed by quantitative RT-PCR (Fig. 8A, bottom left). Analyses of total RNA from unsorted populations of DO11.10 cells indicate that Mek5(A) and erk5 shRNA both decrease, whereas Mek5(D) increases, the expression of endogenous lklf in T cells. The dominant-negative effect of Mek5(A) and the gene silencing effect of shRNA, both ultimately suppressing the ERK5 activity, resulted in the same outcome (repressed lklf expression). In contrast, constitutive activation of ERK5 by Mek5(D) was sufficient to induce lklf expression. These results clearly demonstrate that the levels of lklf strictly correlate with the total ERK5 activity in these cells.
Previously, it was shown that loss of LKLF led to spontaneous activation, whereas enforced expression of LKLF established a resting state in T cells (4, 27). Based on these studies, LKLF has been postulated to play important roles in establishing and maintaining T-cell quiescence. Since ERK5 directly regulated the expression of LKLF in T cells, we further tested to determine whether the activation state of DO11.10 cells expressing Mek5(A), Mek5(D), or erk5 shRNA reflected these apparent differences in the LKLF responses. We compared the relative activation state of these cells by analyzing the expression of CD62L, a surface marker downregulated upon activation in T cells (5). As shown by the overlay of the CD62L profiles, Mek5(A)- or erk5 shRNA-expressing cells (gated on GFP+ cells) exhibited decreased levels of surface CD62L, indicating enhanced activation. In contrast, cells expressing Mek5(D) retained higher levels of CD62L compared to empty virus-infected control cells, indicating that these cells became more quiescent (Fig. 8A).
To determine whether ERK5 modulates the activation state of primary T cells, we used the same retroviruses described above to transduce mouse lymph node T cells. For the purpose of retroviral infection, T cells were first activated in vitro by treatment with antibodies that cross-link CD3 and CD28. Similar to the results seen in DO11.10 cells, expression of Mek5(D) induced, while erk5 shRNA reduced, the levels of the lklf transcripts in these cells (Fig. 8B). Mek5(A) did not appear as efficient as erk5 shRNA in downregulating lklf expression in primary cells (data not shown), and thus we focused on erk5 shRNA to assess its effect on T-cell activation. We analyzed retrovirus-infected T cells for cell size, which increases with activation, and for CD62L and CD25 expression, which are down- and upregulated (5, 38) upon activation, respectively. As determined by flow cytometric analysis, cells expressing erk5 shRNA were larger in size compared to the empty virus-infected control, and they also expressed slightly reduced levels of CD62L (Fig. 8B) while further upregulating the expression of CD25 (data not shown). These results indicate that silencing of the erk5 gene enhanced T-cell activation. Interestingly, although Mek5(D) induced the expression of lklf, increased expression of LKLF was not sufficient to revert the activated phenotype of these primary T cells to the resting state (data not shown), suggesting that induction of LKLF may not be sufficient to establish quiescence in primary T cells.
Since activation of ERK5 led to significant induction of LKLF, which in T cells is known to be highly regulated by T-cell antigen receptor (TCR)- and cytokine receptor-mediated signals (9, 42), we evaluated ERK5 activation in response to stimuli that trigger these signals. To simulate responses of the retrovirally transduced cells, we treated previously activated lymph node T cells with IL-7, which has been shown to activate lklf transcriptional responses, or with antibodies that cross-link TCR/CD3 and CD28. ERK5 activation, as indicated by phosphorylation, was detectable within 45 min and remained activated throughout the two-hour time course after IL-7 stimulation (Fig. 8C). In contrast, activation with anti-CD3 and anti-CD28 led to a faster response (detectable within 15 min), which subsided to the basal level by 2 h. These results clearly demonstrate that both IL-7 and TCR can trigger ERK5 activation, although IL-7 signal induces a more sustained ERK5 response compared to that triggered by TCR engagement.
DISCUSSION
To understand the molecular function of ERK5, a MAPK and a potent transcriptional regulator, we performed comparative gene profile analyses to identify its downstream target genes. Using ERK5-deficient embryos or MEF cells, groups of genes with elevated or reduced expression were found, consistent with our initial hypothesis that ERK5 influences transcription both positively and negatively. We identified bnip3, flt1, and igf-bp3 as genes highly expressed in the absence of ERK5. Bnip3 is a known proapoptotic gene, inducible in response to hypoxic stress (3, 49). The fact that bnip3 is highly elevated in ERK5-deficient embryos is consistent with the reports that apoptotic responses are heightened in the absence of ERK5 (14, 39). We also showed that ERK5 negatively regulates transcription of the bnip3 gene, similarly to its effect on the vegf gene. In fact, our comparative gene profile analyses revealed that many hypoxia-responsive genes were derepressed in the absence of ERK5. Although further studies are required to resolve how ERK5 inhibits transcription, our data suggest that ERK5 acts as a general repressor because loss of ERK5 not only exaggerates hypoxia-inducible response of some genes but also induces several genes normally unresponsive to hypoxia.
We showed that the lung Krüppel-like factor (LKLF) is absolutely dependent on ERK5 for its expression and that ERK5 activates the lklf gene through a mechanism involving MEF2 transcription factors. In contrast to the nur77 gene, for which ERK5-mediated activation of MEF2 is sufficient to induce transcription, induction of the lklf gene through the −141/−103 region requires MEF2 and recruitment of additional factors. This dual requirement represents a novel mode of ERK5-dependent transcriptional coactivation and implies that ERK5 may coordinate multiprotein complexes to regulate transcriptional responses.
Disruption of the lklf gene leads to embryonic lethality at around E12.5 to E14 with severe angiogenic defects, caused at least in part by the failure to induce formation of the tunica media that normally stabilizes blood vessels (26, 27). The lklf gene is expressed primarily in endothelial cells in early developing embryos but is later induced at high levels in the lung and lymphocytes (1). Given the endothelial cell-specific expression of LKLF, the perivascular, tunica media-associated defect in lklf −/ − embryos appears incongruous. However, it is possible that LKLF exerts its effects through transcellular signaling (endothelial-to-perivascular), or that nonendothelial cells can be induced to express LKLF. The latter possibility is further supported by the evidence that endothelial cells can differentiate to become vascular smooth muscle cells in some cases (32a). Thus, one can speculate that LKLF (and ERK5) play roles in trans differentiation of endothelial cells to vascular smooth muscle cells or in promoting survival of such cells. In the Rag2−/− chimeric system, lklf −/− T cells were shown to undergo spontaneous activation and premature apoptosis (26, 27), suggesting that sustained expression of LKLF may be critical for maintaining survival and quiescence of circulating mature T cells. We demonstrated that expression of erk5 shRNA inhibited accumulation of lklf transcripts in T cells and accelerated T-cell activation. Thus, ERK5 plays a central role in differentiation and/or functions of endothelial cells and in T-cell quiescence.
Previously, MEF cells lacking TRAF2 or the p38 MAPK were shown to express reduced levels of LKLF, and p38 was thought to regulate expression of LKLF by TRAF2 (31). However, in our MEF cells, overexpression of TRAF2 or a combination of p38 and MKK6(E) did not activate the lklf(−141/−103) reporter or endogenous LKLF expression, and coexpression of TRAF2 and MKK6(AA) did not repress LKLF expression (S. Sohn, unpublished data). Analysis of erk5 −/ − embryos also indicates that they express normal levels of p38 (data not shown), whereas lklf expression is absent; thus, the presence of p38 is not sufficient to induce the expression of lklf. One explanation for these apparently conflicting results is that expression of LKLF is regulated through a temporally controlled mechanism. In such a model, ERK5 is required at the earliest stage of development, through E10 to E10.5 when ERK5 deficiency results in death. Subsequently, p38 (perhaps downstream of TRAF2) induces LKLF expression, and inability to do so may contribute to embryonic death starting around E11.5, as reported for p38α-deficient embryos (47). This model is consistent with the observations made with regard to LKLF expression in p38 −/ − MEF cells, which were derived from E10.5 to E11.5 embryos (47). In our study, MEF cells were derived from E9.25 embryos (45), representing a considerably earlier stage. Therefore, ERK5 and p38 may both regulate the lklf gene but nonredundantly during embryonic development. These results together might argue for differential wiring of cells at different developmental stages or that distinct stimuli trigger expression of LKLF via two mutually exclusive pathways.
Expression of LKLF is highly regulated in T cells. During T-cell differentiation, LKLF is upregulated at the transition from the CD4+ CD8+ double-positive to the CD4+ or CD8+ single-positive stage (19). Mature naive T cells express sustained levels of LKLF, which rapidly declines after stimulation through the TCR (42). Exposure to cytokines such as IL-2, IL-7, and IL-15, which are known to promote activation and proliferation, reinduce LKLF (9, 42). Therefore, signals through TCR and cytokine receptors exert opposite effects on LKLF expression. Interestingly, both signals activate ERK5, albeit with differences in the activation kinetics. ERK5 activation in response to TCR engagement is rapid and short, whereas the cytokine response is slower but more prolonged. The fact that TCR cross-linking leads to initial disappearance of LKLF suggests a mechanism exists to antagonize the effects of activated ERK5 during the initial phase of T-cell activation or that a transient activation of ERK5 is not sufficient to turn on lklf transcription. In activated T cells, expression of MEK5(D) is sufficient to induce, whereas expression of erk5 shRNA represses, lklf transcription. Since activated T cells produce and respond to many cytokines in an autocrine fashion, the ERK5 pathway in these cells is likely to be engaged in cytokine responses. Significantly, the requirement for ERK5 in the induction of LKLF in T cells points to an important role of ERK5 in the regulation of T-cell activation and long-term survival. In support of this model, inhibition of ERK5 recapitulates the phenotype of T cells lacking LKLF, with increased cell size as well as decreased CD62L expression. How LKLF regulates T-cell activation awaits further studies.
ACKNOWLEDGMENTS
We thank Herbert Kasler for cloning the MSCV-ERK5 and MSCV-ERK5(1-740) constructs, Martin Guerbadot for technical assistance, and Laurel Lenz for reading the manuscript. We are indebted to the following investigators for their generous gifts: Richard Bruick (University of Texas Southwestern) for the bnip3 reporter (Nip3 prom-pGL3 Basic), José Alberola-Ila for retroviral constructs, and Roger Davis (HHMI, University of Massachusetts) for pCDNA3-Flag-MKK6, pCDNA3-Flag-MKK6(glu), and pCMV-Flag-p38 plasmids. We also thank Michael Cooke (Genomics Institute of the Novartis Research Foundation) for the erk5 shRNA sequences and Jamie Geier and Na Xiong for technical advice on the ChIP assays.
FOOTNOTES
- Received 28 March 2005.
- Returned for modification 8 May 2005.
- Accepted 7 July 2005.
- Copyright © 2005 American Society for Microbiology