Glucocorticoid regulation of mouse and human dual specificity phosphatase 1 (DUSP1) genes: unusual cis-acting elements and unexpected evolutionary divergence

Abstract

Anti-inflammatory effects of glucocorticoids (GCs) are partly mediated by up-regulation of DUSP1 (dual specificity phosphatase 1), which dephosphorylates and inactivates mitogen-activated protein kinases. We identified putative GC-responsive regions containing GC receptor (GR) binding site consensus sequences that are well conserved between human and mouse DUSP1 loci in position, orientation, and sequence (at least 11 of 15 positions identical) and lie within regions of extended sequence conservation (minimum 65% identity over at least 100 bp). These were located approximately 29, 28, 24, 4.6, and 1.3 kb upstream of the DUSP1 transcription start site. The homology-based approach successfully identified four cis-acting regions that mediated transcriptional responses to dexamethasone. However, there was surprising interspecies divergence in site usage. This could not be explained by variations of the GR binding sites themselves. Instead, variations in flanking sequences appear to have driven the evolutionary divergence in mechanisms of regulation of mouse and human DUSP1 genes. There was a good correlation between the ability of cis-acting elements to respond to GC in transiently transfected reporter constructs and their ability to recruit GR in the context of intact chromatin. We propose that divergence of gene regulation has involved the loss or gain of binding sites for accessory transcription factors that assist in GR recruitment. Finally, a novel GC-responsive region of the human DUSP1 gene contains a highly unusual element, in which three closely spaced GR half-sites are required for potent transcriptional activation by GC.

FIGURE 1.

Conservation of potential GR binding sites at human and mouse DUSP1 loci. 30 kb of human genomic sequence 5′ of the DUSP1 TSS was aligned with the orthologous mouse genomic sequence using the default settings of the Genome Vista browser (65). Both sequences were scanned for matches to the redundant GR binding site consensus sequence GNACANNNNG (GBS) (37). Matches that are conserved between the two species are indicated below the alignments, and approximate coordinates are given with reference to the TSS of the human sequence. See supplemental Fig. 1 for accurate coordinates of GBS elements.

FIGURE 2.

Deletion analysis of the human DUSP1 promoter. Fragments of the human DUSP1 5′ region were subcloned into the firefly luciferase reporter construct pGL3b. The positions of GBS elements are indicated by vertical bars. GBS-4.6 comprises three closely spaced GBS elements known as GBS-4.6.1, -4.6.2 and -4.6.3. 200 ng of each reporter construct and 100 ng of a Renilla luciferase expression vector were transiently transfected into HeLa cells, which were then treated for 20 h with vehicle (0.1% ethanol) or Dex (100 nm) before harvesting and measuring luciferase activities. For each construct, firefly luciferase activity was normalized to Renilla luciferase activity, and -fold activation in response to Dex was calculated. In this and subsequent figures, mean -fold responses ± S.E. from three independent experiments are shown. ***, p < 0.001; **, p < 0.01; *, p < 0.05; n.s., not significantly different. Unless otherwise indicated, statistical comparisons are against the largest construct, pGL3b−4834.

FIGURE 3.

Characterization of a GC-responsive region 4.6 kb upstream of the human DUSP1 start site. A, Multiz alignment of GRR-4.6 of 28 mammalian species at the UCSC Genome Browser (University of California, Santa Cruz) (66). The positions of GBS-4.6.1, -4.6.2, and -4.6.3 are indicated below the alignment. B, sequences of GBS-4.6.2 of human (Homo sapiens; Hs.), rat (Rattus norvegicus; Rn.), mouse (Mus musculus; Mm.), dog (Canis familiaris; Cfa.), cow (Bos taurus; Bta.), horse (Equus caballus; Eca.), and armadillo (Dasypus novemcintus; Dn.). The conserved GBS sequence GNACANNNNG is indicated by asterisks, and an additional GR half-site is denoted by an arrow. Differences from the human sequence are highlighted. Sequences of mutated versions of GBS-4.6.2 are also shown. C, a 467-bp human genomic fragment centered on GBS-4.6.2 was cloned upstream of the SV40 early promoter in pGL3p. Mutations at GBS-4.6.1, -4.6.2, or -4.6.3 (as indicated in B) were introduced by PCR. Dex responses were assayed as in Fig. 2 and normalized against the response of the empty vector pGL3p. With the exception of pGL3p-GRR-4.6-GBS-4.6.2-m2, all constructs were significantly different from the parental vector pGL3p. Statistical comparisons against pGL3p-GRR-4.6 are indicated. D, orthologous GRR-4.6 fragments from dog, human, mouse, and rat (Cfa., Hs., Mm., and Rn.) were cloned into pGL3p, and responses to Dex were calculated as in C. Statistically significant differences from pGL3p are indicated beside each construct, and additional comparisons are as shown. E, short oligonucleotides containing human or mouse GBS-4.6.2 were subcloned into the firefly luciferase reporter pGL4.26 (for simplicity indicated as pGL4 in the figure). Responses to Dex were calculated as in C. Statistically significant differences from pGL4 are indicated beside each construct, and other statistical comparisons are as indicated.

FIGURE 4.

Identification of a GC-responsive region 1.3 kb upstream of the human DUSP1 start site. A genomic fragment extending from −1426 to −1096 with respect to the human DUSP1 TSS and various deletion derivatives of this fragment were subcloned into pGL3p. Responses to Dex were calculated as in Fig. 3C. The region −1366 to −1237 was identified as a minimal GC-responsive region and is hereafter known as GRR-1.3. Statistical comparisons are first against the parental vector pGL3p and second against the construct containing the minimal GC responsive region −1366 to −1237, which is hereafter referred to as pGL3p-GRR-1.3-Hs.

FIGURE 5.

Characterization of GRR-1.3. A, Multiz alignment of GRR-1.3 of 28 mammalian species at the UCSC Genome Browser (66). The positions of GBS-1.3 is indicated below the alignment. B, alignment of human (Hs.), dog (Cfa.), and mouse (Mm.) GRR-1.3 sequences. The positions of putative C/EBP and GR binding sites (GBS-1.3) are shown below the alignment. GBS-1.3 (in italic type) is a short oligonucleotide containing the GBS-1.3 element and a few conserved flanking residues. Nucleotide changes introduced into the human GRR-1.3 fragment to create the GBSm1 and GBSm2 are shown. C, dog, rat, mouse, and human GRR-1.3 fragments were subcloned into pGL3p, and responses to Dex wre calculated as in Fig. 3C. Differences from the parental vector pGL3p are shown to the right of each bar, and other statistical comparisons are as indicated. D, human and mouse GRR-1.3 fragments or the oligonucleotide GBS-1.3 were subcloned into pGL4.26 (shown as pGL4), and responses to Dex were calculated as in Fig. 3C. Differences from the parental vector pGL4 are shown to the right of each bar, and other statistical comparisons are as indicated.

FIGURE 6.

Differential usage of GRR-4.6 and -1.3 in mouse and human DUSP1 promoters. Site-directed mutagenesis was performed to introduce mutations into the human (Hs.) reporter constructs pGL3b−4834 and −1495. The Δ4.6 mutation is a 28-bp internal deletion that removes GBS-4.6.2 but leaves GBS-4.6.1 and -4.6.3 intact. The m1.3 mutation is a 3-bp substitution of GGG for the ACA motif triplet of GBS-1.3 (identical to m1 in Fig. 5B). Corresponding deletion and point mutation constructs were generated for the mouse (Mm.) DUSP1 promoter. The differences in size of mouse and human promoter fragments are due to internal insertions and deletions at the two DUSP1 loci. Responses to Dex were determined as in Fig. 2. n.s., not significant.

FIGURE 7.

Characterization of a GC-responsive region 29 kb upstream of the human DUSP1 start site. A, alignment of GBS-29 sequences of mouse, rat, human, and dog. This element was first described in the mouse DUSP1 locus (47); therefore, the mouse is regarded as the base sequence, and differences from it are highlighted. Matches to the consensus GNACANNNNG on the top or bottom strand are indicated by asterisks. B, dog, rat, mouse, or human GRR-29 fragments were cloned into pGL3p. pGL3p-GRR-29-Mm.-GBSm was created by mutagenizing both of the ACA triplets within the mouse GRR-29 fragment to GGG. pGL3p-GRR-29-Hs.>Mm. was generated by mutagenizing the sequence GAACATTCGG within human GRR-29 to the corresponding mouse sequence, GAATGTTCAG. This 3-nucleotide change creates a palindromic element with GNACANNNNG matches on both strands. Dex responses were calculated as in Fig. 3C. Differences from the parental vector pGL3p are shown to the right of each bar, and other statistical comparisons are as indicated. C, short oligonucleotides containing human and mouse GBS-29 were cloned into pGL4. Responses to Dex were calculated as in Fig. 3C. Differences from the parental vector pGL4 are shown to the right of each bar, and other statistical comparisons are as indicated.

FIGURE 8.

Species-specific transcriptional responses to Dex in mouse fibroblasts. Dex responses of human and mouse GRR-29, -4.6, and -1.3 fragments were tested in mouse fibroblasts and calculated as in Fig. 3C. Differences from the parental vector pGL3p are indicated above each column, and pairwise comparisons of orthologous mouse and human fragments are also shown. n.s., not significant.

FIGURE 9.

Recruitment of GR to functional GC-responsive regions. HeLa cells (A) or mouse fibroblasts (B) were treated with vehicle (0.1% EtOH) or Dex (100 nm) for 30 min, and then chromatin immunoprecipitation assays were performed using either IgG (open columns) or an anti-GR antibody (filled columns). Immunoprecipitated chromatin fragments were detected by quantitative PCR, and enrichment of specific genomic fragments was calculated as a proportion of the starting material. Measurements were made in triplicate. The figure illustrates means ± S.D. from single experiments, representative of three (HeLa) or two (mouse fibroblast) independent experiments.

FIGURE 10.

Mutation of the GR dimerization motif has distinct effects on different regulatory elements. A, COS-7 cells were transfected with 200 ng of pGL3p (open columns), pGL3p-GRR-1.3-Hs. (filled columns), or pGL3p-GRR-1.3-Hs.-GBSm1 (shaded columns) and with 50 ng of a vector expressing wild type GR or the corresponding empty vector. Responses to Dex were calculated as in Fig. 3C. B, in a parallel experiment, whole cell extracts were prepared from transfected COS-7 cells and Western blotted for GR or for tubulin as a loading control. C, different reporter constructs were transfected into COS-7 cells with wild type GR or GR^dim expression vectors. A value of 100% was assigned to the transcriptional response of each construct in the presence of wild type GR (open bars), and the response in the presence of GR^dim was calculated with reference to this (filled bars). Statistically significant differences between the responses of wild type GR and GR^dim are indicated.