Functional Diversification of Vitamin D Receptor Paralogs in Teleost Fish After a Whole Genome Duplication Event

Abstract

The diversity and success of teleost fishes (Actinopterygii) has been attributed to three successive rounds of whole-genome duplication (WGD). WGDs provide a source of raw genetic material for evolutionary forces to act upon, resulting in the divergence of genes with altered or novel functions. The retention of multiple gene pairs (paralogs) in teleosts provides a unique opportunity to study how genes diversify and evolve after a WGD. This study examines the hypothesis that vitamin D receptor (VDR) paralogs (VDRα and VDRβ) from two distantly related teleost orders have undergone functional divergence subsequent to the teleost-specific WGD. VDRα and VDRβ paralogs were cloned from the Japanese medaka (Beloniformes) and the zebrafish (Cypriniformes). Initial transactivation studies using 1α, 25-dihydroxyvitamin D₃ revealed that although VDRα and VDRβ maintain similar ligand potency, the maximum efficacy of VDRβ was significantly attenuated compared with VDRα in both species. Subsequent analyses revealed that VDRα and VDRβ maintain highly similar ligand affinities; however, VDRα demonstrated preferential DNA binding compared with VDRβ. Protein-protein interactions between the VDR paralogs and essential nuclear receptor coactivators were investigated using transactivation and mammalian two-hybrid assays. Our results imply that functional differences between VDRα and VDRβ occurred early in teleost evolution because they are conserved between distantly related species. Our results further suggest that the observed differences may be associated with differential protein-protein interactions between the VDR paralogs and coactivators. We speculate that the observed functional differences are due to subtle ligand-induced conformational differences between the two paralogs, leading to divergent downstream functions.

Whole-genome duplication (WGD) events are proposed to be a significant force driving vertebrate evolution by providing a novel source of raw genetic material that can be subjected to evolutionary change (1–4). The evolution and divergence of new genes after a WGD may permit the development of more complex gene interactions and networks, leading to novel evolutionary innovations, speciation, and increased adaptability (3–7). Two WGD events (1R and 2R) occurred early in vertebrate evolution (Supplemental Figure 1) (1). A third WGD (3R) occurred in the stem lineage of teleost fish, a division within the vertebrate class Actinopterygii (ray finned fishes) (8, 9) (Supplemental Figure 1). It has been proposed that the 3R event has enabled significant diversification in teleosts (8, 10). For example, teleost fish compose approximately half of all species of vertebrates, whereas less than 50 species are found in pre-3R orders of Actinopterygii (11).

After a WGD, an estimated 50%–90% of gene duplicates become nonfunctional and are lost (12, 13). However, genes involved in signaling and regulatory pathways, such as the nuclear receptor (NR) superfamily, have been preferentially retained in duplicate to maintain pathway integrity (6). NRs are conditional transcription factors that bind to lipophilic signaling molecules, resulting in systematic control and expression of target genes. Such control facilitates cellular responses to endogenous and exogenous signals through the coordination of complex transcriptional processes (14). NRs play an essential role in many physiological and endocrine-mediated processes, including metabolism, reproduction, embryonic development, and cellular proliferation and differentiation. In teleosts, orthologs for all mammalian NR subclasses have been identified (15–17). In fact, teleosts maintain a larger complement of NRs than mammals, likely due to the global retention of NRs subsequent to the 3R event. For example, 48 NRs have been identified in humans (18), compared with 68–70 identified in pufferfishes (Takifugu rubripes and Tetraodon nigroviridis) (15, 16, 19), 70 in zebrafish (Danio rerio) (17), and 71 in the Japanese medaka (Oryzias latipes) (20). Because NRs regulate numerous processes essential to life, NR duplication and divergence may have contributed to the evolutionary success of teleosts. However, the functional role of duplicate NRs remains unresolved in these species.

The vitamin D receptor (VDR; NR1I1) is a member of the NR1I subfamily of nuclear receptors (14). Conventional theory speculates that the vitamin D endocrine system originated in terrestrial animals as a way to control calcium homeostasis in a calcium-limited environment (21). However, this theory has been challenged due to the identification of functional VDRs in numerous aquatic vertebrates including amphibians (21), numerous species of fish (15–17, 22–24), and the sea lamprey (25). The fact that aquatic vertebrates live in an environment that serves as an abundant calcium source suggests that VDR evolved long before the need for strict hormonal control of calcium.

In agreement with the 3R event, two distinct VDR genes (VDRα and VDRβ) have been identified in multiple 3R teleost species (15–17, 22, 23). Evidence indicating that these duplicates are indeed paralogs, and not isoforms, include the observation that VDRα and VDRβ paralogs are shared with multiple species across large evolutionary distances, the observation that they appear to have originated at the same time, and that the paralogs are found within distinct chromosomal loci yet maintain conserved synteny (8, 26). Homology comparisons have revealed that the VDR paralogs maintain a high degree of sequence identity to tetrapod VDRs, although are most similar to VDR paralogs from other teleosts (22). Tissue expression studies revealed both VDR paralogs are expressed in a wide variety of tissues similar to mammalian VDRs; however, VDRα and VDRβ exhibited differential spatial or quantitative tissue expression patterns in each species (15, 22, 27). Although VDR paralogs are highly conserved, studies have identified differential paralog functions, including differences in ligand sensitivities and target gene repertoire, which suggests a divergence of VDR function between paralogs at the molecular level (22, 27). The identification of divergent NR functions between VDRα and VDRβ may reveal important clues driving VDR evolution.

In this study, we build upon an initial observation that VDRα and VDRβ paralogs from the Japanese medaka exhibit differential sensitivities to 1, 25-dihydroxyvitamin D₃ (1, 25D₃) (22). Here we examine multiple steps in the VDR activation pathway to test the hypothesis that VDR paralogs from two distantly related phyla have undergone functional divergence subsequent to the 3R event.

Materials and Methods

DNA constructs

Medaka and zebrafish VDRα and VDRβ sequences were identified through screening the medaka and zebrafish genomes within Ensemble (28). Primers were designed in Primer3 (29) (see Table 1). Full-length VDRα and VDRβ were amplified and cloned into the TA cloning vector pCR2.1 (Life Technologies) and further subcloned into the pSG5 and pVP16 expression vectors as described elsewhere (22). Full-length VDR cDNAs were additionally subcloned into the pET32a expression vector (Novagen) for bacterial recombinant protein production. All constructs consisted of the complete VDR open reading frame including the internal start and stop codons except for pET32a in which the stop codons were removed for the inclusion of 3′ His tags. Sequence data for zebrafish VDRα (KJ925048) and VDRβ (KJ925049) have been submitted to GenBank. Medaka VDRα (EU403115) and VDRβ (EU403116) were submitted previously (22).

Table 1.

VDR Primer Sequences

Name	Vector	R.S.a	F/R	Sequence (5′–3′)
Zebrafish VDRα	pSG5/pVP16	EcoRI	F	GCG AAT TCG CCA TGC TTA CGG AAA ATA GTG CC
		BamHI	R	ATG GAT CCA AAC TAG GAC ACC TCA CTC C
	pET32a	SalI	F	GAT AGT CGA CTG ATG GAT CTG ATG GCC GTG
		NotI	R	GAT TGC GGC CGC ACT GGA CAC CTC ACT CC
Zebrafish VDRβ	pSG5/pVP16	EcoRI BamHI	F	GCG AAT TCA TGG AGT CAG CTG TCA GTA C
			R	ATG GAT CCA GAA AAC TAG GTG ACC TGC C
	pET32a	SalI	F	GAT AGT CGA CTG ATG GAG TCA GCT GTC AGT
		NotI	R	GAT TGC GGC CGC ACT GGT GAC CTG CCC GCC
Medaka VDRα	pSG5	BglII	F	GAA GAT CTA TGG AGT CCA TTA CCG TGA C
		BglII	R	CGA GAT CTC TAT GAC ACC TCG CTG CCG A
	pVP16	SalI	F	CTC GTC GAC TTA TGG AGT CCA TTA CG TG
		HindIII	R	GCT AAG CTT CTC TAT GAC ACC TCG CTG CC
	pET32a	SalI	F	GAT AGT CGA CTG ATG GAG TCC ATT ACC GTG
		NotI	R	GAT TGC GGC CGC ACT TGA CAC CTC GCT GCC
Medaka VDRβ	pSG5	EcoRI	F	GCG AAT TCA TGG AGG CCA CTG TTG TGA G
		BglII	R	CGA GAT CTC TAG GAG ACC TCG CTG CCA A
	pVP16	EcoRI	F	CCA GAA TTC ATG GAG GCC ACT GTT GTG
		HindIII	R	ATT AAG CTT CTC TAG GAG ACC TCG CTG CC
	pET32a	SalI	F	GAT AGT CGA CTG ATG GAG GCC ACT GTT GTG
		NotI	R	GAT TGC GGC CGC ACT GGA GAC CTC GCT GCC

Abbreviations: F, forward; R, reverse.

^aRestriction sites (R.S.) are indicated in bold lettering.

All coregulator expression constructs for transient transactivation studies (human pSG5-SRC1, pSG5-GRIP1, pCDNA-ACTR, pCDNA-RXR_WT, pCDNA-RXR_AF2) and Gal4 bait constructs for mammalian two-hybrid studies (human pM-SRC1_241–386, pM-GRIP1_479–767, pM-ATCR_392–1005, pM-RXR_WT, pM-RXR_AF2) were a gift from Dr Donald McDonnell (Duke University, Durham, North Carolina). The RXR_AF2 mutant is a truncated retinoid X receptor (RXR) that lacks H12, which contains the activation function-2 (AF2) region, because of an internal stop codon (D444Stop). The luciferase reporters XREM-Luc, 5XGal4-TATA-Luc, and pRL-CMV were obtained as previously reported (22).

Transient transactivation assays

Transactivation studies were conducted in HepG2 cells. This cell line exhibits fast growth kinetics and little to no background VDR activity and maintains active and diffusive transport mechanisms for the uptake of both polar and nonpolar ligands and high transfection efficiencies. HepG2 cells were maintained as described previously (22). Cells were seeded in 96-well plates (2.5 × 10⁴ cells/well) and transfected 24 hours later using Lipofectamine 2000 (Life Technologies) following the manufacturer's recommendations. Cells were transfected with 89.7 ng of medaka or zebrafish pSG5-VDRα or pSG5-VDRβ, 19.2 ng of the XREM-Luc reporter (30), and 4.5 ng of pRL-CMV as an internal luciferase control.

Coregulator studies included 18.3 ng of an expression vector containing the complete open reading frame of the coregulator of interest (pCDNA-RXR_WT or RXR_AF2 and/or pSG5-SRC1, pSG5-GRIP1, or pCDNA-ACTR). Media were replaced the following day with complete MEM containing 120 nM 1, 25D₃ (EMD Millipore) in ethanol (<0.1% total solution) for single-dose assays, or 0–120 nM 1, 25D₃ for concentration-response curves. Luciferase activity was measured 24 hours after the exposure using the Dual-Glo luciferase assay system (Promega) following the manufacturer's protocols. Luciferase readings were normalized by cotransfection with pRL-CMV, and VDR response was normalized to an empty pSG5 vector control. To compare VDR transactivation in the presence and absence of coregulators, VDR+ coregulator response was normalized to VDR response in the absence of coregulators. Experiments were repeated at least twice and performed in replicates of four wells. Single-dose assays were analyzed using one-way ANOVAs followed by a Tukey's honestly significant difference (HSD) post hoc test. The EC₅₀, the 95% confidence interval, and the maximal efficacy (E_MAX) of the concentration-response curves were determined via a linear regression analysis using a sigmoidal dose-response calculation with a variable slope. All statistics were conducted in GraphPad Prism 4 (GraphPad Inc).

Mammalian two-hybrid assays

HepG2 cells were seeded into 96-well plates and transfected 24 hours later as described above. Each well contained 33.6 ng pVP16-VDR, 33.6 ng pM-coregulator as a fusion protein containing the yeast Gal4 DNA-binding domain fused to either full-length RXR (pM-RXR_WT, pM-RXR_AF2), or the defined NR box of one of the steroid receptor coactivator (SRC)/p160 coactivators (31): pM-SRC1_241–386, pM-GRIP1_479–767, or pM-ACTR_392–1005. Luciferase reporters included 126.6 ng 5XGal4-TATA-Luc containing response elements for the yeast Gal4 DNA-binding domain and 3 ng pRL-CMV. Controls consisted of transfections containing empty pM, pVP16, or both empty pM and pVP16 expression vectors or ethanol as a vehicle control. Media were replaced the following day with complete MEM containing 120 nM 1, 25D₃ or ethanol control. The cells were tested for luciferase activity 24 hours after the exposure as described above. VDR-coregulator interaction was normalized to VDR in the absence of a coregulator construct. Experiments were repeated at least twice and performed in replicates of four wells. Results were analyzed via one-way ANOVAs followed by Tukey's HSD post hoc test in GraphPad Prism 4.

Electrophoretic mobility shift assays

Recombinant full-length VDR and RXR_WT proteins for EMSAs were expressed and purified as follows. The protein expression constructs pET32a-VDR or pET32a-RXR_WT were transformed into the BL21(DE3)pLysS strain of Escherichia coli following the manufacturer's protocol (Agilent Technologies). Overnight starter cultures (10 mL) were used to inoculate 250 mL LB/amp and grown at 37°C with shaking until the OD₆₀₀ was approximately 0.6. Protein expression was induced by the addition of 1 mM isopropyl-1-thio-β-D-galactopyranoside and 20 μM ZnCl₂, followed by a 3-hour incubation at 25°C and 200 rpm. Cultures were centrifuged at 4000 × g for 20 minutes at 4°C, the supernatant was discarded, and the pellets were stored at −20°C overnight. The QIAexpress Ni-NTA Fast Start kit (QIAGEN) was used to lyse the bacteria and purify the recombinant His-tagged VDR and RXR proteins under native conditions following the manufacturer's protocol. Purified protein concentrations were determined using the average of three A₂₈₀ measurements using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific). Proteins were visualized via Western blot using an anti-His antibody (number 34660; QIAGEN).

DNA-protein binding reactions were carried out using the canonical vitamin D response element (VDRE) (32) (5′-AGCTTCAGGTCAAGGAGGTCAGAGAGC-3′), a mutant form of the canonical sequence (5′-AGCTTCAGAACAAGGAGAACAGAGAGC-3′), or a VDRE containing the DR3 found within the distal promoter region of the CYP3A4-XREM reporter used in transient transfection studies (5′-GCTGAATGAACTTGCTGACCCTCTGCT-3′) (30). Single-stranded 5′-Cy5-labeled and unlabeled oligos were purchased from Integrated DNA Technologies. EMSAs were carried out in a total volume of 20 μL, containing 100 ng of recombinant VDR and 100 ng recombinant RXR_WT where indicated. Receptors were incubated at 25°C for 45 minutes in a binding buffer (100 mM KCl, 10 mM HEPES, 1 mM EDTA, 0.1 mg/mL BSA, 4 μg/mL sonicated salmon sperm, 1.0 mM dithiothreitol, 1% glycerol, and 20 mM MgCl₂) and 100 nM 1, 25D₃ or ethanol as a vehicle control. After the first incubation, 1 pmol Cy5-labeled double-stranded VDRE was added to each reaction, and incubated for an additional 30 minutes. Competition experiments were performed in the presence of 100-fold molar excess of unlabeled wild-type or mutant VDREs. Negative controls include expressed affinity tags isolated from empty pET32a vector stocks, binding reactions run with either VDR or RXR individually, and ethanol as a vehicle control. Protein-DNA complexes were resolved on a 6% nondenaturing acrylamide gel in ice-cold 0.5× buffer (45 mM Tris base; 45 mM boric acid; 1 mM EDTA, pH 8.0) at 100 V for 90 minutes. Gels were visualized on a Storm 865 (GE Healthcare Life Sciences).

Saturation binding analysis

Recombinant VDR expression was conducted as previously described in Cos7 cells, which are void of endogenous VDR and are highly amenable to expression of heterologous nuclear receptors (25). Briefly, cells were maintained in complete DMEM supplemented with 10% fetal bovine serum and passaged every 4–5 days according to established protocols. To prepare lysates, cells were seeded at 3.0 × 10⁶ cells per 150-mm dish, and transfected the following day with 4 μg pSG5-VDR and 4 μg pSG5-RXR_WT along with 16 μg of pBSII vector as carrier DNA, using Lipofectamine 2000 (Life Technologies). Media were replaced 24 hours after the transfection, and 48 hours after the transfection, cells were washed twice with 10 mL Dulbecco's PBS and harvested by trypsinization. The cells were centrifuged for 5 minutes at 1000 rpm at 4°C and resuspended twice in 2 mL ice-cold Dulbecco's PBS. After the second resuspension, the cells were centrifuged and resuspended in 1 mL of an ice-cold buffer of 0.15 M KCl, 1 mM EDTA, 10 mM Tris HCl, 0.3 mM ZnCl₂, 200× dilution protease inhibitor cocktail (EMD Millipore), and 5 mM dithiothreitol (pH 7.6). Cells were sonicated (12 1 sec bursts at 25% power) and centrifuged at 100 000 × g for 30 minutes at 4°C. The supernatant containing the lysate was divided into aliquots and stored at −80°C.

The affinity of 1, 25D₃ for teleost VDRα and VDRβ was assessed as follows: lysate was diluted 1:20 in an ice-cold buffer of 0.15 M KCl, 1 mM EDTA, 10 mM Tris HCl, 0.3 mM ZnCl₂, 200× dilution protease inhibitor cocktail (EMD Millipore), and 5 mM dithiothreitol (pH 7.6). The radioligand [³H]-1, 25D₃ [1, 25-(OH)₂-26, 27-[³H]dimethyl-vitamin D₃, original specific activity 157 Ci/mmol; PerkinElmer] was diluted to 25 Ci/mmol with unlabeled 1 μM 1, 25D₃ and further diluted with ethanol to obtain the desired concentrations (0–1.0 nM). The appropriate concentration of [³H]-1, 25D₃ (10 μL) was added to 200 μL lysate, shaken, and incubated overnight at 4°C. Unbound ligand was removed with the addition of 80 μL of a 0.5% dextran-2.5% charcoal suspension in a buffer of 150 mM NaCl, 15 mM NaN₃, 100 mM anhydrous Na₂HPO₄, 39 mM NaH₂PO₄ · H₂O, 0.1% gelatin (pH 7.6) incubated on ice for 15 minutes with brief shaking every 5 minutes. Samples were centrifuged at 5000 × g for 5 minutes at 4°C, and 200 μL of the supernatant containing the receptor-bound ligand was removed for scintillation counting. Assays were repeated at least twice, with duplicate tubes for each concentration. Total binding was determined by using lysate transfected with both pSG5-VDR and pSG5-RXR_WT, and nonspecific binding was determined with lysate transfected with the empty pSG5 vector in the place of pSG5-VDR. Specific binding was obtained by subtracting the nonspecific binding counts from the total binding counts. Hyperbolic one-site binding curves were fit using GraphPad Prism 4. Reported dissociation constant (K_d) values are the average of three separate curves ± SEM. Variation between the average K_d values of each VDR was tested via a one-way ANOVA followed by Tukey's HSD post hoc test.

Results

VDR transactivation in response to 1, 25D₃

To analyze differential ligand potency and efficacy, concentration-response assays were conducted using a range of 1, 25D₃ concentrations between 0 and 120 nM, using transactivation as an end point. Results demonstrate that EC₅₀ values were consistently in the low nanomolar range for all VDRs tested, and no significant difference between paralogs or between species was detected (P = .6301) (Figure 1). Conversely, significant variation in E_MAX levels was observed between both species and VDRα and VDRβ paralogs (P < .0001), with medaka and zebrafish VDRβ exhibiting significant attenuation in response compared with their corresponding VDRα paralogs (Figure 1).

Figure 1.

Concentration-response curves for the transactivation of medaka (circles) and zebrafish (triangles) VDRα (open) and VDRβ (closed) in response to 0–120 nM 1, 25 D₃. HepG2 cells were transfected with pSG5-VDR, XREM-Luc, and pRL-CMV as an internal control as described in Materials and Methods. Data are represented as the average fold activation normalized to the vehicle control ± SEM (n = 4).

Saturation binding analysis

Innate ligand binding kinetics for each VDR paralog were determined using saturation binding analysis with radiolabeled [³H]-1, 25D₃ (Figure 2). The K_d values were obtained for each VDR and were all within the subnanomolar range (10⁻¹⁰ M), demonstrating high-affinity binding between 1, 25D₃ and each VDR paralog/ortholog tested. There was no observed statistically significant difference in ligand affinities between VDR paralogs or between species (P = .6198).

Figure 2.

Saturation binding analysis of [³H]-1, 25 D₃. Zebrafish VDRα and VDRβ (A) and medaka VDRα and VDRβ (B) are shown. Lysates were prepared from transfected Cos7 cells and incubated with increasing concentrations of [³H]-1, 25D₃ as described in Materials and Methods. Unbound ligand was removed as described. C, The reported K_d for each VDR is the average of three separate experiments ± SEM. Panels A and B are specific binding data from a representative experiment.

DNA binding

EMSAs were conducted to test the protein-DNA interactions between VDR paralogs and well-defined VDREs. For all VDRs tested, a receptor-DNA binding complex was observed only in the presence of RXR_WT (Figure 3A, odd lanes vs even lanes). No binding complex was observed when VDR or RXR_WT was used individually, indicating a dependence on VDR-RXR heterodimerization for DNA binding and a definitive lack of VDR or RXR_WT binding as putative homodimers. To ensure the specificity of the reaction, all EMSAs were repeated in the presence of excess unlabeled wild-type or mutant VDREs (Figure 3B for canonical VDRE and Figure 3C for XREM VDRE). Competition assays using wild type competitors were effectively able to outcompete the VDR-RXR_WT binding to both the canonical VDRE (Figure 3B, lane 1 vs 3, and lane 4 vs 6) and XREM VDRE (Figure 3C, lanes 1 vs 2 and 3 vs 4 for VDRα and lanes 5 vs 6 and 7 vs 8 for VDRβ). Competition assays using an unlabeled mutant VDRE had no effect on VDR binding to the canonical VDRE (Figure 3B, lane 1 vs 2 and lane 4 vs 5).

Figure 3.

EMSAs of recombinant VDRα and VDRβ binding to canonical and noncanonical VDREs. A, Medaka (M) and zebrafish (ZF) VDRα and VDRβ were analyzed for their ability to form DNA binding complexes with both canonical (lanes 5–8) and noncanonical XREM (lanes 1–4) VDREs. Receptors were analyzed both in the presence (+) and absence (−) of recombinant RXR_WT and 100 nM 1, 25 D₃. B, Competition assays with the canonical VDRE using both unlabeled mutant VDRE (canonical VDRE_MUT) and unlabeled wild-type VDRE (canonical VDRE_WT) as competitors. Assays were run both in the absence (lanes 1–3) and presence (lanes 4–6) of 100 nM 1, 25D₃. Lanes 1 and 4 demonstrate DNA binding in the absence of either competitor. Lanes 2 and 5 include 100-fold molar excess of the mutant competitor, and lanes 3 and 6 include 100-fold molar excess of the wild-type competitor. C, Competition assays with the divergent XREM VDRE using an unlabeled wild-type competitor VDRE (XREM VDRE_WT). Lanes 1–4 represent VDRα, and lanes 5–8 represent VDRβ for medaka (M) and zebrafish (ZF). Assays were run both in the presence (lanes 3 and 4 for VDRα and lanes 7 and 8 for VDRβ) and absence (lanes 1 and 2 for VDRα and lanes 5 and 6 for VDRβ) of 100 nM 1, 25 D₃. Even numbered lanes include 100-fold molar excess of the wild-type competitor, whereas odd numbered lanes demonstrate DNA binding in the absence of the competitor.

Partial VDR-DNA binding was observed in the absence of 1, 25D₃ (Figure 3A, lanes 2 and 6; 3B, lanes 1 and 2; 3C, lanes 1 and 5). However, protein-DNA interaction with all VDREs tested was enhanced in the presence of 1, 25D₃ ligand (Figure 3A, lanes 2 vs 4 and lanes 6 vs 8; 3B, lanes 1 vs 4; 3C, lanes 1 vs 3 for VDRα and lanes 5 vs 7 for VDRβ). VDRα paralogs exhibited enriched protein-DNA interactions compared with the VDRβ paralogs with both the canonical VDRE and distal XREM VDRE (Figure 3, A–C).

Coregulator interaction: heterodimerization with RXR

A series of studies were conducted to determine whether VDR transactivation is significantly enhanced by the presence of supplemented RXR. In transient transactivation studies, the addition of full-length RXR_WT resulted in a significant increase in transactivation of all VDRs compared with the activity of VDR alone (Figure 4A, P < .001 for all). Medaka and zebrafish VDRβ exhibited a greater increase in transactivation compared with their respective VDRα paralogs. In each instance in which the RXR_AF2 mutant was substituted for RXR_WT, luciferase activity was significantly attenuated as expected compared with studies with RXR_WT.

Figure 4.

Assessment of VDR-RXR interactions. A, Analysis of overexpressed RXR on VDR transactivation in response to 120 nM 1, 25-D₃. Data are represented as the average fold activation normalized to VDR alone (no RXR) ± SEM (n = 4). Asterisks above bars represent a significant increase in VDR transactivation in the presence of RXR compared to the control (no RXR). ***, P < .001; **, P < .01; *, P < .05. Asterisks above the brackets indicate a significant difference between VDRα and VDRβ. B, Mammalian two-hybrid analysis of VDR-RXR interaction in response to 120 nM 1, 25 D₃. Data are represented as the average fold interaction normalized to VDR alone (no pM-RXR). Asterisks above bars represent either a significant interaction between VDR and RXR compared with the control. ***, P < .001; **, P < .01; *, P < .05. Asterisks above brackets indicate a significant differences in VDR-RXR interaction in the presence and absence of SRC1. The symbol (dagger) indicates that the interaction between medaka VDRβ and RXR was found to be significant with an unpaired t test compared with VDRβ alone: t₆ = 13.71, P < .0001.

VDR-RXR heterodimerization was further confirmed through mammalian two-hybrid assays (M2H). Identical assays were conducted using the mutant RXR_AF2. Data from these studies demonstrate a direct interaction between each teleost VDR tested and pM-RXR_WT in the presence of 1, 25D₃ (Figure 4B). Protein-protein interactions were not observed in the absence of ligand, and VDR-RXR interaction was significantly attenuated with the substitution of pM-RXR_WT with pM-RXR_AF2, highlighting the necessity of the C terminal AF2 domain of RXR for VDR transactivation. The addition of SRC1 significantly increased the interaction between RXR_WT and VDRβ for both species. In contrast, a much smaller increase in interaction was observed between RXR_WT and medaka VDRα in the presence of SRC1, whereas SRC1 did not have an effect on zebrafish VDRα and RXR_WT.

Coregulator interaction: SRC/p160 family of NR coactivators

Protein interaction studies were expanded to include the SRC/p160 family of NR coactivators. This family includes SRC1, glucocorticoid receptor interacting protein-1 (GRIP1), and activator of thyroid and retinoic acid receptor (ACTR). Cotransfection of the SRC/p160 coactivators in transient transactivation assays had varying effects on the VDR paralogs. SRC1 significantly increased VDR activation for each VDR tested (Figure 5, A–D). GRIP1 enhanced only zebrafish VDRα and VDRβ activation but did not affect medaka VDRα or VDRβ. ACTR proved ineffective, exhibiting no enhancement of VDR transactivation.

Figure 5.

VDR transactivation with supplemented SRC/p160 coactivators: analysis of VDR transactivation in the presence of overexpressed SRC1, GRIP1, and ACTR, both in the presence (+) and absence (−) of overexpressed RXR_WT in response to 120 nM 1, 25D₃. Data are represented as the average fold activation normalized to VDR alone ± SEM (n = 4). Asterisks above the bars indicate a significant increase in VDR transactivation compared with the VDR alone control. ***, P < .001; **, P < .01; *, P < .05. Asterisks above the brackets indicate a significant difference in transactivation between the presence and absence of RXR_WT. The letter b indicates cotransfection with both RXR_WT and SRC/p160 coactivators had a greater effect on VDR transactivation compared with VDR + RXR_WT in the absence of an SRC/p160 construct (black bar). The letter a indicates that no difference in VDR transactivation was observed in the presence of both RXR_WT and the indicated SRC coactivator vs. VDR + RXR_WT in the absence of the SRC/p160 coactivator.

Cotransfection of both RXR_WT and the SRC/p160 coactivators significantly enhanced the activation of each VDR tested in the transactivation studies (Figure 5). The level of transactivation achieved for all four VDRs tested were greater with the combination of both RXR_WT and SRC1 compared with either coactivator used individually. However this combination of coregulators appeared to have a preferential effect on VDRβ activation compared with VDRα for both medaka and zebrafish. The enhanced VDR transactivation was additionally demonstrated with the cotransfection of RXR_WT combined with GRIP1 or ACTR in zebrafish; however, these combinations of coregulators did not augment the transactivation of medaka VDR paralogs (Figure 5).

Mammalian two-hybrid assays were conducted to assess the protein-protein interaction between VDR and individual SRC/p160 coactivators (Figure 6, A–D). Given that the cotransfection of both RXR_WT and SRC/p160 coactivators significantly enhanced VDR transactivation, M2H assays were conducted in the presence and absence of RXR_WT to determine whether the obligate heterodimer partner facilitates the interaction between VDR and the SRC/p160 coactivators. First, M2H assays conducted in the absence of RXR_WT demonstrate that both VDRα and VDRβ from both species maintain a strong and significant interaction with SRC1, suggesting that both VDR paralogs directly recruit SRC1 in response to 1, 25D₃ (Figure 6, A–D). Comparatively, only VDRα from both species demonstrated a significant interaction with GRIP1 in the absence of RXR_WT (Figure 6, A and C vs B and D). No protein-protein interactions were observed between ACTR and any of the VDRs tested. The addition of RXR_WT to the SRC/p160 M2H assays had contrasting effects on coactivator recruitment. RXR_WT significantly enhanced VDRβ interaction with both SRC1 and GRIP1, consistent with that observed in transient transactivation assays. Conversely, addition of RXR_WT to the SRC/p160 M2H assays with medaka and zebrafish VDRα resulted in an attenuated response compared with the same assay in the absence of RXR_WT (Figure 6, A vs B and Figure 6, C vs D).

Figure 6.

Analysis of VDR recruitment of members of the SRC/p160 family of NR coactivators in response to 120 nM 1, 25D₃. Data are represented as the average fold interaction ± SEM (n = 4). Data are normalized to VDR in the absence of coregulators. Asterisks above the bars indicate a significant interaction between VDR and SRC-1, GRIP1, or ACTR compared with the VDR alone control. ***, P < .001; **, P < .01; *, P < .05. Asterisks above the brackets indicate a significant difference in VDR-SRC/p160 interaction in the presence and absence of RXR_WT.

Discussion

The duplication and divergence of transcriptional regulators is thought to have played an essential role in the evolution of vertebrate complexity and diversity. Mining of teleost genomes has identified numerous instances of duplicate and paralogous transcriptional regulators. After the discovery of VDRβ, it was determined that it is paralogous to VDRα and represents a novel subfamily of ligand-activated transcription factors within the vertebrate NR1I family (15, 22, 23). To date, however, molecular function and physiological significance of duplicate teleost NRs is based predominantly on structural similarities and orthology to their mammalian counterparts.

Using transient transactivation assays, we first confirmed that VDRα and VDRβ from medaka and zebrafish are capable of transactivation in response to the endogenous VDR ligand, 1, 25D₃. By titrating the concentration of 1, 25D₃, we established an EC₅₀ value for each VDR. EC₅₀ values were highly similar between each VDR tested, indicating that 1, 25D₃ is a highly potent ligand both between (orthologs) and within (paralogs) species. These data are consistent with that observed from other teleost VDR studies that used Gal4-VDR_LBD chimeras to assess VDR transactivation across phylogenetically diverse taxa (22, 33). All VDRs examined to date exhibit at least modest transactivation and comparable EC₅₀ values with 1, 25D₃, including the sea lamprey, a primitive vertebrate (25). This suggests that the association between VDR and 1, 25D₃ occurred early in vertebrate evolution and was highly conserved.

Conversely, the examination of maximal transactivational efficacy (E_MAX) between the VDR paralogs of both species revealed significant differences. We sought to further establish the functional basis for these transactivational differences through examining core NR functions between VDRα and VDRβ in the two teleost species. Because ligand binding is the initiating event of NR activation, we initially investigated the possibility that the observed transactivational differences between VDRα and VDRβ were associated with differential affinity for 1, 25D₃. VDRα and VDRβ from both species exhibited high affinity binding with 1, 25D₃, similar to previously reported values for mammalian VDR proteins. However, statistically significant differences in ligand affinities were not observed between paralogs or species, indicating that differential ligand affinities are likely not the root cause of the observed transactivational disparities. In fact, it is well known that ligand affinity and transactivation efficacy of specific NR ligands are not proportional (34). This phenomenon is likely due to the ability of ligands to induce subtle variations in receptor conformation that elicit different biological responses such as DNA binding and protein-protein interactions. High ligand affinity but low transactivational efficacy of VDR has been observed previously in other aquatic vertebrates. For example, the sea lamprey (Petromyzon marinus) VDR exhibits high affinity binding with 1, 25D₃ (K_d = 10⁻¹⁰ M); however, transactivation of this VDR with 1, 25D₃ is minimal (<4.0-fold) (25).

We next examined whether differences in VDRα and VDRβ transactivation were due to modified NR functions that occur subsequent to ligand binding. We began by assessing the ability of the VDRs to bind to well-characterized VDREs. Canonical VDREs are comprised of a direct-repeat of two hexameric half sites separated by a spacer of three nucleotides (referred to as a DR3). Our results demonstrate that distinct differences in DNA binding occur between the VDRα and VDRβ paralogs, with VDRα exhibiting preferential binding compared with VDRβ. DNA binding differences were conserved between species and occurred with both canonical and noncanonical VDREs. Sequence similarities suggest that the observed DNA binding differences between VDRα and VDRβ are likely not related to sequence divergence in the DNA binding domain (DBD). The DBD sequences of the four VDR paralogs are highly homologous, sharing 95%–96% amino acid sequence similarity (Figure 7A). All eight cysteine residues essential for the zinc finger structure and the recognition helix are identical between VDRα and VDRβ and human VDR (35–37).

Figure 7.

Comparison of VDR paralog functional domains. Amino acid sequence alignment of medaka (M) and zebrafish (ZF) VDRα and VDRβ. Human VDR (H) is included for comparison. Predicted amino acid sequences were aligned using ClustalW. Functional domains and domain elements were determined based on previous work with human VDR discussed in the manuscript. Fully conserved residues are indicated by an asterisk. Partial conservation is indicated by a colon (strong) or period (weak). Lack of residue consensus is indicated by a blank space. For the DNA-binding domain (A), residues that make up the P-box and D-box are indicated with a black box, and helix 1 (the recognition helix) and helix 2 are highlighted in grey. Conserved cysteine residues involved in zinc finger structure are both bold and underlined. For the CTE/hinge region (B), the T-box is highlighted in black with white letters, whereas the A-box is highlighted in grey with black letters. For the ligand-binding domain (C), the estimated α-helices are highlighted in grey. Amino acid residues that serve as contacts for 1, 25 D₃ are indicated in white letters highlighted in black. The residues involved in the charge clamp formation are both bold and italicized. The E1 region within H3 and H4 is indicated with a solid black line above the alignment that includes part of H3 and H4.

Additionally, studies have demonstrated the importance of residues in the C-terminal extension (CTE) for sufficient DNA binding and transactivation (38) (Figure 7B). These residues likely increase the affinity of VDR-DNA interaction by aiding in nonspecific interaction within the minor grove of the DNA backbone (T-box) and impart additional RXR dimerization and sequence specificity to the conserved DBD core region (A-box) (39, 40). Although more sequence variation is observed in the CTE region between VDRα and VDRβ, the basic residues implicated in DNA binding are conserved across all medaka and zebrafish VDR paralogs (32). Mutation studies have also demonstrated that modifications within the LBD may impact DNA binding through alterations in RXR heterodimerization (41, 42). Our DNA binding studies demonstrate a stringent requirement for RXR for binding complex formation. In addition, we demonstrate that DNA binding is significantly enhanced with addition of 1, 25D₃. This suggests that the presence of ligand may favorably alter the conformation of the VDR LBD, resulting in enhanced VDR-RXR interactions and facilitating the interaction with the VDRE (43–45).

To test the impact of RXR on VDR transactivational efficacy, either wild-type RXR (RXR_WT) or a truncated RXR mutant (RXR_AF2) lacking the C-terminal ligand-dependent AF2 region was cotransfected with VDR in transient transactivation studies. We consistently observed an increase in VDR transactivational efficacy in the presence of RXR_WT. The use of the RXR_AF2 mutant attenuated VDR transactivation to background levels and significantly attenuated the interaction between VDR and RXR in M2H studies. Previous studies have demonstrated that the AF2 region of both VDR and RXR are required for vitamin D-dependent gene transcription (46, 47). We speculate that the partial loss of a response with RXR_AF2 is likely due to either an inability to recruit necessary coactivators due to the loss of the charge clamp or to corepressor interactions on sites previously shielded by the RXR AF2 (48). In conjunction with DNA binding data, these results support obligate interactions between both NRs for essential VDR functions in both medaka and zebrafish paralogs.

In addition to RXR, we also studied the effects of the SRC/p160 family of nuclear receptor coactivators on VDR transactivation. Members of the SRC/p160 family directly interact with the AF2 region of NRs and enhance NR-mediated transcription of target genes by both chromatin modification and the recruitment of additional proteins to the transcription complex (31). Although the partnership between VDR and RXR appears to be well conserved, dramatic functional differences between VDRα and VDRβ were observed with the cotransfection of both RXR_WT and the SRC/p160 coactivators in transient transactivation assays. Specifically we observed preferential transactivation of VDRβ paralogs over VDRα in the presence of overexpressed SRC1 and RXR_WT in both species. This transactivational increase was significantly greater than the individual effects of either coregulator used singularly. In addition, results from our M2H assays further support the notion of preferential effects of RXR_WT and SRC1 on VDRβ transactivation.

Examination of key regions within the VDR known for protein-protein interaction with RXR and the SRC/p160 coactivators, including helices 9 and 10, the AF2 and E1 region, and the T box of the CTE, all demonstrate a high degree of conservation between VDRα and VDRβ paralogs (Figure 7). Furthermore, amino acid variation within these regions is usually not conserved in both VDRα and VDRβ paralogs, suggesting that the observed variation may not explain the paralog-specific functional differences. Second, the fact that both VDRα and VDRβ bind 1, 25D₃ with equal affinity indicates that differential ligand affinities are not driving differences in protein interactions between paralogs. Rather, we speculate that receptor-ligand interactions between VDR and 1, 25D₃ may result in different molecular conformations in VDRα vs VDRβ that can differentially influence downstream functions. Mammalian VDR bound to 1, 25D₃ produces a significant conformational change in the receptor from the inactive apo to the active holo state (49). This conformational change appears to be consistent across all mammalian species tested thus far (50, 51) and crystal studies with zebrafish VDRα bound to both 1, 25D₃ and Gemini (a synthetic VDR agonist), reveal a highly similar active conformation compared with human VDR (52). This supports our data indicating a high degree of functional similarity between the two species and provides an initial indication that the VDRα paralog more closely resembles the terrestrial ortholog of this nuclear receptor. Crystal structures of both VDRα/β paralogs with 1, 25D₃ may provide invaluable structural comparisons because the differences observed between the two paralogs may be related to conformational differences and provide additional detail on the evolution of aquatic and terrestrial forms of this important nuclear receptor.

It has additionally been demonstrated that the active conformation of VDR is ligand specific. Jurutka et al (53) have suggested that lithocholic acid induces an alternative active conformation when bound to VDR compared with 1, 25D₃, leading to differential interactions with RXR and other coactivator proteins. These studies further demonstrate that ligand-induced conformational differences are associated with altered protein-protein interactions with coactivators and heterodimerization partners due to suboptimal positioning of helices 9, 10, and 12 (53, 54). These diminished protein-protein interactions ultimately result in significantly reduced VDR transactivation. Although 1, 25D₃ was the only ligand in our study, we speculate that subtle conformational differences between VDR paralogs may lead to the altered coregulator interactions described above. Specifically, the active conformation induced by 1, 25D₃ in VDRβ may be less stable compared with VDRα, resulting in attenuated affinity for RXR and the SRC/p160 coactivators under nonsaturating coactivator conditions. However, enhanced activity in transient transactivation studies and enhanced protein interactions observed with VDRβ may be due to a compensatory effect of overexpressed coregulators. Analogous to our observations, compensatory effects of overexpressed coregulators have been previously observed with a VDR_R391C mutant associated with a form of hereditary vitamin D-resistant rickets (43, 47).

Two possible models may explain the compensatory mechanism of overexpressed RXR and SRC1 on VDR transactivation. One mechanism involves the AF2 regions of both VDR and RXR interacting with different LXXLL motifs of the same SRC/p160 coactivator (47). This bridging effect may help stabilize a less optimal heterodimer. The pattern of one coactivator per heterodimer has been previously observed with other NR heterodimers such as RXR-RAR (55). Alternatively, a second model suggests that each heterodimer partner interacts with a distinct and separate coactivator. A study by Yang et al (56) found that coactivator recruitment to the permissive peroxisomal proliferator-activated receptor-γ-RXR heterodimer is highly ligand specific. For instance, the presence of LG268, an RXR-specific ligand, resulted in protein-protein interaction only between SRC-1 and RXR and not peroxisomal proliferator-activated receptor-γ. Differential coactivator recruitment between heterodimer partners may potentially explain our M2H data with VDRα. The fact that the addition of RXR_WT appears to attenuate VDRα-SRC1 interactions in M2H assays suggests that RXR_WT may be sequestering the SRC1 proteins, such that they are effectively no longer stoichiometrically proportional for the assay.

In summary, we have identified evidence of a functional divergence between VDRα and VDRβ paralogs in teleost fish. Our results suggest that differences in protein-protein interactions between the VDR paralogs and essential coregulators may be driving the observed differential ligand sensitivities between VDRα and VDRβ. Furthermore, this divergence occurred early in teleost evolution because the observed functional differences are highly conserved in two distantly related species. Further studies using VDRs cloned from species that diverged before the 3R event are necessary to gain a better understanding of derived and ancestral VDR functions of the VDR paralogs. Additional studies may also aid in identifying specific evolutionary processes that lead to the observed functional differences.