Morpholino-Mediated Knockdown of ERα, ERβa, and ERβb mRNAs in Zebrafish (Danio rerio) Embryos Reveals Differential Regulation of Estrogen-Inducible Genes

Abstract

Genetically distinct estrogen receptor (ER) subtypes (ERα and ERβ) play a major role in mediating estrogen actions in vertebrates, but their unique and overlapping functions are not entirely clear. Although mammals have 1 gene of each subtype (ESR1 and ESR2), teleost fish have a single esr1 (ERα) and 2 esr2 (ERβa and ERβb) genes. To determine the in vivo role of different ER isoforms in regulating estrogen-inducible transcription targets, zebrafish (Danio rerio) embryos were microinjected with esr-specific morpholino (MO) oligonucleotides to disrupt splicing of the exon III/intron III junction in the DNA-binding domain. Each MO knocked down its respective normal transcript and increased production of variants with a retained intron III (esr1 MO) or a deleted or mis-spliced exon III (esr2a and esr2b MOs). Both esr1 and esr2b MOs blocked estradiol induction of vitellogenin and ERα mRNAs, predominant hepatic genes, but esr2b was the only MO that blocked induction of cytochrome P450 aromatase B mRNA, a predominant brain gene. Knockdown of ERβa with the esr2a MO had no effect on estrogen induction of the 3 mRNAs but, when coinjected with esr1 MO, attenuated the effect of ERα knockdown. Results indicate that ERα and ERβb, acting separately or cooperatively on specific gene targets, are positive transcriptional regulators of estrogen action, but the role of ERβa, if any, is unclear. We conclude that MO technology in zebrafish embryos is an advantageous approach for investigating the interplay of ER subtypes in a true physiological context.

Estrogen receptors (ERs) play a major role in regulating estrogen-dependent processes in reproductive and nonreproductive tissues of both male and female vertebrates (for review, see Refs. ^{1, 2}). ER proteins that mediate the classical (genomic) pathway of estrogen action are ligand-activated transcription factors that bind as dimers to estrogen response elements (EREs) in the promoters of estrogen-responsive target genes, or are tethered to other DNA regulatory elements through protein-protein interactions with transcription factors such as activator protein-1 and steroidogenic factor-1 (1, 3). Like other members of the steroid receptor superfamily, ERs are modular proteins with distinct functional domains (A–F): the constitutively active activation function (AF)1 (A/B) at the N terminus, the DNA-binding domain (DBD) (D) and ligand-binding domain (LBD) (E), and the ligand-dependent AF2 in the LBD domain (E/F) (4). Mammalian species have 2 genetically distinct ER subtypes, termed α and β, encoded by ESR1 and ESR2. The 2 receptors have different but overlapping expression patterns and also differ in their ligand and ERE binding characteristics, coregulator recruitment, transactivation functions, and global gene expression profiles, all of which suggests distinct or selective as well as convergent biological effects (for review, see Refs. ^{5, 6}). Further complicating ER physiology, ERα and ERβ proteins can form homo- or heterodimers and functionally compete or cooperate on consensus or variant EREs in cells where they are coexpressed (for review, see Refs. ^6–8). Much of what is known about ERα and ERβ biology comes from studies with cell lines engineered to express one or both ERs, but significant additional knowledge has been contributed by studies in ERα, ERβ, and ERα/β knockout mice, which display very different phenotypes (9, 10). Paradoxically, results from different laboratories using different gene knockout strategies are contradictory, especially as regards the functions of ERβ and its interactions with ERα. Considering the importance of estrogen signaling in normal development and physiology, the therapeutic potential of ER subtype-selective ligands in pathophysiological conditions, and the need to accurately predict the risk to humans and animals of diverse estrogen-like environmental chemicals, there is considerable interest in understanding the interplay of ERα and ERβ in a normal in vivo context.

With the exception of cartilaginous fish, which have only a single, β-like ER gene (11), separate ERα and ERβ subtypes have been conserved through vertebrate evolution, in itself signifying adaptive value (12). Unique among vertebrates, however, teleost (bony) fish species have a single esr1 but 2 esr2 genes, encoding ERα, ERβa, and ERβb (13), which are thought to be due to a whole genome duplication event in the ancient fish lineage (14). Where duplicates have been retained over evolutionary time, the functions of the ancestral gene are often allocated between the 2 paralogs (subfunctionalization) (15), a feature that could help clarify the actions and interactions of the ERα and ERβ subtypes. A number of studies have attempted to address subtype-specific functions of fish ERs based on ER-dependent reporter gene activity in transfected human or fish cells (16–18) or the measurement of estrogen-inducible transcripts in fish cells using subtype-specific RNA interference (19) or subtype-selective ligands (19–22). Although cell culture experiments indicate what is possible, they do not explain what occurs in a true physiological context. Further, the selectivity of ligands developed for mammalian ER subtypes is not necessarily the same with fish ER, nor can it clearly discriminate between the 2 fish ERβs.

Zebrafish (Danio rerio) embryos have many advantages for addressing this problem. In particular, an ex utero developmental program facilitates experimental observation and manipulation without the complications of maternal effect. Additionally, microinjection of morpholino (MO) oligonucleotides permits observation of the in vivo effects of a temporary knockdown of function by binding and sterically blocking either mRNA splicing or translation of a targeted gene, without the drawbacks of finding or creating a deletion mutant (23, 24). Each of the 3 zebrafish ER genes, ERα, ERβa, and ERβb (formerly ERβ2 and ERβ1) (13, 25–27), has a distinct ligand binding profile, tissue distribution, and developmental program (28–31). The use of zebrafish and other fish species for testing estrogen-like chemicals has generated substantial information on estrogen-response profiles and ER-mediated gene expression changes (32–39).

Among the established estrogen-responsive genes in teleost fish are cyp19a1b, encoding the predominant brain form of cytochrome P450 aromatase B (AroB), which catalyzes the conversion of androgen to estrogen (40, 41), and vitellogenin 1 (vtg), a member of a multigene family encoding yolk proteins and expressed in liver (37, 42, 43). Both cyp19a1b and vtg genes in fish are direct targets of estrogen action, as indicated by at least 1 ERE and several ERE half-sites in their promoters (44–47). Additionally, ERα itself is estrogen inducible (31, 43, 48) from 1 of its 2 promoters (49), indicating its dual role as a mediator and a target of estrogen action. ERα is widely expressed, but the estrogen-responsive transcript is highest in liver. Conveniently, estrogen induction of these 3 mRNAs is measureable between 24 and 48 hours postfertilization (hpf) (30, 39, 41, 50), well before many microinjected MOs are degraded (51, 52).

To test the hypothesis that ERα, ERβa, and ERβb differentially regulate this small panel of estrogen-inducible mRNAs (Vtg, AroB, and ERα), fertilized eggs were microinjected with MOs targeting the exon III/intron III splice junction in the DBD of each of the 3 esr genes. Splice blocking MO have advantages over translation blocking MO in that disruption of mRNA splicing can be confirmed by RT-PCR when isoform-specific antibodies are not available. Also, zygotic mRNAs can be targeted specifically without interference from previously spliced maternal transcripts (24). Additionally, by targeting sequence across the exon-intron junction, a splice-site MO can discriminate among members of the same gene family, even in coding regions with a high degree of sequence identity, such as the DBD of the ERs. To determine the effectiveness, duration, and specificity of the MO knockdown effect, normal and mis-spliced ERα, ERβa, and ERβb transcripts were analyzed using RT-PCR, and the identity of the amplicons was confirmed by sequence analysis. The effects of ER subtype-specific knockdown on basal and estradiol (E2)-induced expression levels of the 3 transcription targets were measured using real-time quantitative (Q)PCR. Our results show that estrogen regulation of gene expression is ER subtype and target gene specific and demonstrate that interactions involving ERα and one or both of the ERβs are already operative in early development and at this phyletic level.

Materials and Methods

Morpholinos

MO oligonucleotides were designed and synthesized by Gene Tools, LLC (Philomath). MOs for esr1, esr2a, and esr2b targeted the splice site between exon III and intron III of their respective pre-mRNA sequences (Table 1). Negative control MOs were the standard control oligonucleotide offered by Gene Tools, which targets human β-globin and is believed not to correspond to any zebrafish gene, and a custom 5-base mismatch MO targeting the esr1 gene. For each gene-specific MO, MOs for the other 2 ER genes served as additional controls. MOs were resuspended upon receipt in 1× Danieau buffer (58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO4, 0.6 mM Ca(NO₃)₂, and 5 mM HEPES [pH 7.6]) at a stock concentration of 2.0 mM and stored at 4°C until use, with aliquots for long-term storage at −20°C. For injection, MOs were diluted to working concentrations in 1× Danieau buffer with 0.005% rhodamine B (Sigma).

Table 1.

Oligonucleotide Sequences

Gene	Oligo	Application	Position		Sequence, 5′-3′	GenBank
			Nucleotides	Exon/Intron
esr1	A-MO	MO	728-a	E III/I III	F-catgtaaaacaggctggtcacCTTG	a
	A-F1	RT-PCR	475–494	E II	F-CTACCTGGATTCCTCGTCCA	AF349412
	A-R1	RT-PCR	971–990	E V	R-CGCTTCTTCCTCTTGTCCTG	AF349412
	A-R2	RT-PCR	ND	I III	R-ttaaaacaaaacccccacca	a
	A-Q1	QPCR	1056–1073	E V	F-GAGCCACCCGCTGTCT	AF349412
	A-Q2	QPCR	1134–1156	E V	R-CGAGTTCTTTGTCAGCCATGT	AF349412
	A-Q3	QPCR	656–680	E III	F-GACTACGCCTCTGGATATCATTACG	AF349412
	A-Q4	QPCR	748–765	E IV	R-TGGTTGGTCGCTGGACAA	AF349412
	MM Ctrl	MO	728-a	E III/I III	catctaaaagaggctcgtgacGTTG	a
esr2a	Ba-MO	MO	849-a	E III/I III	F-agagagtcttacCTTGTATACTC	a
	Ba-F1	RT-PCR	394–413	E II	F-TCCTCTCCATCTCTGCCAGT	AF516874
	Ba-R1	RT-PCR	1061–1080	E V	R-CTCTTGAGACCTCGGACCAG	AF516874
	Ba-R2	RT-PCR	ND	I III	R-tgcctcaacacacctttctg	a
	Ba-Q1	QPCR	931–947	E IV	F-GCCTGCCGACTCCGAAA	AF516874
	Ba-Q2	QPCR	984–1008	E V	R-TTGTTGGTAGCTGCTACGATCTCT	AF516874
esr2b	Bb-MO	MO	673-a	E III/I III	ttgaccatgagcattacCTTGAATG	a
	Bb-F1	RT-PCR	217–236	E II	F-GTGGAGGCCTGTCAGGATTA	AF349413
	Bb-R1	RT-PCR	913–932	E V	R-CCTGAAGGAAAGAGGTGCTG	AF349413
	Bb-R2	RT-PCR	ND	I III	R-tgacatacatgcacgcaaaa	a
	Bb-Q1	QPCR	740–760	E IV	F-CGACTCCGCAAGTGCTATGAA	AF349413
	Bb-Q2	QPCR	803–823	E V	F-ACGATGTCGAGCACCTCGAT	AF349413
cyp19a1b (AroB)	Aro-Q1	QPCR	704–733	E V	F-AAAGAGTTACTAATAAAGATCCACCGGTAT	AF226619
	Aro-Q2	QPCR	832–852	E VI	R-TCCACAAGCTTTCCCATTTCA	AF226619
vtg	V-Q1	QPCR	2752–2773	E XX	F-TGAGACTTATGCCGTGGTCAGA	AF406784
	V-Q2	QPCR	2829–2847	E XX	R-GATGCCTGGGAGTTTTGCA	AF406784
bactin	Act-Q1	QPCR	722–740	E IV	F-CGAGCAGGAGATGGGAACC	AF057040
	Act-Q2	QPCR	805–823	E IV	R-CAACGGAAACGCTCATTGC	AF057040
N/A	Std Ctrl	MO	N/A	N/A	CCTCTTACCTCAgTTACAATTTATA	N/A

ND, not determined; N/A, not applicable; Std Ctrl, standard control; caps, exonic sequence; bold, mismatch bases in control.

^aSee Ensembl for intron sequence.

Zebrafish embryo microinjection and treatments

Adult wild-type zebrafish (Danio rerio) were obtained from a commercial supplier (EkkWill). Fertilized eggs were obtained by natural spawning and incubated at 28.5°C in a 14-hour light, 10-hour dark cycle according to standard protocols in our laboratory as previously described (41, 50, 53). Injected embryos (1–16 cell stage, <2 hpf; Nanoliter Injector, World Precision Instruments) were screened the next day for rhodamine fluorescence, mortality, motility, and early developmental defects. Optimal MO injection concentrations (as determined in pilot experiments) were 0.25 mM, 0.25 mM, and 0.75 mM for esr1, esr2a, and esr2b, respectively, except where specified (all 4.6-nL injection volume). Mismatch and standard MO controls were injected at 0.25 mM. Negative controls were injected with 1× Danieau-rhodamine buffer.

Beginning at 24 hpf, injected embryos, and a corresponding number of uninjected embryos from the same spawning cohort, were cultured in media containing E2 (0.1μM in 0.1% dimethylsulfoxide [DMSO]; Sigma) or vehicle alone and replenished daily. At the end of the specified culture period (48–120 hpf), embryos/larvae were pooled in groups of 15–30, depending on age, in 1.5-mL microfuge tubes, quick frozen on dry ice, and stored at −70°C. In all experiments, embryos/larvae were of mixed sex, because sex is undetermined in zebrafish until approximately 21 days postfertilization (for review, see Ref. 54). The number of independent biological experiments in each treatment condition is indicated in the figure legends.

RNA processing and semiquantitative PCR (RT-PCR)

As described previously (50, 53), pooled frozen embryos were homogenized in TRI Reagent (Sigma). The RNA extracts were treated with DNase I (Roche) and spectrophotometrically checked for quality and quantity. cDNA synthesis was performed using 3.5-μg total RNA template, primed by oligo d(T) (Promega), and reverse transcribed using Superscript II (Invitrogen).

The effects of MO injection on the size and number of ER mRNAs expressed in each sample and experiment were assessed by RT-PCR using subtype-specific primers (Table 1), followed by agarose gel electrophoresis. Primers were designed using the Primer3 or Primer-BLAST websites and synthesized by Invitrogen (Table 1). Cycling conditions were optimized for each primer set.

Sequencing of ER splice products

Product bands from semiquantitative RT-PCR were gel purified using the QIAquick gel extraction kit (QIAGEN), used as template in a second-round PCR under the same conditions, gel purified a second time, and sent for single-extension sequencing (Macrogen USA). Where specified, gel-purified first-round product was cloned using the p-GEM T easy kit (Promega) and sequenced using M13 forward and reverse primers (MWG Operon). Sequencing data were compared with the expected Ensembl sequence using pairwise BLAST or ClustalW alignment. In silico translation was performed using the website http://insilico.ehu.es (55).

Real-time QPCR

QPCR analysis was performed on an ABI Prism 7900HT system (Applied Biosciences), using gene-specific oligonucleotide primers (Table 1) and 2× SYBR Green, as previously described (50, 53). Data were normalized to β-actin (bactin), an appropriate housekeeping gene for developing zebrafish (53). Mean normalized expression (MNE) of each sample was calculated using the Qgene application (56). GraphPad Prism 5 (GraphPad Software) was used for statistical analysis. Data were analyzed by 1- or 2-way ANOVA with Tukey's or Bonferroni's post hoc tests to determine which values differed significantly (see figure legends). Where specified, Student's t test was applied. Significance was set at P < .05.

Animal experimentation

All animal experimentation described in this article was conducted in accord with accepted standards of humane animal care, as monitored by Boston University Institutional Animal Care and Use Committee. As outlined in Institutional Animal Care and Use Committee protocol (project number 07-008), adult animals were maintained as breeding populations to obtain fertilized eggs by natural spawning for these experiments.

Results

Knockdown of ERα mRNA using esr1 MO

The esr1 MO targeted the exon III/intron III splice junction, a region encoding the first zinc finger of the DBD (Table 1 and Figure 1A). To determine the effects on pre-mRNA splicing, RT-PCR was performed using primers positioned in exons II and V (primers A-F1 and A-R1, Figure 1A). RT-PCR of cDNA from uninjected control embryos produced a 519-bp product consistent with the size of normally spliced (N) mRNA. Injection of the esr1 MO dramatically reduced the intensity of the normal ERα cDNA band in both DMSO- and E2-treated embryos, but no variant band was detected (Figure 1B1). We interpret this as indicative of intron retention, because the PCR amplification conditions would not be expected to amplify the very large approximately 6.8-kb intron III. To test this assumption, the exon II forward primer was paired with a reverse primer in the 5′ end of intron III (primers A-F1 and A-R2, Figure 1A). A variant 706-bp product was amplified from MO-injected embryos and the intensity of this band increased with E2 treatment (Figure 1B2). The size of this product indicated that it was amplified from cDNA, not genomic DNA, which would have included the large intron II. Both the normal and the variant band identities were verified by sequencing (Supplemental Figure 1, published on The Endocrine Society's Journals Online web site at http://endo.endojournals.org). In silico translation of the sequence of the intron III-retained ER variant indicates the presence of multiple stop codons within 15–60 codons of the targeted splice junction. If translated, this transcript would be terminated shortly into the intron-retained sequence, resulting in a truncated protein with an incomplete DBD and no LBD. To test the specificity of the ERα knockdown effect, RT-PCR with the ERα primer set A-F1 and A-R1 was performed on samples injected with the esr2a MO (Figure 1B3), the esr2b MO (Figure 1B4), and the esr1 mismatch MO (Figure 1B5). No abnormal expression levels or splice variants of ERα mRNA were seen with any MO injection other than esr1 MO. Data shown in Figure 1 are from 96-hpf embryos injected with 1 mM esr1 MO, but the knockdown effect was essentially the same at 48, 72, and 120 hpf and with 0.25 mM esr1 MO (data not shown).

Figure 1.

MO-mediated targeting of esr1 knocks down expression of wild-type ERα mRNA and induces an intron III-retained splice variant. (A) Diagram of the zebrafish esr1 gene shows translated and untranslated exons (shaded and unshaded boxes, respectively; I–IX), introns (horizontal line), translation start sites (vertical arrows), stop codon (asterisk), DBD, LBD, and approximate locations of the oligonucleotide MO (horizontal bar) and primers used for RT-PCR (F1, R1, and R2) and real-time QPCR (Q1–Q4) analysis (arrowheads). Numbers in italics indicate the number of nucleotides in affected exons/introns. See Table 1 for MO and primer nt positions. (B) Representative agarose gels showing ERα mRNA products, as measured by RT-PCR analysis, in uninjected (C, control) and MO-injected embryos treated from 24 to 96 hpf with DMSO (−) or E2 (+). Each panel (1–5) depicts results of a different MO/RT-PCR primer set. As compared with uninjected controls, knockdown of ERα with esr1 MO followed by RT-PCR with primer set A-F1/A-R1 targeting exons II and V (1) and primer set A-F1/A-R2 targeting exon II and intron III (2) reduces the normal amplicon (519 bp, N) and correspondingly induces an intron-retained variant (706 bp, V). Injection of esr2a MO (3), esr2b MO (4), or a 5-bp mismatch esr1 MO (5), followed by RT-PCR with primers A-F1/A-R1, showed only the normal 519-bp band. See Supplemental Figure 1 for sequence analysis of the splice variants.

Knockdown of ERβa mRNA using esr2a MO

Knockdown of ERβa was accomplished with a MO targeting the exon III/intron III splice junction of esr2a (Table 1 and Figure 2A). RT-PCR amplification of ERβa cDNA with primers Ba-F1 and Ba-R1 in exons II and V (Figure 2A), respectively, showed multiple bands when MO-injected embryos were compared with uninjected controls: a normal-sized band (N, 693 bp), which was substantially reduced in intensity; a prominent variant band (V, 523 bp), corresponding with the predicted exon III-skipped size; and a trace of 1 variant band of intermediate size (V, 578 bp) (Figure 2B1). The Ba-R2 primer positioned within intron III and paired with Ba-F1 in exon II gave no product, indicating no intron retention (data not shown). Sequencing confirmed that the 523-bp variant lacked the entirety of exon III. The intermediate band (578 bp) contained the initial 56 bp of exon III spliced to a cryptic splice site within exon IV (Supplemental Figure 2). In silico translation of the exon III-skipped sequence indicated a frame shift after this splice disruption, followed by multiple stop codons. Thus, if translated, the 523-bp transcript would be terminated shortly into exon IV, resulting in a truncated protein with an incomplete DBD and no LBD. However, in silico translation of the 578-bp variant predicted a product maintaining the normal frame, suggesting that this transcript could be translated to an ERβa protein missing only the first of the 2 zinc fingers (amino acids 152–188) of the DBD. How this product might impact induction of estrogen-inducible mRNAs is difficult to predict or quantify (see Discussion). No abnormal expression levels or splice variants of ERβa were seen with any MO injection other than esr2a (Figure 2, B2 and B3). Data shown in Figure 2 are from 96-hpf embryos injected with 0.25 mM esr2a MO, but knockdown was essentially the same at 48, 72, and 120 hpf (data not shown).

Figure 2.

MO-mediated targeting of esr2a knocks down expression of wild-type ERβa transcript and induces multiple exon III-deletion variants. (A) Diagram of the zebrafish esr2a gene. (B) Representative agarose gels showing ERβα mRNA products as measured by RT-PCR using primer set Ba-F1/Ba-R1 targeting exons II and V. As compared with uninjected controls (C), injection of esr2a MO reduced the normal amplicon (N, 693 bp) and induces a corresponding increase in a major (V, 523 bp) and a minor (V, 578 bp) variant (1). Injection of esr1 MO (2) or esr2b MO (3) showed only the normal 693-bp ERβa RNA product. For embryo treatments and additional abbreviations, see legend to Figure 1. Sequence analyses of variant products are shown in Supplemental Figure 2.

Knockdown of ERβb mRNA using esr2b MO

As above, splicing of ERβb pre-mRNA was targeted with a MO designed to the exon III/intron III splice junction of esr2b (Table 1 and Figure 3A). RT-PCR analysis of the ERβb cDNAs amplified with primers Bb-F1 and Bb-R1 (Figure 3A) showed 2 bands when MO-injected embryos were compared with uninjected controls: a normal size band (N, 717 bp) that was substantially reduced in intensity and a prominent variant band (V, 526 bp) that corresponded with the predicted exon III-skipped size (Figure 3B1). The identity of the variant band was confirmed by sequencing (Supplemental Figure 3). In silico translation of the exon III-skipped sequence indicated a frame shift after the splice disruption, followed by multiple stop codons. Thus, if translated, the variant transcript would be terminated shortly into exon IV, resulting in a truncated protein with an incomplete DBD and no LBD. RT-PCR with the Bb-R2 primer paired with Bb-F1 gave no product, indicating no intron retention (data not shown). No abnormal expression levels or splice variants of ERβb were seen with any MO injection other than esr2b MO (Figure 3, B2 and B3). Data shown in Figure 3 are from 96-hpf embryos injected with 0.75 mM esr2b MO.

Figure 3.

MO-mediated targeting of esr2b knocks down expression of wild-type ERβb transcript and induces an exon III-deletion variant. (A) Diagram of the zebrafish esr2b gene. (B) Representative agarose gels showing ERβb mRNA products, as measured by RT-PCR analysis using primers Bb-F1/Bb-R1. Compared with controls (C) injection of esr2b, MO reduced the normal amplicon (N, 717 bp) and induced a corresponding increase in a single variant (V, 526 bp) (1). Injections with esr1 MO (2) or esr2a MO (3) did not alter the size of the normal 717-bp mRNA product. For embryo treatments and additional abbreviations, see legend to Figure 1. For sequence analysis of the splice variant see Supplemental Figure 3.

Effects of ERα knockdown on estrogen-inducible mRNAs are cumulative, persistent, and gene specific

To test the effects of ERα knockdown on induction of Vtg, AroB, and ERα mRNAs, uninjected and esr1 MO-injected embryos were treated with E2 (0.1μM) or vehicle alone beginning at 24 hpf and collected at intervals up to 120 hpf for QPCR analysis (Figure 4). E2 induction of mRNA expression increased progressively with longer E2 exposures and was gene specific (∼1000-, ∼50-, and ∼10-fold for Vtg, AroB, and ERα at 120 hpf, respectively). Injection of the esr1 MO significantly reduced E2 induction of Vtg and ERα mRNAs, and the knockdown effect persisted up to 120 hpf. By contrast, ERα knockdown had no effect on AroB mRNA induction.

Figure 4.

Effects of ERα knockdown on estrogen-inducible mRNAs are cumulative, gene specific, and persistent. Fertilized eggs were injected with esr1 MO or left uninjected. Beginning at 24 hpf, embryos were exposed to 0.1μM E2 (+E2) or DMSO and collected at 48, 72, 96, or 120 hpf for real-time QPCR analysis of (A) Vtg, (B) AroB, and (C) ERα mRNAs. Each data point shows the log mean fold change (±SEM; n = 3 independent experiments) when E2-treated embryos were compared with their respective DMSO-treated controls. In A–C, data for each mRNA were analyzed separately by 2-way ANOVA for effects of stage of development/E2 exposure and MO treatment. Developmental stage/E2 exposure was significant for all 3 mRNAs (Vtg: F = 3.51, P = .034; AroB: F = 8.16, P = .002; ERα: F = 3.84, P = .03). The MO effect was significant for Vtg and ERα (Vtg: F = 4.77, P = .044; ERα: F = 4.68, P = .046) but not AroB (F = 1.96, P = .18). Asterisks indicate when differences between E2 and E2 + MO samples differed significantly as determined by Bonferroni's post hoc test.

Knockdown of ERα, ERβa, and ERβb mRNAs differentially impacts estrogen-inducible mRNAs

Based on the time-course experiment with esr1 MO and additional time-course experiments with MOs for esr2a (data not shown), we chose a single E2 exposure paradigm (24–96 hpf) to compare the knockdown effects of all 3 gene-specific MOs. Consistent with Figure 4, injection of the esr1 MO significantly reduced E2 induction of Vtg (from >300- to <20-fold) and ERα mRNAs (from >3- to <2-fold) but had no effect on AroB induction, nor did MO injection alter the baseline levels of any of the 3 mRNAs in DMSO controls (Figure 5A). The absence of an effect of esr1 MO on ERα mRNA in DMSO controls, as measured by QPCR (Figure 5A), was surprising, because RT-PCR gels showed a substantial reduction of correctly spliced ERα mRNA (Figure 1B1). Because the principal ERα QPCR primers (A-Q1/Q2, Figure 1A) fall within exon V, downstream of the splicing error, it is possible that these primers were detecting both the normal and the intron-retained forms. To verify this, a second QPCR primer set (A-Q3/Q4, Figure 1A) was designed to target exons III and IV in order to amplify only the normally spliced product. QPCR results comparing the 2 primer sets confirmed that normally spliced ERα mRNA was indeed reduced in response to MO injection: from 0.98 ± 0.16 to 0.23 ± 0.04 fold change (vs DMSO controls) with primer sets Q1/Q2 and Q3/Q4, respectively (n = 5). Worth noting here, detection of both normal and variant splice products with primer set A-Q1/Q2 provides evidence that the reduction of E2-induced ERα mRNA levels after esr1 MO injection is due to a process upstream of splicing (ie, transcription regulation) rather than to MO interference with splicing of E2-induced newly transcribed ERα pre-RNAs. Because our goal with QPCR was to measure the total amount of ERα transcribed (normal plus variant), rather than the amount correctly spliced, we used the A-Q1/Q2 primer set positioned in exon V for all subsequent analysis.

Figure 5.

MO-mediated mis-splicing of (A) ERα, (B) ERβa, and (C) ERβb transcripts differentially impacts E2-mediated induction of Vtg, AroB, and ERα mRNAs. Fertilized eggs were left uninjected or injected with (A) esr1 MO (0.25 mM, n = 2 and 1.0 mM, n = 3 independent experiments were combined, because analysis showed no significant differences in ERα mRNA knockdown or QPCR analysis of target gene mRNAs), (B) esr2a MO (0.25 mM, n = 3), or (C) esr2b MO (0.75 mM, n = 4) and exposed to E2 or DMSO (C) between 24 and 96 hpf. Vtg, AroB, and ERα mRNAs were measured by QPCR and normalized to bactin and expressed as log MNE ± SEM. In A–C, data for each mRNA were analyzed separately for an effect of treatment by 1-way ANOVA. Bars with different letters represent values that differed significantly (P < .05) as determined by Tukey's post hoc test: (A) Vtg (a, b), AroB (c, d), and ERα (e, f); (B): Vtg (a, b), AroB (c), and ERα (d, e); (C): Vtg (a, b), AroB (c, d), and ERα (e, f).

As shown in Figure 5B, knockdown of ERβa with the esr2a MO did not significantly affect baseline or estrogen-inducible expression of Vtg, AroB, or ERα mRNAs. Multiple additional experiments with the esr2a MO and different time-course and E2 treatment paradigms also failed to show significant effects on E2-induced Vtg, AroB, and ERα mRNA expression (see Discussion).

In marked contrast to esr1 and esr2a MOs, knockdown of ERβb with the esr2b MO blocked E2-mediated induction of AroB, Vtg, and ERα mRNAs (Figure 5C).

The absence of effects of any of the 3 MOs on baseline mRNA levels suggests that the 3 target mRNAs measured here are not up-regulated to any significant extent by endogenous estrogens at this point in development.

Combined knockdown of ERα and ERβa mRNAs has no effect on E2-inducible mRNAs but attenuates ERα knockdown effects

To determine whether ERβb alone is sufficient for up-regulation of the 3 E2-inducible mRNAs, esr1 and esr2a MOs were coinjected into fertilized eggs treated with/without E2 up to 96 hpf. RT-PCR confirmed that coinjection of the 2 MOs knocked down their respective ER transcripts, essentially as observed when injected separately (compare Figure 6 with Figures 1B and 2B). Coinjection of esr1 and esr2a MOs did not significantly affect E2 induction or baseline levels of any of the 3 mRNAs. However, comparison of Figures 5A and 6C shows that simultaneous knockdown of ERβa attenuates the knockdown effect of ERα, as measured by Vtg and ERα induction.

Figure 6.

Coinjection of esr1 and esr2a MOs knocks down expression of normally spliced ERα and ERβa mRNAs but has no effect on estrogen-inducible mRNAs. Embryos were treated and analyzed as described in legend to Figure 5, except that esr1 MO and esr2a MO (each 0.25 mM) were coinjected. Representative agarose gels show MO-mediated effects on (A) ERα mRNA amplicons as measured by RT-PCR with primer set A-F1/R1, and (B) ERβa mRNA amplicons as measured with primer set Ba-F1/R1. (C) Combined MO-mediated effects on induction of estrogen-responsive mRNAs were measured by QPCR and expressed as log MNE ± SEM (n = 3 independent experiments). Data in C were analyzed for treatment effects by 1-way ANOVA separately for each mRNA species. Bars with different letters represent values that differed significantly (P < .05) as determined by Tukey's post hoc test: Vtg (a, b), AroB (c, d), and ERα (e, f).

MO injections affect estrogen-inducible ERα but not ERβa and ERβb mRNAs

As described above, each MO was subtype specific in its effect on splicing. However, to examine the possibility that knockdown of 1 ER somehow alters expressed levels of the other 2 ERs, all 3 ER mRNAs were measured by QPCR after injection of each MO with/without E2 using primer sets that amplify both normal and variant transcripts (Table 1 and Figure 7). Consistent with Figures 4 and 5, estrogen induction of ERα mRNA was blocked by esr1 or esr2b MOs (Figure 7). Although ERβa and ERβb were not estrogen responsive, there was a nonsignificant reduction of ERβb mRNA levels after injection of esr1 or esr2b MOs in both DMSO- and E2-treated embryos. This effect was not observed after esr2a MO, suggesting that it is not an artifact of MO injection per se.

Figure 7.

MO injections affect estrogen-inducible ERα mRNA but do not indicate interdependence of ER subtypes. Embryos were uninjected or injected with (A) esr1 MO, (B) esr2a MO, or (C) esr2b MO and exposed to E2 or DMSO (C) as described in Figure 5. QPCR was used to measure ERα, ERβa, and ERβb mRNAs. Values were expressed as MNE ± SEM (n = 3 independent experiments). Note that these data are shown on an expanded arithmetic rather than a log scale, as in previous figures. In each panel (A–C), 1-way ANOVA for effect of treatment was performed separately for each mRNA species. Bars with different letters represent significant differences (P < .05) as determined by Tukey's post hoc test: ERα (a, b), ERβa (c), and ERβb (d).

Control injections are without effect on estrogen-inducible mRNAs

None of the 3 negative controls affected splicing of ERα, ERβa, or ERβb mRNAs (see Figure 1B5; other data not shown), nor did QPCR analysis show significant effects on baseline or E2 induction of Vtg, AroB, or ERα mRNAs (Supplemental Figure 4). However, all 3 control injections resulted in a small but nonsignificant reduction in E2-induced Vtg mRNA, an effect similar to that seen with the esr2a MO (Figure 5) and the esr1 plus esr2a MOs (Figure 6).

Discussion

In studies reported here, we successfully designed gene-specific MOs that interfere with splicing of pre-RNAs encoded by each of the 3 zebrafish ER genes (esr1, esr2a, and esr2b) to knock down expression of ERα, ERβa, and ERβb mRNAs, respectively, and used QPCR analysis of estrogen-inducible mRNAs (Vtg, AroB, and ERα) as evidence of ER subtype-specific actions and interactions in embryos in vivo.

esr1, esr2a, and esr2b MOs knock down their respective normal ERα, ERβa, and ERβb transcripts

MOs designed to target the exon III/intron III splice junction in the DBD of each of the esr genes resulted in mis-splicing of their respective pre-RNAs, without affecting transcripts produced by the other 2 esr genes. Moreover, the specificity of the knockdown effect was seen even when 2 MOs (esr1 and esr2a) were injected simultaneously. The resultant intron III retention (esr1 MO) or exon III deletion (esr2a and esr2b MOs) removed pre-mRNA sequences encoding the DBD and left frame-shifted transcripts, predicted to form truncated proteins lacking the second zinc finger of the DBD, and the LBD and AF2 domains. These splice variants, if translated, predict mutant isoforms that would disrupt not only the classical genomic (ligand and DBD dependent) mechanism of estrogen action but would also fail to dimerize or activate any of the multiple mechanistically distinct pathways of estrogen action mediated by soluble ERs (1). In addition to the major splice variants, traces of normal transcript are detectable with all MO injections, indicating knockdown, not knockout. Also, injection of the esr2a MO results in a minor variant that is lower in abundance than the major exon III-deleted variant. This variant retains a fragment of exon III spliced in-frame to exon IV, predicting an ERβa protein lacking only the first of the 2 zinc fingers of the DBD. Studies in ERα knockout/knock-in mice report that deletion of the first zinc finger interferes with the classical (ERE dependent) but not the nonclassical (ERE independent) actions of estrogen in vivo (57). How remnant normal or variant ER proteins might impact estrogen-inducible mRNAs, for example, by heterodimerizing with transcriptionally active wild-type ERs or interacting with ER-dependent coactivators/repressors is difficult to predict or quantify. Thus, in future studies, it will be important to confirm our results by MO targeting of other areas of the pre-mRNAs, the use of subtype-specific antibodies, or alternative methods, such as rescue constructs (51).

Although we have not determined whether ER proteins are actually formed from the variant transcripts, we interpret changes in estrogen-inducible mRNAs as evidence of functional knockdown of zygotically transcribed ER mRNAs. Not surprisingly, the E2 induction response and the knockdown effect, as measured by QPCR of inducible mRNAs, is relatively weak at 48 hpf (24-h E2 exposure) but persists and increases progressively up to 120 hpf (96-h E2 exposure). This is a time in development when morphogenesis is essentially complete, and all tissue types are represented (58).

ERβb is necessary and sufficient for estrogen induction of AroB mRNA

Results show that knockdown of ERβb, but not ERα or ERβa, completely blocks E2-mediated up-regulation of AroB mRNA. Moreover, simultaneous knockdown of ERα and ERβa transcripts by coinjection of esr1 and esr2b MOs is without effect on AroB induction. This indicates that the remaining ERβb protein is sufficient to mediate estrogen activation of the cyp19a1b gene in vivo. Although a previous study reported that ERα activates a cyp19a1b promoter more strongly than the ERβs (16), this indicates only what is possible when each of the 3 zebrafish ERs are transfected separately into human glial cells. The discrepancy with our results can be explained if ERβb is the only ER subtype actually coexpressed with AroB in the same cells in vivo, at least at this stage of development. To our knowledge, AroB expression has not been definitively colocalized with any of the 3 ER at the cellular level in fish (59). However, as determined by in situ hybridization, AroB, ERβa, and ERβb mRNAs have a similar neuroanatomic distribution as early as 24–48 hpf (30, 60), whereas brain ERα expression does not begin before 15 days postfertilization in zebrafish (26). In fact, Mouriec et al (30) took this spatial and temporal correspondence as circumstantial evidence that AroB induction may be regulated through one or both of the ER betas.

Both ERα and ERβb are required for E2 induction of Vtg and ERα mRNAs

Separate injections of the esr1 and the esr2b MOs demonstrate that ERα and ERβb are both required for E2-mediated induction of Vtg and ERα mRNAs. The functional interaction suggested by these results is consistent with the expression of ERα and ERβb (but not ERβa) mRNAs in zebrafish liver beginning early in development (26, 60). A possible explanation is that ERα and ERβb interact as a heterodimer on the vtg and esr1 promoters or, alternatively, interact as homo- or heterodimers on separate but functionally related chromatin binding sites. Models of action involving cooperation, as well as competition, between the ERα and ERβ proteins in mammalian species have been proposed (for review, see Refs. ^61–63), and a recent study using human breast cancer cells engineered to express ERα, ERβ or both, and ChiP-chip for global gene analysis, described a dynamic interplay between the 2 ER in their selection of chromatin binding sites, with competition, restriction, and site shifting having important implications for the regulation of gene expression by these 2 nuclear receptors (8). Relevant to our findings, this study reported that some genes respond to E2 stimulation only in the copresence of ERα and ERβ.

Previous studies using primary hepatocyte cultures from various fish species, and RNA interference technology or subtype-selective ligands, have implicated ERα and one or both of the ERβs in the autoregulation of ERα and up-regulation of Vtg (18–21), but these models generally envision a 2-step sequence with the ERβs up-regulating ERα, followed by ERα induction of Vtg.

ERβa is dispensable for induction of AroB, Vtg, and ERα mRNAs

Microinjection of the esr2a MO alone has no significant effect on E2 induction of any of the 3 estrogen-responsive mRNAs. However, when coinjected with the esr1 MO, the esr2a MO attenuates the effect of ERα knockdown on Vtg induction. Although the esr2a MO is as effective as the other 2 MOs in knocking down production of the respective normal transcripts, the unexpected cryptic ERβa splice variant that results from esr2a MO injection greatly complicates interpretation of results (see Discussion above). Thus, it may be incautious to draw a definitive conclusion regarding the functions of ERβa without verification using a second MO. Nonetheless, if the observed results are due to the reduction of normal ERβa protein, and not to any aberrant function of a possible variant ERβa, these coinjection results might suggest that ERβa is a weak or negative transcriptional regulator with functions that are masked in the presence of stronger positive regulators, such as ERα and ERβb. A plausible mechanism to explain why ERβb alone is sufficient for induction of Vtg in the absence, but not the presence, of ERβa is that ERβa 1) is a default repressor, 2) normally partners with one or both of the other ERs as a transdominant inhibitor, or 3) competes with the other ERs for occupancy on chromatin binding sites. These mechanisms are reported to account for negative interactions between ERα and ERβ of mammalian species (8, 64). Although there is no direct evidence for heterodimerization of fish ER isoforms, increasing concentrations of largemouth bass ERβ-I (ERβb) or ERβ-II (ERβa) were found to decrease ERα transactivation activity, as measured in a mammalian cell line (HepG2) expressing an estrogen-inducible luciferase reporter (65).

It would be premature at this point to rule out the possibility that ERβa could function as a positive transcriptional regulator later in development, or on the promoters of estrogen-responsive genes not tested here, or that it has acquired other principal functions during the course of evolution (neofunctionalization) (15). Earlier work using MO-mediated knockdown of ERβa in zebrafish embryos examined its role in neuromast development (66–68). Although one of these reports employed an esr2a splice-site MO very similarly positioned as ours, their results focused mainly on ERβa knockdown accomplished through a different, translation start-site targeting MO that was designed to compare maternal and zygotic transcripts (68). Interestingly, microarray analysis of knockdown samples reported in their supplemental data described no known E2-up-regulated genes among those altered by esr2a MO.

Summary and conclusions

In summary, our findings using MO-mediated ER knockdown technology provide strong evidence for ER subtype-specific and interactive roles in the regulation of estrogen-responsive genes during early zebrafish development. They suggest a surprisingly important role for ERβb in the regulation of multiple genes, in different tissues, in addition to confirming the role of ERα together with ERβb in the regulation of Vtg and its own autoregulation. To our knowledge, this is the first report that AroB, a key enzyme in neuroendocrine regulation as well as a valuable marker for endocrine disruption, is induced specifically through ERβb, at least in zebrafish embryos.

Interestingly, the other β-subtype, ERβa, has no positive regulatory role on the gene targets or developmental stages measured in our study. Considering that ERβa has been conserved as 1 of the 2 β-duplicates throughout the fish lineage, it is unlikely that its destiny is to become a nonfunctional pseudogene while the other copy retains the original functions. Perhaps the positive and negative regulatory functions that are accomplished by a single ERβ in mammalian species have been subdivided between the 2 teleostean β-forms.

From these results, we cannot speculate as to the possible functions of membrane-bound ERs such as G protein-coupled receptor 30 (69), nor can we extrapolate findings obtained with administered E2 to estrogen-like environmental chemicals or to maternally derived or in situ-synthesized estrogens. Nonetheless, the feasibility of knocking down 1 or more ER using MO-mediated technology offers a powerful strategy for examining how an array of endogenous and exogenous estrogens could impact the specificity and interplay of soluble and membrane-associated ER subtypes on estrogen-responsive genes, tissue types, and developmental processes early in vertebrate development and reinforces the view that studies in zebrafish embryos can help to advance our understanding of ER biology in a normal in vivo context.