Histone deacetylase activity is necessary for left-right patterning during vertebrate development

Consistent asymmetry of the left-right (LR) axis is a crucial aspect of vertebrate embryogenesis. Asymmetric gene expression of the TGFβ superfamily member Nodal related 1 (Nr1) in the left lateral mesoderm plate is a highly conserved step regulating the situs of the heart and viscera. In Xenopus, movement of maternal serotonin (5HT) through gap-junctional paths at cleavage stages dictates asymmetry upstream of Nr1. However, the mechanisms linking earlier biophysical asymmetries with this transcriptional control point are not known.

To understand how an early physiological gradient is transduced into a late, stable pattern of Nr1 expression we investigated epigenetic regulation during LR patterning. Embryos injected with mRNA encoding a dominant-negative of Histone Deacetylase (HDAC) lacked Nr1 expression and exhibited randomized sidedness of the heart and viscera (heterotaxia) at stage 45. Timing analysis using pharmacological blockade of HDACs implicated cleavage stages as the active period. Inhibition during these early stages was correlated with an absence of Nr1 expression at stage 21, high levels of heterotaxia at stage 45, and the deposition of the epigenetic marker H3K4me2 on the Nr1 gene. To link the epigenetic machinery to the 5HT signaling pathway, we performed a high-throughput proteomic screen for novel cytoplasmic 5HT partners associated with the epigenetic machinery. The data identified the known HDAC partner protein Mad3 as a 5HT-binding regulator. While Mad3 overexpression led to an absence of Nr1 transcription and randomized the LR axis, a mutant form of Mad3 lacking 5HT binding sites was not able to induce heterotaxia, showing that Mad3's biological activity is dependent on 5HT binding.

HDAC activity is a new LR determinant controlling the epigenetic state of Nr1 from early developmental stages. The HDAC binding partner Mad3 may be a new serotonin-dependent regulator of asymmetry linking early physiological asymmetries to stable changes in gene expression during organogenesis.

Despite a bilaterally-symmetrical bodyplan, many animals exhibit a consistent asymmetry in the placement and shape of the heart, viscera, and brain [1]. The wide-spread conservation of laterality, and the consistent linkage of the orientation of the left-right (LR) axis with the dorso-ventral and anterior-posterior axes (in a world that does not distinguish left from right above the quantum level), make LR patterning a fascinating problem [2-4]. In addition to its relevance to basic cell, developmental, and evolutionary biology, laterality presents significant implications for normal physiology and a plethora of clinically-important human syndromes [5,6]. Errors in LR patterning include loss of asymmetry (isomerism), complete inversions (situs inversus), and random placement of individual organs (loss of concordance known as heterotaxia).

It is widely accepted that large-scale LR asymmetry derives from the molecular chirality of subcellular structures [7]. However, at least two main classes of models have been proposed for how this chirality is propagated, amplified, and imposed on multicellular fields during development. One popular model focuses on the net unidirectional extracellular fluid flow achieved during gastrulation by the movement of cilia [8,9].

A different model focuses on much earlier stages, prior to gastrulation, when physiological events leverage asymmetry from the chirality of the intracellular cytoskeleton to set up asymmetrical movement of morphogens through cell fields [10-12]. One such morphogen is serotonin (5HT): a neurotransmitter of clinical relevance that has interesting roles outside the central nervous system [13]. In two vertebrate species (chick and frog), serotonergic signaling has been shown to be required for LR patterning. In the frog embryo, it is known that 5HT accumulates in the right blastomeres in a rapid process dependent on asymmetric voltage gradients across the midline and the presence of open gap junctions through which it traverses [14-16].

In Xenopus embryos, many aspects of this system have been elucidated: the source of the electrophoretic force driving this 5HT gradient has been molecularly characterized [14,17,18], and indeed many aspects are known in enough quantitative detail to allow the whole system to be computationally modeled [19,20]. However, one fundamental question has not been addressed: how does this physiological gradient, occurring in the frog at a time when the zygotic genome is mostly quiescent, couple to the later transcriptional cascade of asymmetrically expressed genes that is known to control organ positioning? Specifically: how does the arrival of 5HT within the right-side blastomeres control gene expression? Well-known 5HT receptor families [21] are not ideal candidates because they are functional on the outside surface of the plasma membrane, while the 5HT arrives through gap junctions, and thus requires an intracellular binding target.

To better understand how intracellular 5HT signaling is transduced into stable gene expression, from early to very late developmental stages, we hypothesized the involvement of epigenetic machinery as a new component of the LR establishment. Such a mechanism is attractive because once epigenetic markers are deposited on the chromatin, they remain stable along successive cell divisions carrying the epigenetic signature of an activated or repressed state of the chromatin. More specifically, we sought to determine whether early manipulation of the epigenetic state of the embryo would affect LR-relevant genes expressed at late developmental stages (e.g., Nr1).

Changes in the state of the chromatin have a key role for the proper control of gene expression during development. Post-translational modifications such as lysine acetylation constitute a code allowing specific interactions between chromatin and DNA binding proteins that ultimately will dictate the status of activation of a gene [22]. These modifications take place at the chromatin level and involve the acetylation of lysines in the amino terminal tail of core histones [23].

HDACs are important players for epigenetic memory control as they act by decreasing the levels of acetylated histones leading to chromatin compaction and repression [24,25]. HDACs play a critical role in regulating gene expression and aberrant levels of protein acetylation have been associated with cancer [26] and a range of neurological diseases such as fragile X mental retardation [27]. Indeed, HDACs blockers can reverse silent heterochromatin to an active conformational structure leading to the normal function of genes that are silent in these pathologies [27]. Despite their prominent role on gene expression, HDACs can not bind to the DNA on their own, but are recruited to specific points on the chromatin via interaction with transcriptional repressive complexes such as those containing Mad proteins [28].

Vertebrate HDACs can be classified into four groups: Class I (HDAC 1,2,3,8), Class IIa (HDAC 4,5,7,9) Class IIb (HDAC 6) and Class IV (HDAC 10,11). Xenopus has a maternal form of HDAC (HDAC) that shows sequence homology to other HDAC [29]. The HDAC activity in early embryos appears to be mainly, if not exclusively, of the HDAC-A type, that is of the vertebrate HDAC I class [30]. Although HDAC function during Xenopus development remains unclear, several studies have demonstrated its importance for vertebrate development. Genetic deletion of HDACs in mice and embryonic stem cells results in problems in proliferation [31,32], while the deletion of hdac1 in zebrafish leads to defects in skeletal and neuronal development [33,34]. Interestingly, in HDAC-null ES cells, only 3% of genes are downregulated and about 5% are upregulated. This suggests that the epigenetic machinery is directed to specific points of the chromatin, and indicates that HDACs could underlie regulation of specific developmental signals [35].

Here, we probe the involvement of HDACs in LR asymmetry establishment. Using several molecular and pharmacological tools, we found that HDAC activity is a new component of LR patterning that acts very early in embryonic development (prior to active zygotic transcription). In addition, we identified the HDAC-interacting protein Mad3 as a new component during LR development in the 5HT signaling pathway. Our results demonstrate how very early physiological signaling can be transduced into a stable pattern of gene expression at late developmental stages, and sheds light on how epigenetic state and chromatin structure orchestrate important events during early stages of axial patterning in animal development.

Plasmids containing X. laevis Mad3 and X. laevis maternal HDAC cDNA were purchased from Open Biosystems (Clone ID: 4175511 and Clone ID: 6862376, respectively) and the DN HDAC was generated by PCR as described [36]. pCS2Vp16 and pCS2Eng were provided by Dr. D. Kessler and the constructs pCS2EngMad3 and pCS2Vp16Mad3 were generated by PCR cloning. To induce ectopic expression of Xenopus Mad3 and HDAC, their coding regions were inserted into the pCS2-flag expression vector, resulting in the addition of the flag-tag to the N-terminus of the coding region of Mad3. Mad3 and HDAC pCS2 plasmids were linearized with NotI to prepare for transcription with the SP6 Message machine kit (Ambion). Mad3-5 mut (Gln125 and Gln161 were replaced by glutamate; Asp145, Asp148 and Asp163 were replaced by asparagine) was generated by PCR mutagenesis with an Agilent kit.

For microinjections, capped, synthetic mRNAs [37], generated using the Ambion mMessage mMachine kit were dissolved in water plus the lineage tracer rhodamine-labeled dextran (RLD) and injected into embryos in 3% Ficoll. 30 minutes post-injection, embryos were transferred to 0.1X MMR and allowed to develop until stage 21 for in situ hybridization or stage 45 for organ placement score. All experimental procedures involving the use of animals for experimental purposes were approved by the Institutional Animal Care and Use Committees (IACUC) and Tufts University Department of Lab Animal Medicine (DLAM) under the protocol number M2008-08.

At stage 45, embryos were anesthetized with 5% tricaine and analyzed for position (situs) of 3 organs: the heart, stomach, and gallbladder [38]. Heterotaxia was defined as reversal in position of one or more organs. Only embryos with normal dorsoanterior development (DAI = 5) were scored to avoid scoring instances of secondary randomization due to errors in the dorso-ventral (DV) or antero-posterior (AP) axial patterning [39], and only clear left- or right-sided organs were scored. Percent heterotaxia was calculated as the number with heterotaxia divided by the number of total scorable embryos, i.e. embryos normal in all other ways. A χ2 test was used for further statistical analysis.

Embryos were collected at different stages of development and fixed in MEMFA for 3 hours at room temperature and used for in situ hybridization as described in Harland (1991). Plasmids containing Xenopus Mad3 and Xenopus HDAC cDNA were purchased from Open Biosystems (Mad3 clone ID: 4175511; HDAC clone ID: 6862376) and cloned in pCS2. The plasmids were linearized and anti-sense probes for in situ hybridization were generated in vitro using DIG labeling mix from Invitrogen.

Batches of embryos were separated into experimental and control groups and exposed to 0.1X MMR (control) or 0.1X MMR containing 100 mM Sodium Butyrate (NaB) (Sigma) during different stages of development. The drug was washed out and the embryos were allowed to develop until stage 21 for Nr1 in situ or until stage 45 for organ placement score.

For Western blottings embryos were collected at stage 7 (64 cells) and homogenized in lysis buffer (100mM NaCl, 20 mMNaF, 50 mM Tris pH 7.5, 5 mM EDTA, 1%NP40, 1% deoxycholate, 1:50 EDTA-free Complete protease inhibitor, Roche). The lysate was centrifuged for 15 minutes at 4°C and the supernatant was collected and frozen. The extracted proteins were subjected to SDS-PAGE and blotted onto a PVDF membrane (BioRad). After blocking with 5% skim milk and 0.1% Tween 20 in PBS, membrane filters were incubated with an anti-Mad3 (NeuroMab clone N129/15.1, supernatant) or anti-acetyl H4 (Milipore 1:500) overnight at 4°C. The membranes were washed in PBT and incubated with secondary HRP conjugated antibody (1:1000, Jackson). Immunosignals were visualized with chemoluminiscence.

For Co-IP assays, embryos were injected at the 1 cell stage with Mad3WT-flag or Mad3-5mut-flag constructs and collected at stage 7 (64 cells) when 10 embryos were homogenized in lysis buffer. A total of 100 μl of embryo lysate was incubated with 2 μg anti-5HT (AB125, Milipore) or rabbit IgG for 3 hours at 4°C followed by incubation with proteinA-agarose for 1 hour at the same temperature. The beads were collected by centrifugation and washed in lysis buffer. The samples were then washed in low salt buffer (50 mM Tris HCl pH 7.5, 0.1%NP40, 0.05% Deoxycholate) and high salt solution (50 mM Tris HCl pH 7.5, 500 mM NaCl, 0.1% NP40, 0.05% Deoxycholate). Proteins were eluted by adding SDS-PAGE sample buffer followed by boiling for 5 minutes. For Western blotting embryo lysates were loaded on 4-12% gradient polyacrylamide gel (Invitrogen) in sample buffer and the proteins were transferred to a nitrocellulose membrane, blocked in 5% non- fat milk in PBT (PBS plus 0.1% Tween 20) for 1 hour at room temperature and incubated overnight with anti-flag (Sigma; 1:1000). Membranes were washed with PBT and incubated with secondary antibody conjugate to HRP. The blotting was developed with a luminescence kit (Pierce) and the images were acquired with a camera coupled to a GBox system (Syngene).

To determine whether NaB treatment was effective, whole embryo lysates were prepared from embryos exposed to NaB at stages 1-7, stages 7-8, and stages 8-9, including also untreated stage-matched controls. The whole lysate from treated and control groups were prepared in the lysis buffer as described above and boiled for 5 minutes. The proteins were separated by SDS-PAGE and transferred to nitrocelulose membranes. Blots were then incubated with anti-acetyl H4 (Milipore) and developed with chemoluminiscence. To determine the relative abundance of acetylated histone H4, membranes were stripped of antibodies in 200 mM glycine buffer pH 2.2 and re-probed with anti-Tubulinα (Sigma, 1:4000). For quantification of acetylated histone H4 protein, the luminosity of individual bands was defined using the histogram function of Photoshop and normalized against Tubulin. Only exposures in the linear phase of detection were used for quantification. SEM was defined from normalized acetylated histone H4 and measured in three independent experiments. Statistical analysis was performed with Student's t-test.

Embryos were exposed to 100 mM NaB from stage 1 to 7, when the drug was washed out and the embryos were allowed to develop in 0.1X MMR until stage 21. The chromatin was then collected and the ChIP was performed following the protocol [40]. The purified chromatin from treated and control groups was then incubated with anti-acetyl H4, anti-acetyl H3, anti-H3K4me2 (Milipore) or anti-rabbit IgG (Milipore) as control. Amplification of the precipitated DNA was carried out in an Applied Biosystems Step One Plus PCR machine, using the standard SYBR green program with an initial melt stage at 95°C for 10 min, followed by 40 cycles of 95°C for 15 sec and 60°C for 1 min. The run was finished by a melt curve from 95°C to 60°C to ensure that no primer dimer artifacts were formed and that cycle threshold (Ct) values represent the desired amplicon. Primer sequences were designed with Primer 3 plus program (http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi).

qPCR primers set used: Nr1 intronic region: Forward 5' - TTCCCTATTGACAGGGGTTG - 3'; Reverse 5'- GCCAAATGTCAAAACACTCG -3'. Nr1 Promoter region: Forward 5' - TCCTTGATGAGGCCATTAGC - 3': Reverse 5' - CAAACAGAGCATTCCCTGAC - 3'. The ΔΔCt method was used to represent the data as in [41] and each experiment was performed in triplicate and combined for further statistical analysis.

One gram of Sepharose 4BCL powder was activated according manufacturer's instructions, washed and equilibrated in 4 mL buffer (50 mM TrisHCl, pH 7.5, 100 mM NaCl) plus 5 mM sodium metabisulfite to conjugate of 5-HT to the resin. Metabisulfite was added to avoid rapid 5HT oxidation at the alkaline pH necessary for the reaction. The 5HT capture was performed in batch and metabisulfite 5 mM was added throughout the whole process. Two 1 mL aliquots of conjugated resin were centrifuged and then loaded with 1 mL of 1 cell embryo lysate prepared in lysis buffer (0.5% SDS; 150 mM NaCl; 5 mM EDTA; 10 mM Tris, pH. 7.6; 2mM PMSF at a final protein concentration of 15 mg/mL) plus 20 mM 5HT (negative control) or 1 mL of 1 cell embryo lysate. The mixtures were incubated overnight at 4 C, after which the resins were washed three times at room temperature for one hour each with 50 mM TrisHCl pH 7.5, 100 mM NaCl at 4C to remove loosely bound proteins. Elution was performed in both cases by rotating the resin for one hour at room temperature with the same buffer plus 20 mM 5HT. The eluates were concentrated by centrifugational filtration with a cutoff of 3 kDa. The proteins from each fraction were subjected to a 10% SDS-PAGE and, after sample preparation and trypsin in-solution digestion, the peptide mixtures from the eluates were analyzed by LC/MS-MS using a Dionex 3000 LC (Sunnyvale, CA) coupled to a linear ion trap mass spectrometer (LTQ, Thermo Fisher, San Jose CA). The raw data collected on the LTQ was analyzed with Sequest software (Thermo Fisher) for protein identification using a database for Xenopus laevis.

The protein sequence of Xenopus Mad3 transcription factor (Q0VH33) was obtained from Swiss-Prot repository (http://www.expasy.ch/sprot). To identify a suitable template for modeling the 3D structure, the sequence was queried against the protein databank (PDB) sequences using Blast search engine [42]. Based on the results from these analysis, the crystal structure of lipocalin AM182, which is a paralog of monotonin, in complex with 5HT [43] was used as a template to model the non-DNA binding region of Mad3, using the homology modeling program MODELLER (ver 9.4, [44]. Ten models were built and ranked using the Modeller's objective function. The best ranking structure was subjected to 5000 energy minimization steps to relieve steric and geometric strains using NAMD (Ver.2) with charmm force field parameters. The 3D structure of 5HT was obtained from the co-crystal structure of AM182, hydrogen atoms were added and the structure was minimized using Amber charges and Amber force field as adopted in MOE program (Ver.2008.10, http://www.chemcomp.com). In the lipocalin structure, 5HT has salt bridge with Asp106 and Ser18 [43]. Among class-A GPCRs (G protein-coupled receptors), all biogenic amines including 5HT have a conserved salt bridge interaction with either an aspartic acid or glutamic acid residue in the third transmembrane region that is required for agonist and antagonist activity of GPCR ligands [45]. Based on this evidence, the proposed site for docking 5HT in Mad3 structure was identified and consisted of Asp163, Asp145, Asp148, Gln125, Gln161. 5HT was docked to the proposed binding site using the docking software Gold (v4.1) [46]. Twenty independent runs were performed to completely sample the ligand conformation and to avoid local minima and all the docked complexes were scored using Goldscore [46] and chemscore [47].

The best ranking complex was minimized using MOE as described above.

For use in binding experiments, Mad3 protein was obtained as a fusion protein with glutathione-S-transferase (GST) by transgenic expression in Escherichia coli and subsequent purification using glutathione (GSH)-agarose affinity chromatography. Briefly, the full Mad3 coding sequence was amplified by PCR using previously cloned Mad3 cDNA from human cell lines as a template. PCR reactions were performed under standard conditions using Pfx DNA polymerase (Invitrogen). PCR products were cloned into pCR-Blunt II-TOPO (Invitrogen) and subsequently subcloned into pGEX-5X-1 (GE Healthcare) using BamHI/SalI sites introduced by the PCR primers. All constructs were confirmed by sequencing. To express GST-Mad3, transformed E. coli BL21 were grown overnight at 37°C, diluted 200-fold in fresh medium and grown at room temperature until OD600 = 0.5-0.8. The culture was induced with 0.1 mM IPTG for 3 hours at room temperature. To express GST, E. coli BL21 were transformed with the empty vector, grown at 37°C to OD600 = 0.5-0.8 and induced with 1 mM IPTG for 1 hour at 37°C. After induction, pelleted cells (6,000 × g, 20 minutes, 4°C) were resuspended in 5 ml BugBuster protein extraction reagent (EMD Biosciences) per gram of wet cell paste and lysed with 100 μg/ml each lysozyme and DNAse I for 20 minutes at room temperature. Lysates were cleared by centrifugation for 20 minutes at 16,000 g and the supernatants were applied directly to a GSH-agarose column. Proteins were purified using standard procedures. Bound proteins were eluted with 25 mM GSH/50 mM Tris pH 8. The eluates were concentrated and the buffer changed to HBS-EP (10 mM Hepes pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% surfactant P20) using an Amicon Ultra-15 device.

Typically, GST was expressed in large quantities in the soluble fraction, while GST-Mad3 was found mostly in inclusion bodies, with a small percentage in the soluble fraction from which it was purified. Protein concentration was determined by the BCA method and the purity of the fractions was confirmed by SDS-PAGE.

Binding experiments were carried out on a Biacore X system (GE Healthcare). Proteins of interest were immobilized on a CM5 chip using the Amine Coupling Kit (Biacore, GE Healthcare). Specifically, two cells on a CM5 chip were activated using a 1:1 ratio of N-hydroxysuccinimide (NHS):1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) at a flow rate of 5 μl/min for 10 min. Recombinant GST and GST-Mad3 proteins were prepared as described above at a final concentration of 50 μg/ml and 10 μg/ml respectively in HBS-EP buffer (GE Healthcare). Prior to injection, 40 μL of protein solution were mixed with 30 μL of 100 mM Glycine pH 2.5, since these were the optimum conditions for capture as determined by pre-concentration experiments. GST was immobilized in the reference cell Fc2 by one 10 μL-injection at 10 μL/min. GST-Mad3 was immobilized in the sample cell Fc1 by five 20 μL-injections at 10 μL/min. Both Fc1 and Fc2 were subsequently blocked with 1 M ethanolamine (pH 8.5) for 7 min at 10 μl/min. A total of 7,962 response units of GST and 6,938 response units of GST-Mad3 were immobilized. Binding of 5HT was carried out at 25°C at 160 μL/min in HBS-EP. The amount of specific analyte bound was monitored by subtracting the response units from the reference cell (GST) from the GST-Mad3 immobilized cell. After each analyte injection, surfaces were regenerated with 50 mM NaOH for 0.25 sec at 160 μL/min. All sensorgram data presented are representative of at least 3 runs. Sensorgrams were aligned to the injection time (defined as t = 0 s) using BiaEvaluation software, and the baselines averaged at response difference = 0 RU. Curve fitting and binding kinetics parameters were calculated using GraphPad Prism Software.

Xenopus embryos express a maternal form of HDAC mRNA that shows sequence homology to other HDACs [29]. The HDAC activity present in early embryos appears to be mainly, if not exclusively, of the HDAC-A type, that is of the vertebrate HDAC I class [30]. Although the biochemical properties of the Xenopus HDAC are well characterized [30], knowledge about its mRNA expression pattern is limited. We processed different stages of early embryos for in situ hybridization. HDAC mRNA was found to be expressed in all animal-pole blastomeres by the 64 cell stage (stage 7) (Figure 1A).

To probe the overall potential involvement of HDACs during Xenopus LR patterning, we injected early embryos with mRNA encoding a well-characterized dominant negative (DN) form of the Xenopus HDAC [36]. Embryos in which the DN HDAC mRNA was injected at the 1 cell stage exhibited a 17-fold increase in heterotaxia compared to controls (17% vs. 1% heterotaxia in control non injected embryos, n = 62, significant at p = 0.002), demonstrating that HDAC is indeed functionally involved in embryonic LR patterning in Xenopus. It should be noted that the penetrance of the phenotype is artificially reduced (compared to those reported in mouse models, which often have midline or other defects) in these experiments by titering down the reagents to sub-optimal levels (because of our stringent requirement of normal dorso-anterior index and purity of LR phenotype). Nevertheless, given the very low background incidence of heterotaxia in un-manipulated embryos, the level of laterality disturbance induced by our manipulations is unmistakably significant. Moreover, when scoring three organs for situs, the maximum level of heterotaxia is 87.5%, not 100%, since random independent assortment of 3 organs will sometimes result in a wild-type phenotype being mistakenly scored when all three organs randomly land in their correct positions.