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

Next we sought to investigate whether the left and right sides of the embryo require different levels of HDAC activity for proper organ situs. To test this hypothesis, we injected embryos at the 4 cell stage on the left (ventral or dorsal) or on the right (ventral or dorsal) side with the constructs HDAC DN or HDAC WT. Indeed, embryos injected in either right blasomere with the HDAC DN, and embryos injected with HDAC WT on the left side, exhibited heterotaxia at stage 45 (Figure 1B). Embryos injected on the dorsal left side with HDAC DN or HDAC WT did not present significant levels of heterotaxia when compared to un-injected controls.

These data are consistent with a role for histone acetylation during LR development and demonstrate a LR-differential dependence on endogenous HDAC activity. Importantly, these data are not consistent with involvement of the HDAC pathway in cilia-driven events in the gastrocoel roof plate (GRP). The ciliated GRP cells derive from the dorsal precursors [48], whereas our results show effects on ventral blastomere descendants (which do not contribute to the ciliated organ); moreover, it is now known that only the left side of the nodal flow is important for asymmetry at neurula stages [49], while our data reveal effects on the descendants of right-side blastomeres.

As Nr1 is the most upstream gene known in the asymmetric cascade of the frog embryo (expressed only on the left side), we processed injected embryos at the stage 21 with a probe for Xenopus Nr1 1 (Xnr-1) in order to determine whether the pattern of Xnr-1 expression was disrupted when early HDAC activity was perturbed. Indeed, embryos injected on the right side with HDAC DN or on the left side with HDAC WT exhibited a loss of the normal asymmetry of Xnr-1 expression (HDAC DN on the right: 50% absence of Xnr-1; HDAC WT on the left: 26% bilateral expression of Xnr-1 with only 5% of control embryos not showing Xnr-1 expression).

These results support a role for HDAC activity in LR establishment and suggest that the epigenetic status of the chromatin may play a key role in determining the normally left-sided expression of Xnr-1.

We next characterized the timing of epigenetic controls of Xnr-1 expression, probing the function of HDAC at the early developmental stages when 5HT signaling takes place [15]. We used a HDAC blocker, Sodium Butyrate (NaB), capitalizing on the ability to use pharmacological reagents at different developmental stages (mRNA injections do not offer temporal control). Xenopus HDAC class I activity present in early embryos has already been shown to be sensitive to HDAC blockers [30].

Embryos at different developmental stages were separated into control groups and experimental groups and the latter were exposed to 100 mM NaB (Figure 2A). Incubation of embryos with NaB induced heterotaxia when exposures occurred between stage 1 and 7 (Figure 2B). Exposure to NaB at any stage after stage 7 did not induce heterotaxia (Figure 2B and Table 1), confirming that the LR-relevant functions of HDAC took place at cleavage stages (under zygotic transcriptional silence and long prior to cilia-mediated flow).

We also processed NaB treated embryos for Xnr-1 in situ hybridization in order to correlate levels of heterotaxia with misexpression of Xnr-1. Interestingly, a high percentage of embryos exposed to NaB lacked Xnr-1 (64% of Xnr-1 absence; Control: 5%). This reproduced the loss of Xnr-1 expression resulting from the HDAC DN mRNA microinjection (that was absence of Xnr-1 expression).

Thus, these results indicate that the time window most sensitive to blockage of HDAC overlaps with the developmental window in which the 5HT pathway was shown to be active [15].

Given that the normal function of HDAC is associated with a loss of immunoreactivity related to acetylated nuclear H4 [36], and as histone H4 is found in the Xenopus egg and early embryos [50,51], we investigated the impact of NaB-induced HDAC blockade on global H4 acetylation levels. Whole protein lysates from embryos exposed to NaB were analyzed by western blotting with a specific antibody against acetylated histone H4. This analysis showed that the histone H4 acetylation levels were markedly increased by 100 mM NaB treatment compared with control groups (Figure 2C). There was a 1.8-fold increase in acetylation levels in embryos exposed from stage 1-7, and a 2.3-fold increase in embryos exposed from stage 7 to stage 8 or from stage 8 to stage 9 (Figure 2C).

These results are consistent with HDAC functioning in the control of histones' acetylation levels during its regulation of LR asymmetry establishment.

Our data indicate that very early epigenetic modifications control expression of Xnr-1 at a much later developmental stage. Thus, we tested the prediction that this epigenetic modification was made directly on the Xnr-1 gene. The Xnr-1 genomic region is composed of 2 main regulatory elements (RE). The RE on the promoter region (PRE) is located 230 bp from the start codon [52]. The second RE is the ASE (asymmetric element) in the intron 1 of the Xnr-1 gene. It is known to drive the asymmetric expression of Xnr-1 in the Lateral Plate Mesoderm (LPM) [53] (Figure 3A).

Because early HDAC activity is necessary for the expression of Xnr-1 at stage 21, we asked whether the deposition of epigenetic markers due to early HDAC blockade could cause Xnr-1's repression later in development. More specifically, we also sought to map which region of the Xnr-1 gene (the promoter or the intronic region that contains the asymmetric element) was a target for the HDAC machinery, and link the deposition of epigenetic markers to the repressed state of Xnr-1 induced by NaB treatment. We performed a Chromatin Immunoprecipitation (ChIP) assay with chromatin from embryos at stage 21 that had been exposed to NaB early during development. We focused on three different markers: acetylated histones H4 and H3 (as HDAC inhibition leads to increased levels of histone acetylation), and H3K4me2, because of the existence of crosstalk between histones' modifications [54] and the fact that H3K4me2 modification is also associated with a repressed state of chromatin [55,56]. The H3H4me2 marker was particularly interesting because it has been associated with the epigenetic memory that precedes transcriptional control by several hours [57], thus being an attractive candidate to transmit an epigenetic memory from early to late developmental stages in absence of transcription.

Embryos were exposed to 100 mM NaB from stage 1 to 7 and allowed to develop in plain MMR medium until stage 21. At this stage the chromatin was isolated as described [40,41]. The ChIP experiment followed by qPCR analysis revealed that the level of acetylation of histones H3 and H4 and the levels of H3K4me2 in embryos exposed to NaB early during development presented increased levels only on the intronic region (Figure 3B), suggesting that higher levels of acetylated H3 due to HDAC treatment led to deposition of di-methylated H3 on the intronic region.

Taken together, these results indicate that NaB treatment during early cleavage stages led to a hyperacetylated status of the chromatin and perturbed LR-relevant gene transcription occurring at late developmental stages.

Our evidence for epigenetic mechanisms operating at cleavage stages controlling much later transcriptional readouts in LR patterning led us to investigate how HDAC activity could be modulated during early development. Our model predicts such controls, because the HDAC mRNA is expressed symmetrically; thus, an asymmetric signal is needed to confer the observed consistent difference in HDAC activity on the L and R sides. Because rightward redistribution of maternal serotonin (5HT) during cleavage stages has been shown to be necessary for the establishment of LR asymmetry [14,16], we sought mechanisms by which 5HT could couple to the epigenetic machinery.

We performed a high-throughput proteomic assay to identify novel 5HT binding proteins (SBP) present in the cytoplasmic fraction that may interact with known epigenetic machinery. A resin conjugated with 5HT was used for affinity capture analysis from whole protein lysate of Xenopus laevis embryos at the 1 cell stage. Two aliquots of 5HT-conjugated resin were loaded with either frog embryo lysate plus 20 mM 5HT (negative control) or frog embryo lysate (test). The eluate from both experiments was subjected to a 10% SDS-PAGE and, after proper sample preparation and trypsin in-solution digestion, the peptide mixtures were analyzed by LC/MS-MS.

The raw data collected from the ion trap mass spectrometer was analyzed for protein identification using a database for Xenopus laevis. The screen identified 11 proteins as possible 5HT interactors. Because we were interested in mechanisms by which intracellular 5HT binding proteins could couple to the cytoplasmic epigenetic machinery (as 5HT is delivered inside the cell, via gap junctions [16]), we focused on candidate proteins with known interactions with HDAC proteins and/or DNA-binding domains. We did not recover the classical 5HT transmembrane receptors in this screen, since we focused our analysis on the cytoplasmic (not the plasma membrane) protein fraction. Indeed, our screen successfully identified 10 candidates with characteristics of cytoplasmic proteins and 1 candidate having a DNA binding domain. This DNA binding-protein was identified as the Max-interacting transcriptional repressor Mad3. Mad3 is involved in the control of the epigenetic state of cells, since it can antagonize transcription by recruiting co-repressor complexes that contain HDAC I [58]. Indeed, the repressive activity of Mad proteins has been shown to be due to HDAC activity and inhibited by HDAC blockers [59]. Mad proteins also interact with Max, and this complex has been implicated in different aspects of cell biology and cancer transformation [60].

A role for Mad3 as a new player in the 5HT signaling pathway during LR patterning requires that Mad3 protein be already present in early embryos. Indeed, Mad3 mRNA is present in the blastomeres at the 4-cell stage by in situ hybridization (Figure 4A). By the 32-64 cell stage (stage 6-7), Mad3 mRNA is found only in the animal cells.

To validate the presence of Mad3 protein in early embryos, as detected on the ion trap mass spectrometer, we performed two different sets of analysis. In the first strategy we used a monoclonal anti-Mad3 antibody to detect the endogenous Mad3 protein in embryos at stage 7 (Figure 4B, lane 1). To confirm the specificity of this anti-Mad3 antibody we injected embryos with the Xenopus Mad3 protein fused to the flag epitope at the 1 cell stage. The lysate was collected at the stage 7 and used for immunoblotting to detect the exogenous Xenopus Mad3. Indeed, the injected Mad3-flag protein was detected migrating at 30 kDa, the expected molecular size for the Xenopus Mad3 protein (Figure 4, lane 4). Taken together these results confirm that Mad3 is present as a maternal mRNA and protein at the early cleavage stages and thus is potentially available for interaction with 5HT and/or HDAC.

To examine a role for Mad3 during LR patterning, we utilized 2 different constructs encoding loss-of-function and gain-of-function reagents for Mad3: Mad3Vp16 (a constitutively active form of Mad3) and EngMad3 (a constitutively repressive form of Mad3) [61].

To probe the requirement of Mad3 between the left and right side of the embryo, the single dorsal or ventral, left or right blastomeres at the stage 3 (4 cells) were injected with mRNA encoding EngMad3 or Vp16Mad3 along with the lineage tracer lacZ mRNA. The embryos were then scored for organ placement at stage 45. Whereas dorsal injections led to considerable toxicity, ventral left or right expression of the EngMad3 (repressive) construct induced heterotaxia at a statistically significant level when compared to control Eng-expressing embryos (Figure 5A). Conversely, embryos injected with Vp16Mad3 in the right ventral blastomere presented significant heterotaxia when compared with Vp16 control injected embryos (Figure 5A).

We also analyzed injected embryos with EngMad3 or Vp16Mad3 by in situ hybridization with a Xnr-1 probe. Indeed, embryos expressing the repressive EngMad3 construct on the left or right side showed Xnr-1 absence, and embryos injected with the activating Vp16Mad3 on the right showed a consistent level of bilateral Xnr-1 expression (Figure 5B).

Thus, Mad3 gain-of-function on left (EngMad3) and a Mad3 loss-of-function on the right (Vp16Mad3) can both control Xnr-1's transcriptional status. Taken together, these results support Mad3 as a new LR determinant, and are consistent with a role for Mad3 as a modulator of Xnr-1 expression.

Next we tested whether Mad3's biological activity is dependent on 5HT. The first strategy was to carry out a binding assay using surface plasmon resonance (SPR) [62]. Recombinant Mad3 protein expressed as a fusion with glutathione S-transferase (Mad3-GST) was immobilized in a Biacore chip and a 5HT solution was then passed over the surface. To subtract possible non-specific binding, we used a reference surface in the same chip where GST alone was immobilized. 5HT showed a high association rate and a low dissociation rate (Figure 6A). Four different concentrations were examined, ranging from 400 μM to 2 mM. Using a 1:1 Langmuir fitting algorithm, we calculated the equilibrium dissociation constant to be in the range of 6.7 to 26 μM (95% confidence). Based on the 5HT levels found in early embryos, 3.11 pmol/egg [15], we measured embryo diameters in order to calculated the endogenous concentration of 5HT. Considering the whole embryo as a sphere (diameter = 1mm), 5HT concentration can then be calculated as 5.7 μM, which is very close to the Kd values determined in vitro. It should be noted that 5HT is not homogeneously distributed in the whole embryo but concentrated in the right blastomeres descendants at the stage 7 [15]; therefore, the local concentrations could be even higher than the ones calculated here, well above the determined dissociation constant. Hence, these in vitro data does suggest that 5HT and Mad3 form a complex in vivo.

The second strategy was to generate a Mad3 mutant lacking 5HT-binding sites. To generate this Mad3 mutant first we mapped the putative 5HT binding sites on Mad3 by modeling the 3-dimensional structure for Mad3 using the crystal structure of a lipocalin AM182, a well characterized 5HT-binding protein, complexed with 5HT [43]. We chose lipocalin because among all known 5HT binding proteins, including class-A GPCRs (G protein-coupled receptors), a conserved salt bridge interaction exists between the amine group and an aspartic acid or glutamic acid residue on the protein [45,63-65]. In lipocalin AM182, this conserved salt bridge is formed between Asp106 and the amine group of 5HT. The structural equivalent aspartic acid in Mad3 is Asp163. Thus the potential binding site for 5HT in Mad3 was derived based on the corresponding binding site residues from lipocalin structure.

To gain insight into the physiological relevance of these putative 5HT binding sites on Mad3, we performed functional experiments with a Mad3 mutant generated based on the structural modeling. Based on the docking mode of 5HT in the Mad3 structure, residues D145, D148, D163, Q125 and Q161 are proposed to form important components of the 5HT binding pocket on Mad3 (see Material and Methods). An evaluation of the binding site residues was performed by designing a Mad3-5mut-flag construct harboring mutations on the five amino acids predicted to be involved in the 5HT putative binding site (Q125E, Q161E, D145N, D148N and D163N). Embryos injected with Mad3-WT-flag at the 1 cell stage developed significant levels of randomization of the heart, gut and gall bladder (heterotaxia) in the absence of other defects and with normal dorsoanterior development at stage 45; in contrast, the Mad3-5mut did not induce this phenotype (Figure 6C, Mad3 WT 17%, p < 0.001, n = 93; Mad3-5 mut 2%, p = 0.7, n = 90).

To confirm that the lack of biological activity presented by the Mad3-5mut-flag was due to the abrogation of the 5HT putative binding site, we performed a Co-IP assay with a 5HT antibody. Because 5HT is evenly distributed in embryos from stage 1 (1 cell) through stage 5 (16 cells), we injected embryos at the 1 cell stage to probe Mad's ability to bind to any available 5HT present in the embryo, and in turn, to test if this ability would be lost in Mad3-5mut injected embryos. Embryos were then injected at the 1-cell stage with Mad3WT-flag or Mad3-5mut-flag and whole embryo lysate was prepared at the stage 7 (64 cells) and incubated with ati-5HT followed by immunoblotting with anti-flag. As predicted, the Mad3-5mut-flag protein was not found in complex with 5HT (Figure 6C, lanes 4,5,6) when compared to the Co-IP performed with Mad3 WT-flag injected embryos (Figure 6, lanes 1,2,3). These results confirm the importance of these 5 residues for the Mad3 biological activity during laterality establishment due to 5HT. We conclude that serotonin binding is a key component of Mad3's ability to functionally participate in LR patterning.

The importance of chromatin remodeling for cancer and developmental biology is increasingly appreciated [66]. The epigenetic state of the cell can be characterized by modifications on histone proteins such as acetylation, methylation, phosphorylation and ubiquitination [67]. One important epigenetic state is characterized by the levels of histone tail acetylation that are tightly controlled by HDACs [68].

Previous work has shown that the Xenopus HDAC protein is found at high concentrations in the oocyte nucleus and its levels remain constant through oocyte maturation, fertilization, and early cleavage stages but decreases after the blastula stage [29]. Measurements of Xenopus HDAC activity indicate that it also remains constant from oogenesis until early embryogenesis and that it is sensitive to HDAC blockers [30], which is consistent with the data we obtained on its role in the early LR pathway. In addition, our study indicates that although HDAC mRNA is expressed in all animal blastomeres, HDAC activity on the right side is important for LR, as HDAC DN injections on the right side led to heterotaxia that was correlated with absence of Xnr-1 expression. In addition, NaB treatment shows that the critical HDAC activity for LR development overlaps with 5HT signaling (right blastomeres) during cleavage stages and takes place before cilia flow (Figure 2). The results of the molecular-genetic loss-of-function (dominant negative construct misexpression) confirm the targeting of the pharmacological reagent (NaB), while the latter offers the opportunity of temporally-limited exposure.

Our results using the ChIP and qPCR analyses indicated that HDAC inhibition led to accumulation of histones H3 and H4 and the epigenetic marker H3K4m2 in the intronic region of Xnr-1. This region is particularly critical for Xnr-1 asymmetric expression and we propose that HDAC activity targets this region of the Xnr-1 gene. Although the H3K4me2 marker has been associated with transcriptional activation when it occupies promoter regions [69], it was also shown to be associated with repression [55]. Interestingly, H3K4me2-enriched regions do not correlate with transcription start sites and have been proposed to control chromatin states over regulatory regions [70]. This feature is particularly interesting as H3K4me2 was found enriched only on the intronic region of Xnr-1 that contains the asymmetric regulatory element. One hypothesis is that, after drug removal, the H3K4me2 marker remained stably tethered on the Xnr-1 intronic region and led to its repression due to the recruitment of zygotic HDACs during late developmental stages.