Xnrs and Activin Regulate Distinct Genes during Xenopus Development: Activin Regulates Cell Division

Background The mesoderm of the amphibian embryo is formed through an inductive interaction in which vegetal cells of the blastula-staged embryo act on overlying equatorial cells. Candidate mesoderm-inducing factors include members of the transforming growth factor type β family such as Vg1, activin B, the nodal-related proteins and derrière. Methodology and Principle Findings Microarray analysis reveals different functions for activin B and the nodal-related proteins during early Xenopus development. Inhibition of nodal-related protein function causes the down-regulation of regionally expressed genes such as chordin, dickkopf and XSox17α/β, while genes that are mis-regulated in the absence of activin B tend to be more widely expressed and, interestingly, include several that are involved in cell cycle regulation. Consistent with the latter observation, cells of the involuting dorsal axial mesoderm, which normally undergo cell cycle arrest, continue to proliferate when the function of activin B is inhibited. Conclusions/Significance These observations reveal distinct functions for these two classes of the TGF-β family during early Xenopus development, and in doing so identify a new role for activin B during gastrulation.


INTRODUCTION
The mesoderm of the amphibian embryo arises through an inductive interaction in which cells of the vegetal hemisphere act on overlying equatorial cells [1]. Of the several mesoderminducing factors that have been discovered, most are members of the transforming growth factor type b family. These include activin [2][3][4], Vg1 [5,6], five nodal-related proteins [7][8][9], and derrière [10]. Although these factors have similar abilities to induce gene expression in isolated animal pole regions, they are differently expressed in the embryo (see above references) and under some experimental conditions have different abilities to exert long-range effects [11,12]. In addition, each exerts different effects at different concentrations [7,13]. The challenge now is to elucidate the individual roles of these proteins within the embryo and to ask how their actions are coordinated.
Some attempts along these lines have been made, and it proves that although each of the factors is essential for normal development, their roles differ. For example, ablation of the maternal transcripts encoding Vg1 causes a reduction in anterior and dorsal development and the down-regulation of genes such as chordin, cerberus and noggin [6]. Of the zygotically-expressed inducing factors, depletion of activin also causes axial defects [3,14,15], although these are less severe than those caused by loss of Vg1, and inhibition of derrière activity causes just posterior defects [10]. Simultaneous inhibition of the activities of all the nodal related proteins, by expression of Cerberus-short, causes loss of mesoderm [16,17] and the down regulation of genes such as Chordin and Pintallavis [18]. The requirements of the individual nodal related proteins have not been studied in detail, although injection of antisense morpholino oligonucleotides directed against Xnr1 causes defects in left-right axis determination [19].
Here we perform microarray analyses of gene expression in embryos in which activin or nodal-related signalling has been inhibited. We find that activin and the nodal-related proteins regulate distinct and almost completely non-overlapping sets of genes, with those regulated by the nodal-related genes tending to be expressed in a more restricted pattern than those regulated by activin. It further proved that the nodal-related proteins often regulate the expression of genes involved in regional specification, while activin particularly regulates genes involved in the control of the cell cycle. Consistent with this observation, we find that inhibition of activin B in the early embryo causes dorsal axial mesodermal cells to fail to exit the cell cycle: the results of others [20][21][22] suggest that it is the continued proliferation of these cells that underlies the gastrulation defects observed in such embryos.

Microarray results
In an effort to understand the different requirements for activin B and the nodal-related genes during Xenopus development, we have carried out microarray analyses of gene expression in embryos in which signalling by the two classes of factor has been disrupted. Activin signalling was blocked using an antisense morpholino oligonucleotide [3], and nodal-related signalling by Cerberusshort, a truncated form of Cerberus [17]. Our microarray slides comprise 10,898 probes designed to recognise sequences derived from a large scale Xenopus tropicalis EST project [23]. These arrays also recognise X. laevis transcripts [24].
For each series of experiments Xenopus laevis embryos from three different spawnings were injected with RNA encoding Cerberusshort (150 pg into each blastomere at the four-cell stage) or with antisense morpholino oligonucleotide MO3 (50 ng into the onecell stage) (samples), or with water or antisense morpholino oligonucleotide mMO1 (50 ng) (controls). These doses of Cerberus-short RNA and MO3 were based on previous work [3,16] and were chosen so as to yield a strong phenotype in which gastrulation was substantially or completely inhibited. In an effort to identify early and perhaps direct targets of activin and the nodal-related proteins, embryos were cultured to stage 10.5 for RNA isolation and some were allowed to develop to later stages to confirm that embryos displayed the expected phenotypes ( Fig. 1A-F). Each microarray slide was hybridised with a 1:1 mixture of sample and control cDNAs, each labelled with a different dye. Each hybridisation was repeated with the Cy3 and Cy5 dyes 'swapped', so that six microarray slides were hybridised for each experiment.
Transcripts recognised by the oligonucleotides were considered to be differentially expressed when (i) they showed at least a two fold difference (sample versus control) in expression levels in at least four out of the six microarrays and (ii) were significantly different (q = 0; see Experimental procedures). In embryos in which activin B signalling was inhibited, 40 oligonucleotides fulfilled these rigorous criteria, of which 8 were down regulated, and in those in which nodal signalling was inhibited, 20 oligonucleotides (representing 18 genes) were differentially expressed, of which 17 were down regulated ( Table 1). The up regulation of Cerberus in the latter experiment is probably due to the introduction of Cerberusshort mRNA into these embryos. Only Sizzled, which encodes an inhibitor of the Tolloid Proteinase [25], was differentially expressed in both types of embryo.
Our experiments identify fewer nodal-regulated genes than the recent microarray study of Sinner and colleagues [26]. This difference probably derives from the facts that Sinner and colleagues harvested embryos at stage 11 rather than 10.5, and defined genes as being differentially expressed if expression levels differed by a factor of 1.4 rather than 2.0. Like Wessely and colleagues, who used a macroarray approach [18], we note that both Chordin and Xsox-17beta are down regulated by Cerberusshort. We also note that some genes that are down regulated following interference with activin signalling, such as Xbra and goosecoid [3], were not identified in the present screen. The most likely explanation for this apparent discrepancy is that the expression of such genes is frequently reduced by only 50% or thereabouts [3], and our criteria for defining genes as being differentially expressed (see above) is so stringent that such differences might be regarded as insignificant. RT-PCR analysis of the RNA samples used on the microarrays confirmed previous observations [3] that the expression of these genes is indeed reduced in embryos in which activin signalling is inhibited (data not shown).

21.5
Restricted [52] Up regulated genes are shown in green and down regulated genes in red. The Table also shows the RT-PCR data plotted in Fig. 1

Real-time RT-PCR validation
Our microarray results were validated by real-time RT-PCR The X. laevis homologues of the X. tropicalis cDNAs recognised by the oligonucleotides (http://informatics.gurdon.cam.ac.uk/cgi-bin/ public.exe) were identified by BLAST searches (Table 1), and PCR primers were designed for the great majority of the transcripts that were considered to be differentially expressed. In the case of the activin B experiment, we were unable to identify X. laevis homologues for six of the cDNAs, and two primer pairs did not yield a product; in the case of the Cerberus-short experiment, X. laevis homologues could not be identified for two cDNAs. Our RT-PCR analysis used the same RNA samples that were used for microarray experiments. Of the genes tested, 80% of those identified in the activin B experiment were confirmed as being differentially expressed, and all of those identified in the Cerberus-short experiment were similarly verified. Bilateral correlation analysis of the results obtained by microarray hybridization and those obtained by real-time RT-PCR showed a Pearson Correlation of 0.848 (p = 0.000) for the activin B experiment and of 0.975 (p = 0.000) for the Cerberus-short experiment (Fig. 1G,H). RT-PCR experiments confirmed that genes regulated by activin signalling are not regulated by nodal-related signalling, and vice-versa (Table 1). Together, these experiments show that activin and the nodal-related genes regulate distinct genes during early Xenopus development.

Classification of genes regulated by activin and nodal-related genes
The expression pattern of each differentially expressed gene was determined from the literature, where possible, or by carrying out in situ hybridisations using Xenopus tropicalis embryos with probes generated by the polymerase chain reaction (PCR). Consistent with the different expression patterns of activin B and of the nodalrelated genes [3,[7][8][9]27], the expression patterns of the genes regulated by the two types of signalling molecules differed (see Table 1). Thus, of the 15 different genes regulated by nodalrelated signalling whose expression patterns we know, all are expressed in a restricted fashion (for example, see Fig. 2A,B), and of the 31 genes regulated by activin B, 28 are expressed ubiquitously (for example, see Fig. 2C-F) and three in a restricted fashion.
Genes were then manually classified according to the annotation of their human homologues (NCBI databases, http://www. ncbi.nih.gov/). Interestingly, this analysis also revealed differences between embryos lacking activin B and those in which nodal related signalling is inhibited (Fig. 2G). In particular, while several of the genes regulated by the nodal-related genes are involved in signal transduction or the regulation of transcription, several of the genes whose expression is affected by lack of activin B activity are involved in cell cycle regulation; this is not the case for embryos in which nodal signalling is inhibited.

Activin regulates cell division in the involuting mesoderm
Both our microarray experiments and our real-time RT-PCR analyses show that down-regulation of activin B, but not loss of nodal-related activity, causes the mis-regulation of genes involved in cell cycle control. One of the effects of the loss of activin B function is a disruption of gastrulation [3], and in this connection we note that the mitotic index of involuting dorsal mesoderm is significantly decreased during gastrulation [28] and that arrest of the cell cycle is required for both bottle cell formation [20] and for convergent extension movements [21,22]. We therefore asked whether loss of activin B affects cell division during early embryogenesis.
Embryos injected with control oligonucleotide mMO1 or specific antisense oligonucleotide MO3 were fixed at the mid gastrula stage and stained using an antibody recognising phosphorylated histone H3, which marks mitotic chromosomes [28]. Inspection of such embryos revealed that the downregulation of the cell cycle that normally takes place in dorsal axial mesoderm does not occur (Fig. 3). In three control embryos stained as sections the mean mitotic index in dorsal axial mesoderm was 0%; in six embryos injected with MO3 the mitotic index was 12.762.7% (mean6standard deviation). Similarly, in a control embryo stained as a whole-mount and then sectioned, the mitotic index was 0%; in an embryo injected with MO3 it was 20%. This failure of the dorsal axial mesoderm to undergo cell cycle arrest is consistent with the observed mis-regulation of cell cycle genes, and it may explain why embryos lacking activin function fail to gastrulate properly [see refs 20-22].

DISCUSSION
Our experiments show that activin B and the nodal-related proteins regulate distinct sets of genes in the early Xenopus embryo. In the future it will be interesting to investigate the molecular basis of this difference. One difference between activin and the nodalrelated proteins is that their expression patterns differ, with activin B being expressed ubiquitously [3,27] and the nodal-related proteins being restricted to the vegetal and equatorial regions of the embryo [7][8][9]. Consistent with these observations, we note that nodal-regulated genes tend to be expressed in more restricted patterns than do activin-regulated genes ( Fig. 2A-F). Another difference is that signalling by the nodal-related proteins, but not activin, requires responding cells to express EGF-CFC family members such as XCR1, 2 and 3 [29][30][31][32]. This difference between activin and the nodal-related proteins may underlie the ability of activin to activate Smad2 earlier than does Xnr1 or derrière [33]. We note that other studies have also noted differences between activin and nodal signalling; for example, continuous treatment of P19 cells with activin causes only transient activation of Smad2 while treatment with nodal causes sustained activation [32].
Of the genes that are exclusively regulated by activin, several have been implicated in cell cycle regulation (Fig. 2G), and embryos that lack activin B function fail to arrest the cell cycle in dorsal axial mesoderm (Fig. 3). These observations indicate that the role of activin B differs from that of the nodal-related proteins in the early Xenopus embryo, and that one of its functions is to control the cell cycle during this critical phase of early Xenopus development. This is of importance because axial mesodermal cells arrest the cell cycle after involution [28], and if they are forced to proliferate, this results in a severe disruption of gastrulation [20][21][22]. Interestingly, we note that the ability of activin to inhibit cell division is not restricted to the early Xenopus embryos; activin also causes cell growth arrest in human breast cancer cells and in human hepatocytes [34,35].
We note no effect of the loss of activin on the cell cycle elsewhere in the Xenopus embryo; there is no acceleration of cell division in the animal hemisphere, for example, in embryos injected with MO3. It is likely that the cell cycle in the dorsal marginal zone is regulated through locally-acting mRNAs or proteins that require activin signalling for their expression or appropriate post-translation modification.
Finally, what do our results say about the role of activin in mesodermal patterning? Although we emphasise here the role of activin in controlling the expression of genes involved in the regulation of the cell cycle, our previous data, confirmed in the course of the present work (data not shown), indicates that in the absence of activin the expression of genes such as goosecoid, chordin and Xbra is reduced by 20-80%, depending on stage and dose of antisense morpholino oligonucleotide [3]. These observations suggest that activin and the nodal-related proteins (together  with Vg1 and derrière) cooperate to specify mesodermal pattern in the embryo, although the results described in this paper argue that the role of activin in this process is less significant than is the role of the Xnrs.

MATERIALS AND METHODS Xenopus embryo manipulations and microinjection
Embryos of Xenopus laevis were obtained by artificial fertilisation, maintained in 10% normal amphibian medium [36], and staged as described [37]. For inhibition of nodal-related protein function, embryos were injected at the one cell stage with 600 pg Cerberusshort RNA [17] or, as a control, water. For inhibition of activin B, embryos were injected with 50 ng antisense morpholino oligonucleotide MO3 [3] or, as a control, mMO1 [3]. Embryos were harvested at stage 10.5 for microarray analysis or stage 12 for immunocytochemistry.
Microarray construction, RNA isolation, labelling and microarray hybridisation These were performed as described [24].

Microarray data analysis
Microarray results were imported into Acuity (Axon) and normalised using Lowess normalisation. Data files were created for points which satisfied the following filter: (Sum of Medians) $500 AND (Flags) $0 AND (%.B532+1SD)$55 OR (%.B635+1SD)$55. This filter eliminates data points flagged as bad by GenePix, or that had the sum of media less than 500, or which had fewer than 55% of pixels above background. Points passing these criteria for at least four out of the six microarrays were used for further analysis. Oligonucleotides were considered to be differentially expressed when they showed at least a two fold difference in expression levels in four out of the six microarrays and had a q value of 0 as assessed by the Significance Analysis of Microarrays software [38]. The microarray datasets were deposited in the GEO data repository (http://www.ncbi.nlm.nih. gov/projects/geo/index.cgi) (accession numbers GSE4771 and GSE4777).

Real time RT-PCR
Differential expression was validated by real-time RT-PCR using the Roche LightCycler 480. Primers specific for ornithine decarboxylase (ODC) were as described [3]; others are listed in Table 2.

In situ hybridisation
This was carried out on embryos of Xenopus tropicalis, essentially as described [39,40]. Probes were made by use of T7 RNA polymerase; substrates were PCR products obtained using T7 and SP6 primers applied to cDNA clones derived from a large scale Xenopus tropicalis EST project [23].