RBMX gene is essential for brain development in zebrafish


  • Enkhjargal Tsend-Ayush,

    Corresponding author
    1. Research School of Biological Sciences, Australian National University, Canberra, Australian Capital Territory, Australia
    • Research School of Biological Sciences, Australian National University, Canberra, ACT 2601, GPO Box 475, Australia
    Search for more papers by this author
  • Lynda A. O'Sullivan,

    1. Centre for Cellular and Molecular Biology, School of Biological and Chemical Sciences, Deakin University, Melbourne, Australia
    Search for more papers by this author
  • Frank S. Grützner,

    1. Research School of Biological Sciences, Australian National University, Canberra, Australian Capital Territory, Australia
    Search for more papers by this author
  • Sara M.N. Onnebo,

    1. Centre for Cellular and Molecular Biology, School of Biological and Chemical Sciences, Deakin University, Melbourne, Australia
    Search for more papers by this author
  • Rowena S. Lewis,

    1. Centre for Cellular and Molecular Biology, School of Biological and Chemical Sciences, Deakin University, Melbourne, Australia
    Search for more papers by this author
  • Margaret L. Delbridge,

    1. Research School of Biological Sciences, Australian National University, Canberra, Australian Capital Territory, Australia
    Search for more papers by this author
  • Jennifer A. Marshall Graves,

    1. Research School of Biological Sciences, Australian National University, Canberra, Australian Capital Territory, Australia
    Search for more papers by this author
  • Alister C. Ward

    1. Centre for Cellular and Molecular Biology, School of Biological and Chemical Sciences, Deakin University, Melbourne, Australia
    Search for more papers by this author


The human RBMX gene was discovered recently through its homology to the spermatogenesis candidate gene RBMY. Its position on the human X chromosome suggests that it may be involved in X-linked mental retardation syndromes. However, to date there is scant information on the in vivo role of RBMX. To address this issue, we have isolated a zebrafish rbmx orthologue and characterized its embryonic expression pattern. Zebrafish rbmx is maternally expressed and then widely expressed in the embryo up to 24 hr postfertilization. In later stages of embryonic development, rbmx transcripts are localized predominantly in the brain, branchial arches, and liver primordium. The function of rbmx during embryonic development was examined by the use of an antisense morpholino targeting rbmx. The rbmx-morphants displayed an underdeveloped head and eyes, reduced body size, defective somite patterning, and absence of jaws. Furthermore, in the absence of functional rbmx, expression of specific markers for the fore- and hindbrain (otx2, krox20) was severely reduced. These studies demonstrate for the first time that rbmx is required for normal embryonic development, in particular of the brain, consistent with a role in X-linked mental retardation. Developmental Dynamics 234:682–688, 2005. © 2005 Wiley-Liss, Inc.


RBMX (RNA-binding motif on the X chromosome, also called hnRNPG) is the X homologue of the candidate spermatogenesis gene RBMY located on the human and mouse Y chromosome (Delbridge et al., 1999; Mazeyrat et al., 1999). The RBMX and RBMY genes have differentiated and evolved separately on the mammalian sex chromosomes, because marsupials and eutherians diverged 180 MYA. RBMY has been identified on the Y chromosome of all therian mammals and RBMX has been found on the X chromosome in human and mouse. Human RBMY is expressed exclusively in the testis (Ma et al., 1993; Elliott et al., 1997), whereas human RBMX is ubiquitously expressed (Soulard et al., 1993) and subject to X inactivation (Lingenfelter et al., 2001).

The human RBMX and RBMY genes encode RNA-binding proteins with approximately 60% homology (Delbridge et al., 1997). RBMY contains a triple repeat of an internal exon compared with RBMX (Delbridge et al., 1999) that results in four SRGY boxes in RBMY (Ma et al., 1993). Both RBMY and RBMX are nuclear proteins, which interact with splice site selection and signal transduction proteins (Elliott et al., 2000; Venables et al., 1999, 2000) and, therefore, are likely to be involved in RNA processing. The RBMX protein contains an 80 amino acid RNA-binding domain at the N-terminus. The C-terminal auxiliary domain of RBMX is composed of a high proportion of serine, arginine, and glycine residues and is posttranscriptionally glycosylated. These features are typical of many RNA-binding proteins, and so presumably enhance the RNA-binding properties of this protein (Weighardt et al., 1996).

The functions of the RBMX and RBMY genes are unknown. The testis-specific expression of RBMY in mammals suggests it is involved in RNA processing during spermatogenesis, and deletions of the Y chromosome that include RBMY lead to azoospermia (Ma et al., 1993). In contrast, RBMX has a widespread expression pattern, suggesting that it may exert its effects on general developmental processes. RBMX is located on the human X chromosome at Xq26 (Delbridge et al., 1999), and on the mouse X chromosome at XA3-XA5 (Mazeyrat et al., 1999). Its position on the human X chromosome, and its widespread expression pattern suggest that it may be involved in one of several multifaceted disorders that have been mapped to Xq26-27 (Zucchi et al., 1999), which include those leading to mental retardation as well as a range of other physical abnormalities. The RBMX gene has also been implicated in human systemic lupus erythematosus (SLE), as dogs with an SLE-like syndrome have circulating autoantibodies to RBMX (Soulard et al., 2002).

To directly investigate the in vivo function of RBMX, we have isolated a zebrafish orthologue and examined its temporal and spatial expression pattern. Antisense morpholino-mediated knockdown of zebrafish rbmx was used to analyse the role of this gene in embryogenesis. Our results provide the first functional evidence for a key role for rbmx in brain development, supporting the putative role of this gene in X-linked mental retardation syndromes.


Identification and Characterization of a Zebrafish rbmx Orthologue

Zebrafish expressed sequence tag (EST) clones with high homology to human RBMX were identified in the NCBI database. Sequence analysis revealed that they represented a single gene with high homology to human RBMX. A zebrafish cDNA clone IMAGp998B0811777Q3 (RZPD, Germany) containing full-length zebrafish RBMX (rbmx) cDNA (GenBank accession no. AJ717349) was used for further experiments. The 1,475-bp gene encodes a putative rbmx protein of 379 amino acids, with 77% identity to human RBMX protein.

Sequence and Evolutionary Relationships Among RBMX Genes

Multiple alignments of the amino acid sequences derived from the RBMX family of genes from human, mouse, marsupial, and fish confirmed that there is extensive conservation across the entire protein, with the N-terminal RNA-binding motif being the most conserved region (Fig. 1a). The genetic distance between the zebrafish rbmx and other known RBM genes calculated from a similar amino acid alignment were used to construct a dendrogram (Fig. 1b). Human and rodent RBMX gene products form one clade, from which marsupial and chicken RBMX gene products are somewhat more distantly related. Another discrete clade contains the fugu and zebrafish RBMX gene products, but shows a highly significant degree of similarity to the RBMX gene family and is distinctly different to the related RBM genes (RBMY, RBMX2, and hnRNPG-T). This finding suggests that RBMX genes in mammals and fish evolved from a common ancestor. Of interest, the fish RBMX genes have longer branches compared with the mammalian homologues, suggesting a higher rate of sequence evolution within the fish lineage. One possibility is that its location on the X chromosome in mammals has selected for a new function that is important and, therefore, highly conserved. Another is that this conservation reflects its new location on the highly conserved X chromosome.

Figure 1.

Molecular relationship of zebrafish rbmx with its homologues. a: Multiple alignment of orthologous RBMX sequences using CLUSTAL X: human (NM_002139), mouse (NM_011252), rat (Ensembl Gene ID: ENSRNOG00000000866), tammar wallaby (AF034741), chicken (Ensembl Gene ID: ENSGALG0000000645), zebrafish (AJ717349) and fugu (Ensembl Gene ID: SINFRUG00000136273). * indicate identities; – in sequence are introduced gaps, while the RNA-binding domain is indicated with a black bar below the sequence. b: Phylogenetic tree constructed using the above sequences (grey box) plus sequences for human RBMX2 (NM_016024), RBMY1A1 (NM_005058) and hnRNPG-T (NM_014469) utilizing the Neighbour-Joining algorithm, with bootstrapping values indicated. c: Structure of human (hu) and zebrafish (zf) rbmx genes, with grey boxes displaying untranslated regions and open boxes exons, with the length of each (in nucleotides) indicated.

The rbmx cDNA sequence was used to identify a fully sequenced bacterial artificial chromosome (BAC) clone CH211-127L15 containing this gene by BLAST searching of NCBI (http://www.ncbi.nih.gov). This strategy was used to characterize the genomic structure of the zebrafish rbmx gene by alignment with cDNA sequences and the application of the GT-AG rule to define intron–exon boundaries. Overall exon–intron structure and size also showed a high degree of conservation between zebrafish and human genes. Furthermore, an EST corresponding to the rbmx gene has been mapped previously to zebrafish chromosome 14 at 23.6 cM in a region showing synteny to the human X-chromosome (Woods et al., 2000). Closer analysis revealed that the rbmx and arhgef6 genes (ENSDARG00000006683, http://www.ensembl.org/) are adjacent in both zebrafish and humans (Xq26.3, data not shown), indicating conservation of synteny relationships as well.

Together these data suggest the zebrafish gene identified is a true orthologue of the human RBMX. The high conservation between the zebrafish rbmx and human RBMX genes, suggests that the zebrafish orthologue represents a worthwhile model for the further characterisation of RBMX gene.

Differential Expression of rbmx mRNA During Embryogenesis

The spatiotemporal expression pattern of rbmx mRNA throughout embryonic development was examined by in situ hybridization using a full-length rbmx antisense riboprobe. Strong ubiquitous hybridization was seen at the one-cell stage and into the gastrula period (Fig. 2a,b), indicating that the rbmx transcript is maternally derived and suggesting a role in early embryonic processes. By the late-segmentation stage of development around 24 hours postfertilization (hpf), expression became more restricted to the anterior part of the embryo, including of the forebrain (telencephalon and diencephalon), midbrain, tectum, cerebellum, hindbrain, and developing eye (Fig. 2c). By 48 hpf and in the early larval period (from 72 hpf), expression of rbmx was observed throughout the head, but more strongly in the forebrain, midbrain, and cerebellum, as well as more weakly in the ganglion cell layer of the retina and the otic vesicles. Expression was also detected in the notochord, branchial arches, liver and gut primordium, and pectoral fin at this time (Fig. 2d–h). By 7 days postfertilization (dpf), expression was limited to the midbrain, predominantly in the tectum, cerebellum, branchial arches, liver, and gut (Fig. 2i). Penetrance appeared not to be an issue because markers of internal structures produced a strong signal at this same stage (data not shown). To control for specificity, embryos of each age were also probed in parallel with a sense probe, which gave no signal at any stage (data not shown).

Figure 2.

Developmental expression pattern of zebrafish rbmx mRNA Whole-mount in situ hybridization with antisense zebrafish rbmx at the following developmental stages: a: one cell (animal pole toward top); b: 8 hpf (lateral view); c: 24 hpf (lateral view); d: 48 hpf (lateral view); e: 48 hpf cross-section of the head; f: 48 hpf (lateral view); g: 72 hpf (lateral view); h: 72 hpf (dorsal view); i: 7 dpf (lateral view). Where applicable, anterior is orientated toward the left. ba, branchial arches; ce, cerebellum; ep, epiphysis; fb, forebrain; gu, gut; hb, hindbrain; li, liver primordium; mb, midbrain; nc, notochord; ov, otic vesicle; pf, pectoral fin; re, retina; te, tectum.

Our data on zebrafish rbmx provides the first insight into the embryonic expression profile of this gene in any vertebrate, because other studies examined RBMX expression only in adult tissues. For example, human RBMX was shown to be expressed in adult brain, heart, liver, and, most strongly, in skeletal muscle (Nasim et al., 2003). In mice, expression is widespread, although the available data are insufficient to make a more detailed description (Mazeyrat et al., 1999). Importantly, strong expression of rbmx was seen in the developing brain from 24 hpf to 7 dpf, suggesting its possible role in zebrafish neural development.

Consequences of rbmx Knockdown on Zebrafish Development

Morpholino-mediated knockdown of genes in zebrafish embryos has been proven to be extremely effective in providing information about gene function in this organism (Patton and Zon, 2001). Due to the paucity of data on the in vivo role of rbmx, the strong conservation of zebrafish rbmx to its orthologues, and its interesting embryonic expression pattern, we used morpholinos to analyse the function of the rbmx gene in vivo. Embryos injected with rbmx morpholinos (“rbmx-morphants”) showed no early developmental defects. This finding means either that the encoded protein has no function at this time or that the high levels of maternally derived transcript or protein overcome the effects of the morpholino. Developmental perturbations were first observed in the pharyngula (24 hpf) and became more prominent in the hatching period (48 hpf; Figs. 3, 4).

Figure 3.

rbmx knockdown in zebrafish. a: Lateral view of wildtype (wt) embryo at 48 hpf. b,c: Lateral view of embryos injected with rbmx morpholino (mo) at 48 hpf. The morphant embryos exhibit a number of phenotypes (arrows) including reduced body size, lack of jaws, delayed reabsorption of the yolk sac, defective somite patterning, curly tail and swollen pericardium. The characteristic defects in brain morphology particularly an inflated tectum and cerebral ventricle are shown in (c). d: In situ hybridization with anti-sense rbmx riboprobe of wildtype (wt) and rbmx morphant (mo) embryos at 72 hpf. e: Partial rescue of morphant (mo+re) phenotype by coinjection of rbmx-morpholino with synthetic, morpholino-resistant rbmx mRNA (lateral view at 72 hpf).

Figure 4.

Characterization of zebrafish rbmx-morphant phenotype. Expression of otx2 (a,d), krox20 (b,e) and pax2.1 (c,f) anti-sense riboprobes at 24 hpf (a,b,c) and 48 hpf (d,e,f) in wildtype (wt) and morphant (mo) embryos, as indicated. Arrowheads indicate aberrant expression patterns of gene markers in morphant embryos. All images except of f are lateral view with anterior to the top. f is a dorsal view with anterior to the top. Dash lines indicate the relative length of the dorsoventral axis of the developing brain.

At 24 hpf, the morphants displayed reduced length of the dorsoventral axis of the developing brain compared with that of wild-type embryos (Fig. 4a,b,c, penetrance of 31%, n > 500, representing five different experiments). At 48 hpf, morphants had characteristic defects in brain morphology such as an inflated tectum and cerebral ventricle (31%, n > 500, Fig 3a,b,c) relative to control embryos. Furthermore, at the same time the morphants displayed small heads and eyes (23%, n > 500), reduced hypothalamic tissue between the eyes (mild cyclopia; 29%, n > 500), reduced body/tail size (24%, n > 500), delayed reabsorption of the yolk sac (25%, n > 500), an absence of jaws (20%, n > 500), and defective somite boundary formation (20%, n > 500; Fig. 3a,b,c,d). The rbmx-morphants also showed impaired motility (data not shown). The majority of these phenotypes were observed in the most severely affected embryos, whereas less severely affected embryos exhibited only a subset of these. Other features, such as a curly tail and swollen pericardium were also consistently observed, although at a lower frequencies (12% and 5% respectively, Fig. 3b,c). In morphants with severe phenotypes, no MHB (mid- and hindbrain boundary) structures were visible (Fig. 3c). Most of the observed perturbations occurred in tissues where rbmx is expressed. However, because no expression was observed in somites, somite defects are more likely to be secondary effects. After 96 hpf, rbmx-morphant embryos died. Of interest, injection of rbmx morpholino drastically reduced expression levels of the endogenous gene (Fig. 3d). This finding could be due to either degradation—although such an effect has not been widely reported before—or, more likely, due to loss of autoregulation and suggests that the morpholino may be acting at both RNA and protein levels to knockdown expression.

The phenotypes observed were pleiotropic and occurred in less than a third of the embryos. Importantly, however, no abnormal phenotypes were observed when the same concentration of standard control morpholino (nonspecific) was injected into wild-type embryos (data not shown). Furthermore, when a synthetic capped rbmx mRNA containing a mutated morpholino-binding site was coinjected with the rbmx morpholino, it substantially rescued the severe and mild morphological defects observed in morphant embryos. For example, coinjection of 50 pg of capped mRNA with 3 ng of rbmx morpholino was sufficient to reduce the incidence of severe head defects by 80% to 6% (n > 300, Fig. 3e). Together, this evidence supports the specificity of the results observed.

To further analyze the brain defects seen in morphant embryos, in situ hybridization was performed with gene markers specific for the fore/midbrain (otx2; Li et al., 1994), hindbrain (krox20; Oxtoby and Jowett, 1993), and presumptive midbrain and MHB (pax2.1; Krauss et al., 1991). The otx2 gene is also crucial for the development of the neural crest cell-derived structures of the head (Kimura et al., 1997). There was identical expression of all three genes pax2.1, krox20, and otx2 in wild-type and morphant embryos at 24 hpf, indicating that at this early stage of development the fore-, mid-, and hindbrain is not affected by knockdown of rbmx (Fig. 4a–c).

However, at 48 hpf, the expression of otx2 was absent in fore/midbrain in rbmx-morphant embryos (Fig. 4d), while in the hindbrain nuclei krox20 expression signal was significantly reduced compared with that in wild-type embryos (Fig. 4e). Furthermore, although the expression level of pax2.1 remained robust, the pattern was clearly changed, such that the posterior fusion of the lateral lines of the expression was reduced (Fig. 4c), which may be due to defects in neural tube closure.

The injection of rbmx morpholino severely affected the anatomical structure of forebrain and ultimately loss of one marker gene (otx2) expression. Structures such as the mid- and hindbrain were less susceptible to the elimination of this transcript, although marker gene expression was perturbed. This finding is unlikely to be due to defective patterning, because early expression profiles are relatively intact. Possibly, the function of rbmx is required to maintain but not initiate the expression of otx2, pax2.1, and krox20 genes. Our observations suggest that the differences in expression patterns of these genes at 24 and 48 hpf, instead, may be due to growth retardation or cell death, which would account to the loss of the forebrain, in particular. The alteration of the expression pattern observed in the brain-marker genes indicates that rbmx may play a role in the later stages of brain development (from 48 hpf), which correlates with its strong expression in these tissues.

A role of rbmx in zebrafish brain development is consistent with the suggestion that the human RBMX gene may be involved in neurological disorders that lead to functional and behavioral deficits such as mental retardation syndromes (Graves et al., 2002). This idea first came from its location on human chromosome Xq26, which made it a prime candidate for several X-linked mental retardation (XLMR) syndromes (Delbridge et al., 1999) and from the observation that mutations in a related RNA-binding gene, FMR1, cause Fragile X Mental Retardation (Weighardt et al., 1996).

More recently, several studies have shown that the RBMX gene product (also called hnRNPG) is an active regulator of splicing machinery in neurodegenerative diseases. First, hnRNPG promotes exon 7 inclusion of the survival motor neuron (SMN) gene. Homozygous loss of this gene causes a common motor neuron disease called spinal muscular atrophy (Hofmann and Wirth, 2002). Second, Tau (microtubule-associated protein) is regulated by an interplay of cis elements and trans factors, including hnRNPG. In humans, exon 10 of the gene is an alternatively spliced cassette, which is adult-specific and whose mis-splicing causes inherited frontotemporal dementia with Parkinsonism (Wang et al., 2004).

Many mental retardation syndromes have characteristic facial and skeletal anomalies, short stature, or delayed development, also consistent with the wide range of phenotypic perturbations in rbmx-morphant embryos. For instance, studies of abnormal neuroanatomy reveal a generalized developmental macrencephaly in young autistic patients (Courchesne et al., 2001). This increase in brain size may be due, in part, to an enlargement of the brain ventricles (Hardan et al., 2001; reviewed by Tropepe and Sive, 2003) similar to that observed in the rbmx-knockdown experiment in zebrafish. Likewise, an intimate coupling of neural, muscle, and notochord development, which was also detected in zebrafish rbmx-morphants, has been shown previously in knockdown experiments of several other genes in zebrafish: including SM22 (Yang et al., 2003), lamb1 (Parsons et al., 2002), and Myf5 (Chen and Tsai, 2002). Understanding the link between these wide-ranging effects remain a key challenge.

Thus, we demonstrate for the first time a role for the rbmx gene in embryonic development, with morpholino-mediated knockdown leading to complex developmental phenotypes, including severe effects on brain development. These experiments encourage us to address whether this gene functions in a similar manner in mammalian neural development and/or brain function. Further characterization of RBMX gene may help to elucidate the role of RNA metabolism in the brain, and the effect on mental function of disruptions of RNA processing.


Zebrafish Husbandry

Wild-type zebrafish (Danio rerio) were sourced from St. Kilda Aquarium (St. Kilda, Victoria, Australia) or MAS Imports (Coburg, Victoria, Australia) and maintained under standard laboratory conditions at 28.5°C as described by Westerfield (1995). The stage of the embryos was determined by morphological features and by hours or days post fertilization at 28.5°C (Kimmel et al., 1995). To suppress pigmentation, embryos were raised in egg-water (2.5% [w/v] Na2HPO4; pH 6.0–6.3) containing 0.003% (w/v) 1-phenyl-2-thiourea (PTU, Sigma).

Cloning and Sequence Analysis

Zebrafish EST clones with high homology to human RBMX (SWISSPROT: P38159) were identified in the NCBI database and sequenced. For further experiments, we used a zebrafish cDNA clone IMAGp998B0811777Q3 (RZPD, Germany) containing full-length zebrafish rbmx cDNA, with its sequence deposited at NCBI (GenBank accession no. AJ717349). The position of intron–exon boundaries were determined by alignment of the zebrafish rbmx cDNA sequence with the corresponding zebrafish WGS traces. Multiple sequence alignments were made using CLUSTALX and phylogenetic trees were then generated from these alignments using the neighbor-joining algorithm and confirmed by parsimony analysis with 1,000 bootstrap replicates (PHYLIP 3.5c software package; Saitou and Nei, 1987; Felsenstein, 1989; Higgins and Sharp, 1988).

Morpholino Oligonucleotides and Synthetic mRNA

A morpholino antisense oligonucleotide (Gene Tools, Corvallis, OR) was designed complementary to the 5′ end of the rbmx coding sequence: 5′-TCCCAGGTCGGTCTGCCTCTGCCAT, with the sequence complementary to the start codon italicised. A standard nontargeted morpholino: 5′-CCTCTTACCTCAGTTACAATTTATA was used as a specificity control. Stock solutions were diluted to working concentrations of 2.0 mg/ml in Danieau solution (58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO4, 0.6 mM Ca(NO3) 2, 5 mM HEPES pH 7.6), before injection with a finely drawn capillary tube into the yolk of one- to four-cell embryos using a micromanipulator (Narishige Science Instruments Laboratories, Tokyo, Japan). A total of 2 ng of morpholino antisense oligonucleotide was injected into each embryo.

For phenotypic rescue experiments, rbmx morpholino was coinjected with 50 pg synthetic rbmx mRNA, which has the morpholino target sequence mutated. This procedure was achieved by introducing silent nucleotide substitutions (lower case) by polymerase chain reaction (PCR) into the zebrafish rbmx construct by PCR with the primer 5′-TCCCtGGaaGaTCgGCgTCaGCCAT-3′, and the resulting PCR product subcloned into pCMV-SPORT6 vector and in vitro transcribed using the mMessage mMashine Kit (Ambion, Austin, TX).

Whole-Mount In Situ Hybridization

To examine the spatiotemporal expression patterns of rbmx, digoxigenin (DIG)-labeled sense and antisense probes were synthesized from rbmx cDNA using an in vitro DIG labeling kit (Boehringer Mannheim, Germany). For antisense probe synthesis, the plasmid was linearized with EcoRI and in vitro transcribed with T7 polymerase. The same plasmid was linearized with NotI, and SP6 polymerase was used for sense RNA synthesis. Embryos were fixed in phosphate buffered saline-buffered 4% (w/v) paraformaldehyde. RNA in situ hybridizations were performed essentially as described by Westerfield (1995). All wild-type and morphant embryos were examined with a DMR microscope (Olympus Optical, Tokyo, Japan) and photographed with a Nikon Coolpix4500 digital camera.


After the whole-mount in situ hybridization the stained embryos were dehydrated with a graded ethanol series to 100%, embedded in paraffin and sectioned at 8 μm. Sections were imaged on a microscope Axioscope (Carl Zeiss Meditec, Jena, Germany) and SPOT digital camera (Diagnostic Instruments, Inc., Sterling Heights, MI).


We thank Duncan Shirley and Simon Yoong for fish husbandry, Jillian Healy for assistance with sectioning, and Dr. Sally Stowe for use of the ANU microscope facility. This work used the resources of FishWorks: Collaborative Infrastructure for Zebrafish Research, funded by the Australian Research Council Linkage-Equipment Infrastructure Grant Scheme. A.C.W. was funded by a Sylvia and Charles Viertel Senior Medical Research Fellowship, S.M.N.O. was funded by a Cancer Council of Victoria Postgraduate Cancer Research Scholarship, and R.S.L. and L.A.O. were funded by Australian Postgraduate Awards.