The C. elegans CBFβ homologue BRO-1 interacts with the Runx factor, RNT-1, to promote stem cell proliferation and self-renewal

CBF, or PEBP2, is a heterodimeric DNA binding complex, composed of anα subunit (CBFα/PEBP2α, now known as Runx) and a βsubunit (CBFβ/PEBP2β). CBFβ has been shown to enhance the DNA binding affinity and stability of Runx proteins and is required for many of their in vivo functions (Adya et al.,2000; Huang et al.,2001). There are three mammalian Runx genes, Runx1, Runx2and Runx3, whereas C. elegans contains a single Runx orthologue, rnt-1 (Kagoshima et al., 2005; Nam et al.,2002; Nimmo et al.,2005). Runx factors act as activators or repressors of transcription, depending on the context in which they bind to DNA(Canon and Banerjee, 2003; Stein et al., 2004), and have central roles in cell fate determination and differentiation during development (reviewed by Coffman,2003). In addition, Runx genes have been postulated to act both as tumour suppressors and oncogenes (reviewed by Cameron and Neil, 2004). CBFβ gene rearrangements are also associated with cancer(Adya et al., 1998; Lutterbach et al., 1999).

RNT-1 (the single C. elegans Runx orthologue), is required for the correct division pattern in the stem cell-like, lateral hypodermal seam cell lineages (Nimmo et al., 2005; Kagoshima et al., 2005)(reviewed by Kagoshima et al.,2007). Seam cells have stem-like properties as they undergo self-renewal and expansion whilst producing differentiated cells. They divide asymmetrically at the beginning of each larval stage (larval stages L1 through L4 - distinct developmental stages separated by a molt), producing another seam cell that will continue to proliferate and a hypodermal nucleus that differentiates and fuses with the main hypodermal syncytium(Sulston and Horvitz, 1977). In addition, at the start of L2 and in a male-specific developmental programme in L3, seam cells undergo expansion via symmetrical division(Sulston et al., 1980). In males, these extra seam cells eventually give rise to ray precursor cells from which the sensory rays are derived(Sulston et al., 1980). Seam stem cells can therefore be regarded as pluripotent as they contribute a number of cell types during postembryonic development. rnt-1 mutant males lack the correct number of sensory rays as a result of variable seam cell division failures during larval development(Nimmo et al., 2005). By contrast, ectopic expression of rnt-1 results in seam cell hyperplasia (Nimmo et al.,2005).

In this report, we show, in contrast to previous suggestions(Adya et al., 2000), that C. elegans does indeed contain a functional CBFβ homologue,BRO-1. bro-1 was originally identified as a likely CBFβorthologue by Lee and colleagues (Lee et al., 2004). We find that bro-1 deletion mutants have a very similar male tail phenotype to rnt-1, suggesting the two genes interact, although we also observe unprecedented RNT-1-independent BRO-1 functions. We find that BRO-1 acts not only to increase the affinity of RNT-1 for DNA but also to dramatically increase the specificity of RNT-1-DNA interactions. Overexpression of bro-1 increases the number of seam cells by causing supernumerary seam cell divisions, as well as preventing asymmetric daughters from adopting the hypodermal fate. Furthermore, when bro-1 and rnt-1 are co-overexpressed, massive seam cell hyperplasia results. This work complements and extends analyses of Runx/CBFβ function in other systems, making C. elegans a premier model system for the further study of these important cancer-associated genes,especially in the context of stem cell lineages.

All strains used were derived from the wild-type (WT) Bristol strain N2. Worm manipulations were performed as previously described(Sulston and Hodgkin,1988).

bro-1 deletion alleles tm1183, tm1229 and tm658were isolated from trimethylpsoralen/ultraviolet (TMP/UV) mutagenesis screens(Gengyo-Ando and Mitani, 2000)by the Japanese deletion consortium (National Bioresource Project for the Experimental Animal `Nematode C. elegans', http://shigen.lab.nig.ac.jp/c.elegans/index.jsp). Deletion strains were backcrossed to WT ten times before analysis.

Worms were mounted for lineage analysis and observed as previously described (Nimmo et al.,2005). Hypodermal nuclei were distinguished from neuronal nuclei on the basis of their appearance: hypodermal nuclei look like `fried eggs'with a large nucleolus, whereas neuronal nuclei are smaller and more granular,with a less distinct nucleolus.

Injections were performed (using 20-100 ng/μl DNA) as previously described (Mello and Fire,1995), using transformation markers rol-6(su1006) or dpy-20⁺ (pMH86). Where appropriate, transgenic arrays were integrated using UV irradiation (Mitani,1995). Integrated strains were backcrossed twice with WT before use.

The bro-1::GFP rescuing reporter construct, pHK196, was made by amplifying a 2.5 kb fragment of cosmid F56A3 (15071-17586 nt) with primers beta-1 (ggaaagGGATCCctcatcgagaaatcagtccaattt cg) and beta-3(gaatctGGTACCcaaatgggaagaccatcgcgtcgaagg) and cloning into the KpnI-BamHI fragment of GFP reporter vector pPD95.79(http://www.addgene.org/Fire_Lab). The bro-1::RFP rescuing construct pHK328 (bro-1::mRFP) was made by substituting monomeric RFP for GFP in pHK196. In addition, a bro-1::DsRed rescuing construct, pAW303 was also made, by the PCR-fusion based method (Hobert,2002), amplifying dsRed from plasmid pHC183 (a kind gift from Neline Kriek) using primers NS45 (atggcctcctccgagaacgtcatc) and NS46(aagggcccgtacggccactagtagg) and bro-1from plasmid pAW272 using primers NS19 (ctcgtaaatcgacacaaatgc) and NS17 (gatgacgttctcggaggaggccat AATG GGAAGACCATCGCG). The sewing reaction was performed using nested primers NS20(tcaaatatgttgcgctgtacg) and NS47 (ggaaacagtt atgtttggtatattggg) and the PCR product cloned into the TOPO-2.1 vector (Invitrogen) to make pAW303. The rnt-1::GFP constructs used in this work, pAW260 and pHK192 have been previously described (Kagoshima et al.,2005; Nimmo et al.,2005). The seam cell-specific marker, SCM::GFP (strain JR667) was used to assay seam cell number (unc-119 (e2498::Tc1III);wIs51[SCM::GFP + unc-119]. A dpy-7::GFP reporter strain(ijIs12) (a kind gift from Iain Johnstone) was used as a hypodermal marker.

PCR of cki-1 for dsRNA synthesis was performed as previously described (Nimmo et al.,2005). dsRNA was synthesised from gel-purified PCR product and injected as previously described (Fire et al., 1998).

Total RNA was extracted from a 100 μl pellet of synchronised L1 larvae using the hot phenol method (Furger et al., 2001). RT-PCR was then performed as previously described(Pocock et al., 2004) using gene-specific primers. rnt-1: RN70 (ctaacgcctgttccagataatac) and RN81(ggagatgataggcatgtagacg). bro-1:RN104 (aaagaacgacaacggaccag) and RN105 (atttcagcatccgtcagtcc). ama-1: RN102 (tgtctcacgcgttcagtttg) and RN103 (aatttccagcactcgaggag).

The Runt domain of RNT-1 and BRO-1 were fused to the C terminus of DsRed1 with linker amino acid sequences (GSTSGSGKPGSGEGSTKPG) in an EBV-based vector,pEB6CAG-MCS (Tanaka et al.,1999). HEp-2 cells in 24-well dishes were transfected with 0.5μg of the plasmid DNAs using TransFectin (Bio-Rad) and cultured in the presence of 1 mg/ml G418 (RUBY-RNT-1RD) or 100 μg/ml Zeocin (RUBY-BRO-1) or both (co-transfection) Cells were used 4 days later for flow cytometry and fluorescent microscopic observation. For the positive and negative control data shown in the supplementary figure (see Fig. S1 in the supplementary material), DNA fragments encoding the Runt domain of Runx2, the mutated version G151R, and CBFβ, were subcloned into pEB6CAG-WR1-SRZ to yield vectors expressing RUBY-wtRunx2-RD, RUBY-G151RRunx2-RD and RUBY-CBFβ.

The wild-type and mutant Runt domains of C. elegans RNT-1 were produced as fusions with N-terminal glutathione S-transferase (GST) and hexahistidine (6xHis) tags. The wild-type Runt domain (amino acids 10-137) was amplified by PCR using the cDNA template yk309f5 and the primers, 5′RD(cgcggtaccCCATGGcatatgagaggatcgcatcaccatcaccatcacggatccatgaccaacgtcttccatcacgttcgg),3′RD (cgcgaattcTGTACAtcattgtggttttggtattctcgcatccc). The mutant Runt domain, containing a missense mutation (I₁₁₂K) corresponding to rnt-1(e1241), was generated by PCR using the primers, 5′RD,3′RD and e1241-5 (cgccgaatgtTtaacgattgtcaaatggaattttcg), e1241-3(gacaatcgtta Aacattcggcgccgatgatggtgg). PCR products were cloned into the NcoI-BsrGI site of the vector pDEST15 (Clontech). BRO-1 was produced as a C-terminal 6xHis-tagged protein. The entire BRO-1 coding sequence was amplified by PCR using the cDNA clone yk211f2 and the primers,N-BRO-1 (cttggtgcatgcCTGCAGacatgaaaagaacgacaacggaccagc) and C-BRO-1(taaaagggatcccCTCGAGaatgggaagaccatcg cgt cgaagg), and was cloned into the PstI-XhoI site of a T7-expression vector derived from pHIT12(Keefe et al., 2001) (H. Tabara, personal communication). All 6xHis-tagged proteins were expressed in the BL21 strain of Escherichia coli.

Since the Runt domain fusion proteins were expressed as inclusion bodies,the harvested cells were incubated in solubilization buffer (0.1 M sodium phosphate, 10 mM Tris, 6 M guanidine-HCl, 1 mM phenylmethylsulfonyl fluoride,10 mM β-mercaptoethanol and 0.1% Nonidet P-40, pH 8.0) and purified on a Ni-NTA resin (Qiagen). Purified proteins were renatured by stepwise dialysis at 4°C against 100 volumes of dialysis buffers (0.1 M sodium phosphate, 10 mM Tris, 50 mM glycine, 2 mM β-ME, 0.1% NP-40 and 20% glycerol, pH 8.0)containing 4 M, 2 M and 0 M guanidine-HCl, respectively, for 4 hours or overnight. BRO-1-6xHis was purified as a soluble protein on the Ni-NTA resin.

For the preparation of the probe and competitors, synthetic oligonucleotides [WT: RUNX-WT-1 (catgactgctAACCGCAgatgac) and RUNX-WT-2(gtacgtcatcTGCGGTTagcagt), and Mut: RUNX-Mut-1(catgactgctAATCGAAgatgac) and RUNX-Mut-2(gtacgtcatcTTCGATTagcagt)] were annealed and extended by a standard Klenow fragment reaction with [α-³²P]dATP, without radioisotope for competitors. The DNA binding reaction (final volume, 10μl) was carried out at 25°C for 30 minutes in EMSA buffer (20 mM HEPES-KOH, 4% Ficoll, 2 mM EDTA, 1 mM DTT, 100 mM KCl, 6% glycerol, 0.2 mg/ml bovine serum albumin, 0.04% Bromphenol Blue and 10 fmol of labelled probe, pH 7.4). The reaction mixture was loaded on a 10% nondenaturing polyacrylamide gel in 0.25×TBE and electrophoresed at 4°C.

The BRO-1 protein sequence is 19% identical (36% similar) to human CBFβ (Fig. 1) (reviewed by Kagoshima et al., 2007). In addition to the low overall homology, it also lacks the conserved region in fly and mammalian CBFβ corresponding to the sequences between the beta sheets, β3 and β6, which might be expected to interfere with the positioning of the Runx interaction domain around β3 and β5(Tahirov et al., 2001)(Fig. 1B). Three deletion alleles of bro-1 are available, tm658, tm1183 and tm1229 (Fig. 1). tm1183 and tm1229 are presumed null alleles and tm1183 has been used throughout this study. Ray and seam cell phenotypes in bro-1(tm1183) animals are shown in Fig. 2. bro-1(tm1183)animals typically lack five rays from each side of the male tail(Fig. 2C,D) and the phenotype is identical in tm1229 (data not shown). Both V- and T-derived rays are variably missing. tm658 has no obvious phenotype (data not shown)but lacks only intronic sequences. The bro-1 ray-loss phenotype is very similar to loss-of-function rnt-1 alleles(Fig. 2B)(Kagoshima et al., 2005; Nimmo et al., 2005) and is completely rescued in transgenic worms carrying a full-length genomic bro-1 construct (pHK193; Fig. 2D). bro-1 animals have fewer seam cells than WT(Fig. 2E). A representative V and T lineage trace of bro-1(tm1183) animals from late L1 up to mid L3 is shown in Fig. 2F,G. Various seam division failures are evident.

In addition, a representative L1 T lineage of bro-1(tm1183)animals is shown in Fig. 2H, in which daughters of the posterior branch T.p often fail to give rise to phasmid neurons. This gives rise to a dye-filling defect (Dyf phenotype) in bro-1 mutant animals (data not shown). Identical T lineage L1 defects in rnt-1 animals have been described previously, and explained as the loss of the characteristic asymmetry of the first T blast cell division(Kagoshima et al., 2005). In WT animals, T divides to give two daughters, T.a and T.p(Fig. 2H). T.a goes on to produce hypodermal daughters whereas T.p normally produces neurons. Thus the two branches of the T lineage are thought of as asymmetric. In rnt-1mutants, it has previously been argued that since the daughters of T.p (the T.px cells) have a hypodermal rather than a neuronal morphology, the asymmetry of the T division is lost (Kagoshima et al., 2005). The same defect was observed in bro-1 mutants(Fig. 2J,L). However, this defect could be a consequence of cell cycle arrest. It has been previously shown that the nuclei of the T.p daughter cells look hypodermal in WT soon after they are born (Fig. 2I)(Sulston and Horvitz, 1977)and we have confirmed their hypodermal characteristics using the hypodermal-specific marker dpy-7::GFP(Fig. 2K). These data suggest that the neuronal fate is normally acquired gradually over time. In WT animals, commitment to subsequent divisions, or at least progress through the cell cycle, may be required to commit to a distinct neuronal fate. In bro-1 (and rnt-1) animals, however, these subsequent divisions do not always occur, and thus hypodermal characteristics are retained. However, in the absence of a suitable neuronal marker, it is difficult to exclude the possibility that RNT-1 and BRO-1 may be required in some other way for the acquisition of neuronal identity, for instance directly, by regulating asymmetry of the T blast cell division, as opposed to indirectly, by programming further cell proliferation.

The phenotype of bro-1 rnt-1 double mutant animals is shown in Fig. 3. These animals have a very similar male tail phenotype to either single mutant(Fig. 3B,E), suggesting that the two genes act in a common pathway. Given their sequence homologies, it is likely that RNT-1 and BRO-1 form a DNA binding complex. However, the L2-L3 V lineage trace shown in Fig. 3Cindicates that the penetrance of division defects is slightly higher in double mutant animals than is the case for either single mutant. In addition, double mutants also have a slightly lower number of seam cells on average. This suggests that there may be aspects of the functions of RNT-1 and BRO-1 that are distinguishable from one another, even though the major functions of RNT-1 and BRO-1 in controlling seam cell lineage development are co-dependent.

Previously, we have shown that seam cell expression of the cyclin-dependent kinase inhibitor cki-1 is up-regulated in rnt-1 mutants, and that cki-1 RNAi partially restores seam cell number in rnt-1mutants (Nimmo et al., 2005). To test whether cki-1 is also a likely downstream target of BRO-1, we performed cki-1 RNAi in a bro-1 mutant background(Fig. 3F). bro-1;cki-1(RNAi) animals were found to have more seam cells than bro-1 mutants alone, with numbers restored almost to WT levels,similar to data previously published for rnt-1. This suggests that RNT-1 and BRO-1 act in a common pathway to repress cki-1 expression in seam cells during key stages of larval development.

The expression pattern of bro-1 in transgenic animals containing a full-length rescuing bro-1::GFP reporter (pHK193) is shown in Fig. 4. Similar expression patterns were observed for a variety of different transgenic alleles, carrying both extrachromosomal and integrated arrays (data not shown). In all larval stages, BRO-1::GFP is expressed in H0-2, V1-6 and T seam progeny. BRO-1::GFP is localised to both the cytoplasm and nucleus, similar to the expression pattern reported for CBFβ in mammalian cells(Tanaka et al., 1997). It seems probable, therefore, that the sub-cellular localisation of CBFβ is conserved. We cannot exclude the possibility, however, that cytoplasmic localisation results from the presence of a GFP fusion, or from overexpression. Faint expression is also observed in hypodermal nuclei, some of which are embryonically derived and which are not therefore simply anterior daughters of the L1 stem cell division containing a perdurance of GFP expression. We also observed expression of BRO-1::GFP in the uterine seam(utse) during late L4 in hermaphrodites(Fig. 4B). This expression is likely to be due to diffusion of cytoplasmic BRO-1::GFP as a result of fusion of the utse with the seam in L4, but also raises the possibility that BRO-1 has a functional role in this tissue.

Intriguingly, we noticed that bro-1::GFP animals contain extra seam cells. For example, an L3 bro-1::GFP-expressing animal is shown in which V1-derived seam cells have over-proliferated(Fig. 4C). In this animal, the anterior progeny of the V1.pa and V1.pp L2 stem cell divisions have not fused with the hypodermal syncytium but instead have remained as seam and undergone further division in L3, giving rise to eight nuclei instead of the expected four.

The expression pattern of BRO-1 in seam cells is very similar to that of RNT-1 (except that RNT-1::GFP is always nuclear), and co-localisation is shown using rescuing bro-1::DsRed (pAW303) and rnt-1::GFP (pAW260)constructs (Fig. 4D). Co-localisation is consistent with the hypothesis that BRO-1 acts together with RNT-1 to control seam cell divisions. Faint BRO-1::RFP and RNT-1::GFP co-localisation is also observed in certain body wall muscle cells(Fig. 4E). Muscle expression of RNT-1 has been previously reported although no functional role has been described (Kagoshima et al.,2005; Nimmo et al.,2005). Finally, we also observed BRO-1::RFP, but not RNT-1::GFP,in certain pharyngeal neurons. In short, BRO-1 and RNT-1 are co-expressed in seam and muscle cells, and BRO-1 is additionally expressed in hypodermal nuclei, certain pharyngeal neurons and the utse.

Closer examination of bro-1 mutant hermaphrodites revealed rupturing at the vulva and eversion of the gonad at the L4 molt in around 20%of hermaphrodites (Fig. 4G,H),suggesting a possible defect in either the uterine-vulval or uterine-hypodermal connection. The penetrance of this phenotype is almost doubled in bro-1(tm1183) rnt-1(tm388) double mutants, but is observed only occasionally in rnt-1 single mutants(Fig. 4H). The origin of the vulval defect is not clear at present. We could see no obvious defects in bro-1rnt-1 double mutants using a variety of markers, including egl-13::GFP and cdh-3::GFP reporters to observe the arrangement of the utse, and utse-seam cell fusion, respectively(Cinar et al., 2003; Pettitt et al., 1996) (data not shown). It is possible that the reduced number of nuclei in the seam and/or hypodermal syncytium in the vicinity of the vulva in bro-1 and bro-1 rnt-1 mutants, somehow acts to weaken the uterine-hypodermal connection. However, the observed expression of BRO-1 in the utse together with the much higher penetrance of vulval and/or uterine defects in bro-1 compared with rnt-1 animals, suggests that BRO-1 may have a specific role in vulval and/or uterine morphogenesis that is, at least in part, independent of RNT-1.

We used electophoretic mobility shift assays (EMSAs) to investigate BRO-1-RNT-1 interaction. The GST-tagged Runt domain of RNT-1 (GST-RD) binds weakly to the consensus Runx DNA binding site AACCGCA(Fig. 5A). However, this interaction lacks specificity, as a mutated competitor probe reduces the band shift as effectively as WT competitor (Fig. 5A). The addition of BRO-1 protein stimulates both the affinity of RNT-1 for the DNA binding consensus, and the specificity of the interaction,as mutated competitor probe fails to reduce the RNT-1-DNA shift in the presence of BRO-1 (Fig. 5A). Next we tested the rnt-1(e1241) mutant protein RNT-1 Runt domain I₁₁₂K (substitution of Lys for Ile₁₁₂). The I₁₁₂ residue of RNT-1 is equivalent to V₁₅₂ of murine Runx1, which is located within the exposed β sheet region of Runx1, known as β10, situated at the Runx1-CBFβ heterodimerisation interface(Tahirov et al., 2001). However, alanine scanning mutagenesis experiments have suggested a role for V₁₅₂ in DNA binding, rather than CBFβ interaction(Li et al., 2003). By contrast, the data in Fig. 5Bshows that RNT-1 I₁₁₂K fusion protein (GST-e1241) retains weak DNA binding activity but does not supershift upon addition of BRO-1, indicating that I₁₁₂ is required for heterodimer formation and therefore high affinity DNA binding.

To confirm that BRO-1 and RNT-1 interact directly, we used the RUBY system. This assay is based on the fact that monomeric DsRed1 is subject to conditional proteolysis and rapid degradation through the ubiquitin-proteasome pathway, but escapes degradation in the associated, dimeric state. Interaction of the proteins of interest brings the linked DsRed1 monomers into close proximity and enables their dimerisation and maturation (J. Tanaka and Y.M.,unpublished). The data in Fig. 5C show fluorescence output from cells expressing either RUBY-BRO-1 or RUBY-RNT-1RD fusions (RD: Runt domain of RNT-1), and from cells co-expressing both fusions. A strong fluorescence signal is only visible when BRO-1 and RNT-1 are co-expressed, indicating that the two proteins form a heterodimer. Neither protein appears to homodimerise. Positive and negative controls for this technique, using Runx2 and CBFβ RUBY fusions, can be seen in Fig. S1 (see Fig. S1 in the supplementary material).

In order to test for possible regulatory interactions between RNT-1 and BRO-1, we examined BRO-1 expression in rnt-1 mutants, and vice versa. The absence of RNT-1 had no effect on either the pattern or level of BRO-1::GFP expression (data not shown). In the converse experiment, however,we observed upregulation of RNT-1::GFP in bro-1 mutant L1 larvae in both seam and muscle cells (Fig. 6B). This result was surprising as in mammalian cells CBFβacts in the opposite way, stabilising Runx proteins by preventing their ubiquitination (Huang et al.,2001). To investigate whether this negative regulation of RNT-1 by BRO-1 was occurring at the level of transcription we analysed animals by RT-PCR and found a corresponding increase in rnt-1 mRNA in bro-1 mutant larvae compared with WT(Fig. 6C). Therefore BRO-1 acts, either directly or indirectly, as a transcriptional repressor of rnt-1.

Previously we reported that overexpression of rnt-1 cDNA from the heat shock promoter at particular stages of development causes an increase in the number of seam cells (Nimmo et al.,2005). We have repeated these experiments using constitutive overexpression of rnt-1 from the integrated rnt-1::GFPrescuing transgenic strain msIs114 and find that this strain also contains extra seam cells (Fig. 7A) due to overexpression of rnt-1 from multiple copies of the gene in transgenic arrays (data not shown). Extra seam cells were observable in a variety of different rnt-1 transgenic strains,containing independent arrays and constructs (data not shown). We therefore also tested constitutive bro-1 overexpression using the integrated rescuing bro-1::GFP strain msIs344 (expressing high levels of bro-1 mRNA as shown in Fig. 7C) and found that this too causes seam cell hyperplasia(Fig. 7B, also discussed earlier, see Fig. 4C). Likewise, extra seam cells could be seen in a variety of different bro-1 multicopy transgenics (data not shown).

We found that overexpression of RNT-1 in bro-1 mutants does not compensate for the reduction in seam cell number(Fig. 7A). In fact, the number of seam cells is slightly lower in rnt-1::GFP bro-1(tm1183) animals compared with bro-1 mutant animals alone. Perhaps RNT-1 has a dominant-negative effect in the absence of BRO-1, possibly by binding to non-specific sites in DNA (as suggested by our EMSA data) and recruiting, and therefore titrating, other transcriptional co-factors away from RNT-1 target sites.

However, the converse experiment, in which BRO-1 was overexpressed in a rnt-1 mutant background (the tm388 rnt-1 allele used is a presumed null), did give rise to animals with extra seam cells(Fig. 7B), suggesting that BRO-1 can function to promote seam cell divisions independently of RNT-1 activity. This, taken together with data showing vulval roles for BRO-1 that are, at least in part, independent of RNT-1, is strongly suggestive of the RNT-1-independent nature of various aspects of BRO-1 function. RNT-1-independent roles for BRO-1 are unprecedented as previous data from other systems always considers Runx/CBFβ solely as a co-dependent DNA binding complex.

Seam cell hyperplasia could be caused either by an increased number of seam cell divisions (increased self-renewal and proliferation) or by changes in fate resulting from a loss of asymmetry. The lineage data shown in Fig. 7C indicate that transgenic animals overexpressing bro-1 do indeed go through extra,unscheduled seam cell divisions but in addition, some of the L2 and L3 asymmetrical stem cell divisions are now transformed to symmetrical divisions resulting in increased self-renewal. The anterior progeny of these stem cell divisions are prevented from acquiring the hypodermal fate and fail to fuse with the hypodermal syncytium. This is a fate change, although in this case the `fate' is to retain the seam characteristic of continued proliferation,giving rise to an extra stem cell division and thus an extra seam cell. In this way BRO-1 (acting, at least partly, together with RNT-1) functions to promote the stem-like (self-renewal) characteristics of seam cell divisions at the expense of the acquisition of the differentiated, hypodermal fate.

We also counted seam cell number in a strain containing both rnt-1and bro-1 integrated arrays. The co-overexpression of RNT-1 and BRO-1 produces massive seam cell hyperplasia, 49 cells on average per side (and up to 70) instead of the 16 expected in WT(Fig. 8A). This extreme seam cell hyperplasia causes lateral expansion of the seam(Fig. 8C,D) and animals containing these seam cell `tumours' are fatter (although shorter) than WT(Fig. 8E,F).

Despite the relatively low level of sequence homology between BRO-1 and CBFβ, we have presented strong genetic and biochemical evidence that BRO-1 is a binding partner for RNT-1, and acts in concert with RNT-1 to regulate seam cell number. Our lineage analysis demonstrates that the ray loss phenotype seen in bro-1 alleles is caused by failed cell divisions in seam lineages, and is very similar to that reported previously for rnt-1 mutants (Nimmo et al.,2005). bro-1 rnt-1 double mutants have similar rates of ray loss to either single mutant, suggesting that a major role of these two genes is to act in a common pathway to promote seam cell proliferation and/or self-renewal, with cki-1 as a probable direct or indirect downstream target.

BRO-1 and RNT-1 are both expressed in seam cells, consistent with the hypothesis that they act in concert to regulate seam cell divisions. The RUBY assay indicates a direct interaction between BRO-1 and RNT-1 and EMSA studies demonstrate that BRO-1 is required for robust binding of RNT-1 to the Runx consensus DNA binding site, consistent with other studies of CBFβfunction (Bushweller, 2000; Li et al., 2003; Nagata and Werner, 2001; Tahirov et al., 2001; Yan et al., 2004). Furthermore, our experiments show that BRO-1 dramatically increases the specificity of RNT-1-DNA binding, thereby extending previous studies of CBFβ/Runx interactions. It will be of interest to investigate whether this finding can be extended to other systems.

BRO-1, as well as RNT-1, promotes extra seam cell divisions when overexpressed, acting to promote self-renewal. Intriguingly, BRO-1 is capable of promoting extra divisions in the absence of RNT-1, thus suggesting that BRO-1 has Runx-independent functionality, an observation confirmed by the enhanced vulval rupture of bro-1 mutants compared to rnt-1mutants. The converse is not true: RNT-1 is only capable of promoting extra divisions in the presence of functional BRO-1, perhaps because of the importance of BRO-1 in influencing the specificity and robustness of RNT-1-DNA interactions. Likewise, the late larval lethality of the rnt-1::GFPtransgene in a bro-1 mutant background (see legend to Fig. 6) may be related to the role we have demonstrated for BRO-1 in increasing the specificity of RNT-1-DNA interactions. In the absence of BRO-1, RNT-1 (already overproduced in a transgenic animal) may bind promiscuously to ectopic sites, thus misregulating gene expression and impairing transcriptional networks.

Our finding that BRO-1 does not appear to be wholly reliant on RNT-1 in order to promote seam cell proliferation is the first indication that CBFβ proteins in general may have Runx-independent functions. Perhaps RNT-1 and BRO-1 regulate transcription as part of the core of a large enhanceosome or repressosome complex, as proposed for the mammalian Runx factors (Carey, 1998), and BRO-1 is able to interact, albeit less efficiently, with this complex even in the absence of RNT-1. Alternatively it may have some other intrinsic activity associated with the control of seam cell proliferation and/or self-renewal.

Co-overexpression of RNT-1 and BRO-1 causes massive hyperproliferation of seam cells, supporting the view that the major function of RNT-1 and BRO-1 is to co-operate as a complex in the transcriptional regulation of genes required to control seam cell number. The resulting hyperplasia distorts the morphology and integrity of the seam, causing it to expand dorsoventrally and invade the surrounding syncytial hypodermis, resulting in fatter but shorter (Dumpy)animals. C. elegans do not exhibit solid somatic tumours as such. The germline is the one other well-established `tumour' system, where failure of mitotic germ nuclei to enter meiosis, caused for example by gain of glp-1 function (Berry et al.,1997; Crittenden et al.,2003), can result in massive mitotic over-proliferation and a concomitant expansion of the gonad. We hypothesise that because of the restriction on body volume exerted by the cuticle, expansion in one direction probably leads to contraction in the other, hence, worms with much increased numbers of seam nuclei resulting form the overexpression of both bro-1 and rnt-1 are shorter and fatter because of lateral expansion, whereas worms with increased syncytial hypodermal ploidy are known to be longer than WT (Flemming et al.,2000), i.e. to expand longitudinally. Perhaps this extreme hyperplasia is as close to a somatic `tumour' that this model organism can get.

One way in which the levels of these genes may be controlled is via negative feedback, and indeed we have discovered that loss of bro-1is associated with increased expression of rnt-1. BRO-1 may either be acting in a RNT-1-independent manner to repress rnt-1 expression, or as part of a rnt-1 autoregulatory feedback loop. It is presumably necessary to limit the amount of functional Runx/CBFβ complexes in order to program the correct cell division pattern of the seam lineage.

There are numerous reports in the literature of Runx genes and CBFβ acting either as proto-oncogenes or tumour suppressors, the nature of which is highly context dependent(Blyth et al., 2005; Cameron and Neil, 2004). It is not clear whether such roles in carcinogenesis arise as a result of defects in cellular differentiation or cell proliferation. The analysis of cell proliferation at the single cell level, which is the hallmark of C. elegans studies, leads us to conclude that BRO-1 and RNT-1 act to promote the self-renewal characteristics of seam stem cell divisions at the expense of the differentiated, hypodermal fate. Thus, our studies would support the view that Runx/CBFβ factors have oncogenic potential.

Rationalising the, sometimes contradictory, effects of Runx/CBFβfactors on cellular development, especially in carcinogenesis, is one of the toughest challenges in understanding the function of these important genes. We have now firmly established C. elegans as a prominent model organism for the study of Runx/CBFβ function, as we can interpret phenotypes in individual cells rather than being forced to rely on analyses of cell populations and tissues, and can work without the interpretative difficulties caused by genetic redundancy that are inherent in other experimental systems where there are multiple Runx genes. One of the interesting ideas to emerge from these studies is that it may not be appropriate to regard BRO-1 simply as a partner subunit for high affinity RNT-1-DNA binding. Although a major role of BRO-1 is to cooperate with RNT-1 to promote seam cell proliferation and/or self-renewal (presumably by stabilising RNT-1-DNA interactions) BRO-1 also increases the specificity of RNT-1-DNA interactions, has a role in the transcriptional regulation of rnt-1, and furthermore appears to have functions in the worm that are, at least in part, independent of RNT-1. It will be of interest to see whether our findings can be extended to other systems.

We thank J. Kajiwara for excellent technical help, K. Shigesada for critical reading of the manuscript, A. Fire, R. Tsein and H. Tabara for vectors and plasmids, and members of the Kohara and Woollard laboratories for comments and discussion on the manuscript. We would also like to thank the CGC and I. Johnstone for strains. Work in Y.K.'s laboratory was supported by a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science and Technology of Japan. Work in A.W.'s laboratory was funded by the UK Medical Research Council, the Association for International Cancer Research and Cancer Research-UK. Work in Y.M.'s laboratory was supported by Grant-in-Aid for Scientific Research on priority areas `System Genomics (18016002)', `Nuclear Dynamics (17050004)' and`G-protein signal (18057003)'.