Development 128, 1299-1312 (2001) 1299 Printed in Great Britain © The Company of Biologists Limited 2001 DEV2653

Meis3 synergizes with Pbx4 and Hoxb1b in promoting hindbrain fates in the zebrafish

Nikolaos Vlachakis, Seong-Kyu Choe and Charles G. Sagerström* Department of Biochemistry and Molecular Pharmacology, and Program in Neuroscience, University of Massachusetts Medical School, Worcester, 55 Lake Avenue North, MA 01655, USA *Author for correspondence (e-mail: [email protected])

Accepted 16 January; published on WWW 22 March 2001

SUMMARY

Many Hox are thought to require Pbx and Meis neurons. This synergistic effect requires that Hoxb1b and co-factors to specify cell identity during embryogenesis. Meis3 have intact Pbx-interaction domains, suggesting that Here we demonstrate that Meis3 synergizes with Pbx4 and their in vivo activity is dependent on binding to Pbx4. In Hoxb1b in promoting hindbrain fates in the zebrafish. We the case of Meis3, binding to Pbx4 is also required for find that Hoxb1b and Pbx4 act together to induce ectopic nuclear access. Our results are consistent with Hoxb1b hoxb1a expression in rhombomere 2 of the hindbrain. In and Meis3 interacting with Pbx4 to form complexes contrast, Hoxb1b and Pbx4 acting together with Meis3 that regulate hindbrain development during zebrafish induce hoxb1a, hoxb2, krox20 and valentino expression embryogenesis. rostrally and cause extensive transformation of forebrain and midbrain fates to hindbrain fates, including Key words: Hindbrain, Mauthner neuron, Homeosis, Embryonic differentiation of excess rhombomere 4-specific Mauthner axis, Hoxb1b, Meis3, Pbx4, Zebrafish

INTRODUCTION DNA-binding (reviewed by Mann and Affolter, 1998). An in vivo role for such complexes is suggested by the effect of Vertebrate hox , like their Drosophila counterparts the dominant negative forms of Hth in Drosophila (Jaw et al., HOM-C genes, play essential roles during embryogenesis. For 2000; Ryoo et al., 1999), by the finding that dimers (e.g. Chang instance, their expression in overlapping domains along the et al., 1997; Knoepfler et al., 1997) and trimers (Ferretti et al., anteroposterior (AP) axis provides a ‘hox code’ that specifies 2000; Shen et al., 1999) can be reconstituted in cell extracts AP positional identity, and changes in hox expression lead and by the observation that Meis, Pbx and Hox binding sites to homeotic transformations in the AP axis, wherein anterior are present in several Hox-dependent promoters (Ferretti et al., structures acquire the character of more posterior structures 2000; Jacobs et al., 1999; Pöpperl et al., 1995; Ryoo et al., (reviewed by Krumlauf, 1994). 1999). Genetic analyses in Drosophila revealed that many HOM-C pbx4, meis3 and hoxb1b are co-expressed in the caudal genes require the extradenticle (exd) and homothorax (hth) hindbrain primordium of the zebrafish embryo and Pbx4, genes for proper function (Rauskolb et al., 1993; Rieckhof et Meis3 and Hoxb1b form complexes in vitro (Vlachakis et al., al., 1997). Homologs of exd and hth are encoded by the 2000). Here we explore the role of these proteins during vertebrate pbx (Kamps et al., 1990; Monica et al., 1991; Nourse zebrafish development and test whether they need to interact et al., 1990; Vlachakis et al., 2000) and meis/prep (Berthelsen to function in vivo. We find that Hoxb1b and Pbx4 act together et al., 1998; Moskow et al., 1995; Nakamura et al., 1996) gene to induce ectopic hoxb1a expression in rhombomere (r) 2. In families, respectively. Recently, the zebrafish lazarus mutation, marked contrast, Hoxb1b and Pbx4 together with Meis3 induce which disturbs segmental patterning in hindbrain and trunk, massive rostral expression of several hindbrain genes (hoxb1a, was cloned (Popperl et al., 2000) and found to encode the hoxb2, krox20 and valentino) and cause anterior truncations, previously reported pbx4 gene (Vlachakis et al., 2000), apparently due to the transformation of rostral (forebrain and suggesting a role for pbx genes also in vertebrate development. midbrain) fates to caudal (hindbrain) fates. This transformation Vertebrate meis genes also likely play a role, as misexpression is extensive enough that we observe excess Mauthner neurons of Xenopus Meis3 leads to abnormalities of the AP axis (normally found in r4) anteriorly. These effects are dependent (Salzberg et al., 1999). on Meis3 and Hoxb1b having intact Pbx interaction domains, Pbx and Meis form dimeric and trimeric complexes with suggesting that they interact with Pbx4 in vivo. Our results also Hox proteins in vitro and these complexes are thought to indicate that Meis3 must interact with Pbx4 to access the modulate Hox activity, primarily by conferring high-specificity nucleus. Since meis3, pbx4 and hoxb1b are co-expressed in the 1300 N. Vlachakis, S.-K. Choe and C. G. Sagerström zebrafish hindbrain primordium during embryogenesis, our HAHoxb1b constructs) or 3 embryos (for MYCMeis3 constructs) results are consistent with complexes containing combinations were run per lane. of Pbx4, Hoxb1b and Meis3 regulating normal hindbrain development. In situ hybridization, immunostaining and immunoprecipitations In situ hybridizations and immunoprecipitations have been described previously (Vlachakis et al., 2000). Immunostaining with 3A10, anti- MATERIALS AND METHODS c- (clone 9E10) and anti-HA (clone 12CA5) was done as described by Hatta (Hatta, 1992). HRP was detected using the TSA- Cloning direct kit (Dupont Biotechnology Systems). Photographs were taken All genes used were derived from zebrafish, all expression constructs with a Leica confocal or an Olympus inverted microscope. Rabbit were in the pCS2+ vector and all constructs were verified by polyclonal anti-Pbx4 antiserum was raised to a peptide containing the sequencing. Meis3, Pbx4, HA-Hoxb1b and MutMeis3 (carries two 13 C-terminal residues of Pbx4 and used at 1:1000 for western blots. point mutations in the homeodomain, Q44→E and N51→A) have been described (Vlachakis et al., 2000, Sagerström et al, 2001). In Fate mapping δhoxb1b, the N-terminal 146 amino acids (aa) were deleted by Embryos were injected with hoxb1b+pbx4+meis3 mRNAs at the 1- digesting pCS2+HAHoxb1b with SmaI/PstI and inserting to 2-cell stage. At early gastrula stage (6.5 hpf; hours postfertilization) oligonucleotide 5′-TTCCCGGGGTAGGCTGCA-3′. In BMNMeis3 animal pole cells of control and injected embryos were labeled with the N-terminal 171 aa were replaced with the FRB (FKBP 12- 1,1′-diooctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate rapamycin binding) domain from FRAP (FKBP 12-rapamycin (DiI) and fate mapping performed as described (Fekany-Lee et al., associated ) (Chen et al., 1995). Primers 5′- 2000) except embryos were fixed at 4- to 5-somite stages. TATATCTAGACTGCTTTGAGATTCGTCGGA-3′ and 5′-GGATG- AATTCATGGACTATAAAGATGACGA-3′ amplified the FLAG- FRB domain which was subcloned via EcoRI and XbaI sites in RESULTS the primers into pCS2+ (pCS2+FLAG-FRB). Primers 5′-GGTCT- AGACGAGAGGGTGGATCTAAATCTGAC-3′ and 5′-GGTCT- Hoxb1b requires an intact Pbx-interaction domain to AGATCAGTGGGCATGTATGTCAAG-3′ amplified aa 172-415 of induce hoxb1a expression in rhombomere 2 the Meis3 ORF from pCS2+meis3. This was subcloned via XbaI Misexpression of Hoxa1 in the mouse leads to ectopic Hoxb1 sites in the primers into pCS2+FLAG-FRB downstream of FRB. MYCFRBMeis3 was generated by subcloning an XhoI/NotI fragment expression in r2 of the hindbrain (Zhang et al., 1994). To test from pCS2+FLAG-FRBMeis3 into pCS2+MT cut with XhoI/NotI. if this is the case also for zebrafish, we expressed Hoxb1b (the For Meis3-VP16 primers 5′-AAAGATATCCCCACCGTAC- zebrafish counterpart to murine Hoxa1; Alexandre et al., 1996; TCGTCAATTCC-3′ and 5′-AAAGATATCTCGACGGCCCCCCC- Amores et al., 1998) ectopically by mRNA microinjection (see GACCGATGTCAGC-3′ amplified the VP16 domain from Materials and Methods). Control injections with lacZ mRNA pCS2+vp16N (Kessler, 1997). This was subcloned via EcoRV sites resulted in normal embryos (Fig. 1Aa,b; Table 1) while in the primers into the SmaI site (1092 bp in Meis3 ORF) of expression of Hoxb1b (Fig. 1Ac,d) resulted in ectopic pCS2+meis3. All point mutations were generated with the expression of hoxb1a (the zebrafish counterpart to murine QuikChange kit from Stratagene: BMHoxb1b (has a substitution in → ′ Hoxb1, normally expressed in r4; Amores et al., 1998; Prince the pentapeptide FDWMK, W186 F) was generated using primer 5 - et al., 1998) in r2 (52%; arrow in Fig. 1Ac), and largely normal GGGGGATTCCTCTTGACTTTCATAAAGTCAAAGGTTGGCGC- 3′, BMM2Meis3 (has two substitutions in the M2 motif, L141→A and expression of hoxb2 (Fig. 1Ad; normally expressed in r3-r5; E142→A) using primer 5′-CGGTTTCATCTATTAGAAGCAGC- Prince et al., 1998). This is consistent with a report AAAGGTTCATGACCTCTGTGATAATTTCTGCC-3′, BMwM2Meis3 demonstrating an ectopic pair of Mauthner neurons (normally (has five substitutions in the M2 motif, I131→A, L134→A, L138→A, found in r4) in r2 following hoxb1b misexpression in zebrafish L141→A and E142→A) using primer 5′-CTGATGATCCAGGCC- (Alexandre et al., 1996). GCTCAAGTTGCACGGTTTCATGCATTAGAAGCAGC−3′ with Adjacent Pbx and Hox binding sites are present in the BMM2Meis3 as a template and BMM1/2Meis3 (has four substitutions murine Hoxb1 enhancer and both sites are required for in the M1 motif, aa 64-67 KCEL→NNSQ and two substitutions expression in a transgenic model (Pöpperl et al., 1995), → → ′ in the M2 motif, L141 A and E142 A) using primer 5 - suggesting that Pbx and Hox proteins might interact to induce GGCTCTGGTATTTGAAAACAATTCACAGCCACTTGCTCACC- murine Hoxb1. Thus, to induce ectopic hoxb1a in zebrafish, 3′ with BMM2Meis3 as a template. NLS BMM1/2Meis3 was generated by cloning oligonucleotide 5′-GATCCCCCGGGATGGCTCC- Hoxb1b might interact with an endogenous Pbx protein. The AAAGAAGAAGCGTAAGGTAAA-3′ into BamHI/ClaI digested most likely candidate is Pbx4, which is expressed broadly in pCS2+MT BMM1/2Meis3. the zebrafish embryo (Vlachakis et al., 2000) and is the predominant Pbx protein at this stage (Popperl et al., 2000). RNA microinjections Expressing Pbx4 together with Hoxb1b did not have a mRNAs were synthesized from NotI-linearized pCS2+ derived significantly different effect than Hoxb1b alone (Fig. 1Ae,f; plasmids, or (for lacZ) from XhoI-linearized pSP6-nucβ-gal, using the Table 1) and Pbx4 alone had no effect (not shown). To test if SP6 mMessage mMachine kit (Ambion) and purified with the RNeasy Hoxb1b interacts with Pbx4, we generated BMHoxb1b, a mini kit (Qiagen). For in situ analysis, 165 pg of each test mRNA was mutant form that is unable to bind Pbx4 (compare lanes 2 and injected and the total amount adjusted to 500 pg by co-injecting lacZ 5 in Fig. 2A), by introducing a single amino acid substitution mRNA. lacZ control injections were done with 500 pg mRNA. For → western analysis and immunostaining, 300 pg mRNA was injected for (W186 F) into the pentapeptide of Hoxb1b (see Materials Meis3 constructs and 450 pg for Hoxb1b constructs. Injections were and Methods). Analogous mutations abolish Pbx binding of at the 1- to 2-cell stage. β-gal staining was performed as described other Hox proteins without altering their DNA binding previously (Blader et al., 1997), except embryos were fixed for 40 (e.g. Knoepfler and Kamps, 1995; Rambaldi et al., 1994). minutes. For western analysis, lysates of 10 embryos (for Meis3 and BMHoxb1b is expressed at levels comparable to wild-type Homeodomain protein interactions in the CNS 1301

Table 1. Meis3 and Hoxb1b require intact Pbx-interaction domains for in vivo function Outcome Ectopic staining/ In situ Injected RNA* Unaffected Ectopic staining‡ truncated axis§ probe lacZ 78/80 (97%) 2/80 (3%) 0/80 (0%) hoxb1a 231/231 (100%) 0/231 (0%) 0/231 (0%) hoxb2 88/89 (99%) 1/89 (1%) 0/89 (0%) krox20 hoxb1b 40/83 (48%) 43/83 (52%) 0/83 (0%) hoxb1a 60/61 (98%) 1/61(2%) 0/61 (0%) hoxb2 hoxb1b/pbx4 135/209 (65%) 74/209 (35%) 0/209 (0%) hoxb1a 219/235 (93%) 15/235 (6%) 1/235 (1%) hoxb2 153/168 (91%) 15/168 (9%) 0/168 (0%) krox20 bmhoxb1b 92/92 (100%) 0/92 (0%) 0/92 (0%) hoxb1a 88/88 (100%) 0/88 (0%) 0/88 (0%) hoxb2 meis3/pbx4 200/221 (91%) 21/221 (9%) 0/221 (0%) hoxb1a 185/197 (94%) 10/197 (5%) 1/197 (1%) hoxb2 90/101 (90%) 11/101 (11%) 0/101 (0%) krox20 hoxb1b/meis3 38/82 (46%) 37/82 (45%) 7/82 (9%) hoxb1a 195/355 (55%) 110/355 (31%) 50/355 (14%) hoxb2 hoxb1b/pbx4/meis3 41/144 (28%) 47/144 (33%) 56/144 (39%) hoxb1a 227/698 (33%) 197/698 (28%) 274/698 (39%) hoxb2 105/358 (29%) 163/358 (46%) 90/358 (25%) krox20 54/150 (36%) 70/150 (47%) 26/150 (17%) valentino bmhoxb1b/pbx4/meis3 68/78 (87%) 10/78 (13%) 0/78 (0%) hoxb1a 63/68 (93%) 5/68 (7%) 0/68 (0%) hoxb2 135/150 (90%) 15/150 (10%) 0/150 (0%) krox20 bmM2meis3 87/87 (100%) 0/87 (0%) 0/87 (0%) hoxb1a 78/78 (100%) 0/78 (0%) 0/78 (0%) hoxb2 hoxb1b/pbx4/bmM2meis3 26/78 (33%) 38/78 (49%) 14/78 (18%) hoxb1a 53/128 (42%) 54/128 (42%) 21/128 (16%) hoxb2 bmwM2meis3 86/86 (100%) 0/86 (0/%) 0/86 (0/%) hoxb1a 85/85 (100%) 0/85 (0%) 0/85 (0%) hoxb2 hoxb1b/pbx4/bmwM2meis3 80/150 (53%) 61/150 (41%) 9/150 (6%) hoxb1a 86/107 (80%) 13/107 (12%) 8/107 (8%) hoxb2 bmM1/2meis3 138/138 (100%) 0/138 (0%) 0/138 (0%) hoxb1a 98/98 (100%) 0/98 (0%) 0/98 (0%) hoxb2 hoxb1b/pbx4/bmM1/2meis3 62/90 (69%) 28/90 (31%) 0/90 (0%) hoxb1a 127/133 (95%) 6/133 (5%) 0/133 (0%) hoxb2 bmNmeis3 57/61 (94%) 4/61 (6%) 0/61 (0%) hoxb1a 72/72 (100%) 0/72 (0%) 0/72 (0%) hoxb2 hoxb1b/pbx4/bmNmeis3 109/125 (87%) 16/125 (13%) 0/125 (0%) hoxb1a 121/133 (91%) 11/133 (8%) 1/133 (1%) hoxb2 106/115 (92%) 9/115 (8%) 0/115 (0%) krox20 nlsbmM1/2meis3 94/94 (100%) 0/94 (0%) 0/94 (0%) hoxb1a 91/91 (100%) 0/91 (0%) 0/91 (0%) hoxb2 hoxb1b/pbx4/nlsbmM1/2meis3 103/144 (72%) 41/144 (28%) 0/144 (0%) hoxb1a 127/128 (99%) 1/128 (1%) 0/128 (0%) hoxb2

*1- to 2-cell stage embryos were injected with mRNAs as listed, fixed at approx. 5- (for hoxb2) or 10- (for hoxb1a, valentino and krox20) somite stages and analyzed by whole-mount in situ hybridization. ‡Embryos with normal morphology but ectopic or distorted expression of marker genes. §Embryos with anterior truncations and ectopic expression of marker genes.

Hoxb1b following microinjection, as assayed by western Meis3+Pbx4 had minimal effect (~90% normal; not shown blotting (compare lanes 2 and 3 in Fig. 2B) and and Fig. 1Bc,d; Table 1). In contrast, Hoxb1b+Meis3 and immunohistochemistry (compare a and b in Fig. 2G). In Hoxb1b+Pbx4+Meis3 resulted in massive ectopic expression contrast to wild-type Hoxb1b, BMHoxb1b expression led of hoxb1a (Fig. 1Be,g) and hoxb2 (Fig. 1Bf,h), anterior to to essentially normal embryos (Fig. 1Ag,h; Table 1). This their normal expression domains (asterisks in Fig. 1Be-h). demonstrates that Hoxb1b requires an intact Pbx-interaction These phenotypes are distinct from those obtained with domain, suggesting that Hoxb1b and Pbx4 interact to activate Hoxb1b alone or Hoxb1b+Pbx4 (Fig. 1Ac-f) and we have hoxb1a expression in zebrafish r2. Since the Hoxb1 regulatory classified them into two groups (Table 1). The least affected elements have been conserved from mouse to pufferfish embryos exhibit ectopic expression of hoxb1a and hoxb2 (Pöpperl et al., 1995) such an interaction might be a general (approx. 30%; Table 1; example in Fig. 1Be) and the most requirement for Hoxb1 induction. severely affected embryos exhibit ectopic expression together with an anterior truncation (e.g. Fig. 1Bg; approx. 10% for Meis3 synergizes with Hoxb1b and Pbx4 to induce Hoxb1b+Meis3 and approx. 39% for Hoxb1b+Pbx4+Meis3; expression of both hoxb1a and hoxb2 Table 1). These data demonstrate that Meis3 synergizes with To test the role of Meis3, we co-expressed it with various Pbx4 and Hoxb1b and that endogenous Pbx4 may be limiting combinations of Pbx4 and Hoxb1b. Meis3 alone and for this effect. 1302 N. Vlachakis, S.-K. Choe and C. G. Sagerström

Fig. 1. Hoxb1b and Meis3 require intact Pbx interaction domains to mediate ectopic expression. (A) Embryos were injected with mRNAs as indicated to the left of each pair of panels and analyzed by in situ hybridization for expression of hoxb1a (left hand panels) or hoxb2 (right hand panels). Arrows in c and e indicate ectopic hoxb1a expression in r2. (B) Embryos were injected with mRNAs as indicated on the left, and scored for expression of hoxb1a (left hand panels) or hoxb2 (right hand panels). Asterisks indicate ectopic hoxb1a (e and g) and hoxb2 (f and h). (C) Embryos were injected with mRNAs as indicated to the left and scored for expression of hoxb1a (left hand panels) or hoxb2 (right hand panels). Asterisk in c and arrows in i, m, q and y indicate ectopic hoxb1a, while asterisk in d and arrows in j and n indicate ectopic hoxb2. Embryos are at the 5- (for hoxb2) or 10- (for hoxb1a) somite stage and are shown in dorsal views with anterior to the left.

Meis3 and Hoxb1b require intact Pbx-interaction Methods) by mutating two Meis N-terminal domains (M1 and domains to mediate the synergistic effect M2) thought to mediate Pbx binding (reviewed by Mann and Meis3 and Hoxb1b cannot interact with each other, but both Affolter, 1998). Since Meis-Pbx binding is not completely bind Pbx4 (Vlachakis et al., 2000), raising the possibility that characterized and mutating M1 or M2 alone may not eliminate they interact with Pbx4 in vivo. In contrast to co-expression of all Pbx-binding (Jaw et al., 2000; Knoepfler et al., 1997), the three wild-type proteins (~70% affected embryos, Fig. we generated several constructs based on a previous report 1Cc,d; Table 1), expression of BMHoxb1b along with Pbx4 (Knoepfler et al., 1997). BMM2Meis3 carries two amino acid and Meis3 led to largely normal expression of hoxb1a and substitutions in M2, BMwM2Meis3 has 5 amino acid hoxb2 (approx. 10% affected; Fig. 1Ce,f), similar to the effect substitutions in M2, BMM1/2Meis3 has the same substitution of expressing Meis3 and Pbx4 in the absence of Hoxb1b (Fig. as BMM2Meis3 plus a four amino acid substitution in M1 and 1Bc,d; Table 1). This suggests that Hoxb1b must interact with BMNMeis3 has had its N terminus replaced by a protein Pbx4 to mediate the synergistic effect in vivo. interaction domain from the unrelated FRAP protein. Each of We next generated forms of Meis3 with reduced Pbx4 these proteins does not bind Pbx4 in vitro (Fig. 2C), but still binding activity (BMMeis3 mutants, see Materials and binds DNA (Fig. 2D) and is expressed at similar levels to wild- Homeodomain protein interactions in the CNS 1303

Fig. 1 1304 N. Vlachakis, S.-K. Choe and C. G. Sagerström

Fig. 2 type Meis3 following microinjection (as assayed by western BMM2Meis3 was slightly less active (Fig. 1Ci,j; Table 1) than blotting in Fig. 2E and immunohistochemistry in Fig. 2G). wild-type Meis3 (Fig. 1Cc,d). In contrast, the activity of Expression of the BMMeis3 mutants alone had no effect (Fig. BMwM2Meis3 was largely abolished, as illustrated by a 1C; Table 1). Following co-expression with Pbx4 and Hoxb1b, pronounced reduction both in frequency (approx. 7% embryos Homeodomain protein interactions in the CNS 1305

Fig. 2. Expression, Pbx4 interaction and subcellular distribution of Meis3 and Hoxb1b. (A) Pbx4 was expressed alone (lane 3) or together with HAHoxb1b (lanes 1 and 2), or HAΒMHoxb1b (lanes 4 and 5) in vitro in the presence of [35S]methionine and either analyzed directly (input; lanes 1 and 4) or first immunoprecipitated with anti- HA antibody (lanes 2, 3, 5). All immunoprecipitations were performed in the presence of an oligonucleotide containing a Pbx/Hox binding site (P/H). (B) Western blot analysis (10 embryos/lane) of uninjected (lane 1), HAhoxb1b- (lane 2), HAbmhoxb1b- (lane 3), or HAδhoxb1b- (lane 4) injected embryos, probed with anti-HA. (C) Pbx4 was expressed alone (lane 4) or together with Meis3 (lanes 1-3), ΒMNMeis3 (lanes 5 and 6), Meis3VP16 (lanes 7 and 8), MYCMeis3 (lanes 9 and 10), MYCBMM2Meis3 (lanes 11 and 12), MYCBMwM2Meis3 (lanes 13 and 14) or MYCBMM1/2Meis3 (lanes 15 and 16) in vitro in the presence of [35S]methionine and either analyzed directly (input; lanes 1, 5, 7, 9, 11, 13, 15) or first immunoprecipitated with anti- Meis antisera (lanes 2, 4, 6, 8, 10, 12, 14, 16) or with preimmune sera (lane 3). Immunoprecipitations were performed in the presence of an oligonucleotide containing a Meis/Pbx binding site (M/P). (D) Meis3 (lanes 1 and 2), ΒΜΝΜeis3 (lanes 3 and 4), Meis-VP16 (lanes 5 and 6), MutMeis3 (lanes 7 and 8), MYCMeis3 (lanes 9 and 10), MYCBMM2Meis3 (lanes 11 and 12), MYCBMwM2Meis3 (lanes 13 and 14) or MYCBMM1/2Meis3 (lanes 15 and 16) were expressed in vitro and incubated with 32P-labeled oligonucleotide containing a Meis/Pbx binding site (M/P; lanes 1, 3, 5, 7, 9, 11, 13, 15) or a random sequence (R; lanes 2, 4, 6, 8, 10, 12, 14, 16). The samples were immunoprecipitated with anti-Meis antisera, resolved on a 5% acrylamide gel, and exposed to X-ray film to detect the presence of labeled oligonucleotides. (E) Western blot analysis (10 embryos/lane) of uninjected (lane 1), meis3- (lane 2), mutmeis3- (lane 3), bmNmeis3- (lane 4) or meis3vp16- (lane 5) injected embryos, or (3 embryos/lane) uninjected (lane 6), MYCmeis3- (lane 7), MYCbmM2meis3- (lane 8), MYCbmwM2meis3- (lane 9), MYCbmM1/2meis3- (lane 10), nlsMYCbmM1/2meis3- (lane 11), or MYCbmNmeis3- (lane 12) injected embryos, probed with anti-Meis antisera. (F) Western blot analysis of lysates of 10 uninjected embryos harvested at either 6 hpf (lane 1) or 14 hpf (lane 2) probed with anti-Meis antisera (top panel) or anti-Pbx antisera (bottom panel). (G) Embryos injected with HAhoxb1b (a), HAbmHoxb1b (b), HAδhoxb1b (c), MYCmeis3 (d,f), MYCmeis3+pbx4 (e), MYCbmM2meis3 (g), MYCbmM2meis3+pbx4 (h), MYCbmwM2meis3 (i), MYCbmwM2meis+pbx4 (j), MYCbmM1/2meis3 (k), MYCbmM1/2meis3+pbx4 (l), nlsMYCbmM1/2meis3 (m), MYCbmNmeis3 (n), or MYCbmNmeis3+pbx4 (o) mRNAs, were fixed at 5 hpf (a-e and g-o) or 13 hpf (f) and immunostained with anti-HA (a-c) or anti-MYC (d-o).

manner (see below). Furthermore, a previous report with truncations vs. 39% for wild-type Meis3) and extent demonstrated that mutating either the M1 or M2 domain of Hth (Fig. 1Cm,n compare with c and d) of ectopic expression. does not completely eliminate Exd binding or activity in vivo, BMM1/2Meis3 (Fig. 1Cq,r) and BMNMeis3 (Fig. 1Cu,v) were while mutating both domains does abolish activity (Jaw et al., essentially inactive when co-expressed with Pbx4 and Hoxb1b 2000). Our results therefore indicate that both Hoxb1b and and gave similar results to embryos injected with Hoxb1b and Meis3 require intact Pbx-interaction domains to mediate their Pbx4 in the absence of Meis3 (Table 1). Our results therefore synergistic effects, suggesting that they interact with Pbx4 in indicate that it is necessary to mutate both the M1 and M2 vivo. domains to abolish all Meis3 activity in vivo. Since these domains are reported to bind Pbx (reviewed in Mann and Meis3, but not Hoxb1b, requires an intact Pbx- Affolter, 1998), this suggests that Meis3 must bind Pbx4 to interaction domain for nuclear access synergize with Pbx4 and Hoxb1b in vivo. Our results also Our results indicate that both Hoxb1b and Meis3 require intact suggest that BMM2Meis3 retains the ability to bind Pbx4 Pbx interaction domains to function in vivo, but the reason for in vivo, although we can not detect this by co- this is not clear. In particular, Meis and Hox proteins may immunoprecipitation in vitro. This is supported by our finding interact with Pbx not only to assemble a transcription that BMM2Meis3 accesses the nucleus in a Pbx4-dependent regulatory complex in the nucleus, but also to access the 1306 N. Vlachakis, S.-K. Choe and C. G. Sagerström nucleus. Although nuclear access of Exd is regulated in proteins contain transcription activation domains at their N Drosophila embryos (Abu-Shaar et al., 1999; Jaw et al., 2000; termini (Di Rocco et al., 1997; Rambaldi et al., 1994), but Meis Pai et al., 1998; Rieckhof et al., 1997), Pbx4 is nuclear (or Prep1) proteins do not appear to contain regulatory domains throughout zebrafish embryos (Popperl et al., 2000). Many (Berthelsen et al., 1998; Jacobs et al., 1999) and Pbx4 belongs Hox proteins contain a nuclear localization signal (NLS; e.g to a Pbx class (Vlachakis et al., 2000) that lacks the C-terminal Harvey et al., 1986) so Hoxb1b would be predicted to be regulatory domain (Asahara et al., 1999; Di Rocco et al., 1997). nuclear in zebrafish embryos. Hth also appears to have a NLS, Hoxb1b may therefore mediate transcriptional activation in our but is unstable in Drosophila embryos in the absence of experiments. It has been previously demonstrated (Di Rocco Exd (Abu-Shaar et al., 1999; Jaw et al., 2000). To test the et al., 1997) that deleting the N terminus significantly reduces distribution of Hoxb1b and Meis3 in zebrafish cells, we injected transcriptional activation by murine Hoxb1 without affecting mRNA of various Meis3 and Hoxb1b constructs and assessed binding to Pbx or DNA and we have generated the analogous their subcellular distribution by immunohistochemistry. We deletion in zebrafish Hoxb1b (δHoxb1b; see Materials and find that Hoxb1b and BMHoxb1b are nuclearly localized (Fig. Methods). δHoxb1b still localizes to the nucleus (Fig. 2Gc) and 2Ga,b), indicating that Hoxb1b need not interact with Pbx4 to is expressed at levels similar to wild-type Hoxb1b following access the nucleus. In contrast, following micro-injection microinjection (compare lanes 2 and 4 in Fig. 2B). Expression Meis3 is cytoplasmic in late blastula stage zebrafish embryos of δHoxb1b by itself led to fewer embryos with ectopic hoxb1a (Fig. 2Gd), but becomes primarily nuclear by early (18%; Table 2) than did expression of wild-type Hoxb1b (52%, somitogenesis stages (Fig. 2Gf). This redistribution of Meis3 Table 1) and the extent of ectopic in the protein correlates with an increase in Pbx4 protein levels affected embryos was significantly reduced (compare Fig. 3e,f that takes place during normal development (Fig. 2F) and is with Fig. 1Ac,d). Similarly, co-expression of δHoxb1b with therefore consistent with Meis3 requiring Pbx4 to access the Pbx4 and Meis3 (Fig. 3g,h) led to less extensive ectopic gene nucleus. To test this directly, we co-expressed Meis3 and Pbx4 expression than co-expression of the three wild-type proteins and found that in the presence of Pbx4, Meis3 is predominantly (Fig. 3c,d) and a reduction in the frequency of affected nuclear even at late blastula stages (Fig. 2Ge). This observation embryos. In particular, the most severely affected embryos also provides an explanation for the finding that endogenous were reduced from approx. 39% to approx. 8% (Table 1 and Pbx4 is limiting for the synergistic effect in Fig. 1B. 2). This demonstrates that the N terminus of Hoxb1b is BMwM2Meis3 (Fig. 2Ci,j), BMM1/2Meis3 (Fig. 2Ck,l) and required for full activity in vivo. BMNMeis3 (Fig. 2Cn,o) remain primarily cytoplasmic even in We next generated a fusion protein where the VP16 the presence of Pbx4, consistent with their having little or activation domain is inserted into Meis3 (Meis3VP16; see no functional activity in vivo. BMM2Meis3 however is Materials and Methods). Meis3VP16 still interacts with Pbx4 cytoplasmic in the absence and nuclear in the presence of Pbx4 (compare lanes 2 and 8 in Fig. 2C), binds DNA, albeit (Fig. 2Cg,h). These results indicate that wild-type Meis3 needs somewhat less efficiently than wild-type Meis3 (compare lanes to interact with Pbx4 in order to access the nucleus and are 1 and 5 in Fig. 2D) and is expressed at levels comparable to consistent with the BMM2Meis3 mutant having residual Pbx4 wild-type Meis3 following microinjection (compare lanes 2 binding activity. These results are also consistent with a report and 5 in Fig. 2E). When Meis3VP16 and Pbx4 were co- demonstrating that Prep1 requires Pbx1 to access the nucleus expressed, we observed the same effects on hoxb1a and hoxb2 in mammalian tissue culture cells (Berthelsen et al., 1999). expression (asterisks in Fig. 3i,j, respectively) as on co- Notably, the nuclear localization of the various Meis3 expression of the three wild-type proteins, and at a similar constructs in response to Pbx4 correlates with their in vivo frequency (70-80% affected; Table 1 and 2). This result activity as assayed in Fig. 1. If Meis3 requires Pbx4 to enter demonstrates that Hoxb1b can be functionally replaced by an the nucleus it may or may not require Pbx4 also to function exogenous activation domain. Taken together these results are in the nucleus. To explore this we co-expressed Pbx4 and consistent with Hoxb1b containing a transactivation domain Hoxb1b with NLSBMM1/2Meis3 (see Materials and Methods). and the synergistic effect of Meis3, Pbx4 and Hoxb1b being NLSBMM1/2Meis3 can access the nucleus independently of mediated by transcriptional activation. In agreement with Pbx4 (Fig. 2Gm) and is expressed at similar levels to wild-type this, co-injecting Meis3VP16 along with Pbx4 and wild- Meis3 following microinjection (Fig, 2E). We find that co- type Hoxb1b gave essentially the same phenotype as expression of NLSBMM1/2Meis3 with Pbx4 and Hoxb1b has Meis3VP16+Pbx4 (Fig. 3k,l; Table 2), although at a higher the same effect as Pbx4+Hoxb1b alone (Fig. 1Cy,z: Table 1), frequency. demonstrating that Meis3 also requires an intact Pbx- To further examine how Meis3, Pbx4 and Hoxb1b function interaction domain to function in the nucleus. to activate transcription we next tested whether Meis3 needs to Taken together with the results in Fig. 1, these results bind DNA in order to synergize with Pbx4 and Hoxb1b. To this indicate that Hoxb1b interacts with Pbx4 in the nucleus, that end we made use of a mutant Meis3 (MutMeis3; see Materials Meis3 interacts with Pbx4 both in the cytoplasm and the and Methods), that cannot bind DNA (Fig. 2D, lane 1 versus nucleus, and that these interactions are all required for Hoxb1b 5), but still interacts with Pbx4 (Vlachakis et al., 2000) and is and Meis3 function. expressed at similar levels to wild-type Meis3 following microinjection (compare lanes 2 and 3 of Fig. 2E). Expression The synergistic effect of Pbx4, Hoxb1b and Meis3 is of MutMeis3 by itself had no effect (Fig. 3m,n; Table 2). mediated at the level of transcriptional activation Expression of Hoxb1b+Pbx4+MutMeis3 resulted in embryos Consistent with their localization to the nucleus, homeodomain showing ectopic expression of hoxb1a (53%, Table 2; asterisk protein complexes are predicted to function as transcriptional in Fig. 3o) and hoxb2 (37%, Table 2; asterisk in Fig. 3p), and regulators (reviewed by Mann and Affolter, 1998). Several Hox the frequency of the most severely affected embryos was lower Homeodomain protein interactions in the CNS 1307

Fig. 3. The effect of Hoxb1b, Pbx4 and Meis3 co-expression is mediated at the level of transcriptional activation. Embryos were injected with mRNAs as indicated to the left and analyzed for expression of hoxb1a (left hand panels) or hoxb2 (right hand panels). Asterisks indicate ectopic hoxb1a (c,i,k,o) and hoxb2 (d,j,l,p) expression. Arrows point to small patches of ectopic hoxb1a (g) and hoxb2 (h). Embryos are at the 5- (for hoxb2) or 10- (for hoxb1a) somite stage and are shown in dorsal views with anterior to the left.

Table 2. The effect of Hoxb1b, Pbx4 and Meis3 co-expression is mediated by transcriptional activation Outcome Ectopic staining/ In situ Injected RNA* Unaffected Ectopic staining‡ truncated axis§ probe δhoxb1b 78/95 (82%) 17/95 (18%) 0/95 (0%) hoxb1a 74/80 (92%) 6/80 (8%) 0/80 (0%) hoxb2 δhoxb1b/pbx4/meis3 98/171 (57%) 59/171 (35%) 14/171 (8%) hoxb1a 113/142 (80%) 19/142 (13%) 10/142 (7%) hoxb2 meis3vp16/pbx4 13/73 (18%) 21/73 (29%) 39/73 (53%) hoxb1a 18/91 (20%) 27/91 (30%) 46/91 (50%) hoxb2 14/86 (16%) 27/86 (31%) 45/86 (53%) krox20 hoxb1b/pbx4/meis3vp16 6/108 (6%) 18/108 (17%) 84/108 (77%) hoxb1a 9/89 (10%) 11/89 (12%) 69/89 (78%) hoxb2 3/60 (5%) 9/60 (15%) 48/60 (80%) krox20 mutmeis3 55/56 (98%) 1/56 (2%) 0/56 (0%) hoxb1a 70/70 (100%) 0/70 (0%) 0/70 (0%) hoxb2 hoxb1b/pbx4/mutmeis3 37/79 (47%) 39/79 (49%) 3/79 (4%) hoxb1a 102/161 (63%) 53/161 (33%) 6/161 (4%) hoxb2 39/73 (53%) 31/73 (43%) 3/73 (4%) krox20

*1- to 2-cell stage embryos were injected with mRNAs as listed, fixed at approx. 5- (for hoxb2) or 10- (for hoxb1a and krox20) somite stages and analyzed by whole-mount in situ hybridization. ‡Embryos with normal morphology but ectopic or distorted expression of marker genes. §Embryos with anterior truncations and ectopic expression of marker genes. 1308 N. Vlachakis, S.-K. Choe and C. G. Sagerström

(~4%; Table 2) than for the three wild-type proteins (~39%; compared to lacZ-injected embryos (Fig. 4Bk), suggesting that Table 1). However, the effect of MutMeis3 is higher than that expansion of hindbrain gene expression anteriorly is of BMM1/2Meis3 or BMNMeis3 co-expressed with Pbx4 and accompanied by a reduction in the expression domains of Hoxb1b, indicating that MutMeis3 retains some in vivo anterior genes. activity. While this could be due to MutMeis3 binding DNA Co-expression of Hoxb1b, Pbx4 and Meis3 also leads to weakly in vivo, we think this unlikely since MutMeis3 does posterior defects in some embryos (e.g. Fig. 1Bg). To explore not bind DNA in vitro. Instead, this result may suggest that this we analyzed expression of myoD (expressed in somites; DNA-binding is not absolutely required for Meis3 to synergize Weinberg et al., 1996) and no tail (ntl; expressed in notochord; with Pbx4 and Hoxb1b in vivo. This would be consistent with Schulte-Merker et al., 1994). lacZ-injected embryos had experiments indicating that Meis and Prep1 can participate as normal somites (Fig. 4B, panels c and e) and notochords non-DNA binding partners in trimeric complexes with Hox (arrowheads in Fig. 4Bg,i,k). Co-expression of Hoxb1b, Pbx4 and Pbx (Berthelsen et al., 1998; Shanmugam et al., 1999; and Meis3 resulted in the majority of the embryos (~85%; Vlachakis et al., 2000). 210/248; Fig. 4Bd,f) having somites of normal appearance, although in approximately one third of these the rows of Co-expression of Hoxb1b, Pbx4 and Meis3 induces somites were slightly displaced (e.g. asterisk in Fig. 4Bd). The ectopic expression of the non-hox genes krox20 and remaining 15% had malformed somites or were missing some valentino somites, but ectopic myoD expression was never observed. To explore the effects of Hoxb1b, Pbx4 and Meis3 in more detail 67% (148/223) of hoxb1b+pbx4+meis3-injected embryos also we analyzed two non-hox genes normally expressed in the had normal notochords (Fig. 4Bh) and the remaining 33% had hindbrain, krox20 (expressed in r3 and r5; Oxtoby and Jowett, ectopic no tail staining (double arrowheads in Fig. 4Bl) that 1993) and valentino (expressed in r5 and r6; Moens et al., 1998). occasionally formed a second notochord (arrowheads in Fig. Co-expression of Hoxb1b+Pbx4 (Fig. 4Ac), or Meis3+Pbx4 (Fig. 4Bj). 4Ae) had little effect on krox20 expression (approx. 90% normal To ensure that Hoxb1b, Pbx4 and Meis3 co-expression did embryos; Table 1), but co-expression of Hoxb1b+Pbx4+Meis3 not have a more profound effect anteriorly because of uneven resulted in massive ectopic expression of krox20 (71%, Table 1; distribution of injected RNAs, we introduced lacZ mRNA as a asterisk in Fig. 4Ag) anterior to its normal expression domain. lineage label together with meis3 and hoxb1b mRNA, and This was dependent on functional Pbx interaction domains of detected β-galactosidase protein by its enzymatic activity. As Hoxb1b and Meis3 as expression of Hoxb1b+Pbx4+BMNMeis3 expected, we found that the distribution of β-galactosidase (Fig. 4Ad) or BMHoxb1b+Pbx4+Meis3 (Fig. 4Af) resulted in varied between embryos, likely explaining the variability in normal embryos (~90%; Table 1). Hoxb1b+Pbx4+MutMeis3 phenotypes reported in Tables 1 and 2, but that there was no resulted in ectopic krox20 expression anteriorly (47%; Table 2; bias in distribution towards the anterior end of the embryo asterisk in Fig. 4Ab) as did Hoxb1b+Pbx4+Meis3VP16 (not (e.g. Fig. 4Ca shows an embryo with strong β-galactosidase shown; 95% Table 2) and Meis3VP16+Pbx4 (Fig. 4Ah; 84% expression posteriorly). Furthermore, ectopic expression of Table 2). Analysis of valentino revealed that this gene was also hindbrain genes was accompanied by β-galactosidase induced ectopically (64%, Table 1; asterisk in figure 3Bh) by co- expression (arrow in Fig. 4Bb points to ectopic krox20 expression of Hoxb1b, Pbx4 and Meis3. Thus, co-expression of expression (purple) overlapping with β-galactosidase staining Hoxb1b, Pbx4 and Meis3 induces the expression of two non-hox (light blue)) confirming that Hoxb1b, Pbx4 and Meis3 mediate hindbrain genes and krox20 induction is regulated similarly to their effects at their site of expression. hoxb2 induction. Thus, we conclude that co-expression of Hoxb1b, Pbx4 and Meis3 has more profound effects anteriorly than posteriorly. Co-expression of Hoxb1b, Pbx4 and Meis3 causes This is phenotypically similar to the outcome of homeotic ectopic expression of hindbrain genes at the mutations and suggests that ectopic expression of these expense of anterior, but not posterior gene proteins may mediate a posterior transformation of anterior expression structures. Since co-expression of Hoxb1b, Pbx4 and Meis3 induces expression of hindbrain genes rostrally, we explored the effect Co- expression of Hoxb1b, Pbx4 and Meis3 causes on anterior gene expression. We found that expression of otx2 an anterior (forebrain and midbrain) to posterior (normally expressed in forebrain and midbrain regions; Li et (hindbrain) transformation of cell fate al., 1994; Mori et al., 1994) was reduced in 66% (84/128) of To determine if co-expression of Hoxb1b, Pbx4 and Meis3 hoxb1b+pbx4+meis3-injected embryos (Fig. 4Bb) and in 41% mediates the transformation of anterior cell fates to posterior (22/53) of hoxb1b+meis3-injected embryos (not shown). These ones, we performed fate mapping of prospective forebrain/ numbers correlate well with the number of embryos exhibiting midbrain cells in control and hoxb1b+pbx4+meis3-injected hindbrain gene expression anteriorly (67% and 49%, embryos. Embryos were labeled with lipophilic dye (DiI) at the respectively; Table 1). Injections with lacZ mRNA resulted in animal pole (the site of prospective forebrain and midbrain embryos with normal otx2 expression (100%; 58/58; Fig. 4Aa). precursors at this stage; Kimmel et al., 1990; Woo and Fraser, To confirm that the reduction in otx2 expression coincides with 1995) at early gastrula stages (6.5 hpf; Fig. 5A). Embryos where the expression of hindbrain markers anteriorly, we performed the injected DiI was localized at the animal pole 2 hours after triple in situ hybridizations for otx2, krox20 and no tail (Fig. labeling, as shown in Fig. 5B, were allowed to develop to the 4Bl). We find that otx2 expression (purple stain, indicated by 4- to 5-somite stage, fixed, photoconverted and analyzed by black arrow in Fig. 4Bl) is reduced and krox20 expression (red whole-mount in situ hybridization for expression of krox20. In stain, indicated by white asterisk in Fig. 4Bl) is expanded control embryos (Fig. 5C; arrow in a and b) labeled cells were Homeodomain protein interactions in the CNS 1309

Fig. 4. Hoxb1b, Pbx4 and Meis3 co-expression causes ectopic expression of hindbrain genes at the expense of anterior, but not posterior gene expression. (A) Embryos were injected with mRNAs as indicated to the side of each panel and analyzed by in situ hybridization by double labeling (a and c) for krox20 (red) and no tail (purple), or by single labeling (b,d,e,f,g,h) for krox20. Asterisks indicate ectopic krox20 expression in b, g and h. All embryos are at 9- to 12-somite stages and are shown in dorsal views with anterior to the left. (B) Embryos were injected with lacZ (a,c,e,g,i,k) or hoxb1b+pbx4+meis3 (b,d,f,h,j,l) mRNAs and analyzed for expression of otx2 (a,b), myoD (c,d), krox20 and myoD (both in purple; e,f), valentino and no tail (both in purple; g,h), hoxb2 and no tail (both in purple; i,j) or by triple labeling (k,l) for otx2 (in purple), krox20 (in red) and no tail (in purple). Arrows indicate diminished otx2 expression (b and l); normal krox20 expression in r3 and r5 (e); normal valentino expression in r5 and r6 (g); normal hoxb2 expression in r3 (i); normal otx2 expression in forebrain and midbrain (k). Arrowheads indicate no tail expression (g-l). Double arrowheads indicate ectopic no tail expression (l). Asterisks indicate displacement of somites (d), ectopic krox20 (f and l), valentino (h) or hoxb2 (j) expression. All embryos are at the 9- to 12-somite stages and are shown in dorsal views with anterior to the left. (C) Embryos were injected with hoxb1b+meis3 along with lacZ mRNA and stained for lacZ (light blue) and krox20 expression (purple). Embryos are at the 10-somite stage and are shown in lateral views with anterior to the left and dorsal up. Arrow in b indicates ectopic krox20 expression. 1310 N. Vlachakis, S.-K. Choe and C. G. Sagerström

Fig. 5. Hoxb1b, Pbx4 and Meis3 co-expression causes an anterior (forebrain and midbrain) to posterior (hindbrain) transformation of cell fate. (A) Schematic outline of fate mapping experiment. Embryos were labeled with lipophilic dye (DiI) at the animal pole at early gastrula stage (approx. 6.5 hpf). (B) Picture of an embryo 2 hours after labeling showing DiI only at the animal pole. (C) DiI-labeled uninjected (a,b) and hoxb1b+pbx4+meis3-injected (c-h) embryos were fixed at the 4- to 5-somite stages, photoconverted and analyzed by in situ hybridization for expression of krox20 (purple staining in a-h). Arrows point to DiI- labeled cells (reddish brown). Embryos are shown in dorsal views with anterior to the left. b,d,f and h (40× objective) show parts of a,c,e and g (20× objective).

Fig. 6. Hoxb1b, Pbx4 and Meis3 co-expression induces ectopic Mauthner neurons. Embryos were injected with hoxb1b/pbx4/meis3 (b,c) mRNAs and stained along with uninjected embryos (a) with 3A10 antibody to reveal Mauthner neurons. Arrows and arrowheads of the same size and color point to cell body and axons, respectively, of the same Mauthner neuron. Asterisks in b indicate axons of neurons whose cell bodies are not distinguishable. All panels are image reconstructions of confocal images and axons were assigned to ectopic Mauthner neurons by analyzing separate images of each stack. present in their very rostral domain only. This domain 26 hpf (Hatta, 1992). Control embryos revealed a single pair of represents the forebrain and midbrain, as illustrated by a gap Mauthner neurons (arrows in Fig. 6a) displaying characteristic between DiI-positive cells (red) and krox20 expression (purple). contralateral axonal projections (arrrowheads in Fig. 6a). In Embryos injected with hoxb1b+pbx4+meis3 also exhibited DiI- contrast, hoxb1b+pbx4+meis3-injected embryos displayed positive cells rostrally, but these cells also expressed ectopic large numbers (at least up to 7) of Mauthner neurons and their krox20 (arrows in Fig. 5Cc-h). Thus, cells that normally give axonal projections (individual Mauthner neurons are indicated rise to forebrain and midbrain acquire a posterior (hindbrain) by arrows of different sizes and colors and their axonal fate following ectopic expression of Hoxb1b, Pbx4 and Meis3, projections by arrowheads, in Fig. 6b and c). Cell bodies and demonstrating that expression of these proteins can mediate axons were identified and traced by analyzing 20-30 confocal posterior transformations of cell fates. sections for each sample, but each panel in Fig. 6 only shows an image reconstruction for each sample. The background Co-expression of Hoxb1b, Pbx4 and Meis3 mediates staining is due largely to melanocytes that were difficult to formation of ectopic Mauthner neurons remove because of the stage and abnormal development of the In order to explore the extent to which differentiation of injected embryos. hindbrain fates was induced rostrally by co-expression of These data reveal that in addition to mediating ectopic Hoxb1b, Pbx4 and Meis3, we examined formation of the expression of hindbrain genes, co-expression of Hoxb1b, Pbx4 Mauthner neurons (a segment-specific, bilateral set of neurons and Meis3 can initiate the differentiation of r4-specific neurons found in r4) in control and affected embryos. In Fig. 6 we used rostrally. Ectopic expression of Hoxb1b alone has been shown 3A10 antibody that specifically stains Mauthner neurons at 24- to induce an extra pair of Mauthner neurons in r2 (Alexandre Homeodomain protein interactions in the CNS 1311 et al., 1996), but co-expression of Hoxb1b, Pbx4 and Meis3 Third, although both Hoxb1b and Meis3 appear to require appears to be more potent in this capacity, since we observe at Pbx4 interaction to be functional in vivo, we do not know least 7 ectopic Mauthner neurons rostrally. whether Hoxb1b and Meis3 interact with the same Pbx4 molecule to form a trimeric complex, or whether they interact with separate Pbx4 molecules to form a pair of dimers. DISCUSSION However, several pieces of data indicate the formation of trimeric complexes. First, both the and hoxb2 r4 Several reports have demonstrated that Hox, Pbx and Meis enhancers contain adjacent Hox/Pbx binding sites and a more binding sites in enhancers of Hox-dependent genes are required distant Meis site, but there is no Pbx site near the Meis site for expression (Ferretti et al., 2000; Jacobs et al., 1999; Ryoo (Ferretti et al., 2000; Jacobs et al., 1999). Consistent with this, et al., 1999). Since Meis, Pbx and Hox can form complexes in DNA fragments containing these sequences support formation vitro (Berthelsen et al., 1998; Ferretti et al., 2000; Jacobs et al., of trimeric Hox/Pbx/Meis complexes, but not of a pair of 1999; Ryoo et al., 1999; Vlachakis et al., 2000) and such dimers, in vitro (Ferretti et al., 2000). Second, a DNA-binding complexes can be isolated from cell extracts (Ferretti et al., mutant Prep1 forms dimers with Pbx that bind DNA only very 2000; Shen et al., 1999), it is possible that these proteins weakly (Berthelsen et al., 1998). Therefore, if Meis3, Pbx4 and function as complexes in vivo. To date, complex formation as Hoxb1b acted as a pair of dimers, a DNA-binding mutant of a requirement for in vivo function is best supported by over- Meis3 (MutMeis3) should not be able to form a functional expression of the Hth N terminus in Drosophila, where it dimer with Pbx4 and should not have any in vivo activity. interferes with the function of endogenous Hth, likely by However, we find that MutMeis3 still functions in vivo, as does preventing Hth and Exd from interacting (Ryoo et al., 1999). a DNA-binding mutant Hth (Ryoo et al., 1999). Here we demonstrate that both Hoxb1b and Meis3 require Lastly, in the only other study of in vivo Meis function intact Pbx interaction domains for the expression of Hox- (Salzberg et al., 1999), misexpression of Xenopus Meis3 by dependent genes in the zebrafish, indicating that Hoxb1b and itself had minimal effect on krox20 and hoxb1 expression, but Meis3 need to form complexes with Pbx4 to function in vivo. nevertheless mediated anterior deletions in Xenopus. This is in contrast with our analysis, where zebrafish Meis3 requires What type of complexes do Hoxb1b, Pbx4 and Meis3 Pbx4 and Hoxb1b for the transformation of anterior fates. form in vivo? Since lineage labeling was not utilized to analyze the deletions Notably, our experiments do not indicate the composition of in Xenopus, we do not know if a distinct mechanism is at work, Hoxb1b-, Pbx4- and Meis3-containing complexes, and several or if some Meis family members may be able to function issues remain to be resolved. First, it is not clear if all independently of Pbx and Hox. complexes contain a Meis family member. We demonstrate that ectopic expression of Hoxb1b by itself induces hoxb1a Co-expression of Hoxb1b, Pbx4 and Meis3 is expression in r2. To perform this function, Hoxb1b needs to sufficient to promote hindbrain differentiation interact with an endogenous Pbx protein (most likely Pbx4 as The effects mediated by Hoxb1b, Pbx4 and Meis3 co- this is the predominant Pbx protein at this stage; Popperl et al., expression are likely to be causally related and to occur in 2000), but does not require exogenous Meis3. While this is sequence. Since murine hoxb1 and hoxb2 have Pbx, Hox and consistent with Hoxb1b and Pbx4 acting in the absence of a Meis binding sites in their enhancers (Ferretti et al., 2000; Meis protein, it leaves open the possibility that an endogenous Jacobs et al., 1999) it is likely that zebrafish hoxb1a and hoxb2 Meis protein is involved. Indeed, the zebrafish prep1 gene are directly induced by Hoxb1b, Pbx4 and Meis3. Ectopic appears to be ubiquitously expressed (N. V. and C. G. S., expression of hoxb2 induces krox20 and valentino expression unpublished) and endogenous Prep1 may interact with Hoxb1b in zebrafish (Yan et al., 1998), suggesting that these genes and Pbx4 in our experiments. However, while the Pbx and Hox may be activated subsequently to hoxb2. Thus, expression of binding sites appear to be required for in vivo expression of Hoxb1b, Pbx4 and Meis3 is sufficient to promote the most Hox-dependent genes (Ferretti et al., 2000; Jacobs et al., differentiation of hindbrain fates, particularly r4 fates, and we 1999; Pöpperl et al., 1995), a Meis/Prep1 binding site is only speculate that they normally perform this function within the required for some genes (Ferretti et al., 2000). Meis proteins caudal hindbrain during zebrafish embryogenesis. may therefore not always be required (at least not as a DNA binding component) for Pbx and Hox proteins to function in We wish to thank members of the Sagerström lab for helpful vivo. discussions, C. Moens for the valentino probe, V. Prince for the hoxb2 Second, different Hox proteins have different effects in vivo, probe, Dan Kessler for the pCS2+VP16N plasmid and Jeffrey but it is not clear if different Meis family members differ Nickerson and Paul Furcinitti for assistance with confocal and digital deconvolution microscopy respectively. This work was supported by functionally. For instance, if Hoxb1b and Pbx4 require Prep1 grant R01NS38183 from the NIH and RPG-00-255-01-DDC from the to activate hoxb1a expression, the synergistic effect we see American Cancer Society. following co-expression of Meis3 could be due to Prep1 being limiting in vivo. In this scenario, Prep1 and Meis3 would be functionally equivalent. An alternative explanation to the synergistic effect is that exogenously supplied Meis3 provides REFERENCES a unique function, perhaps by replacing Prep1, thus playing an Abu-Shaar, M., Ryoo, H. D. and Mann, R. S. (1999). Control of the nuclear instructive role. We favor the latter model, primarily because localization of Extradenticle by competing nuclear import and export published experiments suggest that Prep1 cannot substitute for signals. Genes Dev. 13, 935-945. Homothorax in Drosophila (Jaw et al., 2000). Alexandre, D., Clarke, J. D., Oxtoby, E., Yan, Y. L., Jowett, T. and Holder, 1312 N. Vlachakis, S.-K. Choe and C. G. Sagerström

N. (1996). Ectopic expression of Hoxa-1 in the zebrafish alters the fate of PBX2 and PBX3, new genes with extensive homology to the the mandibular arch neural crest and phenocopies a retinoic acid-induced human proto-oncogene PBX1. Mol. Cell. Biol. 11, 6149-6157. phenotype. Development 122, 735-746. Mori, H., Miyazaki, Y., Morita, T., Nitta, H. and Mishina, M. (1994). Amores, A., Force, A., Yan, Y. L., Joly, L., Amemiya, C., Fritz, A., Ho, Different spatiotemporal expressions of three otx homeoprotein transcripts R. K., Langeland, J., Prince, V., Wang, Y. L. et al. (1998). Zebrafish hox durimg zebrafish embryogenesis. Mol. Brain Res. 27, 221-213. clusters and vertebrate genome evolution. Science 282, 1711-1714. Moskow, J. J., Bullrich, F., Huebner, K., Daar, I. O. and Buchberg, A. M. Asahara, H., Dutta, S., Kao, H. Y., Evans, R. M. and Montminy, M. (1999). (1995). Meis1, a PBX-1 related homeobox gene involved in myeloid Pbx-Hox heterodimers recruit coactivator-corepressor complexes in an leukemia in BXH-2 mice. Molecular and Cellular Biology 15, 5434-5443. isoform-specific manner. Mol. Cell Biol. 19, 8219-8225. Nakamura, T., Jenkins, N. A. and Copeland, N. G. (1996). Identification of Berthelsen, J., Kilstrup-Nielsen, C., Blasi, F., Mavilio, F. and Zappavigna, a new family of Pbx-related homeobox genes. Oncogene 13, 2235-2242. V. (1999). The subcellular localization of PBX1 and EXD proteins depends Nourse, J., Mellentin, J. D., Galili, N., Wilkinson, J., Stanbridge, E., Smith, on nuclear import and export signals and is modulated by association with S. D. and Cleary, M. L. (1990). Chromosomal translocation t(1;19) results PREP1 and HTH. Genes Dev. 13, 946-953. in synthesis of a homeobox fusion mRNA that codes for a potential chimeric Berthelsen, J., Zappavigna, V., Ferretti, E., Mavilio, F. and Blasi, F. (1998). . Cell 60, 535-45. The novel homeoprotein Prep1 modulates Pbx-Hox protein cooperativity. Oxtoby, E. and Jowett, T. (1993). Cloning of the zebrafish krox-20 gene (krx20) EMBO J. 17, 1434-1445. and its expression during development. Nucl. Acids Res. 21, 1087-1095. Blader, P., Fischer, N., Gradwohl, G., Guillemont, F. and Strahle, U. (1997). Pai, C. Y., Kuo, T. S., Jaw, T. J., Kurant, E., Chen, C. T., Bessarab, D. A., The activity of neurogenin1 is controlled by local cues in the zebrafish Salzberg, A. and Sun, Y. H. (1998). The Homothorax homeoprotein embryo. Development 124, 4557-4569. activates the nuclear localization of another homeoprotein, extradenticle, Chang, C. P., Jacobs, Y., Nakamura, T., Jenkins, N. A., Copeland, N. G. and suppresses eye development in Drosophila. Genes Dev. 12, 435-446. and Cleary, M. L. (1997). Meis proteins are major in vivo binding partners Pöpperl, H., Bienz, M., Studer, M., Chan, S. K., Aparicio, S., Brenner, S., for wild-type but not chimeric Pbx proteins. Mol. Cell. Biol. 17, 5679-5687. Mann, R. S. and Krumlauf, R. (1995). Segmental expression of Hoxb-1 Chen, J., Zheng, X. F., Brown, E. J. and Schreiber, S. L. (1995). is controlled by a highly conserved autoregulatory loop dependent upon Identification of an 11-kDa FKBP12-rapamycin-binding domain within the exd/pbx. Cell 81, 1031-1042. 289-kDa FKBP12-rapamycin-associated protein and characterization of a Popperl, H., Rikhof, H., Chang, H., Haffter, P., Kimmel, C. B. and Moens, critical serine residue. Proc. Natl. Acad. Sci. USA 92, 4947-4951. C. B. (2000). Iazarus is a novel pbx gene that globally mediates hox gene Di Rocco, G., Mavilio, F. and Zappavigna, V. (1997). Functional dissection function in zebrafish. Mol. Cell. 6, 255-267. of a transcriptionally active, target specific Hox-Pbx complex. EMBO J. 16, Prince, V. E., Moens, C. B., Kimmel, C. B. and Ho, R. K. (1998b). Zebrafish 3644-3654. hox genes:expression in the hindbrain region of wild-type and mutants of Fekany-Lee, K., Gonzalez, E., Miller-Bertoglio, V. and Solnica-Krezel, L. the segmentation gene valentino. Development 125, 393-406. (2000). The homeobox gene bozozok promotes anterior neuroectoderm Rambaldi, I., Kovacs, E. N. and Featherstone, M. S. (1994). A proline-rich formation in zebrafish through negative regulation of BMP2/4 and Wnt transcriptional activation domain in murine HOXD-4 (HOX- 4.2). Nucl. pathways. Development 127, 2333-2345. Acids Res. 22, 376-382. Ferretti, E., Marshall, H., Popperl, H., Maconochie, M., Krumlauf, R. and Rauskolb, C., Peifer, M. and Wieschaus, E. (1993). extradenticle, a regulator Blasi, F. (2000). Segmental expression of Hoxb2 in r4 requires two separate of homeotic gene activity, is a homolog of the homeobox-containing human sites that integrate cooperative interactions between Prep1, Pbx and Hox proto-oncogene . Cell 74, 1101-12. proteins. Development 127, 155-166. Rieckhof, G. E., Casares, F., Ryoo, H. D., Abu-Shaar, M. and Mann, R. S. Harvey, R. P., Tabin, C. J. and Melton, D. A. (1986). Embryonic expression (1997). Nuclear translocation of Extradenticle requires homothorax, which and nuclear localization of Xenopus homeobox (Xhox) gene products. encodes an Extradenticle-related homeodomain protein. Cell 91, 171-183. EMBO J. 5, 1237-1244. Ryoo, H. D., Marty, T., Casares, F., Affolter, M. and Mann, R. S. (1999). Hatta, K. (1992). Role of the floor plate in axonal patterning in the zebrafish Regulation of Hox target genes by a DNA bound Homothorax/Hox/ CNS. Neuron 9, 629-642. Extradenticle complex. Development 126, 5137-5148. Jacobs, Y., Schnabel, C. A. and Cleary, M. L. (1999). Trimeric association Sagerström, C. G., Kao, B. A., Lane, M. E. and Sive, H. (2001). Isolation of Hox and TALE homeodomain proteins mediates Hoxb2 hindbrain and characterization of posteriorly restricted genes in the zebrafish gastrula. enhancer activity. Mol. Cell Biol. 19, 5134-5142. Dev. Dyn. (in press). Jaw, T. J., You, L. R., Knoepfler, P. S., Yao, L. C., Pai, C. Y., Tang, C. Y., Salzberg, A., Elias, S., Nachaliel, N., Bonstein, L., Henig, C. and Frank, Chang, L. P., Berthelsen, J., Blasi, F., Kamps, M. P. et al. (2000). Direct D. (1999). A Meis family protein caudalizes neural cell fates in Xenopus. interaction of two homeoproteins, homothorax and extradenticle, is essential Mech. Dev. 80, 3-13. for EXD nuclear localization and function. Mech Dev. 91, 279-291. Schulte-Merker, S., van Eeden, F., Halpern, M. E., Kimmel, C. B. and Kamps, M. P., Murre, C., Sun, S. H. and Baltimore, D. (1990). A new Nüsslein-Volhard, C. (1994). no tail (ntl) is the zebrafish homologue of the homeobox gene contributes the DNA binding domain of the t(1;19) mouse T () gene. Development 120, 1009-1015. translocation protein in pre-B ALL. Cell 60, 547-555. Shanmugam, K., Green, N. C., Rambaldi, I., Saragovi, H. U. and Kessler, D. S. (1997). Siamois is required for formation of Spemann’s Featherstone, M. S. (1999). PBX and MEIS as non-DNA-binding partners organizer. Proc. Natl. Acad. Sci. USA 94, 13017-13022. in trimeric complexes with HOX proteins. Mol. Cell Biol. 19, 7577-7588. Kimmel, C. B., Warga, R. M. and Schilling, T. F. (1990). Origin and Shen, W. F., Rozenfeld, S., Kwong, A., Kom ves, L. G., Lawrence, H. J. organization of the zebrafish fatemap. Development 108, 581-594. and Largman, C. (1999). HOXA9 forms triple complexes with PBX2 and Knoepfler, P. S., Calvo, K. C., Chen, H., Antonarakis, S. E. and Kamps, MEIS1 in myeloid cells. Mol. Cell Biol. 19, 3051-3061. M. P. (1997). Meis1 and pKnox1 bind DNA cooperatively with Pbx1 Vlachakis, N., Ellstrom, D. R. and Sagerstrom, C. G. (2000). A novel pbx utilizing an interaction surface disrupted in oncoprotein E2a-Pbx1. Proc. family member expressed during early zebrafish embryogenesis forms Natl. Acad. Sci. USA 94, 14553-14558. trimeric complexes with Meis3 and Hoxb1b. Dev. Dyn. 217, 109-119. Knoepfler, P. S. and Kamps, M. P. (1995). The pentapeptide motif of Hox Weinberg, E. S., Allende, M. L., Kelly, C. S., Abdelhamid, A., Murakami, proteins is required for cooperative DNA binding with Pbx1, physically T., Andermann, P., Doerre, O. G., Grunwald, D. J. and Riggleman, B. contacts Pbx1, and enhances DNA binding by Pbx1. Mol. Cell Biol. 15, (1996). Developmental regulation of zebrafish MyoD in wild-type, no tail 5811-5819. and spadetail embryos. Development 122, 271-280. Krumlauf, R. (1994). Hox genes in vertebrate development. Cell 78, 191-201. Woo, K. and Fraser, S. E. (1995). Order and coherence in the fate map of the Li, Y., Allende, M. L., Finkelstein, R. and Weinberg, E. S. (1994). zebrafish nervous system. Development 121, 2595-2609. Expression of two zebrafish orthodenticle-related genes in the embryonic Yan, Y. L., Jowett, T. and Postlethwait, J. H. (1998). Ectopic expression of brain. Mech. Dev. 48, 229-244. hoxb2 after retinoic acid treatment or mRNA injection: disruption of Mann, R. S. and Affolter, M. (1998). Hox proteins meet more partners. Curr. hindbrain and craniofacial morphogenesis in zebrafish embryos. Dev. Dyn. Opin. Genet. Dev. 8, 423-429. 213, 370-385. Moens, C. B., Cordes, S. P., Giorgianni, M. W., Barsh, G. S. and Kimmel, Zhang, M., Kim, H.-J., Marshall, H., Gendron-Maguire, M., Lucas, A. D., C. B. (1998). Equivalence in the genetic control of hindbrain segmentation Baron, A., Gudas, L. J., Gridley, T., Krumlauf, R. and Grippo, J. F. in fish and mouse. Development 125, 381-391. (1994). Ectopic Hoxa-1 induces rhombomere transformation in mouse Monica, K., Galili, N., Nourse, J., Saltman, D. and Cleary, M. L. (1991). hindbrain. Development 120, 2431-2442.