Early sequential expression of mouse Hox genes is essential for their later function. Analysis of the relationship between early Hox gene expression and the laying down of anterior to posterior structures during and after gastrulation is therefore crucial for understanding the ontogenesis of Hox-mediated axial patterning. Using explants from gastrulation stage embryos,we show that the ability to express 3′ and 5′ Hox genes develops sequentially in the primitive streak region, from posterior to anterior as the streak extends, about 12 hours earlier than overt Hox expression. The ability to express autonomously the earliest Hox gene, Hoxb1, is present in the posterior streak region at the onset of gastrulation, but not in the anterior region at this stage. However, the posterior region can induce Hoxb1 expression in these anterior region cells. We conclude that tissues are primed to express Hox genes early in gastrulation, concomitant with primitive streak formation and extension, and that Hox gene inducibility is transferred by cell to cell signalling.
Axial structures that will later express Hox genes are generated in the node region in the period that Hox expression domains arrive there and continue to spread rostrally. However, lineage analysis showed that definitive Hox codes are not fixed at the node, but must be acquired later and anterior to the node in the neurectoderm, and independently in the mesoderm. We conclude that the rostral progression of Hox gene expression must be modulated by gene regulatory influences from early on in the posterior streak, until the time cells have acquired their stable positions along the axis well anterior to the node.
INTRODUCTION
Hox genes are key players in anteroposterior (AP) patterning, and their crucial role in this process is reflected in their extremely strong conservation among phyla (Gellon and McGinnis, 1998). The conserved chromosomal organisation of Hox genes in clusters (four in vertebrates) is intimately linked to regulatory constraints, which couple Hox gene expression to the progression of embryogenesis. In mammals, the laying down of anterior to posterior structures within the territories patterned by the Hox genes is accompanied by a sequential activation of these genes from 3′ to 5′ in the clusters. As a result, early structures are given an anterior identity with 3′Hox genes as key determinants, while progressively later structures start expressing more 5′ Hox genes and acquire a more posterior identity. The remarkable correlation between the spatiotemporal expression of the genes and their linear order in the Hox clusters has been called spatiotemporal colinearity (reviewed by Duboule and Morata, 1994).
The function of Hox genes in AP patterning in the mouse has been most clearly demonstrated at levels close to the rostral boundary of their definitive expression domains. Homeotic transformations of segmented structures developing from these boundary regions were observed in loss-of-function mutants (Krumlauf,1994). Recently, transient abolition of early colinearity, which is evident during gastrulation, by deletion of a 5′ located Hoxd regulatory region was shown to result in skeletal homeotic transformations in embryos and neonates(Kondo and Duboule, 1999). These results first demonstrated that the concerted colinear control of Hox expression from its onset at primitive streak stages is absolutely required for a correct Hox patterning function.
Most efforts to elucidate the molecular mechanisms that underlie regionalised Hox gene expression have focussed on relatively late developmental stages (reviewed by Deschamps et al., 1999). The few studies addressing the early establishment of the Hox domains have highlighted the difficulty in correlating these early patterns with cell behaviour during morphogenetic movements at gastrulation(Deschamps and Wijgerde, 1993; Gaunt and Strachan, 1994). Hox genes are activated when the primitive streak is almost fully extended, and then in the most posterior (caudal) part of the streak which is generating extra-embryonic mesoderm and not contributing to the embryo proper. The early transcription domains subsequently spread rostrally to reach the anterior part of the streak by an unknown, non-lineage-related mechanism(Deschamps and Wijgerde,1993).
After the early Hox expression domains reach the node region, they continue to spread more rostrally to reach their definitive rostral boundaries in neurectoderm, mesoderm and endoderm in axial and paraxial structures. Previous work suggested that mesoderm acquires its positional information when emerging from the primitive streak (Tam and Beddington, 1987), and Frohman et al.(Frohman et al., 1990)proposed that differential Hox gene expression is established at that moment. The fate map of the presumptive neurectoderm at late gastrulation(Tam, 1989) similarly indicates that the epiblast near the anterior end of the streak contain progenitors of hindbrain and spinal cord; retrospective lineage analysis indicates that spinal cord is laid down at the node, sequentially from anterior to posterior (Mathis and Nicolas,2000). Analysis of the evolution of the early Hox gene expression patterns suggested that the successive rostral boundary regions could be fixed at the node and carried by lineage transmission as the axis was laid down and the node `regressed' (Deschamps and Wijgerde, 1993).
We have investigated the mechanism by which Hox gene expression is initiated and propagated along the streak towards the node, using embryonic explants, and show that the conditions for autonomous Hox expression are already present posteriorly but not anteriorly, at the beginning of gastrulation, more than 12 hours before overt Hox gene expression. We also show that this primed but non expressing posterior tissue can induce Hoxb1 expression in non primed and non expressing anterior streak and epiblast tissue. Second, we examine the possibility that Hox expression boundaries are carried rostral to the node by lineage transmission, and show that this is not so in the neurectoderm, because the precursors that will occupy the future rostral expression boundary region are already anterior to the node when Hox expression reaches the node. No consistent relationship between the Hox gene expression status of cells at the node and the destination of their anteriormost mesoderm descendants was found. We conclude that the rostral progression of Hox gene expression must be modulated by gene regulatory influences from early on in the posterior streak, until the time cells have acquired their stable positions along the axis well anterior to the node.
MATERIALS AND METHODS
Mice
Embryos at E6.5-8.0 (E0.5 is defined as noon on the day of vaginal plug detection after overnight mating) were recovered from (C57BL/6J×CBA) F1 matings, and from a transgenic line containing a Hoxb1-LacZ construct(with 18 kb of genomic DNA from the Hoxb1 locus, described by Marshall et al. (Marshall et al.,1994) on a similar background. This reporter construct perfectly mimics the early and late patterns of the endogenous Hoxb1 expression in mesoderm/neurectoderm and in rhombomere 4(Marshall et al., 1994; Studer et al., 1994). The heterozygous Hoxb1-lacZ embryos analysed were produced by crossing homozygous males with (C57BL/6J×CBA) females.
Embryonic explants
Embryos were isolated from the decidua at the desired gestational stage and Reichert's membrane was removed with tungsten needles(Hogan et al., 1994). Embryos were staged according to morphology (Downs and Davies, 1993), modified for C57BL/6×CBA embryos(Edinburgh Mouse Atlas Project; http://genex.hgu.mrc.ac.uk;K.A.L., unpublished) and according to size(Lawson and Pedersen, 1992). Explants spanning the length of the primitive streak(Fig. 2) were excised with a glass needle while restraining the embryo by the extra-embryonic part with a tungsten needle. The posterior part of the primitive streak (posterior streak region, PSR) was separated from the extra-embryonic tissue at the level of the junction between embryonic and extra-embryonic ectoderm. The anterior part of the primitive streak designated ASR (anterior streak region, see Fig. 2) included the distal part of the embryo, because the anterior part of the streak alone did not grow well in culture. All explants contained the three germ layers and were about 100×100 μm in size, as measured with a micrometer. An intermediate piece between the most proximal and distal pieces of the streak was also taken from older stages ensuring that explants of similar size were taken from embryos of different age. Explants were transferred individually to small depression wells made in bacteriological dishes with a darning needle, covered with 60 μl drops of Dulbecco's modified Eagle's medium (DMEM) plus 15%fetal calf serum (FCS) under mineral oil and further cultured according to Ang and Rossant (Ang and Rossant,1993). Recombinants were made by aggregating two explants. The culture period was 24 hours unless otherwise stated. Growth and survival of the explants were verified by measuring size with a micrometer and viability with Trypan Blue staining.
Embryo culture and cell labelling
The conditions for embryo culture and iontophoretic injection into single epiblast cells were as described(Beddington and Lawson, 1990; Lawson et al., 1991; Perea-Gomez et al., 2001). One epiblast cell/embryo was injected with 7.4% HRP (Horseradish peroxidase,∼1000 U/mg, Boehringer) and 2.6% lysinated rhodamine dextran(103Mr, Molecular Probes) in 0.05 M KCl, as described by Perea-Gomez et al.(Perea-Gomez et al., 2001) for endoderm, except that, for epiblast, depolarising current pulses were applied for 15 to 20 seconds. The fluorescent label served to confirm injection into an epiblast cell and was used to record the position of the cell.
Identification of the position of labelled cells
HRP-containing cells were identified after culture by staining the embryos for 1-1.5 hours with Hanker Yates reagent (Polysciences) as described(Lawson et al., 1991) before fixing with 2.5% glutaraldehyde in PBS, dehydrating, clearing in 1:2 benzyl alcohol: benzyl benzoate (BABB), embedding in glycolmethacrylate (Technovit 1700) and cutting 7 μm serial sections followed by staining with Methylene Blue. The number and position of the labelled cells, in relation to identifiable landmarks along the AP axis, were recorded in embryos in BABB before embedding, and crucial embryonic dimensions noted. The embedded embryos were sectioned in the appropriate orientation to identify the position orthogonal to the midline of labelled cells i.e. DV position in the neurectoderm and whether labelled mesoderm was axial, paraxial or lateral plate.
For comparison of clones in the hindbrain and spinal cord the distance along the midline between the most anterior member of a clone and the boundary between the first and the second somite was measured on a sagittal view of the cleared embryo. The initial axial position of the progenitors was measured along the midline from the anterior junction of epiblast and extra-embryonic ectoderm.
Gene expression
Whole-mount in situ hybridisation was performed as described by Roelen et al. (Roelen et al., 2002).
Digoxigenin-labelled (Boehringer Mannheim) antisense probes were as follows. The Hoxb1 probe was a T7 polymerase transcript from a 800 basepairs (bp) EcoRI fragment(Wilkinson and Krumlauf,1990). The Hoxb8 probes were a 1:1 mix of a SP6 polymerase transcript from a 350 bp 3′ untranslated SacI-KpnI cDNA fragment and a SP6 polymerase transcript from a 420 bp SacI fragment containing the first exon of the gene(Charité et al., 1994). The T/Brachyury specific probe was a T7 polymerase transcript from a 2kb EcoRI fragment (clone pSK75)(Herrmann et al., 1990). The chordin probe is described in Bachiller et al.(Bachiller et al., 2000). Probes were tested on embryos before use on explants.
For β-galactosidase activity, explants were fixed in 1% formaldehyde,0.2% glutaraldehyde (Sigma) in phosphate-buffered saline (PBS) for 5 minutes,rinsed twice in PBS and stained with X-gal as described(Charité et al.,1994).
Statistics
The anterior limits of clones at different stages or in different germ layers at the same stage were compared with the Wilcoxon rank test. The relationship of the anterior limit of clones with progenitor position was obtained by linear regression analysis(Snedecor and Cochran,1967).
RESULTS
Initial activation of Hoxb1 and Hoxb8 in the primitive streak region
The spatiotemporal patterns of activation of the 3′ gene Hoxb1 and the 5′ gene Hoxb8 were compared. Expression of Hoxb1 began in the most caudal part of the primitive streak, at the junction between extra-embryonic and embryonic tissues, at the late midstreak (LMS) stage (E7.0) (Fig. 1A). Dynamic rostral expansion of the transcription domain followed. This domain spread along and lateral to the primitive streak(Fig. 1B-D) and later beyond the node (Fig. 1E,F). Transcription of Hoxb8 started about 12 hours later than Hoxb1 (late neural plate, E7.5, Fig. 1G). The spreading of the expression domains of either Hox gene from posterior to the node region took less than 8 hours. Hoxb1 and Hoxb8 transcripts reached the anterior end of the streak at the late streak early bud (LSEB)/neural plate(NP) stage (Fig. 1D), and early headfold/headfold (EHF/HF) stage, respectively(Fig. 1I). Variation between embryos in the time of arrival of the expression boundary at the node is shown in Table 1 (with Hoxb4included for comparison). During this phase, Hoxb8 transcripts remained mainly restricted to the primitive streak and nascent mesoderm(Fig. 1G-I), while Hoxb1 was transcribed more widely laterally and anteriorly in the mesodermal wings (Fig. 1D-F). This distribution was confirmed in histological sections(Frohman et al., 1990; Deschamps and Wijgerde, 1993). Therefore, late streak nascent lateral and paraxial mesoderm only expresses Hoxb1, while mesoderm born later at the EHF stage expresses both Hoxb1 and Hoxb8. Both ectoderm and mesoderm at the level of the node, but not the axial mesoderm or endoderm express Hoxb1 at the NP stage and both Hoxb1and Hoxb8 at the HF stage.
. | LSOB . | LSEB . | NP . | EHF . | HF . | LHF . |
---|---|---|---|---|---|---|
Hoxb1 | 5 00 | 2 6*0 | 0 10 | 0 04 | 0 06 | 0 01 |
Hoxb4 | 1 00 | 0 31 | 0 05 | 0 07 | ||
Hoxb8 | 4 00 | 4 30 | 0 30 | 0 14 |
. | LSOB . | LSEB . | NP . | EHF . | HF . | LHF . |
---|---|---|---|---|---|---|
Hoxb1 | 5 00 | 2 6*0 | 0 10 | 0 04 | 0 06 | 0 01 |
Hoxb4 | 1 00 | 0 31 | 0 05 | 0 07 | ||
Hoxb8 | 4 00 | 4 30 | 0 30 | 0 14 |
Numbers of embryos at different stages are shown. Regular type, posterior to the node; bold, at the node; italic, anterior to the node.
Anterior expression boundary at posterior edge of node.
Upstream inducing interactions are set much earlier than actual Hox gene expression
Earlier work had made it clear that the spread of the Hox expression domains along the primitive streak was not by proliferative expansion of the initially expressing cell population(Deschamps and Wijgerde, 1993)and did not involve diffusion of inducers from posterior to anterior in the streak (Gaunt and Strachan,1994). In order to investigate whether inducing molecules were involved earlier than at the stages analysed by Gaunt and Strachan for Hoxb4, we analysed the autonomy of expression of one of the earliest Hox genes, Hoxb1, in explants of anterior and posterior streak regions.
We investigated whether Hoxb1 was activated autonomously in embryonic tissues at stages preceding initial gene expression. To do this, we cultured posterior streak region (PSR) and anterior streak region (ASR)(Fig. 2, upper panels)separately at different primitive streak stages, and examined the expression of Hoxb1 and Hoxb1-lacZ. In control experiments, explant culture conditions supported normal Hox and marker gene expression at the different stages, as shown by the maintenance of Hoxb1 and Hoxb8 expression in culture in 100% of PSR and ASR explants from headfold stage embryos (E7.5-7.75): such pieces already express the genes at the time of excision (data not shown). The 24-hour culture period was chosen because this time is sufficient for Hox gene expression to progress from the posterior to the anterior streak region both in vivo and in longitudinally bisected egg cylinders in vitro (Fig. 2, second row and data not shown).
PSR explants excised at different stages between early streak (ES) and late mid streak (LMS) stages expressed Hoxb1(Fig. 3A-D) and Hoxb1-lacZ (not shown) after culture, with the proportion of positive explants rising from 54% for ES explants to 80-100% from the early mid streak(EMS) to the LMS stages (Fig. 3O). Likewise, 80-100% median streak region (MSR) explants from mid streak (MS) and LMS stages had activated the Hoxb1 gene and transgene (Fig. 3E,F,O) after culture. By contrast, ES ASR explants failed to express Hoxb1 and Hoxb1-lacZ (Fig. 3G,O). The proportion of ASR explants expressing the gene rose with increasing age, from 18% at the EMS stage (with a low number of positive cells in this latter case) (Fig. 3H,O), to 36% at the MS stage(Fig. 3I,O), and to 57% for the LMS embryos (Fig. 3J,O). The absence of Hoxb1 expression in the youngest material was not due to inappropriate culture conditions because the explants expressed the Hox-independent genes brachyury (T)(Fig. 3K,L) and chordin(Fig. 3M,N). Hox gene expression could not be induced by increasing the culture period of ES ASR explants to 30-32 hours (data not shown), whereas 100% MS ASR explants were positive after a similar culture period (versus 36% after 24 hours culture,data not shown).
In summary, these results show that Hoxb1 expression can start autonomously in PSR explants cultured from the ES stage onwards. ASR explants were only able to activate Hoxb1 autonomously at the EMS/MS stage and later. This suggests that, early in gastrulation and more than 12 hours before the first Hoxb1 expression appears in the embryo, the underlying molecular genetic interactions have occurred in proximo-posterior embryonic tissues. Alternatively, cell interactions within the explants might set these instructions up in vitro.
The activation pattern of Hoxb8 in explants was also examined. A similar time period (about 12-16 hours) was found to separate permissiveness to `autonomous' activation in explanted tissues(Fig. 4), and effective activation in the intact embryo in vivo(Fig. 1). PSR explants can autonomously activate Hoxb8 expression after culture from the LMS stage onwards (E7.0) (Fig. 4A,B) but not earlier, and the MSR and ASR explants from the late streak early bud (LSEB) stage on (E7.25)(Fig. 4C-F), although Hoxb8 is only expressed in the embryo from late neural plate/head fold (LNP/HF) stages (E7.75) (Fig. 1G-J). The dynamics of activation of Hoxb1 and Hoxb8 in the explant system suggest that, like Hox expression itself,the process which anticipates this expression in the primitive streak takes place sequentially (for the 3′ genes earlier than the 5′ genes) in a proximal (posterior) to distal (anterior) sequence (compiled in Fig. 2, bottom panel).
The proximal posterior region of the primitive streak has Hox-inducing capacity
As the PSR appears to be instructed for Hoxb1 expression before the ASR, and very young ASRs were unable to activate Hoxb1autonomously, we asked whether a PSR explant would induce Hoxb1 in an ES ASR explant when recombined. We combined ASR explants from Hoxb1-lacZ ES embryos with PSR explants from non transgenic embryos at different stages (E6.5 to E7.5) and analysed the lacZ expression after 24 hours of co-culture (Fig. 5A).
Of the ES ASR/PSR recombinates containing PSR explants from ES to HF stage embryos, 92% (11/12) showed lacZ expression, which indicates that Hoxb1 expression was induced in the ASR explant tissues[Fig. 5A (early streak) and Table 2]. Combination with a PSR also strongly increased Hoxb1-lacZ expression in cells of ASR explants from EMS and MS embryos (78% and 100% of positives, respectively),compared with expression seen in ASR explants cultured alone (17% and 27%)(Table 2 and Fig. 3). However, PSR explants were not able to induce Hoxb1-lacZ expression in explants from the extra-embryonic part of ES/EMS embryos, a region that never expresses Hox genes during in vivo development (Table 2). In contrast to the induction observed in ASR/PSR recombinates,no lacZ expression was scored in ASR/ASR control recombinates when the non transgenic ASR was taken from ES to MS embryos (E6.5-E7.0)(Fig. 5B; Table 2). Interestingly though,we observed Hoxb1-lacZ expression in 67% (2/3) of recombinates between transgenic ES ASR explants and the node region of E7.5 headfold stage embryos [Fig. 5B (HF) and Table 2], showing the emergence of a Hoxb1-inducing capacity in late ASRs. The Hox-inductive capacity therefore is present in the PSR early during gastrulation (ES stage, E6.5)until at least the HF stage (E7.5), a time when it can also be identified in the anterior streak region.
Explants and embryonic stage of cutting* . | Proportion of β-gal positive recombinates (number of positives/total number)† . |
---|---|
ES ASR Lac (from Hoxb1-lacZ embryos) | 0 (0/9) |
ES ASRLac + ES PSR | 100 (2/2) |
ES ASRLac + EMS PSR | 80 (4/5) |
ES ASRLac + MS PSR | 100 (1/1) |
ES ASRLac + HF PSR | 100 (4/4) |
ES ASRLac + ES ASR | 0 (0/4) |
ES ASRLac + EMS ASR | 0 (0/2) |
ES ASRLac + MS ASR | 0 (0/1) |
ES ASRLac + HF ASR | 67 (2/3) |
EMS ASRLac | 17 (4/24) |
EMS ASRLac + ES PSR | 60 (3/5) |
EMS ASRLac + EMS PSR | 100 (3/3) |
EMS ASRLac + MS PSR | 100 (1/1) |
EMS ASRLac + ES ASR | 0 (0/4) |
EMS ASRLac + EMS ASR | 0 (0/6) |
MS ASRLac | 27 (6/22) |
MS ASRLac + EMS PSR | 100 (3/3) |
MS ASRLac + EMS ASR | 33 (1/3) |
ES ExtrEmbLac | 0 (0/7) |
ES ExtrEmbLac + ES PSR | 0 (0/3) |
Explants and embryonic stage of cutting* . | Proportion of β-gal positive recombinates (number of positives/total number)† . |
---|---|
ES ASR Lac (from Hoxb1-lacZ embryos) | 0 (0/9) |
ES ASRLac + ES PSR | 100 (2/2) |
ES ASRLac + EMS PSR | 80 (4/5) |
ES ASRLac + MS PSR | 100 (1/1) |
ES ASRLac + HF PSR | 100 (4/4) |
ES ASRLac + ES ASR | 0 (0/4) |
ES ASRLac + EMS ASR | 0 (0/2) |
ES ASRLac + MS ASR | 0 (0/1) |
ES ASRLac + HF ASR | 67 (2/3) |
EMS ASRLac | 17 (4/24) |
EMS ASRLac + ES PSR | 60 (3/5) |
EMS ASRLac + EMS PSR | 100 (3/3) |
EMS ASRLac + MS PSR | 100 (1/1) |
EMS ASRLac + ES ASR | 0 (0/4) |
EMS ASRLac + EMS ASR | 0 (0/6) |
MS ASRLac | 27 (6/22) |
MS ASRLac + EMS PSR | 100 (3/3) |
MS ASRLac + EMS ASR | 33 (1/3) |
ES ExtrEmbLac | 0 (0/7) |
ES ExtrEmbLac + ES PSR | 0 (0/3) |
See Fig. 5 for experimental procedure and legend to Fig. 2for abbreviations. ExtrEmb, explant of extra-embryonic tissues cut on the posterior side. Explants not designated as `Lac' are from non transgenic embryos.
Recombinates were cultured for 24 hours before fixation and X-gal staining procedure.
These results indicate that the posterior part of the primitive streak region is capable of producing signals leading to the induction of Hoxb1 expression in competent anterior streak tissues. Regardless of whether this anterior tissue still represents anterior streak, or is differentiating as neurectoderm and mesoderm, the results imply that rostral spreading of Hox expression from the posterior streak to the node region requires cell-cell interactions.
The spread of Hox expression domains rostral to the node is not lineage related
Rostral spread of Hox expression continues beyond the node(Fig. 1) during a period when the presumptive hindbrain and spinal cord territories are expanding anterior to the node (Tam, 1989) and the node `regresses' while generating spinal cord(Mathis and Nicolas, 2000). Given this coincidence, a plausible mechanism for the continued spread of the expression front along the AP axis is that a cell acquires a Hox code and positional specification while in the node region and its descendants retain it after leaving the node region (Deschamps and Wijgerde, 1993). Descendants remaining (temporarily) in the node region would acquire a new Hox code as the more 5′ genes are expressed there. If this hypothesis is valid, predictions about the final position of the most rostral descendants of cells at the Hox gene expression front at the node at different stages can be made on the basis of the later rostral expression limits of different Hox genes in neurectoderm and mesoderm at E8.5-9.5. Both a general prediction with regard to neurectoderm and mesoderm descendants, and specific predictions can be made.
The general prediction is that the anterior limit of mesodermal clones will be several somite lengths posterior to neurectodermal clones generated in the node region at the same stage (Gaunt et al., 1988; Frohman et al.,1990). Clones labelled with HRP were generated in epiblast at the node region from LS to HF stages (E7.0-E7.7), and the embryos cultured for 1 day (Fig. 6). Some clones were also generated in the axial epiblast anterior to the node at LSEB and older stages. Most (93%) of the clones generated at the node after the LS stage and contributing to neurectoderm were restricted to the ventral half of the neural tube; 78% of the clones contributing to non axial mesoderm anterior to the node were in paraxial mesoderm. The anterior limit of neurectodermal descendants in clones generated at the node was progressively more posterior with advancing initial stage (LS versus LSEB, P<0.01; NP versus HF, P=0.01) with the exception that LSEB and NP did not differ significantly (Fig. 7, upper set). A similar progression was seen in the mesoderm (LS stages versus HF, P<0.01). This trend confirms the sequential addition of neural and mesodermal material from the node region. At no stage was the anterior limit reached by mesodermal clones generated at the node posterior to the anterior limit of neurectodermal clones (Wilkoxon rank test). At the HF stage, mesoderm clones were even slightly more rostral to neurectoderm ones (P=0.05). Clones initiated in the mesoderm layer did not differ from those initiated in the epiblast. Axial mesoderm, which does not express Hox genes in the mouse(Deschamps and Wijgerde, 1993),behaved differently: it remained associated with the node, and therefore relatively posterior (Fig.7). Therefore axial extension from the node progresses at similar speed in neurectoderm and paraxial mesoderm and the general prediction from the hypothesis was not fulfilled.
If Hox codes are acquired at the node, specific predictions about neurectoderm and mesoderm separately are not necessarily dependant on the validity of the general prediction about relative dispersion of neurectoderm and mesoderm from the node region. The specific predictions for neurectoderm,based on later anterior expression boundaries of Hoxb1, Hoxb4 and Hoxb8, and the stage at which expression reaches the node(Table 1), are that the colonising population of rhombomeres 3 and 4 (r3/r4, for the anterior limit of Hoxb1) (Wilkinson et al.,1989) is generated at the node at LSEB/NP stages, that of r6/r7,at the level of the first somite (S1), (Hoxb4)(Gould et al., 1998) is generated at the node at the EHF stage, and that of neurectoderm at the level of S5/S6, in the anterior spinal cord, (Hoxb8)(Deschamps and Wijgerde, 1993;Charité et al., 1998) is generated at the node at the HF stage. Analysis of neurectoderm descendants of epiblast clones generated in the node region (Fig. 7, upper set) and anterior to it (Fig. 7, lower set) showed that the most anterior limit of any neurectoderm clone derived from the node region at LSEB and NP stages was at the level of S1 and S2,respectively (median at S5), and not more anteriorly in r3,r4,r5, as expected from Hoxb1 expression. Contribution to r3,r4,r5 came from the most anterior descendants of clones generated in the node region at the earlier LS stage, and from a region ∼100 μm anterior to the node at the NP stage. Contribution to neurectoderm at the level of S1 by anterior descendants of epiblast near the node came not from LNP/EHF stage embryos as expected from Hoxb4 expression, but from the earlier LS stage and from ectoderm anterior to the node at the LSEB stage. At the HF stage, the most anterior contribution of epiblast at the node region to the neurectoderm was at the level of S7 (median at S9/S10), not at S5/S6 as predicted from the Hoxb8 expression pattern. Contribution to the S5 level by anterior clonal descendants came from the node region of younger LSEB (median at S4/S5)and NP stage (median at S5) embryos as well as from epiblast anterior to the node at NP and HF stages(Fig.7). Therefore cells that will eventually occupy the anterior boundary regions of Hoxb1 and Hoxb4 expression in the hindbrain or Hoxb8 in the spinal cord, are already in positions anterior to the node when the anterior expression boundaries reach the node region(Fig. 8).
The specific prediction with regard to mesoderm is that the precursors of the first somite, which is the anterior limit of Hoxb1 expression in the mesoderm (Murphy and Hill,1991) are present at the level of the node at the LSEB/NP stage,those of S5/6 (Hoxb4) (Gould et al., 1998) at the node at the EHF stage, and those of S10/11(Hoxb8) (Charité et al., 1998; Deschamps and Wijgerde, 1993)at the node at the HF stage. Although the anterior limits of mesoderm clones generated near the node became progressively more posterior with advancing initial stage, there was no general agreement of prediction with the levels at which anterior descendants were found. The anterior limit of two mesodermal clones was in S1 and one was just anterior, as predicted for Hoxb1expression after labelling at the LSEB stage, and two clones derived at the NP stage were in S2, but the anterior limits of the other four clones at LSEB and NP stages were more posterior (median of nine clones at S2). The eight mesoderm clones generated at the HF stage had anterior limits between S4 and S9 (median at S6/7), instead of more posteriorly at S10/11 as expected from the AP level of the definitive anterior expression boundaries of Hoxb8. Therefore, although lineage transmission of Hoxb1expression in paraxial mesoderm has not been excluded, mesoderm generated at the node does not behave as predicted if anterior Hox boundaries were being consistently established by the lineage transmission of specific Hox codes acquired sequentially at the node between LS and HF stages.
Hindbrain and anterior spinal cord elongate both by addition from the node and by internal growth
Sequential addition of material into the hindbrain and spinal cord from the node is indicated by the progressively more caudal position of the anterior limits of clones generated in the node region between LS and HF stages(Fig. 7, upper set). Of the clones contributing to neurectoderm, 45% (14/31) extended to the node, whereas no (0/9) clones generated in the axial region anterior to the node did(Fig. 7, lower set). This is supporting evidence of a self maintaining pool of precursors in the node region for the spinal cord (Mathis and Nicolas, 2000; Mathis et al.,2001) and also for part of the hindbrain.
Comparison of the anterior limits of neurectoderm clones shows a consistency in result in clones generated anterior to the node(Fig. 7, lower set) compared with those generated in the node region(Fig. 7, upper set): the anterior limit is well correlated with the distance from the node of anterior axial progenitors (those anterior to the node and within 55 μm either side of the midline) at all stages. The AP position of the anterior limits of the clones was compared with the position of their anterior axial progenitors by linear regression analysis in order to assess quantitatively whether the neural axis anterior to the node at LSEB to HF stages, representing prospective r3 to spinal cord at the level of S8, is stabilised or is continuing to grow within itself (Fig. 9). The value of the regression coefficient (b) was 2.769(s.e.=0.463, n=9, P<0.001). This value is also significantly greater than 1 (P<0.01), implying that the part of the axis representing precursors of hindbrain and spinal cord to the level of S8, anterior to the node at LSEB to HF stages, increases in length within 24 hours. The axis is therefore extending within itself anterior to the node, and not only at the node by the sequential insertion of new material. In addition,the posterior displacement of the anterior limits of paraxially generated clones relative to anterior axial ones (LSEB and NP, Fig. 7, lower set) suggests that convergence and extension occur in the more lateral ectoderm until at least the NP stage. These results could explain why labelling epiblast just anterior to the node at the HF stage gives clones with anterior limits in the neurectoderm that are more posterior than might have been expected from the shape of the embryo and the space apparently available for the first somites[compare Fig. 1F with Fig. 7 (HF)]. The results also underline the dynamic nature of the relative AP positions of cells that are leaving, and have recently left, the node during a period when Hox expression domains are traversing rostrally through the region.
DISCUSSION
Initiation of Hox gene expression is anticipated by earlier events in the posterior early streak and spreads by cell-cell signalling
Transcriptional initiation of the earliest Hox gene is anticipated by genetic interactions occurring in the posterior streak, much earlier than actual Hox gene expression, perhaps coincident with the generation of the primitive streak. Permissiveness of the presumptive posterior streak region for precociously induced Hoxb1 expression is present at the pre-streak stage(Roelen et al., 2002), also pointing to a link between Hox inducibility and streak formation. Inducibility was present earlier for 3′ than for 5′ genes, sequentially in posterior and anterior streak tissues. The conditions for eventual Hoxb1 expression were not present in the isolated anterior streak of ES stage embryos, and did not develop autonomously in culture. Re-establishing tissue continuity in recombinates restored the expansion of Hox gene inducibility from posterior to anterior tissues. This demonstrates that the mechanism of anterior spreading of Hox expression through the streak(Deschamps and Wijgerde, 1993; Gaunt and Strachan, 1994) (see also Gaunt, 2001) operates by cell to cell signalling, much earlier than suspected, coincident with streak extension. Among possible signalling molecules involved in these interactions,and acting as Hox inducers, are members of the Wnt and Fgf families. Some of these, such as Wnt3 and Fgf8 are strongly expressed in the posteriormost embryonic tissue at prestreak stages, and in the streak at streak stages, and they are required for primitive streak formation and gastrulation(Liu et al., 1999; Sun et al., 1999). Wnt molecules have been shown to induce Hox genes in other contexts in C. elegans (Maloof et al.,1999) and Drosophila(Riese et al., 1997), and loss of function mouse mutants in the Wnt and Fgf pathways exhibit homeotic phenotypes accompanied by decrease in Hox gene expression(Ikeya and Takada, 2001; Partanen et al., 1998). Preliminary data (S.F. and J.D., unpublished) suggest that a Wnt signal is able to induce Hoxb1 in ES anterior explants. In addition, early sequential expression of 3′ to 5′ Hox genes in the primitive streak involves the progressive release of a repression mechanism operating at the level of the cluster from a remote 5′ cis-acting element, as demonstrated for the Hoxd cluster (Kondo and Duboule, 1999).
Neurectoderm cells acquire their Hox code after they have left the node region
The regions of the neural tube that will be positionally specified by the Hox genes are formed from the LS stage onwards by a combination of the sequential addition of cells from the node region(Fig. 7), convergence and extension of paraxial ectoderm anterior to the node(Fig. 7 and data not shown) and subsequent longitudinal extension within the neurectoderm(Fig. 9). Comparison of the clonal behaviour of neurectoderm generated at the node and the spread of Hox gene expression past the node showed that nascent neurectoderm does not acquire and fix its Hox code at the node: cells whose descendants will contribute to definitive anterior boundary regions of the Hox domains (both 3′ and 5′ genes) are already anterior to the node when the waves of Hox expression arrive there. Therefore the early Hox domains have to `catch up' with the cells that will later occupy the rostral boundary domains. In addition, the relative AP positional values in the neurectoderm are changing anterior to the node as the axis elongates and while Hox expression domains are spreading rostrally through this region, implying that positional specification in terms of a stable Hox code must be acquired later. The delay before Hox codes are fixed may correspond to the time required for stabilisation of the relative AP positions of cells in the neurectoderm and mesoderm, when clonal growth in the neurectoderm changes from an AP to a DV mode (Mathis and Nicolas,2000), and when cell mixing stops in unsegmented paraxial mesoderm(Tam, 1988).
Clonal expansion and Hox gene expression in paraxial mesoderm versus neurectoderm
Although the rostralmost neurectoderm descendants of cells at the node when the expression domains arrived there ultimately occupy more posterior positions than the anterior expression boundary of the genes considered, it is not so for the mesoderm. The most anterior mesodermal descendants of epiblast cells near the node when the Hoxb1 expression domain arrived there did, in some embryos, contribute to paraxial positions at or near the anterior boundary region of this Hox gene, and in others were more posterior. By contrast, the most anterior mesodermal descendants of cells near the node when the expression domain of Hoxb8 reached the node were found at positions much more rostral than the expression boundary of this gene. A conclusion from these data is that the regulatory interactions responsible for generating the sequential 3′ to 5′ Hox expression boundaries are different in the neurectoderm and in the mesoderm and must involve down regulation in the mesoderm at least of Hoxb8. This is unsurprising in the light of the recent findings about mesoderm-specific modulation of Hox gene expression in the segmental plate(Zakany et al., 2001). The Hox codes may well be reset by the oscillatory mechanism in the mesoderm descendants of cells near or anterior to the node, after their ingression through the streak.
Axial extension anterior to the node progresses at similar speed in neurectoderm and mesoderm. The offset of rostral expression boundaries of Hox gene expression in the mesoderm compared with the neurectoderm can not be accounted for by germ layer specific clonal distribution of descendants from progenitors around or anterior to the node, but must also result from differential gene regulation including transcriptional induction in the neurectoderm and downregulation in the mesoderm. The anterior progression of the Hox expression domains in both neurectoderm and mesoderm appears to be modulated by gene regulation from early on until at least early somite stages. Transcriptional regulation of Hox genes, although usually studied at later stages, has indeed been shown, in several cases, to depend on germ layer-specific regulatory elements(Gilthorpe et al., 2002; Gould et al., 1998; Marshall et al., 1994; Sharpe et al., 1998).
A single continuous phase of induction of the Hox expression domains between initial transcription and establishment of the rostral expression boundaries
The data suggest that the Hox expression domains are established gradually from the posterior streak to their definitive rostral boundaries anterior to the node, by cell-cell signalling driving transcriptional modulation. Induced by genetic interactions occurring at the time of primitive streak formation and extension, Hox gene expression may continue to spread anteriorly beyond the node under the influence of the same or a similar gene regulation mechanism, and does not rely on proliferative expansion of Hox-expressing cells. The explant and lineage experiments therefore suggest that a single continuing process drives the rostral extension of the Hox domains both posterior and anterior to the node. The node region itself, around which the laying down of tissues along the extending axis is coordinated, would not be specifically involved in instructing newly generated cells as to their AP identity and Hox code, but seems to be passively traversed by the progressing Hox domains.
Although some of the candidate Hox-inducing molecules, such as Wnt3, Wnt3a and Fgf8, could be acting early in the primitive streak, as already discussed,additional inducers might come into action around and anterior to the node during the axial extension phase which we studied. Such possible inducers are Wnt8 (Bouillet et al., 1996),the Cdx transcription factors (van den Akker et al., 2002) and effectors of the oscillatory mechanism of the Notch pathway in the paraxial mesoderm(Zakany et al., 2001). During this period and later, retinoic acid signalling has been shown to sequentially shift the definitive expression boundaries of 3′ to 5′ Hox genes rostrally in the neurectoderm (Marshall et al., 1994; Studer et al.,1994; Gould et al.,1998; Oosterveen et al.,2003). Stabilisation of the Hox expression domains would only take place subsequently, possibly by the epigenetic polycomb and trithorax maintenance system taking over the control of the restricted Hox expression domains, thus putting an end to the rostral spreading of gene expression in the neurectoderm and mesoderm (Yu et al.,1998; Tomotsune et al.,2000; Akasaka et al.,2001). Disruption of the regulatory interactions would, at any stage, lead to altered Hox expression and patterning defects; for example,precocious Hox expression in the primitive streak(Kondo and Duboule, 1999)would result in the disruption of the sequential arrival of the Hox expression domains at the level of the cells to be instructed in the prospective axial and paraxial structures, causing aberrant AP instruction and patterning alteration that cannot be corrected by subsequent regulation.
Acknowledgements
We thank Rob Krumlauf and Heather Marshall for the Hoxb1-lacZtransgenic construct; Wim de Graaff for the generation and maintenance of the Hoxb1-lacZ transgenic mouse line; Jeroen Korving for histology and some in situ hybridisation; Eddy de Robertis, Siew-Lan Ang and Bernard Hermann for probes; Bernard Roelen for initiating S.F. into dissecting early embryos;and Jaap Heinen and Ferdinand Vervoordeldonk for artwork. We also thank Rolf Zeller and his group for three months hospitality (to S.F.) when our laboratory moved. This work was supported by a grant from the French`Association pour la Recherche sur le Cancer' (ARC) (to S.F.) and from the Dutch NWO-ALW (809.38.05) (to J.D.).