Hox-2.3 upstream sequences mediate lacZ expression in intermediate mesoderm derivatives of transgenic mice

In Drosophila, homeobox-containing genes have been identified among the genes controlling pattern formation (reviewed in Gehring, 1987). The discovery of homeobox-containing genes in vertebrate species (Levine et al. 1984; McGinnis et al. 1984) soon stimulated many investigators to set out searching for the function of these genes during mouse development (Colberg-Poley et al. 1985a, 1985b; Hart et al. 1985; Jackson et al. 1985; Joyner et al. 1985; Awgulewitsch et al. 1986; Duboule et al. 1986; Rubin et al. 1986; Wolgemuth et al. 1986; and references in Holland and Hogan, 1988b). Many of the mouse homeobox genes encode proteins that are Antennapedia (Antp)-like. As in Drosophila these genes are clustered. Four clusters of Anip-like homeobox genes have been described in the mouse, all genes within a cluster being transcribed in the same orientation. As soon as sequences could be compared, it became clear that the genes from the four complexes could be aligned, suggesting that these complexes arose by duplication of an ancestral cluster (Hart et al. 1987). Recently, work on sequence comparison, structural organization and differential expression suggested that the vertebrate Hox clusters are true homologs of the insect Antp and Bithorax homeotic gene complexes: in the mouse as well as in Drosophila, a correlation was found between the order of the genes in their cluster and the anteroposterior (AP) position of their rostral expression boundary in the central nervous system (CNS) and in the mesoderm (Gaunt et al. 1988; Duboule and Dollé; 1989; Graham et al. 1989; Dressier and Gruss, 1989); the more 3′ a gene lies in a cluster, the more anterior the rostral border of its expression domain in the CNS and in the mesoderm. In midgestation embryos, the rostral expression boundaries in the CNS are being found at more anterior positions than those in the mesoderm (reviewed in Holland and Hogan, 1988b).

A large amount of work on the expression of mouse Antp-like genes suggests that these genes play a crucial role during pattern formation, like Antp and Bithorax genes do during specification of segment identity in the fly (reviewed in Holland and Hogan, 1988b, and in Dressier and Gruss, 1988). Examination of in situ hybridization results led to the conclusion that the time when Hox gene expression is first detected lies within the interval when position of tissues becomes assigned along the rostrocaudal axis of the embryo (Gaunt, 1987), providing circumstantial evidence that Hox genes specify positional cues for AP axis determination (discussed in Gaunt, 1987). Moreover, experiments by Wilkinson et al. (1989) provided direct evidence for segment-related domains of expression of Hox-2 genes in the developing hindbrain, strengthening the hypothesis that mouse Antp-type Hox genes may have a role in the establishment of positional information during embryogenesis.

Studying the expression pattern of genes throughout development thus can provide clues as to the function of these genes during pattern formation. On the other hand, to study the regulation of the Hox genes themselves both individually and in the context of their cluster is likely as well to lead us towards a better understanding of the gene function(s) and possibly of the mechanisms providing positional signalling along the developing body axis.

We have characterized by in situ hybridization the Hox-2.3 expression pattern in embryos, previously determined by Northern blot analysis (Deschamps et al. 1987b). We also started to characterize the cú-regulatory elements mediating differential Hox-2.3 expression during embryogenesis. Here we describe a Hox-2.3 promoter element capable of driving the luciferase reporter gene in transfected cells and the lacZ gene in transgenic mice. Furthermore, we show that this Hox-2.3 promoter region is able to generate a lacZ staining pattern in transgenic embryos which is very similar to the Hox-2.3 expression pattern in only a subset of Hox-2.3 positive tissues: a regulatory element responsible for expressing the gene in the epithelium of the urogenital system derived from the mesonephric duct is present in the Hox-2.3 5′ flanking region and is thus capable of functioning outside the context of the Hox-2 cluster. The expression pattern of the transgene diverged from that of the endogenous gene in the central and peripheral nervous system, and in somitic and lateral plate mesoderm derivatives, implying that sequences responsible for Hox-2.3 expression in these tissues are located outside the Hox-2.4/Hox-2.3 intergenic region. Further implications for the existence of distinct regulatory elements that mediate Hox-2.3 expression in different embryonic derivatives originating at a similar level along the rostrocaudal axis, and for the genetic organization of individual loci in the clusters, are discussed.

For all embryos used, mid-day after the copulation plug was noticed was taken as day 0.5. Embryos used for in situ hybridization experiments were obtained from a closed outbred colony of Swiss mice. The postimplantation development of this strain lags up to half a day behind that described in Theiler (1972). Embryos used in the lacZ work were produced by mating transgenic animals with C57B16×DBA2 Fl mice. The rate of development of these embryos corresponds to the description in Theiler.

In situ hybridization experiments were carried out on paraffin-embedded tissue sections according to Wilkinson et al. (1987), with the following modifications i) butanol was used instead of toluene before embedding in paraffin wax; ii) 6 pm sections were cut, loaded on a drop of water and dried onto slides which had been pretreated with a 2 % solution of the binding silan TESPA (Sigma) in acetone for 10s, then washed twice in acetone, once in water and dry-baked at 42°C; ii) the hybridization mixture contained 60 HIM dithiothreitol (DTT) instead of 10 mM DTT; iiii) after high-stringency wash and autoradiography, staining was with 0.5 % toluidine blue in water.

The probes used for in situ hybridization were as follows: ³⁵S-labelled antisense RNA probes were synthesized in a direction opposite to that of normal transcription using either T7 or SP6 RNA polymerase. Control (sense) probes gave no specific labelling. One of the three Hox-2.3 antisense probes that have been used in in situ hybridizations experiments was transcribed from a 521 bp sequence extending from the BglH site 43 bp before the end of the homeobox to the BamHl site in the 3′ untranslated region, 98bp upstream from the polyadenylation site (see Fig. 1); a second probe was from the same Ba/riHI site in the 3’ untranslated region to the polyadenylation site 88 bp farther; the third probe was from a 296 bp subclone representing unique Hox-2.3 coding sequences just upstream of the conserved hexapeptide coding sequence (Meijlink et al. 1987), in the first exon. Probes except the shortest one were partially hydrolyzed as described in Wilkinson et al. (1987). No difference in the specificity of the three probes was observed.

Standard recombinant DNA procedures were followed, according to Maniatis et al. (1982). The Hox-2.3 upstream region used in the present work was completely sequenced (Verrijzer et al. 1988, and unpublished but see accession number X06762 from the EMBL Nucleotide Sequence Data Base). Hox-2.3 coordinates are relative to a localized transcription start site (Verrijzer et al. 1988). To generate the Hox- 2.3/luciferase fusion constructs a Sma\ (—1316) to Smal (+207) Hox-2.3 DNA fragment cloned into the pGEM-blue vector was subjected to Bal31 treatment in order to remove the Hox-2.3 ATG and following coding sequences. Sequencing of the resulting clone mapped the 3′ end of the deletion to nt +81. Hox-2.3 sequences (from —1316 to +81) were cut out of the vector using polylinker restriction sites, made blunt by the Klenow DNA polymerase and cloned into the blunt ended Hindi11 site of the promoterless pSV0AL△5′ luciferase containing vector obtained from De Wet et al. (1987). This generated the Sml.3L construct. Sc2L was made by adding Hox-2.3 5′ sequences to the Sml.3L plasmid using the SacI (—2104) to Sad (—169) Hox-2.3 subclone. Deletion from the Sod site (—169) to +81 generated clone Sml.3 DL. S10.2L only contains Hox-2.3 sequences encompassing —207 to +81. pSV2AL△5′ was a construct in which the luciferase gene is driven by the SV40 early promoter and enhancer; this plasmid was generously provided to us by Dr Subramani (De Wet et al. 1987). Control construct LK1.5L was made by cloning the 1.5Kb Kpnl fragment from the clindlts857 lambda phage DNA (Gibco-BRL) in front of the luciferase gene from the pSV0L△5′ vector.

The Hox-2.3/lacZ construct was made the following way: Hox-2.3 sequences between the Smal site (—1316) and the Smal site (+207) were inserted at the blunt-ended Sall site of pMoMuLVnlslacZLTRAEnh. This vector is the same as described in Bonnerot et al. (1987), except for the 3′LTR from which the enhancer was deleted as in Linney et al. (1984).

C1003 EC cells (McBurney, 1976) and the differentiated Fib9 cell line derived from it (A. Piersma and C. Mummery, unpublished) were grown in DMEM/F12 medium supplemented with 7.5% FCS. Retinoic acid (RA;10⁻⁶M) treatment was as described in Deschamps et al. (1987a, 1987b) and was for up to one day starting one day after the DNA had been applied to the cells. The transfection protocol was a modification of the standard calcium phosphate coprecipitation method (Graham and Van der Eb, 1972): per 6cm diameter tissue culture dish, 10μg supercoiled DNA was added to 250 μ1 Hepes-buffered saline solution (42 mM Hepes, 275 mM NaCl, 10 mM KC1, 1.4 mM Na₂HPO₄ and 10 mM dextrose, pH 7.05). While mixing, 250μ1 of 250 mM CaCl₂ was added and the precipitate was allowed to form at room temperature for 25 min. Cells were washed with the Hepes-buffered saline solution diluted twice and directly covered with the DNA co precipitate. After 30 min at room temperature, 4 ml of medium at 37°C was added, and incubation was continued overnight in a CO₂ incubator. The next morning, the DNA containing medium was replaced by fresh medium and incubation was continued for one day before harvesting the cells.

In most experiments, DNA of a lacZ gene under the control of a house keeping gene promoter (hydroxy-3-methylglutaryl-coenzyme A reductase promoter, characterized by M. Meth-ali and R. Lathe, personal communication), was used as an internal control: 5μg of HMG/lacZ DNA (a generous gift from Dr R. Kothary) was coprecipitated with the 5 μg luciferase construct DNA, and /8-galactosidase and luciferase activities were measured in the same extracts.

Luciferase activity was measured in transfected cell extracts according to De Wet et al. (1987) with the following modifications: cells were washed with PBS, and directly harvested in 0.5 ml extraction buffer (100 mM potassium phosphate [pH 7.8], 10mM dithiothreitol) by scraping. Cells were lysed by three cycles of freezing in liquid nitrogen and thawing at 37°C. Membranes and cell debris were pelleted out by centrifugation at 4°C. Protein concentration of each extract was determined using the Biorad assay reagent. Up to 75 μl sample of each extract was added to 265μl of a reaction mixture containing 90μ1 of 100 mM glycylglycine, 90 μl of 20 mM ATP and 54 μ1 of 100 mM MgSO₄ in a small test tube and the tube was placed in a Lumac 3M luminometer. The reaction was started with the injection of 100 μl 1 mM luciferin (Boehringer, Mannheim). Light emission was displayed and recorded every second for 10 s. This way of following the peak light emission and the time course of the reaction proved to be equally accurate as using a chart recorder.

In the cotransfection experiments, β-galactosidase activity was measured in the extracts prepared for the luciferase assay. To 180μl extract, 20μl of a 1M sodium phosphate buffer (pH6.8) containing 0.1M KC1, 10mM MgSO₄ and 0.5M β mercapto-ethanol was added; the reaction was initiated by adding 40 μ of the β-galactosidase substrate analog orthoni-trophenyl-β-D-galactopyranoside (ONPG, Sigma) at 4mgml⁻¹. Incubation was for 1 to 3h at 37°C. The reaction was stopped by increasing the pH with 100 μl of 1 M Na₂CC>3. After a short centrifugation to get rid of any possible precipitated materials, the colorimetric assay was read in a spectrophotometer at a 420 nm wavelength.

Linear Hox-2.3/lacZ DNA insert was excised from vector sequences by Pstl restriction enzyme digestion, separated on an agarose gel and purified by the glass powder method (Vogelstein and Gillespie, 1979). A 6μgml⁻¹ DNA solution was injected into the male pronucleus of zygotes from C57B16×DBA2 F1 females mated to 129/Sv males (about 800 copies of linearized molecules per zygote). In vitro manipulations and oviduct transfers were according to standard methods. Transgenic mice carrying the Hox-2.3/lacZ DNA were identified by tail DNA analysis and bred further to establish transgenic lines. The number of integrated copies of the transgene was determined by Southern analysis using a lacZ probe. As a control, DNA from cells in culture harboring respectively one and three lacZ proviruses was treated in parallel.

Embryos and adult kidneys were stained as whole mounts as follows: fixation was with 4% paraformaldehyde in PBS at 4°C during 20 min (9.5 and 10.5 day embryos) or 60 to 80 min (12.5 and 14.5 day embryos); embryos at 14.5 days and later were first intracardially perfused in order to improve penetration of the fixative. Fixed tissues were washed with PBS during 30 min at 4°C, two to three times depending on their size. Staining was carried out at 30°C usually overnight in a solution of X-gal (Sigma) at a final concentration of 0.4 mg ml⁻¹ made from a 40 mg ml⁻¹ stock in DMSO, with 4IDM K₃Fe(CN)₆, 4mM K₄Fe(CN)₆.6H₂O, 2mM MgCl₂ in PBS.

After staining, tissues and embryos were rinsed with PBS and kept at 4°C in a 30% sucrose PBS solution until being photographed as whole mounts and subsequently frozen in OCT compound (Miles laboratory) for cryostat sectioning at 30 pm. Sections were viewed after further staining by incubating them overnight in X-gal solution at 30°C. No or a very light counterstaining of the sections with phloxine was performed. Staining of non-transgenic embryos and adult kidneys either as whole mounts or in sections never showed any detectable β-galactosidase activity in the organs we were concerned with.

To get insight into the molecular mechanism accounting for Hox-2.3 expression, we set out to test Hox-2.3/ reporter gene constructs and deletions thereof in in vitro cell systems. As the firefly luciferase system was reported to be remarkably sensitive (De Wet et al. 1987), we constructed chimeric transcription units containing the luciferase gene and presumptive Hox-2.3 regulatory elements (Fig. IB). We introduced these constructs by transfection into cells in culture where Hox-2.3 is expressed or inducible by retinoic acid (RA). In undifferentiated EC cells Hox-2.3 mRNA does not accumulate; high levels of Hox-2.3 transcripts are found, however, when these cells are treated with RA (Deschamps et al. 19876; Meijlink et al. 1987). Induction of homeobox gene mRNA accumulation in EC cells is RA-dependent and is not required for cell differentiation (Deschamps et al. 1987a), in keeping with the hypothesis that RA used in vitro may be an effector that mimics, if not is, a physiological inducer of Hox genes in vivo. We used C1003 EC cells, C1003 cells treated with RA, and a differentiated fibroblast-like derivative of C1003, Fib9 (A. Piersma and C. Mummery, unpublished) as we found that the latter constitutively expresses Hox-2.3 at a high level. Luciferase activity was measured two days after transfection (Fig. 1B and Table 1). In C1003 EC cells, construct Sml.3L that contains almost the entire intergenic region between Hox-2.4 (Kongsuwan et al. 1989) and Hox-2.3 (Meijlink et al. 1987, 1989) 5′ of the luciferase gene (Fig. 1A and B) gave rise to about 100 times more luciferase activity than a promoterless construct (pSV0ALA5’) and to about 30 times more activity than a control construct containing a randomly chosen 1.5 kbp of λ phage DNA 5′ of the luciferase gene, all these constructs being derived from the same basal plasmid.

The Sml.3L construct contains the transcription start site previously mapped by SI protection experiments (Verrijzer et al. 1988; Meijlink el al. 1989). A3′ deletion of the Hox-2.3 fragment driving the luciferase gene in the Sml.3L construct - generating construct Sml.3DL - gave rise to a more than 10-fold drop in activity (Fig. IB). A construct containing only the proximal Hox-2.3 portion of Sml.3L (S10.2L) gave virtually no activity (Fig. IB), suggesting that both DNA fragments, — 1316 to —169, and —207 to +80, contain sequences needed for transcription activity. To investigate whether sequences upstream from the Smal site and thus inside the Hox-2.4 transcription unit were involved in Hox-2.3 transcriptional regulation 772 more nucleotides upstream of the Smal site were included in a new construct (Sc2L, Fig. IB). No important increase in transcription was noticed in Sc2L versus Sml.3L (Fig. 1B). Similar results were obtained for all constructs in C1003 EC cells and in Fib9 cells (Table 1).

We wanted to relate the Hox-2.3 promoter activity to that of a standard promoter in the cell systems used, and therefore we tested the SV40 early promoter/ luciferase construct pSV2AL△5′ (De Wet et al. 1987) in parallel with the above described Hox-2.3/lucife rase constructs (Table 1). This made clear that the Hox-2.3 promoter is weak compared to the SV40 early promoter in these experiments.

When luciferase was assayed in extracts of C1003 transfected cells treated for up to one day with RA, enzyme activity was increased by approximatively the same factor two to three for the Hox-2.3/luciferase constructs as for the controls (Table 1); cotransfection with a lacZ construct (kindly provided by R. Kothary) in which lacZ is under the control of the promoter of the ‘house keeping’ gene, hydroxy-3-methylglutaryl-coen-zymeA reductase, (HMG; M. Mehtali and R. Lathe, personal communication), and staining of duplicate culture dishes with X-gal enabled us to establish that RA increased the number of lacZ expressing cells at least threefold. Moreover in similar cotransfection experiments where the luciferase values were normalized using the β-galactosidase catalytic activity measured in the same extracts, no significant difference between Hox-2.3/luciferase activity in the presence or in the absence of RA could be observed (not shown). This leads us to the conclusion that the RA effect observed in the luciferase transfection experiments (Table 1) is due to a facilitation either of DNA transfection or of expression of transfected DNA in general, or both. Given the fact that the RA treatment used induces Hox-2.3 transcripts to accumulate at least 50 to 100 times in C1003 cells (Deschamps et al. 1987b), we conclude that this regulation by RA is not mediated by sequences present on the Hox-2.3/luciferase hybrid genes in this in vitro system.

While a relevant in vitro system can be expected to allow studies to be made on the basal transcription regulatory elements of Hox genes, the hypothetical involvement of these genes in cell position signalling during embryogenesis is to be approached by in vivo experiments such as testing Hox/reporter gene constructs in transgenic mice. Before studying the expression pattern of Hox-2.3/lacZ transgenes in embryos in vivo, we performed in situ hybridization experiments to establish how the Hox-2.3 endogenous pattern is set up and evolves during embryogenesis. We knew from RNA analysis in dissected 13.5-day embryos that gene expression at this stage is the strongest in the spinal cord, starting rostrally at a level posterior to that of Hox-2.1 taken as a control (Deschamps et al. 1987b). Establishment and evolution of the boundaries of Hox-2.3expression along the AP axis within the central and peripheral nervous system and within the mesoderm, as well as a detailed description of the expression pattern in these tissues throughout embryogenesis will be submitted separately (RV and ID, in preparation). Attention in the present communication will particularly focus on Hox-2.3 expression in intermediate mesoderm derivatives where faithful expression of our Hox-2.3/ lacZ construct was detected.

The first sign of Hox-2.3 expression in the nephrogenic area was detected by in situ hybridization experiments in 9.5-day embryos (about 20 somites, Theiler stage 14). At this stage Hox-2.3 expression is restricted along the AP axis: analysis of transverse and sagittal sections revealed that the posterior part of the embryo was labelled, while the cephalic and heart region did not express the gene (not shown). This was true as well in day 10.5 embryos (Theiler stage 16; Fig. 2A,B showing a 32-somite embryo). Expression was clearly detected in the spinal cord and in somitic, lateral plate (stomach primordium) and intermediate (mesonephric) mesoderm. Mesonephric duct and tubule epithelium specifically displayed a high density of silver grains (Fig. 2C,D). At Theiler stage 17 (35–39 somites), the ureteric bud emerges as an évagination from the mesonephric duct and will then start branching to form the collecting tubules of the metanephric or definitive kidney. In 12.5-day embryos (Theiler stage 20), the mesonephric duct and metanephric kidneys gave a very strong Hox-2.3 hybridization signal (Fig. 3A,B). The degenerating mesonephric tubules were less intensely labelled (not shown). Transcripts were localized in the epithelium of the mesonephric ducts and tubules as was the case in 10.5 day embryos, and in the epithelium of the branching ureteric tree (collecting tubules of the metanephros) and ureters (Fig. 3A,B). Fig. 3 also shows that the spinal cord expresses Hox-2.3 at a very high level, the spinal ganglia being labelled as well although at a lower level. One day later (Theiler stage 21), the ureteric bud induces the surrounding mesenchyme to condense and form the comma-shaped and s-shaped bodies (reviewed in Rugh, 1968 and Saxen, 1987). High amounts of Hox-2.3 transcripts were observed in the collecting tubules in 14.5 day embryos (Theiler stage 22) (Fig. 4A-D). We cannot exclude that weak Hox-2.3 expression also occurs in the comma-shaped and s-shaped tubuli. The epithelium of the mesonephric (Wolffian) ducts and of the ureters gave a high Hox-2.3 signal, both in 12.5- and 14.5-day embryos. Evidence for Hox-2.3 expression in the Müllerian - paramesonephric - duct (future oviduct and uterus in the female) was obtained with 14.5-day sections containing both Wolffian and Müllerian ducts (as shown in Fig. 4A,B). High Hox-2.3 expression was observed all along the spinal cord, as was the case in 12.5-day embryos, while the 14.5-day spinal ganglia gave a lower signal (not shown). High expression of Hox-2.3 persists in adult kidneys and spinal cord as has been reported earlier (Meijlink et al. 1987).

In order to test whether Hox-2.3 upstream sequences containing a S1-mapped transcription start site and shown to possess promoter activity in vitro are able to mediate gene transcription in vivo, and are involved in differential gene regulation during embryogenesis, the sequences between Hox-2.4 and Hox-2.3 (Fig. 1A) were coupled to the E. coli lacZ gene and used to generate transgenic mice. The lacZ gene used was a modified version encoding nuclear-targeted β-galactosidase (Kalderon et al. 1984; Bonnerot et al. 1987) (Fig. 1C and see Materials and methods). Six transgenic founder animals were identified by Southern blotting experiments using tail biopsy DNA (not shown) and were bred further. Four of these transgenic lines did not express the transgene. Transgenic offspring from line 23 (harboring two copies of the transgene) and from line 21 (carrying about ten copies of the transgene) were analyzed extensively for β-galactosidase staining at various stages of embryogenesis. From whole-mount staining of mid-gestation embryos, it was obvious that Hox-2.3/lacZ expression was very strong in the developing kidneys (Fig. 5A showing a 14.5-day embryo from line 23).

β-galactosidase-positive tissues were detected around day 9.5 (Theiler stage 15) along the nephrogenic cord (Fig. 5B showing a 25-somite embryo, and 5C showing a 30-somite embryo from line 23). Labelled cells were found ventro-lateral to the somites, from the level of the anterior limb bud on to more caudal regions. In day-10.5 (Fig. 5D) and day-11.5 (Fig. 5E) embryos, the mesonephric (Wolffian) duct, the ureteric bud and developing metanephros, and the ureters were clearly stained. At day 12.5 and 14.5 p.c., (Theiler stage 21 and 23, respectively) embryos possessed stained cells in the epithelium of the mesonephric ducts, in that of the branching ureteric tree (metanephric collecting tubules) and in that of the ureters (Fig. 5F and G). Ureters and mesonephric ducts were labeled all the way down to their junction with the urogenital sinus (future bladder). Examination of histochemically stained tissue sections from 14.5 day old transgenic embryos (Fig. 5K,L) led us to the conclusion that the epithelium of the collecting tubules, mesonephric ducts and ureters was selectively labelled, whereas the epithelium induced from the surrounding mesenchyme by the ureteric buds (comma- and s-shaped bodies which will develop into renal tubules, and glomerular (Bowman’s) capsules [reviewed in Saxen, 1987]) did not show any fi- galactosidase activity (Fig. 5K,L). No lacZ expression was observed in the Müllerian ducts. Examination of this and other sections allowed us to conclude that expression of the transgene takes place exclusively in the epithelium derived from the mesonephric duct. From embryonic day 15 on and in postnatal male mice the epididymis, vas deferens and seminal vesicle (all derived from the mesonephric duct) were positive for lacZ expression (Fig. 5H,I). Adult kidneys were stained as whole mounts, revealing a strong signal along the collecting tubules running from the pelvis radially throughout the medulla and cortex (Fig. 5J). Hox-2.3/ lacZ expression in the urogenital system was thus maintained throughout embryogenesis and in transgenic adults. A pattern identical in time and space was observed in the urogenital system for the second transgenic line analyzed (not shown, and Fig. 51 as an example). The level of Hox-2.3/lacZ expression in line 21 was lower than in line 23 in all positive tissues. Variation of the expression level of a transgene in independent lines, irrespective of the number of integrated copies is a well-known phenomenon.

Aside from the β-galactosidase staining pattern that reflected endogeneous Hox-2.3 expression in mesonephric duct-derived structures of the urogenital system, β-galactosidase-positive cells were also visible in the nervous system, although according to a pattern different from that of Hox-2.3. In embryos from line 23 at 11.5 days, lightly stained cells were visible in the mesencephalon, in the anterior and posterior myelencephalon and along, as well as in the neural tube. At day 12.5, embryos from this line showed weak transgene expression in the nervous system as described for day 11.5. X-gal-stained sagittal and transverse sections enabled us to analyze the distribution of the stained cells in the spinal cord: two longitudinal columns of labelled cells were observed to run parallel to the spinal canal all along the spinal cord. The dorso-ventral position of these columns of cells was medio-lateral to the spinal canal (not shown). In day 14.5 embryos, label was still observed in the spinal cord, apparently decreasing posteriorly to the cervical region. All spinal ganglia were lightly stained at this stage (see Fig. 5A). In embryos from line 21, lacZ expression at 11.5 day is detected in a few widespread cells in the mesencephalon while rare labelled cells were also detected in the spinal cord.

Transgene expression that is consistent neither with Hox-2.3 expression nor between lines 23 and 21 probably reflects an effect of the integration sites of the transgene: in line 23, label was also detected at the basis of the third visceral arch and later in the epithelium of the olfactory pits. In line 21, a few β-galactosidase-positive cells were present on the surface of the muzzle in 14.5 day embryos. As position effects are frequently observed on the expression of transgenes (Allen et al. 1988; Khotaryeia/. 1988; Gossler et al. 1989), it was not surprising to observe some integration site-dependent Hox-2.3/lacZ expression in our transgenic mouse lines.

Like all the mouse homeobox genes of the Antp type studied so far, Hox-2.3 is expressed according to a temporally and spatially restricted pattern in the embryonic central and peripheral nervous system and in derivatives of the axial, lateral plate and intermediate mesoderm (Deschamps et al. 19876; Graham et al. 1989; this paper; RV and JD, unpublished results).

We addressed the question of the molecular basis of this differential genetic control, since identification of regulatory elements influencing individual Hox genes within a cluster might shed some light on their interrelationship and on their function. We demonstrate in this communication that Hox-2.3 upstream sequences containing a S1-mapped transcription start site and encompassing 1.3 kbp 5′ to this site are active in driving transcription of a reporter gene in transfected cells and in transgenic mice.

These Hox-2.4/Hox-2.3 intergenic sequences contain a promoter mediating transcription of the luciferase gene in transfected cells in vitro. A similar level of transcription was observed in cells expressing (Fib9) or not expressing (C1003 EC) Hox-2.3 at a high level. As no difference could be detected either upon measuring luciferase in EC cells with or without RA, we conclude that the Hox-2.3 upstream sequences present on the luciferase constructs might define a promoter that is constitutively active at a low level in a transient assay in the in vitro cell systems used thus far.

The use of a chimeric transcription unit composed of Hox-2.3 upstream sequences encompassing this promoter, and the bacterial lacZ gene to generate transgenic mouse lines, has enabled us to demonstrate that this Hox-2.3 region is responsible for differential gene expression in vivo. The Hox-2.3 sequences used contain cis-acting control sequences mediating selective lacZ expression in the epithelium of the urogenital system derived from the mesonephric duct, a site of high Hox-2.3expression. These sequences are likely to be responsible for the regulation of Hox-2.3 in the corresponding mesodermal tissue. The fact that the expression pattern of the transgene differs from that of the endogenous gene in somitic and lateral plate mesoderm and in neurectoderm suggests that the transgene is lacking sequences that account for Hox-2.3 expression in these tissues. A conclusion from these observations is that distinct regulatory elements are used in different subsets of cells expressing Hox-2.3.

The vertebrate urinary apparatus arises from the intermediate mesoderm, as a nephrogenic column that shows a markedly metameric character. The most anterior part, the pronephros remains very rudimentary and without functional significance. The mesonephros, lying caudally to the pronephros, is composed of a number of renal vesicles, the nephrons, connected to the mesonephric or Wolffian duct. The metanephros, which will replace the mesonephros as functional excretory organ, arises partly from the mesenchyme of the posterior part of the nephrogenic cord, and partly from the ureteric bud growing out from the mesonephric duct and subsequently branching extensively to form the collecting tubules. The metanephric excretory units, the nephrons, develop from the metanephric mesenchyme after an inductive signal has been emitted by the tips of the branching ureteric tree (reviewed in Hamilton and Mossman, 1972, and Saxen, 1987). In the male embryo, the mesonephric duct, originally concerned with urinary excretion, partially degenerates and eventually becomes part of the genital tract - epididymis, vas deferens, seminal vesicle. In the female embryo, the primitive excretory duct undergoes degeneration and does not contribute to the reproductive tract.

In the meso- and metanephros, Hox-2.3 and Hox-2.3/lacZ appear to be expressed at a high level in the ureter cell lineage: the epithelium of the mesonephric ducts and their derivatives, ureters and collecting tubules of the metanephros. Although some expression of Hox-2.3 in the epithelium induced from the metanephric mesenchyme cannot be excluded, the level would be weak compared to that of the N-myc protooncogene, the transcription of which has been specifically associated with early differentiation stages of the induced tubules (Mugrauer et al. 1988).

One can only speculate about the physiological meaning of persistent Hox-2.3 expression in mesodermal cells derived from the mesonephric duct from 9.5 day on throughout development and in adults. In agreement with the proposition that homeobox gene products may serve as positional cues during vertebrate development (for recent reviews, see Holland and Hogan, 19886, Dressier and Gruss, 1988, and Akam, 1989), Hox-2.3-expressing cells belong to structures, the origin of which lies in the same rostrocaudal domain along the anteroposterior axis. The nephrogenic cord arises at a level within the Hox-2.3 expression domain defined in the spinal cord and somitic mesoderm, as the mesonephric tubules appear at a level posterior to somite 11 in a 24 somite embryo, and the metanephros at the level of somite 26 in 34 somite embryos (Torrey, 1943), and as the most rostral somite to express Hox-2.3 is somite 11 or 12 at similar stages of development (RV and JD, unpublished results). Hox-2.3/lacZ positive cells might have received a positional signal early during embryogenesis, with which Hox-2.3-expression would correlate; gene expression would be maintained through the subsequent cell generations either as a nonerased memory of the original ‘label’, or because Hox-2.3 plays an additional and independent role in the urogenital epithelium in embryos and in adults.

Hox gene expression in the kidney is a recurrent observation: in all cases where Hox transcript level and localization have been analyzed by in situ hybridization in meso- and metanephros, cells in both these structures have been shown to be positive for mRNA accumulation (Holland and Hogan, 19886 as a review; Dressier and Gruss, 1989; Dollé and Duboule, 1989; Galliot et al. 1989; Bogarad et al. 1989). This situation is not unexpected since the rostral boundaries of Hox gene expression in the mesoderm occupy an AP position anterior to, or within the meso- and metanephric region (Duboule and Dollé, 1989; Dressier and Gruss, 1989). Interestingly, comparing the anterior expression boundaries of Hox-2.3 (this paper), Hox-5.2 and Hox-5.3 (Dollé and Duboule, 1989) in the 12.5-day mesonephric tissue makes clear that these boundaries are displaced relatively to each other along the AP axis, as are the boundaries found in the somitic mesoderm. This observation suggests an involvement of Hox gene expression in positional signalling within the mesonephric duct derivatives. Interesting as well is the fact that the cell type where Hox transcripts accumulate in the meso- and metanephros also differs depending on the gene: for instance, genes like Hox-1.4 (Galliot et al. 1989), Hox-1.5 (Gaunt, 1988), Hox-5.3 (Dollé and Duboule, 1989) are expressed homogeneously in the meso- and metanephric mesenchyme with no enhancement in tubules, whereas Hox-2.1 (Jackson et al. 1985; Krumlauf et al. 1987; Holland and Hogan, 1988a), Hox-5.2 (Dollé and Duboule, 1989) and Hox-2.3 (this paper) are expressed particularly strongly in tubular epithelium.

Another major site of Hox-2.3 expression from 8.5 day on throughout embryogenesis is the rostrocaudally defined domain in the nervous system, the posterior myelencephalon, the spinal cord and the dorsal root ganglia being positive for transcript accumulation in mid-gestation embryos (Deschamps et al. 19876; Graham et al. 1989; this paper; RV and JD, in preparation). The transgenic lines examined did not give a j5-galacto-sidase-staining picture similar to the Hox-2.3 expression pattern in these structures. The conclusion is that Hox-2.3 gene expression in the nervous system operates via a mechanism involving (additional) sequences located outside the transgene. One possibility is that a neurospecific element is required for proper high expression in the nervous system and is not present on the transgene. Alternatively, sequences responsible for the establishment of the AP-restricted Hox-2.3 expression domain in both neurectoderm and mesoderm might not be present on the transgene, and expression of Hox-2.3 and Hox-2.3/lacZ in the mesonephric duct-derived epithelium would correlate with an independent tissuespecific function of Hox-2.3 in these structures.

An interesting observation results from the comparison of the j5-galactosidase-staining pattern of the two mouse lines examined in detail: localized expression of Hox-2.3/lacZ in a subset of cells widespread in the mesencephalon of both families of embryos suggests that this feature is inherent to the transgene. This was not due to the reporter gene, as a number of lacZ transgenic lines with other promoters do not show β-galactosidase-positive cells in these structures (CB and JFN, unpublished results). This property might result from the insensitivity of the transgene to a negative regulation normally modulating Hox-2.3 expression in mesencephalon cells, as no Hox-2.3 transcripts accumulate in the mesencephalon that is located outside the Hox-2.3 expression boundaries. The fact that primary cultures of embryonic mesencephalon cells respond to RA treatment by an abundant accumulation of Hox-2.3 transcripts whereas, for example, limb bud mesenchymal cells do not (Deschamps et al. 1987b) is in keeping with Hox-2.3 being turned on in the CNS by a mechanism different from that in certain other tissues. This hypothesis of the absence of a negative regulation in mesencephalon cells may also apply to the weak expression of the transgene in the spinal cord and spinal ganglia that also differs from the endogenous pattern.

Altogether, we conclude from our Hox-2.3/lacZ experiments that regulatory sequences selectively mediating Hox-2.3 expression in derivatives of the intermediate mesoderm are present on the transgene, and that sequences involved in Hox-2.3 expression in cells lying at a similar rostrocaudal level in the spinal cord, in spinal ganglia and in somitic and lateral plate mesoderm derivatives are located elsewhere. Therefore, the regulatory mechanism leading to Hox-2.3 expression in cells derived from the intermediate mesoderm would not be identical to that employed by cells from somitic and lateral plate mesoderm and from the neurectoderm. A similar conclusion as to the existence of distinct regulatory elements upstream of an Antp-type Hox gene was drawn by Zakany et al. (1988), who showed that Hox-1.3 most proximal 912 bp mediate expression of a Hox-1.31lacZ transgene in a specific region of the embryonic spinal cord, while expression in the Hox-1.3-positive mesodermal tissues (Dony and Gruss, 1987; Dressier and Gruss, 1989) does not take place. An important difference between the situations concerning Hox-1.3 and Hox-2.3, is that Hox-2.3/lacZ expression is observed in the urogenital system of the transgenic embryos/mice whereas the Hox-1.3/lacZ transgene is transcribed in the spinal cord. This difference reveals that the regulatory element(s) directing gene expression in the spinal cord, positive for all Hox genes, is (are) not always located directly upstream of the gene to be controlled; in the case of Hox-2.4 and Hox-2.3, such an element might be located upstream of, or within Hox-2.4 and influence both genes. It is also possible that the putative ‘kidney element’ present upstream of Hox-2.3 controls several genes of the cluster. Future experiments will test these speculations.

An analysis of the spatial and temporal activity of the promoter of Hox-1.1, the homolog oí Hox-2.3 in cluster 1, has recently been performed using Hox-1.1/lacZ transgenic mice (Püschel et al. 1990). 3.6 kbp sequences upstream and 1.7 kbp sequences downstream from Hox-1.1 were fused to lacZ and shown to direct gene expression according to a pattern identical to that of Hox-1.1 at early stages of development (7.5 to 8.5 days). At later stages, the patterns diverged in their caudal expression boundary and in the tissue distribution of the transcripts. Clearly, the Hox-1.1 sequences present on the construct are endowed with a regulatory potential different from that of the 1.3 kbp Hox-2.3 upstream region. For a direct comparison between Hox-1.1/lacZ and Hox-2.3/lacZ, differences in the constructs should be born in mind which might be causally related to the difference in expression pattern of the transgenes: Hox sequences present on Hox-1.1/ lacZ extend farther upstream than those on Hox-2.3/ lacZ do, and they contain intragenic and downstream sequences. In addition, the fact that there is no counterpart to Hox-2.4 in the Hox-1 cluster (Duboule and Dollé, 1989; Graham et al. 1989) makes it very likely that the 5’ flanking regions of the two members of the Hox-1.1 subfamily extensively differ from each other. Duplication of the putative ancestral gene cluster (reviewed in Akam, 1989), as it gave rise to alterations within the clusters, thus might have provided the gene family with an additional variety of regulatory mechanisms, possibly enriching the repertoire of cellular instructions during embryogenesis.

We wish to thank Drs A. Graham and R. Krumlauf for introducing us to the in situ hybridization technique. C. K. is thankful to Dr M. Rassoulzadegan for teaching her the art and skill of generating transgenic mice. We also acknowledge Dr S. Subramani for sending us the luciferase clones, Drs M. Mehtali and R. Lathe for allowing us to use their cloned HMG promoter and Dr R. Kothary for providing us with the HMGlacZ plasmid. Many thanks are due to M. Hameleers for constructing the Sml.3L plasmid, and to Dr K. Lawson for stimulating discussions and for reading and criticizing the manuscript. We also would like to thank Dr P. Ekblom and Dr G. Mugrauer for their help in interpreting the gene expression patterns in developing kidney. This work was supported by grants from the ‘Association Française contre les Myopathies’, the ‘Association pour la Recherche sur le Cancer’, and the ‘Institut National de la Recherche Médicale’ to J.-F.N.