While constructing a cDNA library of human embryos, we have isolated a clone homologous to jumonji, a mouse gene required for neural tube formation. We have determined the complete coding sequence of the human homologue (JMJ) and deduced the amino acid sequence of the putative protein. We show here that human and mouse jumonji putative proteins are homologous and present 90% identity. During human embryogenesis, JMJ mRNAs are predominantly expressed in neurons and particularly in dorsal root ganglion cells. They are also expressed in neurons of human adult cerebral cortex. In view of these observations, we propose JMJ as a candidate gene for developmental defects of the central nervous system in the human. The human JMJ gene maps at position 6p24–6p23.
Although the genes involved in numerous human inherited diseases have now been cloned, only a few are known which are associated with malformations occurring during embryogenesis. One reason is likely to be the relative lack of information available on normal human embryogenesis. This is particularly true for gene expression during this process. The studies of developmental processes in lower organisms (such as Drosophila and mouse) have contributed significantly to the understanding of human birth defects (reviewed in 1) and, thus, the characterization of human genes homologous to genes in lower species involved in abnormal development remains an important goal. That is why we have undertaken the construction (2) and normalization (3) of a human embryonic library. We have sequenced several clones from our library and isolated a clone (clone Y4) exhibiting a high degree of similarity with a mouse gene (named jumonji) that is required for normal morphogenesis of the neural tube (4). The mouse jumonji gene is conserved among vertebrates and shows homology with the human retinoblastoma binding protein-2 (5) and with the putative protein encoded by the human gene XE 169 (6) and thus with the mouse homologues smcx and smcy (7,8). Neural tube defects are among the most common malformations in mammals; we thus decided to characterize further the human homologue (JMJ) of the mouse jumonji gene. We have determined the complete nucleotide sequence of the clone Y4 and further deduced the sequence of the complete coding region of JMJ mRNA. The human and mouse putative proteins are highly homologous. The pattern of expression of the JMJ gene was also analysed using in situ hybridization. We observed a prevalent expression in differentiated neural tissues. Human JMJ maps at position 6p23–6p24.
Cloning of the human jumonji cDNA
We have constructed a cDNA library of human embryos collected after elective termination at 10–12 weeks of embryonic development (P. Jay, J. L. Bergé-Lefranc, in preparation). Several randomly selected clones were sequenced and compared with database sequences using FASTA (9) and BLAST (10) programs. A 1300 bp cDNA clone (clone Y4) homologous to mouse jumonji cDNA was identified. Comparison of the sequences of the clone Y4 with those of the mouse jumonji mRNA indicated that this clone was homologous to the 5′ end of jumonji mRNA including the first ATG sequence that was assigned as the initiation codon in the mouse sequence. In order to characterize further the human JMJ mRNA, we searched the human EST databases for clones exhibiting sequence homology with the 3′ portion of mouse jumonji mRNA. We observed the presence of these sequences in an EST of a human cDNA library prepared from fetal spleen and fetal liver tissues (Clone 239750 IMAGE consortium) (11). Two primers were prepared in order to PCR amplify the remaining 3′ coding sequences of JMJ mRNA. The sequence of the 3′ primer was designed from the sequence of the homologous EST, and the sequence of the 5′ primer was deduced from the sequence of the clone Y4. RT-PCR reactions were performed and the sequence of the amplified fragment (3 kb long) was determined. Analysis of the composite sequence indicated the presence, in JMJ mRNA, of a 3600 nucleotide open reading frame (ORF). The human and mouse ORFs show 87% nucleotide identity and 90% identity of the putative peptides (Fig. 1). Two regions, in the mouse protein have significant homology with human retinoblastoma binding protein-2 (5) and with the protein encoded by human gene XE 169 (6) and these are conserved in human jumonji protein (amino acids 544–1044 in human jumonji peptide sequence) exhibiting 96% identity.
Expression profile of the human jumonji gene
The cDNA clone Y4 was hybridized to Northern blots of various adult and fetal human tissues. Figure 2 illustrates the results obtained with fetal tissues. A single faint band of ∼6.5 kb was detected in all the tissues tested, indicating ubiquitous expression of the JMJ gene. Similar results were obtained on Northern blots of adult tissues (not illustrated). In order to analyze the expression profile of JMJ more accurately, we determined its expression using in situ hybridization (Fig. 3). Sense and antisense probes were prepared from clone Y4 and hybridized to frozen sections of embryos and adult brains. As expected from the results of Northern blot experiments, a low level of expression was detected in all the tissues analysed. However, in slices of embryos, a significantly high level of expression was observed specifically in dorsal root ganglia neurons (Fig. 3a–c), whereas spinal cord neuroepithelium did not exhibit any specific labelling, nor did cartilage tissue of vertebra anlage. The high level of expression was maintained in the cerebral cortex neurons from adult human brains (Fig. 3d): JMJ mRNAs were detected in neuronal cell bodies and in the proximal part of apical dendrites. Nuclear areas always appeared unlabelled.
Chromosomal localization of the human jumonji gene
JMJ was mapped using in situ hybridization of a human jumonji cDNA clone on chromosome preparations (12). In the 100 metaphase cells examined after in situ hybridization, there were 288 silver grains associated with chromosomes and 55 of these (19.1%) were located on chromosome 6; the distribution of grains on this chromosome was not random: 37/55, 67.3% of them mapped to the p24–p23 region of the short arm of chromosome 6 (Fig. 4).
We report here the characterization of a human cDNA homologous to the mouse jumonji cDNA. Convincing data indicate that the jumonji gene is required for the normal morphogenesis of the neural tube in mouse embryos and that jumonji is conserved among vertebrates (4). The characterization of the human homologue of such a gene is therefore important, since mutations in JMJ may underlie defects in human development. The amino acid sequence of the jumonji putative product is conserved between human and mouse species, leading to the assumption that human jumonji protein is involved in neural tube formation in human embryos. Only small differences were observed between human and mouse jumonji gene products. At position 267–278, eight amino acids out of 11 are different between mouse and human. At position 432–437, six amino acids are deleted in mouse when compared with the human sequence. Two other small differences are located ∼50–100 amino acids from this region. If we exclude the above regions, the identity between mouse and human jumonji putative proteins reaches a mean value of 95%. This value was observed in two regions of mouse jumonji putative protein that are homologous to the human retinoblastoma binding protein-2 (5) and to the product of human XE 169 gene (6). This high level of sequence conservation suggests that human and mouse jumonji proteins have a similar function. However, our results strongly suggest that the function of the jumonji product is not restricted, at least in humans, to the formation of neural tube: (i) using northern blot experiments, we identified the expression of JMJ in all the adult and fetal tissues tested. (ii) we have analysed the expression of JMJ in 10- to 12-week-old human embryos using in situ hybridization. We have observed that JMJ was strongly expressed in differentiated neural tissues (i.e. dorsal root ganglion neurons) but not in less differentiated neural tissues such as the embryonic spinal cord. This pattern of expression can be related to that observed in mouse embryos: Takeuchi et al. (4) reported that jumonji expression coincided with the timing of migration and functional differentiation of granule cells. These results indicate that high levels of jumonji gene expression coincide with neuronal differentiation and could suggest a selective expression in neural crest derivatives. However, this is deduced from the analysis of a single stage of development, and the availability of human embryos from different stages will be required to provide a more complete analysis of JMJ expression, as has already been performed in the mouse. Results from northern blot experiments do not indicate a high level of expression of the jumonji gene in fetal and adult brains. However, using in situ hybridization, we detected this expression in neurons of the cerebral cortex of adult brain. This indicates that JMJ expression persisted in adulthood at least in cerebral cortex neurons. Analysis of the function of the jumonji product in neuronal differentiation is complicated by homozygous lethality (4). The existence of human pathological models should provide insights into jumonji function during neuronal differentiation.
The human JMJ maps on the short arm of chromosome 6 at position 6p24–6p23, where linkage to morphological defects has been suggested. Linkage of a non-syndromic orofacial cleft to 6p23 was first described by Eiberg et al. (13), and similar findings have been reported in more recent studies (14,15). Previous work suggested the linkage of two other malformations to the HLA locus on distal 6p: an atrium defect of secumdum type (16) and paralysis of the laryngeal adductor showing autosomal dominant inheritance (17). Morphological aberrations related to neural tube closure defect, such as spina bifida, have also been linked to the HLA locus (18). Although this linkage has not been observed consistently, the multifactorial inheritance of neural tube defects is now generally accepted (reviewed in 19). Our results establish JMJ as a candidate gene for some of these defects. However, the absence of definitive data linking such defects to markers in 6p means that this conclusion must be viewed with some caution. Whatever the involvement of JMJ in these specific cases, its predominant expression in differentiated neural tissue of embryo and adult and the observed effects of the null mutation generated in mouse jumonji strongly suggest that JMJ is required for the normal development of neural and neuronal tissues.
Materials and Methods
Human embryos ranging from 10 to 12 weeks post-ovulation were obtained from elective terminations at Hôpital Carremeau, Nimes. The embryonic tissue was only used with maternal consent. Human adult temporal cortex was obtained during surgical treatment of focal epilepsy. Our planned research has been approved by the CNRS ethical committee (COPE of CNRS, April, 30, 1993).
cDNA library screening and characterization of human jumonji cDNA
A cDNA library was constructed from the embryos using random unidirectional primers (Stratagene). The cDNA was cloned in λ-ZAP vector according to the manufacturer's recommendations. A cDNA clone containing sequences homologous to the 5′ end of mouse jumonji mRNA was obtained after large-scale sequencing of cloned cDNA. The 3′ sequence of mouse jumonji cDNA was compared with those of the more recent human EST databases. A homologous EST was found. Two primers were prepared: a 3′ primer (5′-TTAGAGCTAGTGCAAAGACAGCT-3′) designed from the sequence of this EST and a 5′ primer (5′-TGACGAAGGGGGCTGTCACAT-3′) whose sequences derived from that of the cloned human jumonji cDNA. These primers were used to amplify by RT-PCR the 3′ region of human jumonji mRNA. The amplified fragment was sequenced directly by primer walking, or cloned in TA cloning vectors and further sequenced. Each strand of the fragment was sequenced at least once using both methods.
Northern blot analysis and in situ hybridization
Human adult and fetal Northern blots obtained from Clontech were hybridized to the human jumonji cDNA clone, washed and exposed according to the manufacturer's recommendations.
A probe for in situ hybridization experiments was generated by shortening the size of the Y4 clone. Briefly, two primers were designed from the sequence of this clone (positions: 899–919 and 1251–1271 in the JMJ cDNA sequence). A 372 bp fragment was amplified by PCR and inserted in both orientations in pGEM 3 vector (Promega, Madison). cRNA probes and in situ hybridization were performed as follows (20): antisense and sense cRNA probes were produced by in vitro transcription using SP6 RNA polymerase after linearization of the plasmid DNAs using the restriction enzyme XbaI. cRNA probes were labelled with digoxigenin-11-UTP (Boehringer Mannheim) as suggested by the supplier. Labelling efficiency was estimated using serial dilutions of the labelled probes spotted and fixed onto a nylon membrane (21). After detection using polyclonal sheep anti-digoxigenin Fab fragments conjugated to calf intestinal alkaline phosphatase, spot optical densities (OD) were measured using an image analysis system. The amount of incorporated digoxigenin-11-UTP was compared with a standard labelled RNA (Boehringer Mannheim).
Cryostat serial sections (14 µm) of human tissues were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4 at room temperature for 30 min, then rinsed in the same buffer. They were dehydrated in ethanol, then Nissl stained or stored at -80°C prior to use. Pre-hybridization and hybridization were performed as already described (20). Detection was carried out using polyclonal sheep anti-digoxigenin Fab fragments, conjugated to calf intestinal alkaline phosphatase (Boehringer Mannheim) diluted 1:500. Colour development was performed using nitroblue tetrazolium salt and 5-bromo-4-chloro-3-indoylphosphate as suggested by the supplier (Boehringer Mannheim). Slides were then mounted in Mowiol (Calbiochem) and stored in a dark box at 4°C until observation.
In each experiment, controls included: (i) slides pre-digested with RNase A (10 µg/ml, Boehringer Mannheim) in 300 mM NaCl, 10 mM Tris-HCl pH 7.5 and 5 mM EDTA at room temperature for 10 min prior to the pre-hybridization step; and (ii) slides without any probe (controls for non-specific binding of antibodies).
Gene mapping by in situ hybridization
In situ hybridization was carried out on chromosome preparations obtained from phytohaemagglutinin-stimulated human lymphocytes cultured for 72 h. 5-Bromodeoxyuridine was added for the final 7 h of culture (60 µg/ml of medium) to ensure a post-hybridization chromosomal banding of good quality. The human jumonji cDNA clone was tritium labelled by nick translation to a specific activity of 1×108 d.p.m./µg. The radiolabelled probe was hybridized to metaphases spreads at a final concentration of 25 ng/ml of hybridization solution as previously described (10). After coating with nuclear track emulsion (KODAK NTB2), the slides were exposed for 18 days at +4°C, then developed. To avoid any slipping of silver grains during the banding procedure, chromosome spreads were first stained with buffered Giemsa solution and the metaphases were photographed. R-Banding was then performed by the fluorochrome-photolysis-Giemsa (FPG) method and metaphases rephotographed before analysis.
We are indebted to Mike Mitchell for helpful discussions and critical reading of the manuscript. This work was supported in part by grants from GREG issued to Jean-Louis Bergé-Lefranc and Philippe Berta.