Genesis and prevention of spinal neural tube defects in ... - Development

Development 121, 681-691 (1995) Printed in Great Britain © The Company of Biologists Limited 1995

681

Genesis and prevention of spinal neural tube defects in the curly tail mutant mouse: involvement of retinoic acid and its nuclear receptors RAR-β and RAR-γ Wei-Hwa Chen1,*, Gillian M. Morriss-Kay1,† and Andrew J. Copp2 1Department of Human Anatomy, University of Oxford, South Parks Road, Oxford OX1 3QX, UK 2Division of Cell & Molecular Biology, Institute of Child Health, University of London, London WC1N

1EH, UK

*Present address: Department of Obstetrics and Gynecology, Tri-Service General Hospital, Taipei, Taiwan †Author for correspondence

SUMMARY A role for all-trans-retinoic acid in spinal neurulation is suggested by: (1) the reciprocal domains of expression of the retinoic acid receptors RAR-β and RAR-γ in the region of the closed neural tube and open posterior neuropore, respectively, and (2) the preventive effect of maternally administered retinoic acid (5 mg/kg) on spinal neural tube defects in curly tail (ct/ct) mice. Using in situ hybridisation and computerised image analysis we show here that in ct/ct embryos, RAR-β transcripts are deficient in the hindgut endoderm, a tissue whose proliferation rate is abnormal in the ct mutant, and RAR-γ transcripts are deficient in the tail bud and posterior neuropore region. The degree of deficiency of RAR-γ transcripts is correlated with the severity

of delay of posterior neuropore closure. As early as 2 hours following RA treatment at 10 days 8 hours post coitum, i.e. well before any morphogenetic effects are detectable, RARβ expression is specifically upregulated in the hindgut endoderm, and the abnormal expression pattern of RAR-γ is also altered. These results suggest that the spinal neural tube defects which characterise the curly tail phenotype may be due to interaction between the ct gene product and one or more aspects of the retinoic acid signalling pathway.

INTRODUCTION

(Gruneberg, 1954; Copp, 1985) and this abnormality is counteracted by RA when administered at 10 days 8 hours post coitum (p.c.), leading to normalisation of neuropore closure and prevention of spinal neural tube defects (Chen et al., 1994). This preventive effect of RA differs from that of other environmental influences such as food deprivation in utero, hyperthermia in vitro (Copp et al., 1988a), and treatment with the cytotoxic drugs hydroxyurea and mitomycin-C (Seller and Perkins, 1983, 1986), which appear to prevent spinal neural tube defects via non-specific growth retardation. The effect of RA, in contrast, appears specific and does not involve alteration of growth or other aspects of development (Chen et al., 1994). The preventive effect of RA may, therefore, be mediated by a specific interaction with its nuclear receptors. The effects of retinoids on growth, differentiation and development are mediated by two groups of nuclear receptors which belong to the steroid/thyroid hormone receptor superfamily: the retinoic acid receptors (RARs) and a distantly related group of retinoid X receptors (RXRs) which are activated by 9-cis retinoic acid and form heterodimers with several other nuclear receptors including RARs (Mangelsdorf et al., 1994, and references therein). The RA-activated receptor complexes transactivate target genes through binding sites known as retinoic acid response elements (RARE). RARs are capable of up-reg-

Neural tube defects are among the commonest of human congenital anomalies. Most cases are thought to result from a combination of genetic predisposition and environmental trigger factors (Carter, 1969), although few of the environmental causes and none of the genetic factors have been identified in humans to date. Mouse embryos homozygous for the mutation curly tail (ct) develop spinal neural tube defects in 54% of cases (Chen et al., 1994) and display several features characteristic of the defects in humans (Embury et al. 1979). The curly tail mutant provides an opportunity, therefore, for experimental studies on the embryonic development of spinal neural tube defects. Retinoic acid (RA) is a physiologically active metabolite and a potent teratogen which, when administered at teratogenic doses, can induce a variety of congenital defects including neural tube defects (Kalter and Warkany, 1961; Wiley, 1983; Yasuda et al., 1986; Tibbles and Wiley, 1988). By contrast, at low doses (5-10 mg/kg body weight), RA has been found to reduce the incidence of spinal neural tube defects significantly in curly tail mice (Seller et al., 1979; Seller and Perkins, 1982; Chen et al., 1994). Posterior neuropore closure is delayed in ct/ct embryos that develop spinal neural tube defects

Key words: curly tail, mutant, spinal cord, spina bifida, neurulation, neural tube defects, posterior neuropore, retinoic acid, retinoic acid receptors, mouse embryo, in situ hybridisation

682

W.-H. Chen, G. M. Morriss-Kay and A. J. Copp

ulating or down-regulating the transcription of target genes, which include the HoxB cluster (Simeone et al., 1990), Hoxa1 (Langston and Gudas, 1992), genes for extracellular matrix molecules such as laminin B1 (Vasios et al., 1989), enzymes modifying extracellular proteins, including collagenase and stromelysin (Pan et al., 1992; Nicholson et al. 1990), and transforming growth factor-β1 (Salbert et al., 1993). There are three types of RARs: RAR-α, RAR-β and RARγ, each of which is encoded by a gene on a different chromosome (Mattei et al. 1991). Studies of the spatiotemporal pattern of expression of RARs in neurulating mouse embryos have revealed reciprocal expression domains of RAR-γ and RAR-β in the developing trunk region: RAR-γ is expressed specifically in the open posterior neuropore region, whereas RAR-β is excluded from this region and is expressed only in the trunk and caudal hindbrain regions following neural tube closure (Ruberte et al., 1991). This pattern is maintained as the trunk elongates, until the posterior neuropore closes at the end of primary neurulation. RAR-γ transcripts cannot be detected in any part of the nervous system later in development (Ruberte et al., 1993). Since this reciprocal pattern of RAR-γ/RAR-β expression correlates spatiotemporally with the development and closure of the posterior neuropore, it raises the possibility that RAR-γ and/or RAR-β may play a role in posterior neuropore closure. We decided, therefore, to examine the patterns of expression of RAR-γ and RAR-β in the posterior neuropore region of mutant curly tail embryos, in which neuropore closure is genetically disturbed. The effect of RA treatment on RAR-γ and RAR-β expression was also examined in order to investigate the molecular basis of the action of RA in preventing spinal neural tube defects.

MATERIALS AND METHODS Collection of embryos Curly tail mouse pregnancies were produced by 2-hour timed matings, and embryos were collected for in situ hybridisation at early 10 days post coitum (p.c.). It has been shown that, at the 27-29 somite-stage, the majority of ct/ct embryos with small neuropores (category 1/2) do not develop spinal neural tube defects when cultured in vitro. In contrast, more than 90% of embryos with severely enlarged neuropores (category 4/5) develop spinal neural tube defects. By using this classification of posterior neuropore length, we can distinguish between prospective normal (unaffected) and abnormal (affected) ct/ct embryos prior to the appearance of gross congenital malformations (Copp, 1985). In the present study, only embryos with 27-29 somites, classified according to posterior neuropore length, were chosen to compare expression patterns of RAR-γ and RAR-β in the caudal trunk region. Both random-bred PO (Pathology, Oxford) and inbred CBA/Ca embryos were used as non-mutant controls. Intraperitoneal injection of RA Pregnant females received either 5 mg/kg of freshly prepared RA (alltrans-retinoic acid; Sigma) or the equivalent volume of vehicle (peanut oil; Sigma) by intraperitoneal injection at 10 days 8 hours p.c. which is the optimal time for RA prevention of spinal neural tube defects (Chen et al., 1994). Embryos were removed from the uterus 2 to 4 hours later. During this period, 1-2 somites are added. Only embryos with 27-33 somites were used for in situ hybridisation, since our previous work had shown that embryos with 26-31 somites show the maximal response to RA in terms of normalisation of posterior

neuropore closure (Chen et al., 1994). Sections of RA-treated and vehicle-treated embryos were mounted on the same slide for in situ hybridisation. In situ hybridisation riboprobes were transcribed using T7 polymerase (Boehringer Mannheim) from cDNAs encoding whole mouse RARβ and RAR-γ subcloned in an antisense orientation in the Bluescript and pSG5 vectors, respectively. These cDNAs detect all isoforms of RAR-β and RAR-γ (Zelent et al., 1989; Kastner et al., 1990). The in situ hybridisation procedures were performed as described by Wilkinson and Green (1990) with the following modifications. After fixation, serial dehydration, orientation, embedding and sectioning at 6 µm, one ct/ct embryo with a small neuropore (category 1/2, unaffected) and one with a severely enlarged neuropore (category 4/5, affected) were mounted together on each TESPA- (Sigma) subbed slide in order to ensure comparison under the same experimental conditions. After rehydration, pretreatment and redehydration, the sections were hybridised with riboprobe (2.5×104 cpm/µl) at 50°C for 16-18 hours. The sections were washed twice under high stringency conditions (65°C for 30 minutes) and then treated with RNAse (Sigma) for 30 minutes. The slides were dipped in photographic emulsion (Ilford), and exposed for 14 days after drying. They were developed in D19 (Kodak), counterstained with Harris’ Haematoxylin (BDH), dehydrated through a series of graded alcohols, cleared in xylene (BDH), mounted with DPX (BDH), and examined with a Leitz Laborlux S microscope under both bright- and dark-field illumination. Photographs were taken with a Wild Leitz camera using Fujichrome 64T film. 35S-labelled

Computerised image analysis A computerised image analyser (Kontron) was used in order to compare the levels of RAR-γ transcripts between non-mutant normal embryos and ct/ct embryos with different posterior neuropore sizes. Sections were examined at 200× magnification under dark-field illumination, and the microscope image was converted to a digital computer image with grey-scaled (0=black, 255=white) pixels in real time. In order to maintain a uniform optical environment for all measurements, magnification and light intensity were fixed and focus level was fine tuned, with the grey-scaled value standardised to a standard slide before measurement. The intensity of labelling in both neuroepithelium and mesoderm was directly measured on the digital image of serial transverse sections by drawing an area within each region of the neuroepithelium and mesoderm. Values were compared after subtraction of the background signal, which was determined from an area adjacent to the section on the same slide. Labelling intensities, expressed as grey values, were plotted against position along the craniocaudal body axis measured in microns from the caudal end of the posterior neuropore (0 µm). Regression lines were computed using the Sigmaplot graphics programme (version 4, 1991, Jandel, Corte Madera, CA). Three category 1/2 ct/ct embryos, two category 3 ct/ct embryos, three category 4/5 ct/ct embryos, and two non-mutant embryos were used for comparison of labelling intensities following hybridisation with the RAR-γ probe. Five vehicle-treated and five RAtreated ct/ct embryos hybridised with RAR-γ probe were also compared. An attempt was also made to quantify RAR-β expression in the hindgut endoderm, but the area of the sectioned hindgut is so small that the measurements showed a wide variation in grey value; the results are therefore not shown.

RESULTS Reciprocal expression patterns of RAR-β and RAR-γ in the posterior neuropore region In situ hybridisation studies of longitudinal sections of early

RA and RA receptors in neural tube defects day-10 embryos confirmed the findings of Ruberte et al. (1991) that there is a reciprocal pattern of expression of RARγ and RAR-β in the developing trunk region. In regions where the neural tube was closed, the hindgut and neuroepithelium showed high levels of RAR-β expression, whereas expression was not detected above background levels in caudal tissues. RAR-β transcripts were least abundant in the tissues of the open posterior neuropore region and in the tail bud (Fig. 1A,C). In contrast, adjacent sections of the same embryos hybridised with the RAR-γ probe showed highest levels of expression in the tail bud, and in the neuroepithelium and mesoderm of the open posterior neuropore region (Fig. 1B,D). This mutually exclusive pattern of expression of RAR-β and RAR-γ was seen in both ct/ct embryos with normal phenotype and in non-mutant CBA embryos (compare Fig. 1A,B with 1C,D). A craniocaudal sequence of RAR-β expression is disturbed by the ct mutation Expression of RAR-β was examined further using serial transverse sections of early day-10 ct/ct embryos (the same stage as shown in longitudinal sections in Fig. 1). In embryos with normal neuropores (category 1/2), RAR-β was not expressed in the tail bud or posterior neuropore region (Fig. 2B). Moving cranially through serial sections, RAR-β expression was first encountered in the hindgut endoderm and in the ventral mesoderm, whereas the recently closed neural tube expressed only low levels of RAR-β (Fig. 2C). At even more cranial levels (Fig. 2D), expression in the hindgut was intense, the neural tube also began to show strong RAR-β expression, while the somites remained negative. Thus, there is a craniocaudal sequence of RAR-β expression in the trunk of early day-10 embryos. Since the trunk develops in a cran-

683

iocaudal sequence, this means that the temporal sequence of expression of RAR-β during caudal tissue differentiation is first hindgut endoderm and ventral mesoderm, and then neural tube. Non-mutant embryos (Fig. 2E) and unaffected ct/ct embryos with small neuropores (category 1/2; Fig. 2C) both showed RAR-β expression in the hindgut endoderm at the level of the most recently closed neural tube. In contrast, severely affected ct/ct embryos (category 4/5) did not show RAR-β transcripts in the hindgut endoderm in sections (Fig. 2G) cut at the same distance from the caudal extremity (where the large sized neuropore was still open), although it was detected at more cranial levels (Fig. 2H). Four pairs of unaffected (neuropore category 1/2) and severely affected (neuropore category 4/5) ct/ct embryos were examined, and all showed this contrasting pattern of transcript localisation. Thus, the appearance of RARβ transcripts in the hindgut endoderm is delayed in affected ct/ct embryos, which are destined to develop spinal neural tube defects. Downregulation of RAR-γ expression in the neuropore region of affected ct/ct embryos The labelling intensities of RAR-γ transcripts in the open posterior neuropore region were found to be lower in affected ct/ct embryos with severely enlarged neuropores than in unaffected ct/ct embryos with small neuropores, which were mounted on the same slide. To determine whether these differences may have been due to sectioning the posterior neuropore region at different neuraxial levels, we made further detailed comparisons between sections from the same level of the neuraxis. Compared with unaffected ct/ct embryos with small neuropores (Fig. 3B-D), affected ct/ct

Fig. 1. Longitudinal sections, viewed under dark-field illumination, of the caudal trunk region of a category 1/2 (normal) ct/ct embryo with 27 somites (A,B), and a nonmutant control embryo with 27 somites (C,D). (A,C) Sections hybridised with the RAR-β riboprobe, show highest levels of expression in the closed neural tube (nt) and hindgut (hg), whereas expression is low in the posterior neuropore (pn) and tail bud (tb); (B,D) Sections hybridised with the RAR-γ probe, show the reciprocal pattern in which expression is high in the posterior neuropore and tail bud and low in more rostral regions. Scale bar, 210 µm.

684


embryos with severely enlarged neuropores (Fig. 3F-H) consistently showed decreased abundance of RAR-γ transcripts, in both the neuroepithelium and mesoderm, in sections through the caudal (Fig. 3B,F), middle (Fig. 3C,G) and cranial parts (Fig. 3D,H) of the open posterior neuropore region. However, whether the labelling intensities of RARγ transcripts in the posterior neuropore region are lower in unaffected ct/ct embryos with small neuropores (Fig. 3BD) than in non-mutant control embryos (Fig. 3E) could not be easily determined by microscopic observation. In order to make a quantitative comparison of RAR-γ expression in the mutant (normal and abnormal) and non-mutant (normal) embryos, labelling intensities in serial transverse sections were directly measured using a computerised image analyser. The data were plotted graphically against craniocaudal position along the body axis (Fig. 4). In non-mutant embryos, there was a graded distribution of RAR-γ transcripts along the craniocaudal axis with maximal expression throughout the open posterior neuropore, and with progressively lower levels cranial to the

Fig. 2. The expression pattern of RAR-β in the caudal trunk region of 10.5-day curly tail and non-mutant embryos. (B-D) An unaffected (category 1/2) ct/ct embryo with 29 somites in a series of transverse sections cut at different levels (as indicated in diagram A). (B) RAR-β was not expressed in the posterior neuropore region. (C) Cranial to the posterior neuropore, RAR-β transcripts are present in the endoderm of the hindgut (hg) and the ventral mesoderm surrounding the hindgut, but only at a low level in the neuroepithelium (ne). The presomitic mesoderm (m) does not express RARβ. (D) At a level cranial to C, RAR-β transcripts are present at even higher levels in the hindgut, the neuroepithelium of the closed neural tube is more strongly labelled, but the somites (s) are negative. (E) Section from a 28 somite-stage non-mutant embryo cut at the same level as C, as indicated in diagram A, and showing a similar expression profile, with expression strongest in the hindgut. (G,H) Sections from an abnormal 28 somite ct/ct embryo with severely enlarged neuropore (category 4/5), cut at equivalent levels to C and D (see diagram F). In G, RAR-β expression in the hindgut endoderm is much lower than in the normal embryos (compare with C and E) whereas more cranially, in H, hindgut expression is slightly higher, although still much less intense than at this level in a normal embryo (section D). Neuroepithelial expression is low throughout the caudal region of the abnormal embryo (G,H). Scale bar, 130 µm.

posterior neuropore (Fig. 4A,A′). In ct/ct embryos, expression of RAR-γ was lower in both the neuroepithelium and mesoderm of the open posterior neuropore region than in nonmutant embryos, the difference showing a direct correlation

RA and RA receptors in neural tube defects with the size of the neuropore (Figs 4B-D,B′-D′). Interestingly, even the unaffected ct/ct embryos (Fig. 4B) differed from the non-mutant embryos (Fig. 4A) in that the constant, high level

685

of RAR-γ transcripts was not maintained throughout the open posterior neuropore region, but declined in the cranial part of the neuropore.

Fig. 3. The expression pattern of RAR-γ in the open posterior neuropore region of a 26 somite-stage non-mutant embryo (E) and 28 somitestage ct/ct embryos with different degrees of delayed posterior neuropore closure as shown in the diagram (A). The category 1/2 (unaffected; B-D), and category 4/5 (severely enlarged neuropore; F-H) ct/ct embryos were mounted on the same slide. The diagram (A) shows the shape of both category 1/2 and category 4/5 neuropores, and indicates the levels of sections B-H. Sections from caudal, middle, and cranial levels of the posterior neuropore of category 1/2 and category 4/5 embryos were compared. The category 1/2 embryo showed higher levels of RAR-γ transcripts in all sections, compared with the category 4/5 embryo. Abbreviations as in Fig. 2. Scale bar, 120 µm.

686


RA and RA receptors in neural tube defects Expression patterns of RAR-β and RAR-γ in the posterior neuropore region of ct/ct embryos after RA treatment There was a striking effect of RA treatment on expression of RAR-β in the hindgut endoderm. Transverse sections of the hindgut at the level of the most recently closed neural tube showed that ct/ct embryos, as early as 2 hours after RA treatment at 10 days 8 hours p.c., exhibited expression of RAR-β in the hindgut endoderm (Fig. 5D), closely resembling that of untreated non-mutant embryos (Fig. 5B). Vehicle-treated ct/ct embryos, matched for somite-stage and posterior neuropore size, did not express RAR-β in the hindgut endoderm at this level of the body axis (Fig. 5C). A similar, specific upregulation of RARβ in the hindgut endoderm was also observed at a more caudal level, beneath the open posterior neuropore (Fig. 5F), where RAR-β is not expressed in untreated embryos (Fig. 5E). A total of five pairs of RA-treated and vehicle-treated ct/ct embryos was studied with a similar result in each case. Thus, expression of RAR-β in the hindgut endoderm is induced at a more caudal level, within the region of active neuropore closure, in RAtreated ct/ct embryos compared with vehicle-treated controls. RAR-β upregulation induced by 5 mg/kg RA was therefore specific for the hindgut endoderm and did not affect the neuroepithelium or paraxial mesoderm of the open posterior neuropore region (Fig. 5D,F). These observations indicate that the delayed expression of RAR-β in the hindgut endoderm of ct/ct embryos is quickly restored by RA to a normal pattern of expression without ectopic expression of RAR-β in other tissues. Compared with vehicle-treated ct/ct embryos (Fig. 6A), the expression profile of RAR-γ in both the neuroepithelium and mesoderm was also altered in RA-treated ct/ct embryos (Fig. 6B) as early as 2 hours after RA treatment. The altered expression profile of RAR-γ was complex in that both downregulation and upregulation of RAR-γ transcripts could be observed at different craniocaudal levels. The most striking feature was an upregulated peak of RAR-γ transcripts in both the neuroepithelium and mesoderm (Fig. 6B,B′) within the open posterior neuropore region, which was observed in all five RA-treated ct/ct embryos examined. DISCUSSION The results presented here indicate that delayed posterior Fig. 4. The expression profile of RAR-γ in both the neuroepithelium and mesoderm along the craniocaudal axis of the caudal trunk in, (A) a non-mutant control embryo; (B) a small neuropore ct/ct embryo (category 1/2); (C) a moderately enlarged neuropore ct/ct embryo (category 3); (D) a severely enlarged neuropore ct/ct embryo (category 4/5). The vertical dotted line indicates the position of posterior neuropore closure. The smoothed regression lines (regression order = 5) were drawn using SigmaPlot software. Arrows: position of sections illustrated in A′-D′. No ct/ct embryos maintain the normal (consistently high) expression profile of RAR-γ within the open posterior neuropore as seen in the non-mutant embryo (A), and the extent of abundance of RAR-γ transcripts in both the neuroepithelium and mesoderm correlates well with the size of the neuropore (B-D). (A′-D′) Representative transverse sections of the embryos hybridised with the RAR-γ probe, and used for the grey value measurements shown graphically in A-D. The position of each illustrated section was at the same level of the body axis. Abbreviations as in Fig. 2. Scale bar, 125 µm.

687

neuropore closure in severely affected ct/ct embryos correlates with abnormal patterns of expression of both RAR-β and RARγ in the tissues adjacent to the neuropore. Maternal administration of 5 mg/kg RA at 10 days 8 hours p.c. alters both morphogenesis (Chen et al. 1994) and gene expression in the caudal tissues. As early as 2 hours after RA treatment, i.e. before the onset of RA-induced normalisation of the posterior neuropore, RAR-β expression is upregulated specifically in the hindgut endoderm and the abnormal expression profile of RAR-γ in the posterior neuropore region is also altered. Three days later, the incidence of spinal neural tube defects is reduced by 50% in ct/ct embryos (Chen et al., 1994). This sequence of events supports the idea that the mechanism of RA prevention of spinal neural tube defects is mediated by changes in the expression of RAR-β and RAR-γ. Role of RAR-γ in spinal neurulation RAR-γ hybridisation shows a graded distribution of transcripts in both the neuroepithelium and mesoderm along the craniocaudal axis of the caudal trunk. The transcript level is maximal in the tail-bud and the whole region of the open posterior neuropore, diminishing sharply at levels cranial to the posterior neuropore. The two regions with the highest expression of RAR-γ thus correspond to the site of active primary neurulation and to the site of neural induction for secondary neurulation. A role for RAR-γ in primary neurulation events is consistent with the correlation between normalisation of the posterior neuropore induced by low dose RA treatment of ct/ct embryos at 10 days 8 hours p.c., reported in our previous study (Chen et al., 1994), and RA-induced upregulation of the deficient RAR-γ expression within the posterior neuropore region shown here. Furthermore, in ct/ct embryos, the extent of reduction of RARγ transcript abundance in the open posterior neuropore region correlates with the severity of delayed posterior neuropore closure. Upregulation of RAR-γ in the tail bud of ct/ct embryos could also normalise secondary neurulation, and thereby mediate the reduced incidence of tail flexion defects, observed in our previous study (Chen et al., 1994). RAR-γ may therefore be involved in both primary and secondary neurulation, and in the mechanism by which RA can influence these processes. The phenotype of RAR-γ null mutant embryos would seem to provide evidence contrary to this interpretation of our results: RARγ−/− embryos do not develop neural tube defects, showing only minor homeotic transformations of some cervical and thoracic vertebrae, which occur with variable frequency (Lohnes et al., 1993). The minor nature of these skeletal abnormalities is surprising, since RAR-γ is expressed throughout the developing precartilaginous and cartilaginous skeleton (Ruberte et al., 1990), and suggests that there is some functional redundancy between RAR-γ and other RARs, especially RAR-α, which is expressed almost ubiquitously at this stage (Ruberte et al., 1991), and shows high levels of expression in the tail bud mesenchyme (E. Ruberte, personal comunication). The response of RAR-γ−/− embryos to teratogenic levels of RA is, however, very significant: exposure of wild-type embryos to 100 mg/kg RA at day 8.5-9.0 causes spina bifida and truncation of the axial skeleton, as well as craniofacial abnormalities and homeotic transformations of the axial skeleton (Kessel and Gruss, 1991; Kessel, 1992); RARγ−/− embryos treated in the same way show only the craniofacial component of this response, having a normal trunk and tail

688


Fig. 5. The expression pattern of RAR-β after RA treatment: transverse sections of non-mutant and ct/ct embryos, at a level cranial to the posterior neuropore (B-D) and a level mid-way along the posterior neuropore (E,F) as shown in the diagram (A). (B) A non-mutant embryo with 28 somites, showing RAR-β expression in the hindgut and adjacent mesoderm; (C) a vehicle-treated ct/ct embryo with 31 somites, showing no RAR-β expression at the same level of the body axis; (D) a ct/ct embryo with 31 somites and a similar posterior neuropore size to that of the embryo in C, 2 hours after RA administration: RAR-β is expressed strongly in the hindgut but not in the neural tube at this level. (E) A vehicle-treated ct/ct embryo with 31 somites, showing no RAR-β expression; (F) a ct/ct embryo with 31 somites and a similar posterior neuropore size to the embryo in E, 3 hours after RA treatment: RAR-β is now expressed in the hindgut endoderm (arrowhead), beneath the open posterior neuropore. Abbreviations as in Fig. 2. Scale bar, 130 µm.

(Lohnes et al., 1993), indicating that RAR-γ is required to mediate the mechanism by which RA causes spina bifida and major vertebral anomalies. Our results similarly suggest that the action of 5 mg/kg RA on posterior neuropore closure in ct/ct embryos is mediated, at least in part, by RAR-γ. The fact that upregulation of RAR-γ within the posterior neuropore region of ct/ct embryos is only partial, and is not restored to the normal pattern, correlates well with the observation that RA does not prevent all cases of spinal neural tube defects, but significantly lowers the incidence. Rapid upregulation of RARγ has also been reported in the lungs of retinol-deficient rats

exposed to RA (Haq et al., 1991). RA may act to regulate RARγ expression via the RAR-γ2 isoform which has a RARE in its promoter region (Lehmann et al., 1992), and is the most ubiquitously expressed RAR-γ isoform (Lohnes et al., 1993). Role of RAR-β in spinal neurulation Since RAR-β is not expressed in the neuroepithelium until after neural tube closure, it seems unlikely to play a role in the intrinsic movements of this tissue during neurulation. Indeed, Ruberte et al. (1991) suggested a role for RAR-β in neuroblast maturation following neural tube closure. In the present study,

RA and RA receptors in neural tube defects

689

Fig. 6. The expression profile of RAR-γ in the caudal trunk region of ct/ct embryos, 3 hours after maternal administration of vehicle only (A) and RA (B). The two embryos had 31 somite pairs and a similar posterior neuropore size. The expression profile of RAR-γ in both neuroepithelium and mesoderm was changed by RA, an upregulated peak within the open region being the major alteration. The vertical dotted line indicates the position of posterior neuropore closure. The regression lines (regression order =8) were drawn using SigmaPlot software. (A′) Transverse section of a vehicle-treated ct/ct embryo, and (B′) an RA-treated ct/ct embryo, both with 31 somite pairs and a similar posterior neuropore size, hybridised with the RAR-γ probe. The sections were cut at equivalent levels of the body axis, which is indicated by an arrow in the expression profile graphs. The level of RAR-γ in all tissues is higher in the RA-treated embryo (B′) than in the vehicle-treated one (A′). Scale bar, 125 µm.

we have shown that RAR-β is also expressed specifically in the hindgut endoderm beneath the site of active neural tube closure at the posterior neuropore. Its absence from this region in ct/ct embryos with an enlarged posterior neuropore suggests that RAR-β deficiency here may be of significance in relation to the ct phenotypic defect. Moreover, when low dose RA is applied, we see a specific upregulation of RAR-β in the hindgut, coincident with normalisation of posterior neuropore closure. This correlates well with our previous finding that the ct defect is expressed primarily as a low proliferation rate of the hindgut endoderm and notochord (Copp et al., 1988b).

There is good evidence for an association between RA and both growth and differentiation. It has recently been shown that RA stimulates embryonal stem cell growth at low concentrations (10−10-10−9 M), whereas higher concentrations (10−810−6) are inhibitory (Pijnappel et al., 1993). It has also been reported that RAR-β2, which contains a RARE in its promoter, is rapidly expressed in response to the RA-induced endodermal differentiation of embryonal carcinoma cells (Zelent et al., 1991). If RAR-β plays a role in the control of cell proliferation in the hindgut endoderm, the specific upregulation of RAR-β in this tissue could normalise its proliferation defi-

690


ciency, thereby correcting the growth imbalance between the dorsal neural and ventral non-neural tissues in the caudal trunk region. In turn this could prevent the excessive ventral curvature that has been shown to mediate the effect of the growth imbalance in ct/ct embryos (Brook et al., 1991), thereby enhancing posterior neuropore closure. Clearly, 5 mg/kg RA is sufficient to affect morphogenesis in ct/ct embryos, but a distinction must be made between this effect and RA-induced teratogenesis. Teratogenic levels of RA (20-100 mg/kg) cause ectopic expression of RAR-β, and can cause neural tube defects (Rossant et al., 1991; OsumiYamashita et al., 1992). In contrast, the effect of low dose RA on RAR-β expression in ct/ct embryos is specific to the tissues in which it is deficient in untreated ct/ct embryos, and is correlated with a reduced incidence of spinal neural tube defects (Chen et al., 1994). Possible molecular basis of the ct defect Our observations suggest that the transcription of RAR-β and RAR-γ in the posterior neuropore region is modulated, directly or indirectly, by the ct gene product, which could itself be a transcription factor with a tissue-specific distribution. This hypothesis is consistent with the fact that while the capacity to develop neural tube defects depends on the presence of homozygosity for the ct gene, the RA-preventive effect apparently involves RARs. RAR-β and RAR-γ map to mouse chromosomes 14 and 15, respectively, and so are unrelated to ct, which has been localised to distal chromosome 4 (Neumann et al., 1994). Three possible mechanisms may be envisaged (Fig. 7). The first is a direct protein-protein interaction between the ct gene product and RAR, affecting RAR activity in a manner analogous to that shown by RAR/cJun heterodimers (YangYen et al., 1991; Pfahl et al., 1992). The second possible mechanism is an upstream transcription effect. For instance, the ct product could influence the transcription of RARs, directly via binding to a RAR promoter (a protein-DNA interaction), or indirectly via binding to some transcriptional coactivators (an indirect protein-protein interaction), which are required for RAR transcription and expressed in a cell-specific manner, analogous to the adenovirus E1A-like activity in embryonal carcinoma cells (La Thangue and Rigby, 1987; Stunnenberg, 1993). Thus disturbance of ct gene function could lead to misregulation of RARs in specific tissues, leading to abnormal development of these tissues. A third possibility is that the ct gene product may affect RA metabolism, perhaps via altered levels of cellular retinol binding protein (CRBP I), causing neural tube defects through local deficiency of endogenous RA in the posterior neuropore region. The most common isoforms of RAR-β and RAR-γ in embryos at this stage of development are RAR-β2 and RAR-γ2, respectively (Mendelsohn et al., 1994; Lohnes et al., 1993, and references therein). Both of these isoforms are RA-inducible, via a RARE in their promoter regions. Deficiency of endogenous RA in the posterior neuropore region might therefore be the reason for deficient expression of these two genes in the posterior neuropore tissues, and raised RA levels following administration of 5 mg/kg RA could be responsible for their upregulation. It is clear that the normal functions of the ct gene are related to the developmental patterns specific to the posterior

Fig. 7. Three possible mechanisms by which the ct product could cause local downregulation of RAR-γ in the neuroepithelium and mesoderm of the posterior neuropore, and of RAR-β in the hindgut endoderm. For detailed explanation, see text.

neuropore and tailbud regions. In this study we have shown that within these regions, ct gene activity either affects, or is mediated by, the molecular interactions of RA and its nuclear receptors RAR-β and RAR-γ, and/or the processes of retinoid metabolism. We thank Professor Pierre Chambon for providing cDNAs for mouse RAR-β and RAR-γ and Ann Stanmore for advice on computerised image analysis. W. H. C. is a clinical scientist of the Tri-Service General Hospital, the National Defence Medical Centre, Taipei, Taiwan, and was supported by a grant from Taiwan, Republic Of China. This work was supported by the Imperial Cancer Research Fund (A. J. C.), The Human Frontier Science Program (G. M. M.-K.), and as part of a Multicenter Agreement for Studying Neural Tube Defects in Mutant Mice funded by the National Institutes of Child Health and Human Development, USA, through Cooperative Agreement HD 28882 (A. J. C.).

REFERENCES Brook, F. A., Shum, A. S. W., van Straaten, H. W. M. and Copp, A. J. (1991). Curvature of the caudal region is responsible for failure of neural tube closure in the curly tail (ct) mouse embryo. Development 113, 671-678. Carter C. O. (1969). Spina bifida and anencephaly: A problem in geneticenvironmental interaction. J. biosocial Sci. 1, 71-83. Chen W.-H., Morriss-Kay G. M. and Copp A. J. (1994). Prevention of spinal NTD in the curly tail mouse mutant by a specific effect of retinoic acid. Dev. Dyn. 199, 93-102. Copp A. J. (1985). Relationship between timing of posterior neuropore closure and development of spinal neural tube defects in mutant (curly tail) and normal mouse embryos in culture. J. Embryol. exp. Morph. 88, 39-54. Copp A. J., Crolla J. A. and Brook F. A. (1988a). Prevention of spinal neural tube defects in the mouse embryo by growth retardation during neurulation. Development 104, 297-303. Copp A. J., Brook, F. A. and Roberts, H. J. (1988b). A cell-type-specific abnormality of cell proliferation in mutant (curly tail) mouse embryos developing spinal neural tube defects. Development 104, 285-295. Embury S., Seller M. J., Adinolfi M and Polani P. E. (1979). Neural tube defects in curly-tail mice. I. Incidence, expression and similarity to the human condition. Proc. R. Soc. Lond. Biol. 206, 95-107.

RA and RA receptors in neural tube defects Gruneberg H. (1954). Genetical studies of the skeleton of the mouse VIII. curly-tail. J. Genet. 52, 52-67. Haq R. U., Pfahl M. and Chytil F. (1991). Retinoic acid affects the expression of nuclear retinoic acid receptors in tissues of retinol-deficient rats. Proc. Natl. Acad. Sci. USA 88, 8272-8276. Kalter H. and Warkany J. (1961). Experimental production of congenital malformation in strains of inbred mice by maternal treatment with hypervitaminosis A. Amer. J. Path. 38, 1. Kastner, P., Krust, A., Mendelsohn, C., Garnier, J. M., Zelent, A., Leroy, P., Staub, A. and Chambon, P. (1990). Murine isoforms of retinoic acid receptor gamma with specific patterns of expression. Proc. Natl. Acad. Sci. USA 87, 2700-2704. Kessel, M. (1992). Respecification of vertebral identities by retinoic acid. Development 115, 487-501. Kessel, M. and Gruss, P. (1991). Homeotic transformations of murine vertebrae and concomitant alteration of Hox codes induced by retinoic acid. Cell 67, 89-104. La Thangue N. B. and Rigby P. W. J. (1987). An adenovirus E1A-like transcription factor is regulated during the differentiation of murine embryonal carcinoma stem cells. Cell 49, 507-513. Langston A. W. and Gudas L. J. (1992). Identification of a retinoic acidresponse enhancer 3′ of the murine homeobox gene Hox-1. 6. Mech. Dev. 38, 217-228. Lehmann J. M., Zhang X. -K. and Pfahl M. (1992). RAR-γ2 expression is regulated through a retinoic acid response element embedded in Sp1 sites. Mol. Cell. Biol. 12, 2976-2985. Lohnes, D., Kastner, P., Dierich, A., Mork, M., LeMeur, M. and Chambon, P. (1993). Function of retinoic acid receptor γ in the mouse. Cell 73, 643-658. Mangelsdorf D. J., Umesono, K. and Evans R. M. (1994). The retinoid receptors. In The Retinoids: Biology, Chemistry, and Medicine, 2nd ed., (eds. M. B. Sporn, A. B. Roberts, and D. S. Goodman), pp. 319-349, New York: Raven Press Ltd. Mattei M. -G., Riviere M., Krust A., Ingvarsson S., Vennstrom B., Islam Q., Levan G., Kautner P., Zelent A., Chambon P., Szpirer J. and Szpirer C. (1991). Chromosomal assignment of retinoic acid receptor (RAR) genes in the human, mouse, and rat genomes. Genomics 10, 1061-1069. Mendelsohn, C., Mark, M., Dolle, P., Dierich, A., Gaub, M. P., Krust, A., Lampron, C. and Chambon, P. (1994). Retinoic acid receptor β2 (RARβ2) null mutant mice appear normal (submitted). Neumann, P. E., Frankel, W. N., Letts, V. A., Coffin, J. M., Copp, A. J. and Bernfield, M. (1994). Multifactorial inheritance of neural tube defects: Localization of the major gene and recognition of modifiers in ct mutant mice. Nature Genet. 6, 357-362. Nicholson R. C., Mader S., Nagpal S., Leid M., Rochette-Egly C. and Chambon P. (1990). Negative regulation of the rat stromelysin gene promoter by retinoic acid is mediated by an AP1 binding site. EMBO J. 9, 4443-4454. Osumi-Yamashita N., Iseki S., Noji S., Nohno T., Koyama E., Taniguchi S., Doi H. and Eto K. (1992). Retinoic acid treatment induces the ectopic expression of retinoic acid receptor β gene and excessive cell death in the embryonic mouse face. Dev. Growth Differ. 34, 199-209. Pan L., Chamberlain S. H., Auble D. T. and Brinckerhoff C. E. (1992). Differential regulation of collagenase gene expression by retinoic acid receptors-,α, β and γ. Nucl. Acids Res. 20, 3105-3111. Pfahl M., Zhang X. -K., Lehmann J. M., Husmann M., Graupner G. and Hoffmann B. (1991). Molecular mechanisms of retinoic acid action. In Retinoic Acids in Normal Development and Teratogenesis (ed. G. MorrissKay), pp. 51-64, Oxford: Oxford University Press. Pijnappel W. W. M., Hendriks H. F. J., Folkers G. E., van den Brink C. E., Dekker E. J., Edelenbosch C., van der Saag P. T. and Durston A. J.

691

(1993). The retinoid ligand 4-oxo-retinoic acid is a highly active modulator of postional specification. Nature 366, 340-344. Rossant J., Zirngibl R., Cado D., Shago M and Giguere V. (1991). Expression of a retinoic acid response element-hsplacZ transgene defines specific domains of transcriptional activity during mouse embryogenesis. Genes Dev. 5, 1333-1344. Ruberte E., Dolle P., Krust, A., Zelent, A., Morriss-Kay G. and Chambon, P. (1990). Specific spatial and temporal distribution of retinoic acid receptor gamma transcripts during mouse embryogenesis. Development 108, 213-222. Ruberte E., Dolle P., Chambon P. and Morriss-Kay G. (1991). Retinoic acid receptors and cellular retinoid binding proteins II. Their differential pattern of transcription during early morphogenesis in mouse embryos. Development 111, 45-60. Ruberte E., Friederich V., Chambon P. and Morriss-Kay G. (1993). Retinoic acid receptors and cellular retinoid binding proteins III. Their differential transcript distribution during mouse nervous system development. Development 118, 267-282. Salbert G., Fanjul A., Piedrafita F. J., Lu X. P., Kim S. -J., Tran P. and Pfahl M. (1993). Retinoic acid receptors and retinoid X receptor-α downregulate the transforming growth factor-β1 promoter by antagonizing AP-1 activity. Mol. Endocrinol. 7, 1347-1356. Seller M. J., Embury S., Polani P. E. and Adinolfi M. (1979). Neural tube defects in curly-tail mice. II. Effects of maternal administration of vitamin A. Proc. R. Soc. Lond. Biol. 206, 95-107. Seller M. J. and Perkins K. J. (1982). Prevention of neural tube defects in curly-tail mice by maternal administration of vitamin A. Prenatal Diagnosis 2, 297-300. Seller M. J. and Perkins K. J. (1983). Effect of hydroxyurea on neural tube defects in the curly tail mouse. J. Craniofac. Genet. Dev. Biol. 3, 11-17. Seller M. J. and Perkins K. J. (1986). Effect of mitomycin C on the neural tube defects of the curly-tail mouse. Teratology 33, 305-309. Simeone A., Acampora D., Arcioni L., Andrew P. W., Boncinelli E., Mavilio F. (1990). Sequential activation of Hox2 homeobox genes by retinoic acid in human embryonal carcinoma cells. Nature 346, 763-766. Stunneberg, H. G. (1993). Mechanisms of transactivation by retinoic acid receptors. BioEssays 15, 309-315. Tibbles L. and Wiley M. J. (1988). A comparative study of the effects of retinoic acid given during the critical period for inducing spina bifida in mice and hamsters. Teratology 37, 113-125. Vasios G. W., Gold J. D., Petkovich M., Chambon P. and Gudas L. J. (1989). A retinoic acid-responsive element is present in the 5′ flanking region of the laminin B1 gene. Proc. Natl. Acad. Sci. USA 86, 9099-9103. Wiley M. J. (1983). The pathogenesis of retinoic acid-induced vertebral abnormalities in golden Syrian hamster fetuses. Teratology 28, 341-353. Wilkinson D. G. and Green J. (1990). In situ hybrdization and the threedimensional reconstruction of serial sections. In Postimplantation Mammalian Embryos: A Practical Approach (eds A. J. Copp and D. L. Cockroft), pp. 155-171, Oxford: IRL Press. Yang-Yen H-F, Zhang X-K, Graupner G., Tzukerman M., Sakamoto B., Karin M. and Pfahl M. (1991). Antagonism between retinoic acid receptors and AP-1: Implication for tumor promotion and inflammation. New Biologist 3, 1206-1220. Yasuda Y., Okamoto M., Konishi H., Matsuo T., Kihara T. and Tanimura T. (1986). Developmental anomalies induced by all-trans retinoic acid in fetal mice: I. Macroscopic findings. Teratology 34, 37-49. Zelent A., Mendelsohn, C., Kastner, P., Garnier, J. -M., Ruffenach, F., Leroy, P., Krust A. and Chambon P. (1991). Differentially expressed isoforms of the mouse retinoic acid receptor β are generated by usage of two promoters and alternative splicing. EMBO J. 10, 71-81. (Accepted 22 November 1994)