letters to nature

Received 5 June; accepted 18 August 1997. gene to be shown to be required in vivo for the genesis of granule 1. Lebacq-Verheyden, A.-M., Trepel, J., Sausville, E. A. & Battey, J. in Handbook of Experimental cells, and hence the predominant neuronal population in the Pharmacology, Vol. 95/II. Peptide Growth Factors and their Receptors II (eds Sporn, M. B. & Roberts, A. B.) 71–124 (Springer, Berlin, Heidelberg, 1990). . 2. Spindel, E. R., Giladi, E., Brehm, P., Goodman, R. H. & Segerson, T. P. Cloning and functional The mammalian cerebellum consists of deep centrally located characterization of a complementary DNA encoding the murine fibroblast bombesin/gastrin- neurons, referred to as the deep nuclei, and a peripheral cortex. The releasing peptide receptor. Mol. Endocrinol. 4, 1956–1963 (1990). 3. Battey, J. F. et al. Molecular cloning of the bombesin/gastrin-releasing peptide receptor from Swiss 3T3 cortex contains two principal neuronal subtypes, the Purkinje cells cells. Proc. Natl Acad. Sci. USA 88, 395–399 (1991). and the cells of the internal granular layer (IGL). The IGL neurons, 4. Corjay, M. H. et al. Two distinct bombesin receptor subtypes are expressed and functional in human lung carcinoma cells. J. Biol. Chem. 266, 18771–18779 (1991). which eventually constitute most of the neurons in the cerebellum, 5. Wada, E. et al. cDNA cloning, characterization, and brain region-specific expression of a neuromedin- are derived from peripherally located cells in the external germinal B-preferring bombesin receptor. Neuron 6, 421–430 (1991). layer (EGL) which migrate postnatally along radial glia1,5. The 6. Fathi, Z. et al. A novel bombesin receptor subtype selectively expressed in testis and lung carcinoma cells. J. Biol. Chem. 268, 5979–5984 (1993). mouse cerebellar anlage is specified at embryonic-day 9 (E9) just 7. Gorbulev, V., Akhundova, A., Buchner, H. & Fahrenholz, F. Molecular cloning of a new bombesin anterior to the area where closure of the neural tube is incomplete. receptor subtype expressed in uterus during pregnancy. Eur. J. Biochem. 208, 405–410 (1992). 8. Ohki-Hamazaki, H., Wada, E., Matsui, K. & Wada, K. Cloning and expression of the neuromedin B The neuroepithelium of the ventricular zone in the receptor and the third subtype of bombesin receptor genes in the mouse. Brain Res. 762, 165–172 gives rise to the precursors of the neurons of the deep nuclei and (1997). Purkinje cells6. The precursors of the EGL are derived at E13–E15 9. Bray, G. A. & York, D. A. Hypothalamic and genetic obesity in experimental animals: An autonomic and endocrine hypothesis. Physiol. Rev. 59, 719–809 (1979). from a structure named the rhombic lip (also known as the germinal 10. Frederich, R. C. et al. Leptin levels reflect body lipid content in mice: Evidence for diet-induced trigone), which is located at the posterior edge of the cerebellar resistance to leptin action. Nature Med. 1, 1311–1314 (1995). 1,5,7 11. Maffei, M. et al. Leptin levels in human and rodent: Measurement of plasma leptin and ob RNA in anlage (Fig. 1). The proliferating cells of the rhombic lip, which obese and weight-reduced subjects. Nature Med. 1, 1155–1161 (1995). are committed to become EGL neurons, disperse rostrally over the 12. Schwartz, M. W., Peskind, E., Raskind, M., Boyko, E. J. & Porte, D. Jr Cerebrospinal fluid leptin levels: surface of the cerebellar anlage, where they establish the EGL. Relationship to plasma levels and to adiposity in humans. Nature Med. 2, 589–593 (1996). 13. Wu, J. M., Nitecki, D. E., Biancalana, S. & Feldman, R. I. Discovery of high affinity bombesin receptor Proliferation of these precursors continues in the EGL subtype 3 agonists. Mol. Pharmacol. 50, 1355–1363 (1996). until postnatal day 15 (P15). The EGL cells migrate inwardly to give 14. Boer, P.H. et al. Polymorphisms in the coding and noncoding region of murine Pgk-1 alleles. Biochem. Gen. 28, 299–308 (1990). rise to the IGL, starting at birth until day P20, when maturation of 1,5,7 15. Yanofsky, R. L., Fine, M. & Pellow, J. W. A mutant neomycin phosphotransferase II gene reduces the the cerebellum is complete . resistance of transformants to antibiotic selection pressure. Proc. Natl Acad. Sci. USA 87, 3435–3439 The Drosophila atonal gene is essential for the generation of the (1990). 16. Nabeshima, Y. et al. Myogenin gene disruption results in perinatal lethality because of severe muscle chordotonal organs, the proprioreceptors, and the R8 photorecep- defect. Nature 364, 532–535 (1993). tor precursors in the eye4,8. Adult flies that lack atonal function do 17. Hooper, M., Hardy, K., Handyside, A., Hunter, S. & Monk, M. HPRT-deficient (Lesch–Nyhan) mouse 8 embryos derived from germline colonization by cultured cells. Nature 326, 292–295 (1987). not have photoreceptors, are uncoordinated and tend not to fly . 18. Yamamoto, K. & Kikuyama, S. Radioimmunoassay of prolactin in plasma of bullfrog tadpoles. Expression of Math1, the mouse atonal homologue, is highly Endocrinol. Jpn 29, 159–167 (1982). restricted to a few neurons in the and dorsal neural 19. Bouillaud, F., Weissenbach, J. & Ricquier, D. Complete cDNA-derived amino acid sequence of rat 2,3 brown fat uncoupling protein. J. Biol. Chem. 261, 1487–1490 (1986). tube . Math1 is expressed in the precursors of the EGL in the

Acknowledgements. We thank J.-I. Miyazaki for ES cells and training regarding ES cells; A. F. Parlow for rhombic lip starting at E13, and later in the EGL of the developing mouse growth hormone and polyclonal monkey anti-rat GH serum; K. Wakabayashi for goat anti- cerebellum (Fig. 1). Math1 expression ceases during the inward monkey IgG serum; H. Ohno for rat UCP1 probe; T. Nishikawa for providing Animex Auto; N. M. Le Douarin, K. Mikoshiba, R. S. Petralia and J. Smith for critical reading of the manuscript. This work was migration phase of the granule cells to the IGL. To investigate the supported in part by research grants from the Ministry of Education, Science, Sports and Culture, the role of Math1 in cerebellar development, a targeted deletion of the Ministry of Health and Welfare, the Science and Technology Agency of Japan, the Japan Health Science m1 Foundation. Math1 gene (Math1 ) was generated by using embryonic stem (ES) cell technology (Fig. 2a). Heterozygous mice were intercrossed to Correspondence and requests for materials should be addressed to H.O.-H. (e-mail: [email protected]) or K.W. (e-mail: [email protected]). obtain mice homozygous for the Math1 deletion (Fig. 2b). Absence of a hybridization signal with DNA from homozygous mutant mice, when probed with an internal probe spanning the basic-helix–

Math1 is essential for genesis ab AQ of cerebellar granule neurons MC SC Nissim Ben-Arie*†, Hugo J. Bellen*‡k, Dawna L. Armstrong§, IV Alanna E. McCall†, Polina R. Gordadze†, Qiuxia Guo§, Martin M. Matzuk*‡§ & Huda Y. Zoghbi*† k IC Cb IST Cb EGL Departments of *Molecular and Human Genetics, †Pediatrics, ‡Cell Biology, NE NE CP EGL §Pathology, and kHoward Hughes Medical Institute, Baylor College of Medicine, IV URL/GT URL/GT Houston, Texas 77030, USA PO LRL CP ...... MO The cerebellum is essential for fine motor control of movement and posture, and its dysfunction disrupts balance and impairs IO control of speech, limb and eye movements. The developing cerebellum consists mainly of three types of neuronal cells: Figure 1 The developing cerebellum at E16.5. The sagittal (a) and coronal (b) granule cells in the external germinal layer, Purkinje cells, and views show the localization of the cerebellum (Cb) between the inferior colliculus neurons of the deep nuclei1. The molecular mechanisms that (IC), (PO), mesencephalon (MC) and medulla oblongata (MO). The germinal underlie the specific determination and the differentiation of matrix, which gives rise to EGL precursors, is the rhombic lip, with an upper and a each of these neuronal subtypes are unknown. Math1 (refs 2, 3), lower component (URL, LRL, respectively). The proliferative zone at the URL, the the mouse homologue of the Drosophila gene atonal 4, encodes a germinal trigone (GT), has three prongs: EGL, neuroepithelium (NE), and choroid basic helix–loop–helix transcription factor that is specifically plexus (CP). Granule cell precursors originate at the URL/GTand migrate (arrow) expressed in the precursors of the external germinal layer and to form the EGL. The LRL will contribute to the medullar precerebellar nuclei, their derivatives. Here we report that mice lacking Math1 fail to including the inferior olive (IO). Math1 is expressed in granule precursors in GT form granule cells and are born with a cerebellum that is devoid of and in emerging EGL (shaded area). AQ, aqueduct of Sylvius; IST, isthmus; IV, an external germinal layer. To our knowledge, Math1 is the first fourth ventricle; SC, superior colliculus. Adapted from ref.1.

Nature © Macmillan Publishers Ltd 1997 NATURE | VOL 390 | 13 NOVEMBER 1997 169 letters to nature loop–helix (bHLH) domain, confirmed that the mutation is a null a allele (data not shown). A H RI RV H A A X A RI Mice heterozygous for the Math1 deletion are viable, fertile and Math1 locus appear normal. Genotyping of newborn mice from heterozygous intercrosses demonstrated that the targeted allele did not result in RI RV X A embryonic lethality. The observed ratio of genotypes fitted that RI H PGKhprt MC1-TK expected for a fully penetrant, mendelian recessive inheritance Targeting pattern, with 28 (25.2%) wild types, 55 heterozygote (49.5%) and B BN N vector 28 (25.2%) homozygote (null) mice. Math1-null mice are born A H RI RV X A RI alive, but fail to breathe, become cyanotic, and die a few minutes RI H Targeted PGKhprt after birth. We observed no morphological abnormalities in the Math1 locus lung (data not shown). Furthermore, the lack of expression of 1 kb 5' probe B 3' probe Math1 in the lungs, combined with the expression pattern in the brainstem2, suggest a central mechanism for this respiratory failure. The structure and histology of the cerebellum of Math1-null mice b is abnormal compared to their littermates. The most peripheral cell layer in the cerebellum, the EGL, is normally about eight cells thick at E18.5. These rapidly proliferating cells are the granule cell 8.5 kb (WT) precursors and express Math1 abundantly2. Math1-null mice com- 6.4 kb (mutant) pletely lack the EGL, as shown by haematoxylin-and-eosin staining (Fig. 3a, b), and cresyl violet staining (data not shown), and have a cerebellum that is reduced in size. To analyse the changes in the Figure 2 Targeted disruption of the Math1 locus and generation of Math1-null cytoarchitecture of the cerebellum, we stained the two principal mice. a, Homologous recombination between the 3.7-kb and 3.1-kb fragments in neuronal populations in the cerebellum, the Purkinje cells (Fig. 3c, the targeting vector and the genomic locus, resulted in the replacement of 3.2 kb d) and the granule cells (Fig. 3e, f ). RNA in situ hybridization using (black box) which include the entire coding region (hatched box). b, Tail DNA from RU49, a zinc-finger transcription factor that labels the EGL, the newborn littermates obtained by heterozygous intercrosses was digested with 9 migrating granule cells and the IGL , demonstrates the absence of EcoRI and HindIII and hybridized with the 3Ј diagnostic probe. The three pups the EGL and migrating granule cells in the cerebella of Math1-null homozygous for the deletion died a few minutes after delivery, whereas the four mice (Fig. 3e, f ). In contrast to the lack of granule neurons, the wild-type (WT) and five mice heterozygous for the deletion were viable. Purkinje cells and the neurons of the deep nuclei are still present in mutant mice, as demonstrated by staining with anti-calbindin antibody (Fig. 3c, d) and Nissl staining (data not shown). However, because of the absence of the EGL, the Purkinje cells, which are normally located beneath the EGL, are now localized at the periphery of the cerebellum and do not form a distinct and organized layer. In addition, a significant subpopulation of Purkinje cells fails to migrate from the central area of the cerebellum into the periphery (Fig. 3c, d). Staining with glial fibrillary acidic protein10 (GFAP) and nestin11, which both label glial cells, demonstrates that there is not excessive gliosis, and that the radial glia are localized properly (data not shown). These results show that Math1 is essential for the formation of the EGL but not the Purkinje cells, the neurons of the deep nuclei, or the glial cells. In Math1-null mice, the absence of the EGL also leads to the lack of foliation of the cerebellum typically observed in normal embryos starting from E18 (Fig. 3a, b). Possible causes for the missing EGL in mutant mice include failure in fate determination of granule precursors at the rhombic lip, aberrant migration of precursors to form the EGL, or degenera- tion of the neurons already localized at the EGL. The granule neuron precursors originate at E13–E15 in the proliferation zone of the rhombic lip1,7. At E14, the EGL precursors initiate migration from the rhombic lip over the surface of the cerebellar primordia. At this stage, Math1 is expressed in the granule precursors localized at the rhombic lip and in the emerging EGL2. As shown in Fig. 4a–d, Math1-deficient mice contain fewer cells in the rhombic lip than do control embryos. In addition, there is no migration of cells out of the rhombic lip to form the EGL. Bromodeoxyuridine (BrdU) labelling demonstrates the absence of proliferating cells in both Figure 3 Cerebellar abnormalities in E18.5 Math1-null embryos. a, b, Midsagittal sites that normally expand the pool of granule cells: the rhombic lip sections through the cerebellum stained with haematoxylin and eosin. The and the cells that have already migrated to the EGL (Fig. 4e, f ). cerebellum in wild-type (a) mice is well developed and foliated (asterisks), but is Hence, the lack of the cerebellar EGL in Math1-null mice is evident smooth in Math1 null mice (b). The external germinal layer (EGL) is absent and the early on and is caused either by the absence of the granule cell layer (PCL) less organized, with some PC located deep in the cerebellum, precursors or by the inability of these cells to proliferate at the as revealed by anti-calbindin staining (c, d). Parasagittal sections of the cerebellum of rhombic lip. Thus, the agenesis of granule cell precursors in Math1- wild-type (e) and mutant (f) embryos were hybridized with the granule-cell marker null mice is a much earlier event than the postnatal apoptotic RU49. Wild-type EGL is strongly positive for RU49, whereas the cerebellum of 12,13 degeneration of the EGL seen in weaver mice . Math1-null mice is negative. Original magnifications: a–d, ×100; e, f, ×40.

Nature © Macmillan Publishers Ltd 1997 170 NATURE | VOL 390 | 13 NOVEMBER 1997 letters to nature

It has been suggested that inferior olive neurons are derived from the rhombic lip14, which consists of an upper and a lower component (Fig. 1) and that it is the lower lip that gives rise to the neurons of the inferior olive1. Math1 is only expressed in the germinal trigone that is part of the upper lip and not the lower lip2 (Fig. 1). The inferior olive neurons are present in Math1-null mice and appear normal (Fig. 5), consistent with Math1 expression in the upper rhombic lip. This finding supports the histological and functional division of the rhombic lip into an upper and a lower lip, each contributing to the generation of different types of neurons that migrate to separate destinations. As Math1-null mice die shortly after birth from respiratory failure and Math1 is expressed in the hindbrain, we wanted to establish whether there is selective loss of neurons implicated in the control of respiration, including those of the nucleus tractus solitarius, nucleus ambiguus, dorsal vagus nucleus and trigeminal ganglion. These brainstem nuclei were identified and analyzed for neuronal density and morphology. No differences between Math1- null mice and littermate controls were noted (Fig. 5). The lack of morphological defects suggests that Math1 has different roles in the brainstem and cerebellum, or that Math1 is required in other unidentified neuronal cells involved in respiratory control. We have shown that the Math1 gene is required for generation of the EGL cells and their derivatives. Although ϳ30 mutations in mice cause cerebellar defects and ataxia15–17, none of these muta- Figure 4 Absence of the EGL in E14.5 Math1 null mice due to lack of proliferation tions causes a phenotype like that of Math1-null mice. For example, of granule precursors. a–d, Generation of the EGL by cells migrating from the the developmental mutant reeler causes a defect in cell migration in rhombic lip (haematoxylin and eosin stain). In coronal sections from wild-type the cerebellum, cerebral cortex and hippocampus, rather than a 18 brain (a, c), the EGL is identified as cells migrating out from the rhombic lip (RL) defect in cell specification or proliferation . In the cerebellum of over the cerebellar primordium (CbP). In Math1 null mice (b, d), the rhombic lip is weaver mutant mice, proliferation of EGL precursor cells is normal, thinner and lacks EGL precursors migrating rostrally. e, f, The absence of the EGL but failure of granule cell migration from the EGL to the IGL causes 19 precursors in Math1 null mice is supported by the lack of proliferating cells in the degeneration of this neuronal population . rhombic lip of these mice, as demonstrated by BrdU staining. Proliferating cells The cerebellum is derived from the posterior end of the mesen- are seen at the rhombic lip and EGL of wild-type mice (e), but not in Math1 null cephalon and from the metencephalon. The expression of wnt-1 at mice (f). Original magnifications: a, b, ×40; c, d, ×200; e, f, ×100. the mesencephalon–metencephalon junction stabilizes the expres- sion of the engrailed homologues En-1 and En-2, thereby specifying the cerebellar anlage20. Absence of Wnt-1 or of En-1 and En-2 causes a loss of cerebellar components, including the deep nuclei, Purkinje cells and IGL21–26. In the absence of Math1, the cerebellum develops but there is selective loss of only the EGL precursors and granule neurons. Other defects, like the ectopic localization of the Purkinje cells and the lack of foliation are most likely to be secondary to the absence of EGL. Mice lacking Mash1, the murine homologue of the Drosophila proneural gene complex achaete-scute, suffer from loss of olfactory neuronal progenitors and are arrested in the differentiation of sympathetic neuronal precursors27. Therefore, Mash1 and Math1 are both essential for the generation of neurons through progression of a differentiation programme, possibly in selected progenitors. On the basis of our results, we propose that Math1 plays a role in cerebellar granule fate specification or proliferation. This role is functionally similar, although not iden- tical, to the role of its homologue, the Drosophila atonal gene4,8. In both mice and fruitflies, loss of function leads to the absence of specific neuronal cell types early in neurogenesis. However, Math1 is expressed in cells that are already committed to be neural tissue, whereas atonal plays a role in the decision of epidermal rather than neuronal cell fate. Ⅺ ......

Methods Figure 5 Inferior olive nuclei are normal in Math1 null mice. Coronal sections from Targeted deletion of Math1. The genomic locus was targeted by homologous brains of Math1 heterozygous (a, c) and null mice (b, d) at E18.5 were stained with recombination using two isogenic flanking fragments of 3.7 kb (EcoRI–EcoRV) haematoxylin and eosin. The inferior olive (IO) nuclei in the medulla oblongata and 3.1 kb (XbaI–ApaI), which were isolated from a mouse 129/SvEv genomic (MO), which arise from the lower rhombic lip (LRL) appear normal in both animals DNA library (Stratagene). A human hypoxanthine–guanine phosphoribosyl- (a–d). Arrows indicate the nucleus tractus solitarius (NS) and nucleus ambiguus transferase (hprt) and the herpes simplex virus thymidine kinase gene were (NA), which appear normal in both animals (a, b). This is in contrast to the lack of used as positive and negative selection markers, respectively. The linearized the external germinal layer (EGL), and the smaller upper rhombic lip (URL) seen in targeting construct was electroporated into hprt-negative AB2.1 embryonic Math1 null mice (b). Original magnifications: a, b, ×40; c, d, ×100. stem (ES) cells and selection was achieved using HAT (hypoxanthine,

Nature © Macmillan Publishers Ltd 1997 NATURE | VOL 390 | 13 NOVEMBER 1997 171 letters to nature

aminopterin and thymidine) and FIAU (1-(2Ј-deoxy-2Ј-fluoro-b-D-arabino- 13. Wood, K. A., Dipasquale, B. & Youle, R. J. In situ labeling of granule cells for apoptosis associated DNA fragmentation reveals different mechanisms of cell loss in developing cerebellum. Neuron 11, 621–632 furanosyl)-5-iodouracil). The targeted alleles can be identified by the presence (1993). of a 14.6-kb fragment instead of 10 kb, when DNA is digested with ApaI and 14. His, W. Die Entwicklung des menschlichen Rautenhirns vom Ende des ersten bis zum Beginn des probed with a 5Ј diagnostic probe, and by a fragment of 6.4 kb instead of 8.5 kb, dritten Monats. I. Verla¨ngertes Mark. Abhandlung der ko¨niglicher sa¨chsischen Gesellschaft der Wissenschaften, mathematische-physikalische Klasse 29, 1–74 (1891). when digested with EcoRI and HindIII and probed with an external 3Ј probe. 15. Heintz, N., Norman, D., Gao, W.-O. & Hatten, M. in genome Maps and Neurological Disorders (eds Homologous recombination was detected in 74% (130/175) of the ES clones Davies, K. E. & Tilghman, S. M.) 19–45 (Cold Spring Harbor Laboratory Press, New York, 1993). 16. Mouse Genome Database 3.1 (Mouse Genome Informatics, The Jackson Laboratory, Bar Horbor, screened. Two independent ES cell lines heterozygous for the targeted Math1 Maine (http://www.informatics.jax.org/), 1996). allele (AT37-G6 and AT37-G12) were injected into blastocysts and gave rise to 17. Frontiers in Bioscience 2.6 (http://www.bioscience.org/knockout/knochome.html) (1997). chimaeric mice, which transmitted the deletion through the germline. 18. Goffinet, A. M. Events governing organization of postmigratory neurons: studies on brain develop- ment in normal and reeler mice. Brain Res. Rev. 7, 261–296 (1984). Histology, immunohistochemistry and in situ hybridization. Pregnant 19. Rezai, Z. & Yoon, C. H. Abnormal rte of granule cell migration in the cerebellum of weaver mutant females were killed and embryos removed quickly into cold PBS. Extra- mice. Dev. Biol. 29, 17–26 (1972). 20. Joyner, A. L. Engrailed, Wnt and Pax genes regulate –hindbrain development. Trends Genet. embryonic membranes were discarded, and the yolk sac was collected for 12, 15–20 (1996). DNA extraction. Embryos or newborn mice were fixed overnight in either 21. McMahon, A. P. & Bradley, A. The Wnt-1 (int-1) proto-oncogene is required for development of a Bouin’s fixative or 4% paraformaldehyde in PBS at 4 ЊC, dehydrated and stored large region of the mouse brain. Cell 62, 1073–1085 (1990). − 22. Thomas, K. R. & Capecchi, M. R. Targeted disruption of the murine int-1 proto-oncogene resulting in at 20 ЊC. Embryos were embedded in paraffin wax, and sections of 7 ␮m were cut. severe abnormalities in midbrain and cerebellar development. Nature 346, 845–850 (1990). Sections were stained in Mayer’s haematoxylin and eosin or cresyl violet. Isotopic 23. Thomas, K. R., Musci, T. S., Neumann, P. E. & Capecchi, M. R. Swaying is a mutant allele of the proto- 35 28 oncogene Wnt-1. Cell 67, 969–976 (1991). in situ hybridization with RNA probes labelled with [ S]UTP has been described . 24. McMahon, A. P., Joyner, A. L., Bradley, A. & McMahon, J. A. The midbrain–hindbrain phenotype of Non-radioactive in situ hybridization using the digoxigenin system Wnt-1-/Wnt-1- mice results from stepwise deletion of engrailed-expressing cells by 9.5 days (Boehringer Mannheim) was done according to ref. 29, but with paraffin postcoitum. Cell 69, 581–595 (1992). 25. Millen, K. J., Wurst, W., Herrup, K. & Joyner, A. L. Abnormal embryonic cerebellar development and embedding. Sections were dewaxed (Histoclear, National Diagnostics), rehy- patterning of postnatal foliation in two mouse Engrailed-2 mutants. Development 120, 695–706 drated through a graded series of ethanol concentrations, and processed (1994). 26. Wurst, W., Auerbach, A. B. & Joyner, A. L. Multiple developmental defects in Engrailed-1 mutant mice: according to the protocol. Templates for cRNA synthesis were as follows: an early mid-hindbrain deletion and patterning defects in forelimbs and sternum. Development 120, Math1 (PCR-amplified coding region) and RU49 (0.847-kb EcoRI/PstI frag- 2065–2075 (1994). ment, a gift from N. Heintz). For immunohistochemical staining, paraffin 27. Guillemot, F. et al. Mammalian achaete-scute homolog 1 is required for the early development of olfactory and autonomic neurons. Cell 75, 463–476 (1993). blocks were cut at 5 ␮m, dewaxed in xylene, antigen-retrieved by microwaving 28. Albrecht, U., Eichele, G., Helms, J. & Lu, H. in Visualization of Gene Expression Patterns by in situ in citrate buffer, blocked and incubated with the appropriate antibodies in PBS Hybridization (ed. Daston, G. P.) 23–48 (CRC Press, Boca Raton, Florida, 1997). 29. Schaeren-Wiemers, N. & Gerfin-Moser, A. A single protocol to detect transcripts of various types and containing 10% normal goat serum for several days at 4 ЊC. Primary antibodies expression levels in neural tissue and cultured cells: in situ hybridization using digoxigenin-labelled were detected using the avidin–biotin complex system (ABC, Vector Lab). cRNA probes. Histochemistry 100, 431–440 (1993). Antibodies used were as follows: anti-calbindin-D (1:1,000; Sigma), anti-GFAP Acknowledgements. We thank B. A. Antalffy for technical assistance; P. R. Cox and J. Dong for their help; (1:1,000; Dako) anti-nestin (1:150, a gift from R. McKay). and R. R. Behringer, K. A. Mahon and B. Hassan for critical reading of the manuscript. H.Y.Z. is an investigator and H.J.B. is an associate investigator of the Howard Hughes Medical Institute. This work was For evaluation of brainstem cranial nerve nuclei, coronal sections (6 ␮m) supported by grants from the NIH/NINDS and by the Baylor Mental Retardation Research Center. were cut from E18.5 fetuses and alternative sections were stained with Correspondence and requests for materials should be addressed to H.Y.Z. (e-mail: hzoghbi@bcm. haematoxylin and eosin, cresyl violet (Nissl staining) and one-step trichrome tmc.edu). (Gomori’s staining). Cranial nerve nuclei were identified and scored independently by two individuals. Criteria for the examination were: localization of each cranial nerve nucleus, symmetry, shape and size, and neuronal density. Impaired mast cell-dependent BrdU staining of proliferating cells. Pregnant females were injected with − BrdU (5-bromodeoxyuridine) (50 mg kg 1 body weight from a stock solution natural immunity in − of 10 mg ml 1 in 0.1M Tris-HCl, pH 7.5) and killed two hours later. Embryos were dissected in cold PBS, fixed in 70% ethanol, dehydrated, embedded in complementC3-deficientmice paraffin wax and cut at 7-␮m sections. These were treated with 2M HCl for 30 min at 37 ЊC to denature the DNA partially, then neutralized in 0.1M sodium Andrey P. Prodeus*, Xiaoning Zhou*, Marcus Maurer†, borate, pH 8.5, for 10 min, as recommended by the supplier of the anti-BrdU Stephen J. Galli*† & Michael C. Carroll* antibodies (Novocastra Lab). BrdU-labelled cells were detected using the ABC * Departments of Pathology, Harvard Medical School and † Beth Israel Deaconess system. Medical Center, Boston, Massachusetts, 02115, USA ...... Received 19 May; accepted 7 August 1997. The complement system is widely regarded as essential for normal 1. Altman, J. & Bayer, S. A. in Development of the Cerebellar System: In Relation to its Evolution, Structure, and Functions (CRC Press, Boca Raton, Florida, 1996). inflammation, not least because of its ability to activate mast 2. Akazawa, C., Ishibashi, M., Shimizu, C., Nakanishi, S. & Kageyama, R. A. Mammalian helix–loop– cells1–5. However, recent studies have called into question the helix factor structurally related to the product of Drosophila proneural gene atonal is a positive transcriptional regulator expressed in the developing nervous system. J. Biol. Chem. 270, 8730–8738 importance of complement in several examples of mast cell- 6–9 (1995). dependent inflammatory responses . To investigate the role of 3. Ben-Arie, N. et al. Evolutionary conservation of sequence and expression of the bHLH protein Atonal complement in mast cell-dependent natural immunity, we exam- suggest a conserved role in neurogenesis. Hum. Mol. Genet. 5, 1207–1216 (1996). 10,11 4. Jarman, A. P., Grau, Y., Jan, L. Y. & Jan, Y. N. atonal is a proneural gene that directs chordotonal organ ined the responses of complement-deficient mice to caecal formation in the Drosophila peripheral nervous system. Cell 73, 1307–1321 (1993). ligation and puncture12, a model of acute septic peritonitis12,13 that 5. Hatten, M. E. & Heintz, N. Mechanisms of neural patterning and specification in the developing cerebellum. Annu. Rev. Neurosci. 18, 385–408 (1995). is dependent on mast cells and tumour necrosis factor-␣ (TNF-␣). 10,11 6. Hallonet, M. E. & Le Dourain, N. M. Tracing neuroepithelial cells of the mesencephalic and We found that C4- or C3-deficient mice were much more metencephalic alar plates during cerebellar ontogeny in quail-chick chimaeras. Eur. J. Neursci. 5, sensitive to caecal ligation and puncture than wild-type (WT) 1145–1155 (1993). 7. Hatten, M. E., Alder, J., Zimmerman, K. & Heintz, N. Genes involved in cerebellar cell specification controls (100% versus 20% in 24-h mortality, respectively). C3- and differentiation. Curr. Opin. Neurobiol. 7, 40–47 (1997). deficient mice also exhibited reductions in peritoneal mast cell 8. Jarman, A. P., Sun, Y., Jan, L. Y. & Jan, Y. N. Role of the proneural gene, atonal, in formation of Drosophila chordotonal organs and photoreceptors. Development 121, 2019–2030 (1995). degranulation, production of TNF-␣, neutrophil infiltration and 9. Yang, X. W., Zhong, R. & Heintz, N. Granule cell specification in the developing mouse brain as clearance of bacteria. Treating the C3-deficient mice with purified defined by expression of the zinc finger transcription factor RU49. Development 122, 555–566 (1996). C3 protein enhanced activation of peritoneal mast cells, TNF-␣ 10. Debus, E., Weber, K. & Osborn, M. Monoclonal antibodies specific for glial fibrillary acidic (GFA) protein and for each of the neurofilament triplet polypeptides. Differentiation 25, 193–203 (1983). production, neutrophil recruitment, opsonophagocytosis of bac- 11. Tohyama, T. et al. Nestin expression in embryonic human neuroepithelium and in human teria and resistance to caecal ligation and puncture, confirming neuroepithelial tumor cells. Lab. Invest. 66, 303–313 (1992). 12. Smeyne, R. J. & Goldowitz, D. Development and death of external granular layer cells in the weaver that the defects were complement-dependent. These results pro- mouse cerebellum: a quantitative study. J. Neurosci. 9, 1608–1620 (1989). vide formal evidence that complement activation is essential for

Nature © Macmillan Publishers Ltd 1997 172 NATURE | VOL 390 | 13 NOVEMBER 1997