Proc. Nail. Acad. Sci. USAVol. 82, pp. 7782-7786, November 1985Neurobiology

Ontogeny of electrically excitable cells in culturedolfactory epithelium

(neurogenesis/cell culture)

D. SCHUBERT,* W. STALLCUP, M. LACORBIERE, Y. KIDOKORO, AND L. ORGEL

The Salk Institute, P.O. Box 85800, San Diego, CA 92138

Communicated by W. Maxwell Cowan, July 2S, 1985

ABSTRACT A primary system has been developed inwhich it is possible to study the production of electricallyexcitable neuron-like cells from a precursor population ofolfactory epithelial cells. Rat nasal epithelium was dissociatedand placed in culture. The initial surviving cells are flat andciliated and contain glial fibrillary acidic protein (GFAP). After3-5 days electrically excitable cells appear that contain neuron-specific enolase but not GFAP. These round cells originate bymeans of the differentiation of the GFAP-positive flat cell to around cell, followed by the division of the round cell. There-fore, neuron-like cells can be derived from cells that synthesizeGFAP.

The olfactory epithelium ofmost animals contains three typesof cells-receptor, sustentacular, and basal. The receptorcells contain sensory receptors on their epithelial surface andextend their axons directly to the brain. These neuronsoccupy about 70% of the total epithelial volume. Thesustentacular cells, whose function is unknown, but whichare thought to resemble glial or Schwann cells in the nervoussystem, are interdigitated between the receptor cells. Under-lying the receptor and supporting cells are the basal cells (see,for example, refs. 1 and 2). When the olfactory axons in miceare lesioned, the receptor cells degenerate and are replacedthrough the mitosis of the basal cells and the subsequentdifferentiation of the daughter cells into receptor cells (3, 4).The question of whether or not the basal cells in normal adultolfactory epithelium are able to divide and give rise tosensory neurons was answered by autoradiographic tech-niques following pulse labeling of animals with [3H]thymi-dine. The results in mice showed that most of the isotopiclabeling of nuclei occurred in the basal cell compartment (1).The labeled nuclei later appeared in receptor cells and, to alesser extent, in sustentacular cells. These results, andsimilar data in frogs (2), strongly suggest that receptor andsustentacular cells are continuously renewed by the divisionof basal cells. Since the olfactory epithelium is continuallyproducing neurons from a less-differentiated cell type, itwould be an excellent system in which to study neurogenesisif it were possible to reproduce this event in culture. Ideally,one would like to have primary cultures in which nerve cellsare produced from their precursors as well as a method forestablishing permanent cell lines of the precursor and prog-eny populations. The experiments outlined here describe aprimary culture system in which the formation of electricallyexcitable cells can be examined.

MATERIAL AND METHODSCell Cultures. Olfactory epithelia were dissected from

2-month-old Sprague-Dawley rats. The tissue was dissoci-ated with 20% pancreatin (GIBCO) for 45 min at 37°C in

Hepes-buffered Vogt-Dulbecco modified Eagle's medium.The dissociated cell suspension was washed twice withmodified Eagle's medium containing 10% cadet calf serum(Biocell Laboratories), plated on 60-mm plastic tissue culturedishes (Falcon) at about 106 cells per dish, and cultured at370C. A variety of other dissociation procedures, sera,substrata, and culture media was tested; none was effectivein producing the cell populations described below. Macro-phage cultures were prepared by the method of Wahl et al.(5). B65 is a rat neuroblastoma, B15 is a rat glioma, and C1is a mouse lymphoid cell line (6, 7).

Immunohistochemistry. Rabbit anti-glial fibrillary acidicprotein (GFAP) was obtained from A. Bignami, mousemonoclonal anti-rat T200 protein was obtained from R.Hyman, and monoclonal anti-S100 and anti-neuron-specificenolase was from B. Boss. For staining with anti-T200,cultures were incubated with the primary antibody for 30 minat room temperature. The cultures were then washed andexposed for an additional 30 min to a rhodamine-labeled goatantibody against mouse immunoglobulin (Tago, Burlingame,CA). Finally, the cultures were washed, air dried, and fixedwith 95% ethanol for 10 min at -20°C. For labeling withanti-GFAP, cultures were first fixed with 95% ethanol for 10min at -20°C, dried, and then incubated for 30 min with theprimary antiserum. After washing, the cultures were exposedto a fluorescein-labeled goat antibody against rabbit immu-noglobulin for an additional 30 min, followed by morewashing and a final air drying. For double-labeling the cellswere first stained with anti-T200, fixed, and then stained withanti-GFAP. Staining with anti-S100 and anti-nerve-specificenolase was done according to published procedures (8).Time-Lapse Photography. Time-lapse photography was

done on a Zeiss Universal microscope equipped with aCohu-4400 video camera and a Panasonic time-lapse videorecorder. The cells were cultured on glass coverslips instandard medium and maintained at 37°C and at the properpH with a Bachofer culture chamber.Thynudine Labeling. Cells were labeled in complete growth

medium for 4 hr with 2 ,Ci (1 Ci = 37 GBq) of [3H]thymidineper ml and fixed with ethanol, and the presence of isotopi-cally labeled nuclei was determined by autoradiography. Atleast 500 cells were examined under each growth condition.

Carnosine Synthesis. To measure the rate of carnosinesynthesis, cells were labeled for 6 hr with [14C]histidine (2,uCi/ml) and [3H]leucine (10 ,Ci/ml) in Vogt-Dulbeccomodified Eagle's medium containing 1% of the normalamount of histidine. The cells were then washed once withthe culture medium and the amount of [3H]leucine incorpo-rated into protein was determined by trichloroacetic acidprecipitation of an aliquot of cells. The remaining cells wereextracted with ethanol and the amount of [14C]histidineincorporated into carnosine was determined by high-voltage

Abbreviations: GFAP, glial fibrillary acidic protein; Bt2cAMP,N6,02'-dibutyryladenosine 3',5'-cyclic monophosphate.*To whom reprint requests should be addressed.

7782

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Janu

ary

5, 2

020

Proc. Natl. Acad. Sci. USA 82 (1985) 7783

electrophoresis (9). The data are expressed as the ratio ofisotope incorporated into carnosine to that of total proteinsynthesis.

Electrophysiology. Intracellular recording of the membranepotential was performed with conventional procedures (10).The microelectrodes were filled with 4 M potassium acetateand had a resistance of about 80 MC. The bath saline had thefollowing salt composition (in mM): NaCl, 142.6; KCI, 5.6;CaCl2, 10; Hepes/NaOH, 5 (pH 7.4). The experiments werecarried out at room temperature (20-220C). Successful re-cordings were made from 17 round cells; all gave similarresults.

RESULTSCulture System. Primary cultures of olfactory epithelia

were made from juvenile Sprague-Dawley rats. The epitheliawere dissociated and plated into tissue culture dishes con-taining modified Eagle's medium and 10% newborn calfserum. After 3 days in culture the majority of the plated cellswere dead, but small patches of cells were attached to theculture dish substratum (Fig. 1A). The majority of these cellshad actively beating cilia on their upper surface. After 5 dayssome of the cells within the patches divided and migrated outfrom the initial clumps, frequently forming a semiconfluentmonolayer of very large (100-150 Am in diameter) flat cells(Fig. 1B). If N6,02'-dibutyryladenosine 3',5'-cyclic mono-phosphate (Bt2cAMP) was added to this cell population aretraction ofthe plasma membrane and cytoplasm around the

_~~~~~~~*

FIG. 1. Olfactory epithelium primary cultures. The olfactoryepithelium of a 2-month-old rat was dissociated and placed in cultureand growth was followed by phase microscopy. The smallest roundcells are 20 urm in diameter. (A) Three days after plating. (B) Fivedays after plating. (C) Eight days after plating. (D) Round cellsseparated from flat cells of an 8-day culture and replated in a Petridish.

cells' cytoskeleton was initiated within 10 min (Fig. 2A) andwas complete within 2 hr (Fig. 2B). A similar cAMP-inducedcytoplasmic "shrinkage" has been reported in cultured glialcells (11, 12).Also at 5 days a population of loosely attached phase-bright

round cells appeared in the cultures. These increased in numberfor several days (Fig. 1C). By gentle trituration it was possibleto remove the majority of the round cells from the attached flatcells, thus isolating a morphologically homogeneous populationof round cells (Fig. ID). Several weeks after they were pro-duced, some of the round cells attached to the substratum andgrew short neurite-like processes (Fig. iD Inset).

Electrophysiology and Neurotrans itter Synthesis. Twoproperties of olfactory neurons are electrical excitability andthe ability to synthesize the f-alanyl-L-histidine dipeptidecarnosine, which is thought to be the primary neurotrans-mitter in this cell type (13). When the round cells wereexamined by electrophysiological techniques, they werefound to be able to generate action potentials following thepassage of a depolarizing current pulse (Fig. 3). It has beenimpossible to obtain stable, reproducible recordings from theflat cell population. To assay the rate of carnosine synthesisrelative to total protein synthesis in the small amounts ofavailable cells, cultures were labeled with both [14C]histidineand [3H]leucine, and the amount of each incorporated intocarnosine and total protein, respectively, was determined.Table 1 shows that the round cell population synthesizessignificantly more carnosine than the flat cells. The ratneuroblastoma, glioma, and mouse lymphoid cells synthe-sized relatively little of the dipeptide.Immunological Properties of Cells. The ability of the round

cell population to produce action potentials suggests that theyare neuronal. However, since lymphoid cells such as mac-rophages are also able to generate action potentials (14), cellswere stained by fluorescent antibody techniques for a varietyof nervous system-specific antigens. These included GFAP,which is specific for astrocyte intermediate filaments (15),S100, which is found in some glia (16), and neuron-specificenolase, which is uniquely associated with neurons in pri-mary cultures (8). The initially attached cells, as well as thehomogeneous population of flat cells that grew out fromthem, stained with antiserum against GFAP (Fig. 4) but notwith anti-S100 or anti-neuron-specific enolase. The round cellpopulation stained with neither anti-GFAP nor anti-S100(Fig. 5D). The round cells did, however, stain with anti-neuron-specific enolase (Fig. SB). The round cells did notstain with IgG coupled to the fluorochrome, indicating thatthey did not possess Fc receptors, which are characteristic ofmacrophages. To further rule out the possibility that theround cells were macrophage-like, both nasal epitheliumcultures and bona fide peritoneal macrophage cultures werestained with a monoclonal antibody against rat T200, anantigen that is found on all lymphoid-derived cells (17).Although the rat macrophages stained intensely with thisantiserum, the round cells in the nasal epithelium cultures didnot stain. In contrast, the flat cells in the epithelium culturesdid stain with anti-T200. This label occurred in the same cellsthat contained GFAP (Fig. 4). Peritoneal macrophages didnot stain with anti-GFAP. These immunological data showthat the round cell population does not have the surfacecharacteristics of macrophages, whereas the flat cells con-tained both the glia marker, GFAP, and the T200 antigen.

Origin of the Round Cells. The ontogeny of the round cellswas examined by three methods-radioautography using[3H]thymidine, time-lapse photography, and the use ofmicrocultures containing single cells (Table 2). All threetechniques showed that the round cells were derived from theflat cells by means of the differentiation of the flat cell into around cell, which was then able to divide three to five times,giving rise to 8-32 additional round cells. When cells from the

Neurobiology: Schubert et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

5, 2

020

7784 Neurobiology: Schubert et al.

FIG. 2. Alteration of flat cell morphology byBt2cAMP. Bt2cAMP (0.5 mM) was added to 5-daycultures (Fig. 1B) and the attached, flat cells were

photographed at 10 min (A) and 2 hr (0) after the additionof the reagent.

attached cell population of 5-day cultures, which containedonly flat, GFAP-positive cells, were plated at 1 cell per wellin Terasaki dishes in 20 Aul of culture medium and the cell wasfollowed as a function of time, three classes of results were

obtained. In a quarter of the cases the individual cell died. Inhalf ofthe 20 examples the flat cell became round and dividedseveral times to give rise to 8, 16, or 32 round cells. No furtherdivisions were ever observed. In 4 cases the flat cell dividedonce, one of the sister cells became round, and it divided,leaving 8, 16, or 32 round cells and 1 flat cell in the microwell.Finally, in one culture the flat cell divided twice, and 2 flatcells gave rise to round cells, leaving 2 flat and z20 roundcells. These results were compatible with' time-lapse studies,which showed that both the flat cells and the round cells candivide, and with data obtained from radioautographic studies.When flat cells of 5-day cultures were pulse-labeled for 4 hrwith [3H]thymidine and the cultures were examined forisotopically labeled'nuclei, nuclei ofthe flat cells were labeledas well as those of the round cells that appeared in thecultures 1-2 days later (Table 2). Thymidine labeling ofhomogeneous round cell populations indicated that they werealso synthesizing DNA. These data show that both GFAP-positive flat cells and the electrically excitable round cells arecapable of dividing and that the round cell population isderived from the flat cells.

DISCUSSIONWhen adult rat olfactory epithelium is dissociated and placedin cell culture, some of the cells attach and create a popula-

20 ms0.2 nA20 mV

FIG. 3. Action potential recorded from a round cell. The actionpotential was evoked by passing depolarizing current through therecording electrode using a bridge circuit. The upper trace shows thecurrent pulse as well as the zero potential level. The lower trace is

the membrane potential. The resting membrane potential was -76

mV and the input resistance was 160 MQ.

tion of flat cells, which, in turn, give rise to loosely attached,phase-bright round cells (Fig. 1). The population of flat cellsis relatively homogeneous, as defined by their sensitivity toBt2cAMP (Fig. 2) and the presence of GFAP and T200antigens (Fig. 4), whereas the round'cell population iselectrically excitable (Fig. 3) and contains a protein onlyfound in nerve cells, neuron-specific enolase (Fig. 5).Through a series of experiments involving radioautography,time-lapse photography, and single-cell cultures, it has beenfound that the round cells are derived from the flat cells bymeans of the differentiation of a flat cell into a round one andits subsequent division (Table 2). These results suggest thatround, electrically excitable neuron-like cells can be derivedfrom a GFAP-positive precursor population.The origin of neuroblasts within the neuroepithelium of the

developing central nervous system has been debatedthroughout the last century. Neurons and glia may share a

common dividing percursor (Fig. 6, cell A; cells B and C neednot divide) or they may originate directly from distinctdividing cell populations (Fig. 6, cells B and C). His (18)suggested that the proliferative zone contains two distinct celltypes, one that gives rise to neurons and the other that givesrise to glia (Fig. 6, cells B and C). In contrast, Schaperproposed that the neuroepithelium is initially a uniformpopulation of cells that is able to generate both nerve and glialcells (refs. 19 and 20; Fig. 6, cell A). Electron microscopicstudies of the neuroepithelium of fetal monkeys have shownthat mitotic cells containing the glial intermediate filament

Table 1. Carnosine synthesis by cells in culture['4C]Carnosine/

Cell type [3H]proteinThree-day culture 2.1Five-day culture 1.8Nine-day flat cells 1.0Nine-day round cells 8.2B65 neuroblastoma 0.4B15 glioblastoma 0.5C1 lymphoid cell 0.2

Cells were labeled for 6 hr with [14C]histidine and [3H]leucine. Theradioactivity in cellular carnosine was determined by high-voltageelectrophoresis and amount of [3H]leucine incorporated into proteinwas determined by trichloroacetic acid precipitation. The data werecorrected for overlap and counting efficiencies and are expressed asthe ratio of isotope in carnosine relative to isotope in total protein.The experiment was performed three times with similar results.

Proc. Natl. Acad. Sci. USA 82 (1985)

Dow

nloa

ded

by g

uest

on

Janu

ary

5, 2

020

Proc. Natl. Acad. Sci. USA 82 (1985) 7785

FiG. 4. Staining of flat cells with anti-GFAPand anti-T200. Five-day cultures were double-labeled with fluorescein anti-GFAP and rho-damine anti-T200 and were examined by fluores-cence microscopy. (a) Phase-contrast. (b) Anti-GFAP. (c) Anti-T200.

protein GFAP are intermingled with mitotic GFAP-negativecells (21). On the basis ofthese results it was argued that thereare separate stem cells for neurons (the GFAP-negative cells)and for glia. In contrast, by using S100 protein and nerve-specific enolase as specific markers for glia and nerve,respectively, it has been shown by immunohistochemicaltechniques that both marker proteins appeared in the samecells of the mouse neuroepithelium during early stages ofdevelopment (Fig. 6, cell A) but were uniquely localized tonerve or glia in mature animals (22). In addition, a clonal stemcell line, called RT4-AC, has been described that is capableof producing both nerve and glial phenotypes (23). In thiscase, the parental cell line synthesizes the S100 protein andis also weakly electrically excitable (Fig. 6, cell A). With timein culture, these cells spontaneously generate clones ofmorphologically distinct cells that are either S100 positiveand electrically passive (Fig. 6, cell C, glial-like) or S100negative and electrically active (Fig. 6, cell B, nerve-like).Another clonal precursor cell line, called F7, has beenisolated from the mouse hypothalamus; by manipulation ofthe culture conditions it is possible to convert this cell line to

FIG. 5. Staining of round cells with anti-nerve-specific enolaseand anti-S100. Eight-day cultures were labeled with rhodamineanti-nerve-specific enolase or rhodamine anti-S100. (A) Phase-contrast, anti-nerve-specific enolase. (B) Fluorescence, nerve-speci-fic enolase. (C) Phase-contrast, anti-S100. (D) Fluorescence, anti-S100.

cells that have exclusively glial or neuronal properties (24).Phenotypically stable clonal cell lines have also been isolatedfrom the rat central nervous system (6), and many of theselines have the properties ofboth nerve and glia. For example,some clonal lines have both Na+ and K+ channels charac-teristic of neurons but also have surface antigens normallyassociated with glial cells (25, 26). These cell lines probablyrepresent a phenotype that is arrested in the normal devel-opmental lineage by the transformation event. The clonal celllines with intermediate phenotypes were used to generateantisera that recognize incompletely differentiated neural cellpopulations in primary cultures and in nerve-producingteratoma cells (27, 28). Collectively, the data outlined aboveshow that individual stem cells in the germinal layer can havetraits associated with both neurons and glia and that they cangive rise to both phenotypes. Our findings on cultured cellsfrom the nasal epithelium are consistent with these observa-tions.The data derived from cultured olfactory neuroepithelium

show that GFAP-positive cells are'capable of dividing andthat at least a subset of this population can give rise to roundneuron-like cells. The criteria used to identify the round cellsas neuron-like were their ability to generate action potentialsand the fact that they can be stained with a monoclonalantibody against neuron-specific enolase. Since macro-phages are able to generate action potentials under someconditions (14), it was critical to rule out the possibility thatthe round cells were simply macrophages. It is unlikely thatthe round cells are macrophages for the following reasons. (i)They do not stain with an antiserum (anti-T200) that recog-nizes rat macrophages and other lymphoid cell types. (ii) Theround cells do not bind fluorescently labeled IgG to their

Table 2. Cell division within the two olfactoryepithelium populations

[3H]Thy-midine,* Time-lapse% labeled photog-nuclei Terasaki dishest raphy*

Flat cells 3.5 No division, 10 cells Cells divide(5 days) One division, 4 cells

Two divisions, 1 cellRound cells 5.1 Three divisions, 5 cells Cells divide

(7 days) Four divisions, 8 cellsFive divisions, 2 cells

*Cells were labeled for 4 hr with [3H]thymidine and processed forradioautography.

tCells were plated individually in 20 Al ofculture medium in Terasakidishes.*Time-lapse photography was done under normal cell culture con-ditions using a video recording system.

Neurobiology: Schubert et al.

I

Dow

nloa

ded

by g

uest

on

Janu

ary

5, 2

020

7786 Neurobiology: Schubert et al.

I

*N%

Neuron

i

T

p.,l

I

Glial cell

c

A

FIG. 6. Generation of nerve and glial cells. A signifies a dividingprecursor to glial and nerve precursors. The intermediate nerve (B)and glia (C) may or may not divide.

stuface, indicating that they do not have Fc receptors, whichare characteristic of imacrophages. (iii) The round cells growin suspension, whereas macrophages tend to attach to thesubstratum. (iv) If the. round cells are macrophages, theymust be the differentiated product of gi cell that containsGFAP, for the round cell precursor is GFAP positive.Cultures of identified rat peritoneal macrophages do notcontain OFAP-positive cells. (v) The round cells synthesizesignificantly more carnosine than lymphoid cells. It is there-fore unlikely that the round cells are of lymphoid origin.

In contrast to the round cells, the flat cell precursors arestained with both anti-T200 and anti..GFAP. To our knowl-edge, the presence of the T200 antigen in nonlymphoid cellshas not been reported previously, but it is not uncommon forlymphoid cell-surface molecules to be associated with cells ofthe nervous system (see,. for example, ref. 29). AlthoughGFAP is generally thought to be a specific marker fordifferentiated astrocytes (30), it has been found in a dividingcell population in the ventricular zone of the mammaliancerebral cortex (21) and in the ependymal lining of postnatalrat spinal cord (31). In addition, it has been observed in theepithelium of the mouse lens, a nonneuropal tissue (32), andthe supportive cells of the pineal gland (33). The pinealsupportive cells are analogous to the sustentacular cells of theolfactory epithelium in that they surround the (neuronal)photoreceptor cell. It follows that the presence of GFAP ina cell is no longer sufficient to identify it as a fully differen-tiated glial cell.Although the differentiation of putative olfactory receptor

cells in organ cultures of embryonic rat olfactory epitheliumhas been described (34), there is a paucity of published workon the culture of mammalian olfactory tissue. This is prob-ably due to the difficulties associated with dissociating thetissue and maintaining high cell viability. In our preparations

cell viability is very poor, but the cell culture conditions usedapparently select for cells that are able to divide in culture andthat are able to give rise to a cell type that is electricallyexcitable, contains neuron-specific enolase, and is capable ofsynthesizing the olfactory neuron dipeptide carnosine. Al-though it is not possible at present to directly relate thephenotypes of the cultured cells to those in vivo, our datashow that a cell that has the phenotype of a neuroblast can bederived from a dividing precursor population that containsthe glial intermediate' filament protein GFAP.We thank Dr. M. Cowan for showing us the dissection of rat nasal

epithelium and Drs. M. Cowan, J. H. Steinbach, S. Halegoua, andJ. Levine for their critical reading of the manuscript. This work wassupported by grants from the Muscular Dystrophy Association andthe National Institutes of Health.

1. Moulton, D. G. (1975) Ann. N. Y. Acad. Sci. 237, 52-61.2. Graziadei, P. P. C. (1973) Tissue Cell 5, 113-131.3. Nagahara, Y. (1940) Jpn. J. Med. Sci. 5 5, 165-199.4. Graziadei, P. P. C. & Metcalf, J. F. (1971) Z. Zellforsrh.

Mikrosk. Anqt. 116, 305-318.5. Wahl, L. M., Wahl, S. M., Mergenhagen, S. E. & Martin,

G. R. (1974) Proc. Natl. Acad. Sci. USA 71, 3598-3601.6. Schubert, D., Heinemann, S., Carlisle, W., Tarikas, H.,

Kimes, B., Patrick, J., Steinbach, J. H., Culp, W. & Brandt,B. L. (1974) Nature (London) 249, 224-226.

7. Horibata, K. & Harris, A. (1970) Exp. Cell Res. 60, 61-77.8. Haan, E. A., Boss, B. D. & Cowan, M. W. (1982) Proc. Natl.

Acad. Sci. USA 79, 7585-7589.9. Pisano, J. J., Wilson, J. D., Cohen, L., Abraham, D. &

Udenfriend, S. (1960) J. Biol. Chem. 236, 499-554.10. Kidokoro, Y. (1980) Dev. Biol. 78, 231-241.11. Steinbach, J. H. & Schubert, D. S. (1975) Exp. Cell Res. 91,

449-453.12. Goldman, J. E. & Chiu, F. (1984) Brain Res. 306, 85-95.13. Margolis, F. L. (1974) Science 184, 909-911.14. McCann, F. V., Cole, J. J., Guyre, P. M. & Russell, J. (1983)

Science 219, 991-993.15. Bignami, A., Eng, L., Dahl, D. & Uyeda, C. (1972) Brain Res.

43, 429-435.16. Zomzely-Neurath, C. E. & Walker, W. A. (1980) in Proteins

of the Nervous System, eds. Bradshaw, R. A. & Schneider,D. M. (Raven, New York), 2nd Ed., pp. 1-57.

17. Sunderland, C., McMaster, W. & Williams, A. (1979) Eur. J.Immunol. 9, 155-159.

18. His, W. (1889) Abh. Math. Phys. Kl. Saechs. Ges. Wiss. 15,313-372.

19. Schaper, A. (1897) Arch. Entwicklungsmech. Org. 5, 81-132.20. Schaper, A. (1897) Science 5, 430-431.21. Levitt, P., Cooper, M. L. & Rakic, P. (1981) J. Neurosci. 1,

21-39.22. DeVitry, F., Picart, R., Jacque, C., Legault, L., Dupouey, P.

& Tixier-Vidal, A. (1980) Proc. Natl. Acad. Sci. USA 77,4165-4169.

23. Tomozawa, Y. & Sueoka, N. (1978) Proc. Natl. Acad. Sci.USA 75, 6305-6309.

24. DeVitry, F., Delaunoy, J. P., Thibault, J., Buisson, N.,Lamando, N., Legault, L., Delasalle, A. & Dupouey, P. (1983)EMBO J. 2, 199-203.

25. Arner, L. S. & Stallcup, W. B. (1981) Dev. Biol. 83, 138-145.26. Wilson, S., Baetge, E. E. & Stallcup, W. B. (1981) Dev. Biol.

83, 146-153.27. Stallcup, W. (1981) Dev. Biol. 83, 154-165.28. Darmon, M., Stallcup, W. B. & Pittman, Q. J. (1982) E~xp. Cell

Res. 138, 73-78.29. Lesley, J. F. & Lennon, V. A. (1978) Brain Res. 153, 109-120.30. Eng, L. (1980) in Proteins of the Nervous System, eds.

Bradshaw, R. A. & Schneider, D. M. (Raven, New York), 2nded., pp. 85-117.

31. Bullon, M. M., Alvarez-Gago, T., Fernandez-Ruiz, B. &Aguirre, C. (1984) Dev. Brain Res. 14, 129-133.

32. Hatfield, J. S., Skoff, R. P., Maisel, H. & Eng, L. (1984) J.Cell BiQI. 98, 1895-1898.

33. Higley, H. R., McNulty, J. A. & Rowden, G. (1984) BrainRes. 304, 117-120.

34. Farbman, A. I. (1977) Anat. Rec. 189, 187-200.

Proc. Natl. Acad. Sci. USA 82 (1985)

B

Dow

nloa

ded

by g

uest

on

Janu

ary

5, 2

020