The Neuronal Stem Cell of the Olfactory Epithelium

at a particular level. These studies suggest that OE ABSTRACT: The vertebrate olfactory epithelium neuronal progenitors—which are in close physical (OE) is a system in which behavior of neuronal pro-proximity to ORNs—can ‘‘read’’ the number of dif- genitor cells can be observed and manipulated easily. ferentiated neurons in their environment and regulate It is morphologically and functionally similar to em-production of new neurons accordingly. Putative neu- bryonic germinal neuroepithelia, but is simpler in that ronal stem cells of the OE have been identiﬁed in vitro, it produces large numbers of a single type of neuron, and studies of these cells indicate that ORNs produce the olfactory receptor neuron (ORN). The OE is ame-a signal that feeds back to inhibit neurogenesis. This nable to tissue culture, gene transfer, and in vivo surgi-inhibitory signal may be exerted at the level of the cal approaches, and these have been exploited in ex-stem cell itself. Recent studies to identify this signal, periments aimed at understanding the characteristics as well as endogenous stimulatory signals that may of OE neuronal progenitor cells. This has led to the be important in regulating OE neurogenesis, are also realization that the ORN lineage contains at least discussed. q 1998 Wiley Sons, three distinct stages of proliferating neuronal progeni-205, 1998 tor cells (including a stem cell), each of which repre-Keywords: sents a point at which growth control can be exerted. growth factors; morphogenetic studies in vivo have is trophins; C; transgenic mouse of ORNs the ‘‘globose’’ basal cells, which lie ORNs) them. The globose basal cells are a heterogeneous popula-of rodent. tion, consisting of neuronal colony-forming cells (CFU), ORNs in the OE on the MASH1-expressing neuronal progenitors (M1), and im-bulb die mediate neuronal precursors (INP). These cell types their roles in olfactory neurogenesis cells such on explants Cy3-indirect evidence for the a neuronal attempt to isolate and these demarcates cells the stage; this stage appears to be interposed between am-neuronal stem cell (tentatively identiﬁed as the neuronal colony-forming cell or neuronal CFU) and the concern-neuronal precursor (INP). INPs are the last proliferating cells in the ORN lineage: they undergo one or more Plac-rounds of division, after which their progeny differentiate ing into NCAM-expressing ORNs. ORNs are postmitotic cells.

In the mammalian nervous system, proliferation of that neurons be produced, but that they be produced neuronal progenitor cells and differentiation of their in precise numbers in specific locations. How, and progeny into neurons-the process known as neuat what cellular stage(s), is this growth control exrogenesis-appears to be under tight control: Proerted in the developing nervous system? Recent evigenitor cell proliferation occurs rapidly at first and dence from studies of neurogenesis in the olfactory then more slowly, and in most regions, permanently epithelium (OE) indicates that such regulation may ceases toward the end of development (Kauffman, be exerted at the very earliest stages in the neuronal 1968; Caviness et al., 1995). To build a properly lineage, perhaps at the level of the neuronal stem functioning nervous system, it is important not only cell itself. The OE of the nasal cavity can be divided into three major compartments, illustrated in Figure 1 : adei, 1973;Moulton, 1974;Graziadei and Monti Graziadei, 1978), as the lifespan of ORNs in animals protected from environment (respiratory) insult is dramatically lengthened (Hinds et al., 1984). It is in its capacity to replace lost neurons in the adult that the OE differs from most other regions of the nervous system, but this ability to regenerate its characteristic differentiated cell type is a functional similarity between the OE and regenerating tissues such as skin, liver, muscle, or blood. As with those other tissues, the presence of stem cells in the OE has long been hypothesized to be the basis of its regenerative capacity.
In vivo studies performed by a number of laboratories have contributed to the realization that neurogenesis in the OE is a highly regulated process, serving to maintain the number of ORNs at a particular level, and such studies have provided support for the existence of a neuronal stem cell in the epithelium. As already discussed, in normal animals, ORNs in the OE on the side that innervated that MASH1-expressing neuronal progenitors (M1), and imbulb die (Costanzo and Graziadei, 1983;Holcomb mediate neuronal precursors (INP). These cell types and et al., 1995). As ORNs die and the OE degenerates their roles in olfactory neurogenesis are discussed in the (decreases in thickness), neuronal progenitor cells text. in the epithelium proliferate and replace the lost ORNs (Schwartz-Levey et al., 1991;Gordon et al., 1995). partment, which contains the primary sensory neurons [olfactory receptor neurons (ORNs)] that Such in vivo anatomical studies have made it clear that the hypothesized stem cell is likely to transmit the sensation of smell from the nose to the olfactory bulb of the brain. ORNs can be distin-reside in the basal compartment of the OE in adult vertebrates (Moulton and Fink, 1972; Graziadei, guished from other cell types in the epithelium by their expression of the neural cell adhesion molecule 1973; Graziadei and Monti Graziadei, 1979;Camara and Harding, 1984;Hinds et al., 1984;(NCAM), antibodies to which recognize all postmitotic neurons in the epithelium (Calof and Chikara-Mackay-Sim and Kittel, 1991;Schwartz-Levey et al., 1991;Gordon et al., 1995). Retroviral lineage ishi, 1989;DeHamer et al., 1994); ORNs that have fully matured express the cytoplasmic protein olfac-analyses have confirmed the existence of neuronal progenitors in basal OE and have also contributed tory marker protein (OMP), a molecule that is unique to these cells (Margolis, 1980). to the idea, derived from developmental and immunohistochemical studies of OE, that both horizontal Like the keratinizing epithelium of the skin, the OE is in direct contact with the environment, and basal cells and sustentacular cells are not part of the neuronal lineage (Klein and Graziadei, 1983; ORNs are continually lost owing to disease or injury. This is likely to be the cause of the ongoing Mulvaney and Heist, 1971;Matulionis, 1976;Caggiano et al., 1994). However, these types of studies turnover of ORNs accompanied by compensatory neurogenesis observed in adult animals (e.g., Grazi-alone have not been adequate to define precursor-  DeHamer et al., 1994). That multiple divisions of INPs occur in FGF-2 was demonstrated Although the OE was perceived initially as a relatively simple neuroepithelium, ongoing studies of conclusively using sequential labeling with bromodeoxyuridine (BrDU) and 3 H-TdR to specifically neurogenesis have revealed that the ORN lineage is much more complex than was previously appreci-mark ORNs that had been generated as a result of two successive rounds of cell division in culture. ated. In our laboratory, this understanding has evolved during the course of ongoing in vitro stud-We found that the incidence of such double-labeled ORNs is four to five times greater in FGF-2-treated ies of molecular regulation of neurogenesis in mouse OE (Calof et al., 1995;Mumm et al., 1996).
cultures than in controls (DeHamer et al., 1994). Importantly, FGFs do not alter the fate of INPs; Initially, explant cultures of OE purified from embryonic day 14-15 (E14-15) mouse embryos were virtually all INPs still give rise to ORNs, even with FGFs present. Thus, as committed neuronal progen-developed (Calof and Chikaraishi, 1989). To facilitate these studies, we identified antibody markers itor cells capable of undergoing a limited number of divisions in response to exogenous factors, INPs that recognize two major cell types in the OE, in vitro and in vivo: antibodies to NCAM, which mark fit the description of ''transit amplifying cells,'' a term used to describe intermediate-stage progenitors postmitotic, differentiated ORNs; and a commercial antiserum to keratins, which marks the horizontal in other lineages (e.g., the hematopoietic lineage) (Hall and Watt, 1989;Potten and Loeffler, 1990;basal cells (Calof and Chikaraishi, 1989). In OE explant cultures, the pieces of explanted epithelium DeHamer et al., 1994). Further studies revealed additional complexity in adhere to appropriate culture substrata (Calof and Lander, 1991), and within hours a population of the ORN lineage: In this work, we were interested in understanding why mice with disruption of both cells sorts out from the main body of the explant (in which keratin-expressing horizontal basal cells alleles of the gene encoding Mammalian Achaete Scute Homolog 1 (Mash1) show a near total ab-remain) and begins to migrate out onto the substratum (Calof and Chikaraishi, 1989; Calof and sence of ORNs, as well as certain autonomic and enteric neurons (Guillemot et al., 1993). Using ex-Lander, 1991). Virtually all of these migratory cells are either NCAM / ORNs or proliferating cells that plant cultures of OE obtained from wild-type animals, we found that MASH1 is expressed by neu-incorporate [ 3 H]thymidine ( 3 H-TdR). Over the course of 2 days in culture, the ORNs and proliferat-ronal progenitor cells at a distinct stage in the ORN lineage, a stage upstream of INPs (Gordon et al., ing cells continue to migrate, during which time nearly every proliferating cell divides once to give 1995). Evidence supporting this conclusion came partly from observing the rapid disappearance of rise to two daughter cells that both differentiate into NCAM-expressing ORNs. This process was fol-MASH1 / cells from OE explant cultures; this disappearance appears to reflect the division of lowed by pulse-labeling cells with 3 H-TdR and following their acquisition of the ORN-specific NCAM MASH1 / cells to generate MASH1-negative INPs (Gordon et al., 1995). Additional studies in which marker (Calof and Chikaraishi, 1989;DeHamer et al., 1994). These and other experiments have dem-olfactory bulbectomy was used to induce neurogenesis in adult OE in vivo provided data on how onstrated that neurogenesis occurs efficiently in OE cultures, and served to identify the direct progenitor changes in the numbers and proliferative states of MASH1 / cells correlate with changes in overall of ORNs, which we have called the immediate neuronal precursor (INP) (Fig. 1) (Gordon et al., 1995). Instead, the kinetics of proliferation of MASH1 / cells indicate To test whether neurogenesis could be restored in Mash1 0 / 0 OE, a Mash1 retroviral expression vector that they, like INPs, act as neuronal transit amplifying cells (Gordon et al., 1995). In addition, was constructed by inserting the cDNA for mouse Mash1 (Guillemot and Joyner, 1993) into LIGNS, MASH1 / cells, like INPs, are found in the basal compartment of adult OE (Fig. 1).
a retroviral vector giving high titers and levels of expression due to retention of endogenous retroviral To see if we could gain further insights into the characteristics of neuronal stem cells in the OE, gag sequences (Lillien, 1995). In preliminary experiments, cultures of OE explants from Mash1 /// , we analyzed OE in E14-15 Mash1 0 / 0 embryos to determine why absence of Mash1 gene function / / 0 , and 0 / 0 embryos were plated onto feeder layers of growth-arrested Mash1-LIGNS C2 (virus-pack-might result in cessation of ORN development. We found that few NCAM / ORNs ever appear in aging) cells; as a control, half of the tissue from each embryo was plated onto C2 cells packaging Mash1 0 / 0 embryos, suggesting that the lack of ORNs at birth apparently reflects a defect in ORN the LIGNS vector minus the Mash1 insert. Explants were grown for 28 h and then scored for NCAM production, rather than ORN survival. Although the OE of Mash1 0 / 0 embryos is abnormally thin, it immunoreactivity to determine the level of neurogenesis. Exposure of Mash1 0 / 0 OE to the Mash1-still contains a substantial number of NCAM 0 cells (presumptive neuronal progenitors). Using TUNEL LIGNS virus resulted in a twofold increase in the percentage of explants showing a high level of neu-staining for DNA fragmentation as a criterion for apoptotic cell death (Gavrieli et al., 1992), we rogenesis (''high NCAM score''), while exposure to the Mash1-LIGNS virus had no apparent effect found that apoptosis is strikingly elevated in the OE of E14 Mash1 null embryos ( In the hematopoietic system, if multipotent pro-stained for both NCAM (an ORN-specific marker) and TUNEL (to detect apoptotic cells), we found genitors are rescued from cell death by overexpression of the bcl-2 gene, they go on to differenti-that the cell death taking place in Mash1 0 / 0 OE is not in the few NCAM / cells that escape the genetic ate in the absence of added growth factors (Fairbairn et al., 1993). If the ultimate function of lesion. Rather, virtually all (98.7%) apoptotic cells in Mash1 0 / 0 OE are NCAM-negative cells (Calof MASH1 is to regulate expression of genes required for neuronal progenitor survival, then by analogy et al., 1996). The data suggest that in the absence of Mash1 function, neuronal stem cells produce neu-with the hematopoietic system, it may be possible to bypass the Mash1 null phenotype by ''artificially'' ronal progenitor cells (which normally would be MASH1 / ), but most of these progenitor cells die supporting the survival of Mash1 0 / 0 progenitors. To test this hypothesis, we generated an LIGNS-without generating ORNs.
These findings predict that at least some progeni-based retroviral vector to express bcl-2, a protooncogene that blocks apoptosis in a variety of cells, tors of ORNs (presumably neuronal stem cells, and possibly their immediate progeny) are present and including neurons (Hockenbery et al., 1990;Davies, 1995). C2 cells that package the bcl2-LIGNS vector mitotically active in Mash1 0 / 0 OE. If so, it should be possible to rescue neurogenesis by expressing a were generated; growth-arrested feeder layers made from them; and OE explants from Mash / / / , / / 0 , Mash1 transgene in progenitor cells of Mash1 0 / 0 OE. To accomplish this, we made explant cultures and 0 / 0 embryos were plated onto these feeders and grown and analyzed as described above. Expressing of OE from embryos derived from mating Mash1 heterozygotes (homozygous null animals die around Bcl2 in Mash1 0 / 0 OE had an even greater positive effect on neurogenesis than did expressing MASH1 the time of birth), and used NCAM expression by cells in the explants as a criterion for neurogenesis.
[ Fig. 3(A,D)]: three times as many Mash1 0 / 0 explants had a high NCAM score when exposed to OE explants from Mash1 0 / 0 embryos grow and are viable, but after a day in culture, only Ç4% (3.6 the bcl2-LIGNS virus as did when exposed to control virus. These preliminary findings suggest that { 3.5%) have generated NCAM / ORNs, whereas at the same time, Ç95%   is sufficient to restore neurogenesis, although the colleagues recently showed that Math4C/neuro-genin1, which like Mash1 encodes a basic helix-possibility that this effect might result from enhanced proliferation by Bcl2-overexpressing pro-loop-helix transcription factor, is expressed by a population of progenitor cells in the OE in vivo. This genitors has yet to be ruled out (A. L. Calof, P. C. Rim, and J. Shou, unpublished results). Despite this progenitor cell population is separate from Mash1expressing progenitors, but the data indicate that caveat, however, these studies reinforce the hypothesis that neuronal stem cells are still present in neurogenin1 expression is genetically downstream of MASH1, since neurogenin1 is not expressed in Mash1 0 / 0 animals. In addition, they suggest that a major function of MASH1 in the ORN lineage may the OE of Mash1 0 / 0 embryos (with the exception of a small ventrocaudal domain of OE that appears be to control (directly or indirectly) the expression of genes that play roles in regulating the survival to be Mash1 independent). Thus, these authors have postulated that neurogenin1 is expressed by the of neuronal progenitor cells.
Among the downstream targets of MASH1 may INP, an interesting hypothesis that remains to be tested in vitro (Cau et al., 1997 (3300 rad), confluent C2 feeder cell layers producing either control, Bcl2, or Mash1 retroviruses. To distinguish the genotype of the OE tissue used in each culture, DNA was prepared from unused tissue from each embryo and primers to the neo coding region in the Mash1 targeting vector (present only in Mash1 nulls and heterozygotes) and to endogenous Mash1 (present only in wild types and heterozygotes) were used to amplify DNA in PCR reactions. After 28 h in vitro, cultures were fixed in 10% formalin in phosphate-buffered saline for 10 min, then reacted with H28 monoclonal anti-NCAM (DeHamer et al., 1994) and visualized with horseradish peroxidase-conjugated goat anti-rat immunoglobulin G (Chemicon; 1:300 dilution). Explants that possessed ú10 NCAM-expressing neurons were given a ''high NCAM'' score, and the percentage of total explants receiving a high NCAM score was calculated for all cultures from animals of a given genotype. (A) Exposing Mash1 null OE to either Bcl2 or Mash1 retroviruses resulted in a partial restoration of neurogenesis (''High NCAM Score''; 0/0, black bars) compared to heterozygous and wildtype OE (//0, ///) , and increased neurogenesis two-to three-fold over that seen with control virus. Another class of molecules likely to be down-may be an important developmental event regulating progenitor cell survival in the OE. stream targets of Mash1 is receptors for neuronal Thus, in vivo and in vitro analyses of wild-type trophic factors. Our previous studies had indiand mutant mice have led to the realization that cated that neurotrophin-3 ( NT-3 ) , the neurothe ORN lineage is much more complex than had trophin ligand for the tyrosine kinase receptor previously been appreciated, with at least two dis-( TrkC ) , promotes survival of embryonic OE neutinct types of proliferating transit amplifying proronal cells in vitro and that TrkC is expressed in genitors-the MASH1 / progenitor and the INPneuronal cells, particularly ORNs, in neonatal OE interposed between the ORN and the stem cell  . Because TrkC expression (Gordon et al., 1995). In addition, these studies is detectable in the OE at such an early age, and suggest that an important consequence of lineage because NT-3 has been shown to promote survival complexity may be that it provides opportunities for of neuronal progenitor cells in other tissues ( Bircell number to be regulated at each cell stage in the ren et DiCicco-Bloom et al., 1993) , lineage (Calof, 1995. we examined TrkC expression in OE of E14 -15 Mash1 0 / 0 embryos. As shown in Figure 4, there is widespread TrkC immunoreactivity in wild-type OE, but none in OE of Mash1 0 / 0 littermates, sug-  , 1994).

WHAT IS THE IDENTITY OF THE OE
These results indicate that a small fraction of explants (5-8%) undergoes continuous ORN production for up to 4 days in vitro in the presencebut not in the absence-of FGF-2. The data are suggestive that explants in this ''outlier'' fraction contain within them rare cells of high proliferative potential, cells that are absent in the majority of explants, but which are revealed when explants are anti-NCAM-treated petri dishes were used to selectively remove postmitotic ORNs from a dissociated neuronal cell fraction containing only NCAM / two ways to promote neurogenesis in the OE: by promoting proliferation of INPs (discussed above), ORNs and NCAM 0 progenitors (Calof and Lander, 1991;Holcomb et al., 1995). This procedure typi-and, at longer periods in culture, by increasing proliferation and/or survival of rare early ORN progen-cally yields a preparation that consists of ú96% NCAM-negative, presumptive ORN progenitors itors, putative neuronal stem cells (DeHamer et al., 1994). To demonstrate these longer-term effects of (Mumm et al., 1996). Quantitatively, this purified progenitor cell frac-FGFs on OE neurogenesis, OE explants were grown for 48-96 h in the presence of FGF-2 or no growth tion turns out to contain mostly INPs, as was expected, since INPs constitute the vast majority of factor, and exposed to 3 H-TdR for the last 24 h of culture. Labeling indices were determined for each neuronal progenitor cells in the OE (Calof and Chi karaishi, 1989;DeHamer et al., 1994;Gordon et al., explant by counting the total number of 3 H-TdRlabeled cells (assessed by emulsion autoradiogra-1995). Indeed, most of these purified progenitors behave as neuronal transit-amplifying cells when phy) surrounding each explant; this number was normalized to an average explant size (Fig. 5).
they are plated in isolated culture on defined substrata in serum-free medium: Under these condi-These data were plotted as frequency histograms to reveal the percentages of explants with labeling tions, ú95% of purified progenitors rapidly give rise to differentiated, NCAM-expressing ORNs, all indices falling within different ranges. As Figure 5 shows, at 72 and 96 h in culture, FGF-2 allowed a of which then die within a few days, presumably owing to the absence of target tissue or appropriate reproducible, small subset of OE explants (5-8%  Gordon et al., 1995;Holcomb et al., 1995;Mumm et al., 1996). In an effort to neurons and then die. However, after 6-7 days in vitro in the coculture condition, about 1 in 1000 enhance survival and/or proliferation of neuronal progenitors, we turned to culturing them over mono-progenitors goes on to develop into morphologically homogeneous, proliferative colonies, as shown in layers of feeder cells isolated from the stroma that lies subjacent to the OE, and with which OE progen- Figure 6. Despite the existence of variation in cell morphology among colonies [ Fig. 6(A-D)], cells itors are normally associated in vivo. For these experiments, OE progenitors were purified from within a colony always have a similar morphology, which suggests that colonies are derived from clonal Rosa26 mice, which express a lacZ transgene in every cell (Friedrich and Soriano, 1991); this en-expansion; this was supported by experiments in which a number of colonies were inspected periodi-abled us to use X-gal histochemistry or immunocytochemistry using antibodies to bacterial b-galaccally for several days, during which time progressive increases in colony size were seen (Mumm et tosidase, to detect progenitors and their progeny when cultured on stroma derived from outbred CD-al., 1996). One of these colony types, accounting for about one fourth of the colonies that appear in 1 mice (Mumm et al., 1996).
When this is done, the majority of purified these cultures, contained tightly clustered cells which possess obvious neurites [ Fig. 6(D)]. Indeed, when cultures were stained with antibodies to NCAM, only this colony type was found to contain NCAM-expressing ORNs among its cells [Fig. 6(E)]. To prove that these neuronal colonies actively produce neurons, pulse-chase 3 H-TdR incorporation experiments were performed in progenitor/ stroma cocultures to label neurons being generated from 6 to 7.5 days in vitro. As shown in Figure 7, many NCAM-expressing neurons (red) in neuronal colonies were also 3 H-TdR labeled (yellow) in such experiments, indicating that ORN production is on- Figure 5 Effects of FGF-2 on distribution of explant labeling indices at late times in culture. OE explants were cultured in the presence or absence of FGF-2 (10 ng/ mL) for either 48, 72, or 96 h and were exposed to 3 H-TdR (0.1 mCi/mL) during the last 24 h of culture.

Figure 7
Continual neurogenesis in neuronal colonies at 7 days in vitro. Cocultures of purified OE progenitors over growth-arrested stroma were pulsed for 10 h with 1 mCi/mL 3 H-TdR at 6 days in vitro, then chased with cold thymidine (50 mM) for 24 h prior to fixation and processing for NCAM immunoreactivity and autoradiography. (A) A neuronal colony is shown in a double exposure for NCAM immunoreactivity (red) and silver grains indicating 3 H-TdR incorporation. Silver grains appear green if present in an NCAM 0 cell, yellow if present in an NCAM / cell. Some dividing cells have generated postmitotic, NCAM / neurons (arrowhead), but a significant number NCAM 0 , 3 H-TdR / cells remain (arrow). For clarity, the same colony, exposed for NCAM immunoreactivity alone, is shown in (B). The results indicate that neurons continue to be generated as late as day 7 in culture by proliferating progenitors present in neuronal colonies, and undifferentiated progenitors continue to be generated as well. going in these colonies for at least a week in culture. stem cell, the neuronal lineage of the OE is diagrammed in Figure 8. Significantly, 3 H-TdR-labeled cells that were NCAM negative were also observed in such neuronal colonies (Fig. 7), indicating that these colo- al., 1996).

CELLS
Our current hypothesis is that neuronal colonies Experimental up-regulation of neurogenesis in OE arise from a neuronal colony-forming cell or unit has been a useful paradigm for generating hypothe-(CFU) similar to CFUs that are thought to be indicses concerning cell interactions that regulate this ative of stem cells in hematopoietic development process. As discussed previously, surgical removal (Dexter and Spooncer, 1987). We find that neuronal of the olfactory bulb on one side of the brain (unilatcolonies are able to produce differentiated ORNs as eral olfactory bulbectomy) of an adult rodent causes late as 13 days in vitro (J. Shou, P. C. Rim, and ORNs in the OE on that side to rapidly undergo A. L. Calof, unpublished results), and the existence cell death; the peak of apoptosis is approximately of proliferative, undifferentiated cells within them 2 days following surgery . has been demonstrated as late as 7.5 days in vitro Neuronal progenitor cells in the ipsilateral OE re- (Mumm et al., 1996). These findings are consistent spond by increasing proliferation, which reaches a with the presence within neuronal colonies of propeak at about 5-6 days postbulbectomy (Schwarzgenitors with both the capacity for self-renewal and Levey et al., 1991;Gordon et al., 1995). There is the ability to give rise to a lineage whose end point marked degeneration of the ORN cell layer, which is a terminally differentiated cell. These are defining can be quantified by measuring epithelial thickness, characteristics of stem cells (Lajtha, 1979;Hall and and this degeneration follows a time course similar Watt, 1989;Potten and Loeffler, 1990).
to that of progenitor cell proliferation: Epithelial In some permanently renewing systems, such as thickness is at a minimum at about 5 days postbulthe hematopoietic system, stem cells also are charbectomy (Costanzo and Graziadei, 1983; Schwartzacteristically rare (Dexter and Spooncer, 1987). In Levey et al., 1991;Holcomb et al., 1995). As new this light, it is interesting to note that our initial ORNs are generated in the ipsilateral OE, epithelial analysis of neuronal stem cell activity in purified thickness increases again to approximately 70% of progenitor/stroma cell cocultures yielded an estimate of 1 in 3600 progenitors for neuronal CFU frequency (Mumm et al., 1996). Significantly, this is in the range of the estimate for stem cells derived from studies of long-term explant cultures grown in FGF-2 (there the estimate was about 1:2500; see above). It is also on the order of that observed for spleen CFUs isolated from bone marrow (Ç1 in 3800) (Spooncer et al., 1985), although somewhat lower than estimates of stem cell frequency in epidermis or fetal liver (Jordan et al., 1990;Jones and Watt, 1993). Since the procedures required for purifying progenitors could easily cause damage, and it is virtually certain that we have not yet optimized  (DeHamer et al., 1994;Gordon et al., 1995). Placrounds of division, after which their progeny differentiate ing the neuronal colony-forming cell (neuronal into NCAM-expressing ORNs. ORNs are postmitotic cells.
CFU) as our current best candidate for the neuronal 8p45 1984 / 8p45$$1984 06-29-98 16:58:07 nbioal W: Neurobio MASH1 / progenitor, and INP), it seems possible that the signal for bulbectomy-induced neurogenesis could act on an early progenitor cell, perhaps even the stem cell: The peak in proliferation at 5-6 days postbulbectomy that is observed would then simply reflect expansion of transit-amplifying cells (MASH1 / cells and/or INPs) in response to the early mitogenic stimulus. We have taken advantage of the purified progenitor/stroma coculture system to test these models in tissue culture. Using the frequency of neuronal colony formation as an assay for neurogenesis, we per- week, processed for X-gal histochemistry, and then scored for each colony type. The results demonstrated that the addition of ORNs decreases the inci-its original value, and progenitor cell proliferation decreases, but to a level that is elevated over that dence of neuronal colonies significantly, whereas addition of stroma cells has no significant effect on seen in the contralateral (control) OE (Schwartz-Levey et al., 1991;Gordon et al., 1995; frequency of neuronal colony formation. Data from typical experiments are illustrated in Figure 10. . The temporal relationship of ORN death, cell frequency of neuronal CFU formation is Ç1 in 3600 of plated purified progenitors under normal progeni-degeneration, progenitor cell proliferation, and generation of new ORNs suggests two models for how tor/stroma coculture conditions, but when an excess of neurons is added, this frequency drops 2.5-fold, cell interactions in OE could regulate progenitor cell proliferation. These models are diagrammed in Figure 9. In model 1, the feedback inhibition model, proliferation of progenitor cells following death and degeneration of the ORNs overlying them is the result of an interruption of an ORN-derived negative feedback signal that inhibits progenitor cell proliferation. This model is consistent with the observation that maximum proliferation is associated with maximum ORN loss (i.e., maximum epithelial degeneration, at 5-6 days postbulbectomy) and that an elevated level of proliferation is maintained in chronically bulbectomized animal (where a reduced number of ORNs is present in the OE). In model    et al., 1996). interesting because there is some evidence to suggest that olfactory Schwann cells, the cells that en-Two pieces of evidence indicate that this is a specific effect on the development of neuronal colonies: sheath ORN axons, originate from progenitor cells located within the OE itself (Klein and Graziadei, First, no significant effect on the numbers of nonneuronal colony types is seen when excess differen-1983;Chuah and Au, 1991;Calof and Guevara, 1993). However, we think it unlikely that neuronal tiated ORNs are plated along with purified progenitors (Mumm et al., 1996). Second, no inhibitory colonies and spindle-shaped (putative olfactory Schwann cell) colonies arise from the same cell, as effect is observed when an excess of stroma cells is plated with progenitors [ Fig. 10(B)], indicating colonies of mixed neuronal and spindle-shaped cells have never been observed in our progenitor cell/ that inhibition cannot be explained as a consequence of added cells simply depleting the medium of nutri-stroma cocultures, suggesting that the neuronal CFU may be directed toward an exclusively neuronal (as ents. It is also worth noting that addition of excess ORNs to OE progenitor cell cultures results in a opposed to glial) fate. Thus, the results of in vitro experiments suggest decrease in the number, rather than the size, of neuronal colonies (Mumm et al., 1996). If the signal that differentiated ORNs produce a signal that feeds back to inhibit production of new neurons by their produced by differentiated ORNs acts upon cells at a very early progenitor stage-for example, neu-own progenitors, and could explain the surge in neurogenesis in the OE in vivo that occurs as a ronal stem cells-causing them to die or to stop producing the downstream neuronal transit ampli-consequence of ORN death following bulbectomy: This increase in neurogenesis could be due to loss fying cells that generate ORNs, then this would be the expected outcome.
of an inhibitory signal normally produced by living ORNs. Uncovering the mechanisms that underlie Thus, although model 2 cannot yet be excluded, the results of these studies are most consistent with feedback regulation of neuron production in the OE may be important for understanding the regulation model 1 (Fig. 9), since differentiated ORNs appear to produce a signal that inhibits development of of neuronal production in other parts of the nervous system. neuronal CFUs (Mumm et al., 1996). Interestingly, indirect evidence that the rate of proliferation of cells in the basal OE is regulated by the number of neurons in the epithelium has been provided by

CANDIDATE MOLECULAR
several anatomical studies (Mackay-Sim and Patel, 1984;Mackay-Sim et al., 1988;Weiler and Farb-STEM CELL man, 1997). A similar regulatory mechanism has been suggested in larval Xenopus retina, where ablation of cells in vivo by intraocular injection of 6-As discussed above, tissue culture assays of neurogenesis in the OE are useful for testing the behavior hydroxydopamine or the excitotoxin kainic acid results in preferential production of new cells of the of putative OE neuronal stem cells and for determining what signals regulate proliferation and survival appropriate type (Reh and Tully, 1986;Reh, 1987).

REGULATORS OF THE NEURONAL
One other possibility that should be considered of these cells. In particular, there are two important questions that need to be addressed: First, what are is that the presence of excess differentiated ORNs may cause neuronal CFUs to adopt another fate.
the positive factors that allow the OE to continue to generate ORNs on an ongoing basis? Presumably, Such an idea is not without precedent: Shah and colleagues (1994) showed that neural crest stem among such factors will be found those which maintain and/or stimulate proliferation of the neuronal cells are biased toward a glial, rather than neuronal, fate when grown in glial growth factor (GGF) and stem cell, as well as other factors that may act locally to stimulate proliferation and differentiation that crest-derived neurons express both GGF protein and mRNA, suggesting that neuron-derived GGF of cells later in the neuronal lineage. Second, what is the identity of the neuron-derived inhibitory sig-affects the fate choice of this type of neural stem cell. In our studies, we noted that addition of excess nal that acts as a negative regulator of neurogenesis in the OE? In theory, it is this same factor or factors ORNs to OE progenitor cell cultures resulted (in addition to the decrease in the number of neuronal that causes a decrease in the number of neuronal colonies that develop when ORNs are added back colonies) in an increase in the number of colonies of spindle-shaped cells (which resemble olfactory in the neuronal colony-forming assay, and this assay should be useful for identifying it. A true under-Schwann cells morphologically) in two of four experiments, although this result was not statistically standing of the regulation of neurogenesis in the OE should encompass a molecular identification of both significant (Mumm et al., 1996). This finding is 8p45 1984 / 8p45$$1984 06-29-98 16:58:07 nbioal W: Neurobio positive and negative regulatory signals and deter-of an autocrine loop in the OE, in which neuronal progenitor cells respond to a growth factor that they mination of their mechanisms of action.
Proliferation of OE neuronal progenitor cells is themselves make. Equally intriguing is the question of the identity presumably regulated by the activity of mitogenic signals produced in or near the cells on which they of the neuron-derived inhibitory signal that appears to function to suppress OE neurogenesis in vitro act, and so until recently, we were puzzled by the fact that while FGFs, and in particular FGF-2, had (Mumm et al., 1996). Interestingly, in initial experiments to characterize this signal, we have found been shown by us to have profound stimulatory effects on both INPs and presumed OE neuronal that ORNs that have been killed by freezing and thawing still retain the ability to inhibit neuronal stem cells in vitro (DeHamer et al., 1994), there were no convincing data demonstrating expression colony formation, whereas ORNs that have been boiled do not (Calof et al., 1997b). These findings of a member of the FGF family within the OE itself. However, molecular cloning of the Fg f-8 gene, to-suggest that ORNs inhibit neuronal colony formation not through a complex cell-cell interaction, gether with in situ hybridization studies of Fg f-8 expression patterns in early embryos, has recently but more likely through the release of one or more heat-labile factors that they already contain. This provided a possible solution to this puzzle: Reports from several labs indicate that Fg f-8 mRNA is pres-result, together with the known growth inhibitory actions of members of the TGF-b superfamily of ent in mouse embryos in nasal pits, the primordia of the OE, and that expression is detectable as early polypeptide growth factors (Massague and Polyak, 1995) and a recent report demonstrating expression as E9 (Crossley and Martin, 1995;Heikinheimo et al., 1994;Mahmood et al., 1995).
of Bmp-4 mRNA in the olfactory placode (Wu and Oh, 1996), led us to examine the possibility that These findings prompted us to perform in situ hybridization experiments to examine the expres-BMPs exert effects on OE neurogenesis. When tested in the colony-forming assay for neu-sion of Fg f-8 in the OE of later-stage mouse embryos and adult mice. Our initial findings indicate ronal stem cell activity (Mumm et al., 1996), BMPs 2 and 4, members of the Decapentaplegic subfamily that Fg f-8 continues to be expressed in the OE at E14-when the rate of neurogenesis is high-and of TGF-bs, were found to specifically inhibit formation of neuronal colonies, but had no effect on other is also present in a small fraction of cells in the basal compartment of adult OE-where neurogenesis is colony types (Shou et al., 1997(Shou et al., , 1998b. Since these two BMPs are highly related structurally and act on ongoing, albeit at a much lower levels than those observed in the embryo (Calof et al., 1997a; Shou the same serine-threonine kinase receptors (Kingsley, 1994;Massague, 1996), we examined the pat-et al., 1998a). Thus, FGF-8 is likely to be produced and available to ORN progenitors undergoing neu-tern of expression of both factors by in situ hybridization in developing and adult mouse OE. These rogenesis, and may be an endogenous signal that stimulates proliferation of progenitors within the studies demonstrated that Bmp-4, but not Bmp-2, is expressed in neuronal layers of the OE both during OE, during development and possibly in adults as well. Interestingly, in E14 OE explant cultures Fg f-development and in the adult (Shou et al., 1997(Shou et al., , 1998b. Taken together, these data suggest that 8 is expressed in NCAM-negative, non-neuritebearing migratory cells, at least some of which are BMP4 expressed by ORNs constitutes at least part of the neuron-derived inhibitory signal detected in capable of incorporating BrDU, suggesting that it is expressed by neuronal progenitor cells (Calof et our neuronal colony-forming assay. Moreover, experiments varying the time at which BMP-4 is al., 1997a; Shou et al., 1998a). We have also found that recombinant FGF-8 shares with FGF-2 two ma-added to colony-forming assays suggest that BMP4's inhibitory action may be exerted primarily jor actions on OE neuronal progenitors: It stimulates proliferation of INPs, and it promotes extended neu-on neuronal colony-forming progenitors, the presumptive neuronal stem cells of the OE (Shou et rogenesis in long-term OE explant culture assays to detect stem cell activity in rare explants (DeHamer al., 1997(DeHamer al., ). et al., 1994Calof et al., 1997a;Shou et al., 1998a). Taken together, these findings suggest that FGF-8 may be an endogenous stimulatory factor that

REMAINING ISSUES AND FUTURE DIRECTIONS
promotes continual neurogenesis in the OE in vivo, possibly owing to an effect on OE neuronal stem cells. In addition, the provocative finding that Fg f-Despite considerable progress, much remains unresolved concerning the identity and developmental 8 appears to be expressed by neuronal progenitors suggests that Fg f-8 expression may be the basis regulation of the neuronal stem cell of the OE.