Genesis of Olfactory Receptor Neurons In Vitro: Regulation of Progenitor Cell Divisions by Fibroblast Growth Factors

Summary Olfactory receptor neurons are produced continuously in mammalian olfactory epithelium in vivo, but in explant cultures neurogenesis ceases abruptly. We show that in vitro neurogenesis is prolonged by fibroblast growth factors (FCFs), which act in two ways. FGFs increase the likelihood that immediate neuronal precursors (INPs) divide twice, rather than once, before generating neurons; this action requires exposure of INPs to FCFs by early Gl. FGFs also cause a distinct subpopula-tion of explants to generate large numbers of neurons continually for at least several days. The data suggest that FGFs delay differentiation of a committed neuronal transit amplifying cell (the INP) and support proliferation or survival of a rare cell, possibly a stem cell, that acts as a progenitor to INPs.


Introduction
The mammalian olfactoryepithelium (OE) is uniquely suited to studies of how neurogenesis is controlled.  Klein et al., 1989;Heuer et al., 1990;Maisonpierre et al., 1990;Reid et al., 1990;Yeh et al., 1991;Schecterson and Bothwell, 1992 (D) Same field as (C), dark field. Bar, 50 pm. (E) Growth factor screening assays. Bars represent mean proliferation factor + SD for growth factors at optimum tested concentrations (FGFI, 100 ngiml; FGFZ, 1 nglml; FCF4, IO ngiml; FCF7, 10 rig/ml; NGF, BDNF, and NT-3, 50 rig/ml; CNTF, PDGF-AA, PDCF-BB, TCFBI and TCFBZ, 10 rig/ml; EGF and TGFa, 20 rig/ml. FGF4 and FGF7 were each also tested at 1 and 100 rig/ml and gave effects similar to those observed at 10 nglml [data not shown]). Proliferation factors were calculated as follows: for each explant, the explant labeling index was calculated as the number of migratory cells with silver grains over their nuclei divided by the area of the explant (measured using NIH Image 1.52). The proliferation factor is the ratio of the mean explant labeling index in a given condition to that of controls (no added growth factor) grown on the same day. Percent of error for these ratios (the square root of the sum of the squares of percent errors [from SEMI of the two labeling indices being compared) averaged -20%. ANOVAfollowed by Dunnett's test (for multiple comparisons against a single control; Glantz, 1992)  were grown in FGFP (IO nglml) and pulsed from t = 4 to 6 hr with BrdU (1:5000), then either fixed or chased with unlabeled TdR. At the indicated times, cultures were pulsed for 2 hr with SH-TdR (5 pCi/ml) and fixed. Times on the abscissa represent the interval between the ends of the BrdU and 3H-TdR pulses. (Inset) The data from (A) are compared with a predicted curve based on a model of cell cycle phase lengths, the parameters of which were adjusted to provide a close fit to the experimental data (for model derivation, see Experimental Procedures). Predicted values are S = 8 hr, G2 + M = 3.5 hr, total cell cycle length = 17 hr, and 12% of precursors divide twice. The predicted curve is defined by a quotient, the numerator of which is the sum of two components: the first has a value of (L,, -t)/L, for t < L,. and 0 fort > L,,; the second has the value 0 for t < L, -L,,; 2f(L,. -L, + t)/k for L, -L,, < t < L,; 2f(L,. + L -t)/L for L, < t < L, + L,,; and zero for t > L, + L,.. The denominator of the quotient has a value of LJk fort < LGZM + Q; a value of (L,, + t -LCZM -Q)/ k for LGZM + Q < t < LCZM + Q + L,.; and a value of 2LJk for t > LCIM + Q + L,.. In these equations, t is the time interval between BrdU and 'H-TdR pulses; L, is the total length of the cell cycle; l-5, = f-5 + L,"IS. -2Q, where L, is the length of S phase, L,.I,. is the length of the BrdU pulse (2 hr), and Q is the interval over which a cell must be exposed to BrdU or 'H-TdR to become detectably labeled; LC2,., is the combined length of G2 and M phases; and f is the fraction of BrdU+ cells that undergo a second cell cycle. (B) Cultures were pulsed with BrdU from t = 0 to 6 hr, chased with unlabeled TdR for 15 hr, pulsed with SH-TdR from t = 21 to 45 hr, then fixed and processed.
FGF2 (10 rig/ml) was added to a series of separate, duplicate cultures at 6 hr intervals from 0 to 30 hr in culture.
Dotted lines illustrate the anticipated pro- Overall, the data in Figure  6 and Figure  7  were grown continuously in FCFZ and pulsed with SH-TdR (0.1 BCilml) from 48 to 70 hr, then chased with unlabeled TdR (50 PM) for 12 hr and fixed and processed for N-CAM immunocytochemistry and autoradiography. SH-TdR labeling indices were calculated (as in Figure 6) for 100 explants in 5 cultures. Thedistributionoflabelingindicesinthisexperimentwassimilar to that observed at t = 72 hr with no chase (Figure 66) FGFZ, the FGFfamily memberthat was studied in detail, had two actions: the first was to increase the number of INPs that underwent two, rather than one, rounds of cell division prior to giving rise to ORNs; the second action was to maintain, in a distinct, small fraction of explants, the continued proliferation or survival of neuronal precursors for as late as 72-96 hr in culture (the longest time tested).
The conclusion that FGF2 causes some INPs to undergo an additional cell division in vitro is supported by the following observations: FGF2 causes a delay in the onset of N-CAM expression by the progeny of INPs (Figure 3), FGF2 does not alter the 3H-TdR pulselabeling index of INPs at early times in culture (Table  I), and FGF2 increases by 3-to 5-fold the fraction of ORNs that are double-labeled by short pulses of BrdU and 3H-TdR administered 12 hr apart (Figure 4; Figure  5B), an interval sufficiently long so that double labeling implies traversal of two successive S phases in vitro ( Figure 5A). In addition, it was noted that, in the presence of FGF2, the migratory cells labeled by 3H-TdR administered between 24 and 48 hr were sometimes found in small clusters of 4 or more cells ( Figure  IB); such clusters would be expected if some INPs had divided twice and their clonal descendants had not migrated far from each other.
The conclusion that FGFs allow neurogenesis to be maintained in a small fraction of OE explants is based on the analysis of the number of migrating cells per explant that could be labeled by3H-TdR at increasingly late times in culture ( Figure 6). It was observed that, in the presenceof FGF2,5%-8% of explantscontinued to produce large numbers of dividing, migratory cells long after proliferation had virtually ceased in the remaining explants. In addition, a large proportion of the proliferating migratory cells produced by such explants gave rise to N-CAM+ neurons (Figure 7).

What
Is the Mechanism of Action of FGFs?
The early effects of FGFs (i.e., those observed between 24 and 48 hr) are most likely a direct action on INPs, since they are seen in dissociated cultures that are free of basal cells (Table IE). The fact that these effects can be accounted for by an increase in the fraction of INPs that undergo a second cell division in vitro (Figure 4; Figure 5), rather than exiting the cell cycle and expressing N-CAM (Figure 3), suggests that the mechanism of action of FGFs may be to repress terminal (neuronal) differentiation, thereby making further divisions possible. The data in Figure 5 also suggest that this action of FCFs must be exerted on INPs by early Cl of their cell cycle, a time at which commitment to terminal differentiation would be expected to occur (Soprano and Cosenza, 1992 Figure   4; Figure  5B). were fixed in Omnifix II, permeablized in 0.1% Triton X-100 in PBS, treated with 2 N HCI in water for 15 min, and incubated overnight in anti-BrdU ascites (1:500). BrdU staining was visualized with Texas red goat anti-rat IgG (Jackson; 1:50) or, in cultures grown on plastic dishes, with biotinylated rabbit anti-rat IgG (2.5 pgiml; Vector) followed by avidin-horseradish peroxidase (Vectastain ABC-peroxidase kit). Cultures incubated in 3H-TdR were either fixed and processed for immunocytochemistry as specified above or simply fixed in 3.7% formaldehyde/S% sucrose in PBS. Coverslips were dehydrated and dipped in NTB2 emulsion diluted I:? in water, then exposed at -85OC. Cultures pulsed for 2 or 6 hr were exposed for 2 days; those pulsed for 24 hr pulse were exposed for 7or 8 days. Slides were then developed in D-29 developer, and nuclei were stained with Hoechst 33258 (bisbenzimide; 1 pgiml).
Estimations of Cell Cycle Parameters The predicted curve in Figure 5 (inset) was obtained by fitting data to a simple model, in which the fraction of BrdU' ceils that are 3H-TdR+ was calculated for each time point by dividing a predicted numberofdouble-labeledcells byapredicted number of BrdU'cells.
For any population starting with an arbitrary number of cells in the cell cycle (and having the same cell cycle kinetics), the denominator is determined by the length of the cell cycle, the length of the BrdU pulse, the duration of BrdU incorporation required for detectable labeling, and the length of the G2 + M phases of the cell cycle (Nowakowski et al., 1989). The numerator is the sum of two components: cells that are double-labeled because they were exposed to 'H-TdR before having left the S phase during which they incorporated BrdU and cells that are double-labeled because they were undergoing a second S phase at the time of the 3H-TdR pulse. Calculating thesecomponents is straightforward and requiresonly introducing a parameter representing the fraction of BrdU+ cells that enter a second cell cycle (rather than becoming postmitotic). In addition, it was assumed that the duration of 'H-TdR incorporation required for detectable labeling was similar to that required for detectable labeling by BrdU (neither parameter had a substantial impact on the output of the model).