Within each sympathetic ganglion, a number of different types of postganglionic neurons are present, defined by the target they innervate in the periphery (Jänig, 2006; Cane and Anderson, 2009). While the general mechanisms that regulate the neuronal differentiation of neural crest progenitor cells are well described (Francis and Landis, 1999; Goridis and Rohrer, 2002; Nishi, 2003; Glebova and Ginty, 2005; Howard, 2005; Apostolova and Dechant, 2009), it is not clear how, or even when, a postganglionic neuron becomes committed to innervating a particular target tissue. In the best-studied example, that of the cholinergic postganglionic neuron innervating the sweat glands of rodents, the target tissue seems to have a major role in regulating the phenotype of the mature neuron. For instance, the postganglionic neurons that first innervate the rodent sweat gland are phenotypically noradrenergic but, under the influence of a neuropoietic cytokine released from the sweat gland, the neurons suppress gene expression characteristic of noradrenergic neurons and activate genes characteristic of cholinergic neurons (Guidry and Landis, 1998b). Furthermore, all developing sympathetic neurons seem capable of responding to sweat glands in this way. Sweat glands transplanted to the flank skin of rats force a cholinergic change on any sympathetic neuron that innervates them (Schotzinger and Landis, 1988), while cultured rodent-sympathetic neurons also respond to co-cultured sweat glands, sweat gland extracts, or medium conditioned by co-cultured sweat glands by adopting a cholinergic phenotype (Stevens and Landis, 1990; Rao et al., 1992; Rohrer, 1992; Habecker and Landis, 1994; Habecker et al., 1995, 1997). Another sympathetic target tissue, the rat pineal gland, can also alter the chemical phenotype of sympathetic postganglionic neurons that innervate it in an ectopic location (Anderson et al., 2002). However, these experiments leave open the possibility that, while the target tissue can influence the phenotype of sympathetic postganglionic neurons, the target is normally innervated by neurons that were destined to innervate a limited number of target tissues from an early age.
This idea is supported by a recent study (Makita et al., 2008), which demonstrated that the initial projection of sympathetic mouse postganglionic axons (at E12.5) along the external carotid artery depends on endothelin receptor A (EdnrA) expressed on a subset of neurons in the ganglion. The major target of the axons projecting on this route is the salivary glands. Thus, the expression of EdnrA seems to mark a functionally distinct sub-group of postganglionic neurons at an early embryonic age, well before target contact.
Other features of developing neurons may foreshadow the eventual function of the neuron. For instance, throughout the central nervous system, the “birth date” of a neuron (the time of the final cell division of the neuroblast) is often highly correlated with the neurotransmitter, connectivity, and exact position that the neuron takes up in the mature nervous system. A striking example of this is in the development of the mammalian cortex, where the birth date of the neuron is highly correlated with the cortical layer, and hence is also correlated with the neuron's eventual function (Epstein et al., 1992). There is also a correlation between birth date and neuron function in the retina (Cepko et al., 1996) and cerebellum (Altman and Bayer, 1997). It is unknown whether different sub-types of sympathetic neuron differ in their time of exit from the cell cycle.
In this study, we used the thymidine analogue, bromodexyuridine (BrdU), to pulse-label developing sympathetic ganglia in utero in rats at a range of embryonic ages. After the sympathetic neurons had innervated their targets and adopted a mature phenotype, different classes of sympathetic neuron were identified by retrograde labelling and/or immunohistochemistry to neuropeptides. Our study shows that there are significant differences in the peak birth date of some classes of sympathetic postganglionic neurons.
A pulse of BrdU delivered to the dam results in labelling of all cells in S-phase at the time of delivery, including cells in the embryos (Taupin, 2007). Any subsequent divisions of the cell after the initial BrdU labelling dilute the BrdU so that progeny of labelled cells become undetectable. Cells that undergo their final S-phase at the time of the BrdU pulse retain a strong BrdU signal and represent cells withdrawing from the cell cycle. BrdU is retained in a cell for periods exceeding 60 days in the absence of cell division (Miller and Nowakowski, 1988), and so in the current study a sympathetic neuron found to retain BrdU in ganglia examined more than 5 weeks after BrdU injection was deemed to have been born on the day of BrdU injection. Note that we counted any neuron as BrdU-labelled if any BrdU-immunoreactivity could be detected, without attempting to distinguish heavily labelled cells from lightly labelled cells. The latter may indicate BrdU-labelled cells that have undergone more than one division. The inclusion of any cells that undergo more than a single division would have a tendency to extend the distribution in time towards younger ages, but we have not attempted to correct for this, assuming that it will affect all counts equally.
BrdU-labelled cells were present in the superior cervical and stellate ganglia of 5-week-old rats after injections of BrdU at E12.5, E14.5, E15.5, E16.5, E17.5, E18.5, or E20.5. BrdU-labelled cells were rare in the ganglia after injections on E12.5 and E20.5, but were abundant following injections at intervening ages. Nearly all labelled cells within sympathetic ganglia had the histological characteristics of postganglionic neurons. Cells with the characteristics of satellite glia (small cells located between principal neurons) were only found on E20.5 and then were very rare. This suggests that satellite glia generated before E20.5 (Callahan, 2008) must continue to divide.
Retrograde Labelling With Fast Blue and NPY Immunohistochemistry
The results of the retrograde tracing studies confirmed previous studies (Fig. 1A–I). Most neurons (93% of a total of 803 neurons) labelled from the anterior chamber of the eye are neuropeptide Y-immunoreactive (NPY-IR) and have previously been shown to innervate the iris (iridomotor neurons, Zhang et al., 1984; Björklund et al., 1985; Grkovic et al., 1999). Most neurons (92% of 1,041 neurons) labelled from the submandibular salivary gland lacked NPY-IR and have previously been shown to innervate the secretory acini (secretomotor neurons, Lundberg et al., 1988). Most neurons (91% of 1,158 neurons) labelled from the brown fat lack NPY-IR and have been identified as innervating the brown fat, where they regulate thermogenesis (thermomotor neurons, Cannon et al., 1986; Grkovic and Anderson, 1997). Injections into the skin labelled 1,249 neurons with roughly equal numbers with NPY-IR (44%, 552 neurons) and without NPY-IR (56%, 697 neurons). Neurons lacking NPY-IR projecting to the skin have been shown to innervate the piloerector muscle (pilomotor neurons, Schotzinger and Landis, 1990), while those with NPY-IR innervate the cutaneous blood vessels (cutaneous vasoconstrictor neurons, Schotzinger and Landis, 1990). Injections into skeletal muscle would normally be expected to label muscle vasoconstrictor neurons, which in the rat have previously been shown to express NPY-IR (Pernow et al., 1987; Grkovic and Anderson, 1997). However, after injections of Fast Blue into the masseter muscle only 54% of the neurons expressed NPY. Examination of sections taken through the masseter muscle and stained for tyrosine hydroxylase, a marker of sympathetic nerve terminals and NPY, revealed that all terminals around skeletal muscle blood vessels were NPY immunoreactive, but that the larger blood vessels were accompanied by large nerve trunks containing sympathetic preterminal axons (data not shown). We assume that the large proportion of sympathetic cell bodies labelled by Fast Blue that lacked NPY-IR after injections into the masseter muscle reflects labelling of preterminal sympathetic axons of non-vascular function running in nerves passing through the masseter and destined for other targets in the head. For the analysis reported below, we have identified only the NPY-IR neurons (526 cells) labelled from the masseter muscle as muscle vasoconstrictor neurons.
Finally, we used immunoreactivity to CGRP alone (without retrograde labelling) to identify cholinergic neurons in the stellate ganglion. Among sympathetic postganglionic somata in the rat stellate ganglion, immunoreactivity to CGRP is restricted to the subset of cholinergic neurons projecting to the sweat glands (Landis and Fredieu, 1986; Anderson et al., 2006). In contrast, the subset of cholinergic neurons projecting to the periosteum (Asmus et al., 2000, 2001) lacks CGRP, but are uniquely marked by the presence of a pericellular network of preganglionic nerve terminals that are immunoreactive to CGRP (Anderson et al., 2006).
Relationship of BrdU Retention to Functional Class
In 5-week-old rats injected with Fast Blue, retrogradely-labelled sympathetic neurons immunoreactive for BrdU were present in both the stellate and superior cervical ganglia after BrdU injections into the dam on E14.5 to E18.5 inclusive (Fig. 1A–I). Only one retrogradely-labelled neuron (labelled from the salivary gland) was BrdU-immunoreactive after BrdU injections on E12.5 while retrogradely-labelled neurons immunoreactive for BrdU were never seen after injections on E20.5. BrdU-labelled neurons were also very rare amongst the neurons lacking colocalised retrograde tracer at both E12.5 and E20.5. The same pattern of BrdU labelling was seen in postganglionic neurons identified by their content of CGRP-immunoreactivity (sudomotor neurons) or by their association with pericellular baskets of CGRP-immunoreactive terminals (defining them as cholinergic periosteal neurons). The frequency distributions of the occurrence of BrdU-labelled neurons across all ages of BrdU injection for each class of sympathetic neuron labelled is shown in Figure 1J.
Table 1. Estimated Time of Peak Withdrawal From the Cell Cycle for Sympathetic Neurons (± Standard Error of the Mean)a
Lines join means that are not significantly different.
Cubic curves fitted to each distribution (Fig. 1K) were used to identify the peak time of cell withdrawal for each type of neuron and to define the order of cell cycle exit for the different classes of neuron (Table 1). Statistical analysis of the eight distributions was carried out using a general linear model comparing the curves generated using a cubic regression equation fitting each set of data points. The model indicated that, overall, the shape of the curves for the different cell types was not identical and at least one pair of curves was statistically different. A post-hoc pairwise analysis of the peaks calculated from the fitted curves using calculated confidence intervals showed that the pilomotor and sudomotor peaks were not significantly different from each other, that the peaks for iridomotor and periosteal neurons were not significantly different from each, and that the thermomotor, secretomotor, and muscle vasoconstrictor neurons were each significantly different from all other classes of neuron (Table 1). Note that the standard error for the peak of the cutaneous vasoconstrictor neuron curve could not be calculated with the algorithm used, due to the flatness of the distribution. As the peak value for this curve lay between the peaks of the pilomotor and sudomotor neurons, which were not significantly different from each other, we have assumed that the peak of the cutaneous vasoconstrictor neurons is not significantly different from either of these.
Statistical analysis of the proportion of total neurons of each functional class labelled with BrdU injected between E13.4–E18.5 was also performed (Table 2). In particular, the total proportion of cholinergic sudomotor neurons (87%) and cholinergic periosteal neurons (120%) labelled with BrdU was greater than for any of the other (noradrenergic) classes of neurons, which were all between 46 and 67%. When the total proportion of cells labelled with BrdU at all ages injected for each class of neuron were compared using a chi2 test, the null hypothesis was rejected (P > 0.0001, chi2 = 133.03, critical value, 14.07). The proportions that were significantly different (Table 2) were identified with a Tukey-type multiple comparison (Zar, 1984). A significantly higher proportion of periosteal neurons was labelled with BrdU than for all other classes. Of the remainder, the proportion of sudomotor neurons labelled with BrdU was significantly higher than for cutaneous vasomotor, iridomotor, thermomotor, and pilomotor neurons, but not for secretomotor and muscle vasoconstrictor. All remaining classes of neurons did not significantly differ from each other in the total proportion of cells labelled with BrdU. These data show that cholinergic sympathetic neurons have an increased probability of retaining BrdU following injection of BrdU in the mid embryonic period compared to other functional classes of sympathetic neurons.
Table 2. Statistical Analysis of the Proportion of Total Neurons of Each Functional Class Labelled Following Single Injections of BrdU at E13.5 to E18.5a
Lines join proportions that are not significantly different.
The two key findings of this study are that (1) different types of rat sympathetic postganglionic neuron withdraw from the cell cycle at overlapping, but different, times, and (2) some classes of sympathetic neuron differ in the retention of BrdU following single pulses of BrdU delivered between E14.5 and E18.5.
The peak time of withdrawal from the cell cycle for sympathetic postganglionic neurons was calculated to be between E15.25 and E17.0 for all classes of sympathetic neuron examined, but was significantly different for different functional classes of neuron. The statistical method used to estimate the variability in the peaks may have led to an overestimate of the significance of small differences in peak withdrawal time, but even if differences of less than half a day between peaks are ignored, there are at least three groups of neurons with different times of peak withdrawal from the cell cycle. A large group of neurons peak between E15.27 and E15.80 (Group 1, consisting of pilomotor, cutaneous vasomotor, sudomotor, iridomotor, periosteal, and thermomotor neurons). While we calculated that there were significant differences in timing within this group, they all peaked within half a day of each other. Secretomotor neurons (Group 2) peaked more than half a day later than the latest of Group 1, at E16.35, and pilomotor neurons (Group 3) were the last neurons born by over a further half day, at E16.99. It may be necessary to inject BrdU at more frequent intervals to reliably resolve differences in time of peak withdrawal less than half a day.
While the function of a sympathetic neuron appears to be correlated with its birth date, it should be noted that the same outcome could be generated by differential cell death of a pool of neurons born simultaneously. Sympathetic postganglionic neurons normally undergo substantial cell death (Oppenheim, 1991) around and immediately after birth, which is between the time of BrdU injection and the time of analysis. For the observed pattern, cell death would have to selectively kill early born cells for some functional classes of sympathetic neuron and the later born classes for other classes. While this must remain a possibility, neuronal birth date is related to the eventual function of neurons in a range of different places in the nervous system without there being any evidence for differential cell death (Lawson and Biscoe, 1979; Epstein et al., 1992; Cepko et al., 1996; Altman and Bayer, 1997). In the absence of any such evidence, we have discounted the possibility of differential cell death in the following discussion.
What is the relationship of the peak time of withdrawal from the cell cycle (E15.27–E16.99) to the time of target contact of the different classes of sympathetic neurons examined here? The earliest reported contact with a target tissue by sympathetic neurons in the rat is the innervation of the lacrimal gland at E15 (Rubin, 1985). All other reports from a range of tissues suggest that the initial contact is much later, and that most sympathetic neurons first contact their target around the time of birth. For instance, De Champlain et al. (1970), using formaldehyde-induced fluorescence of noradrenaline, reported that the majority of tissues seemed to be contacted shortly before or at the time of birth. In another study, Vidovic et al. (1987) reported that up to one third of the sympathetic axons innervating the iris were present at the time of birth, with the rest arriving postnatally. It thus seems likely that the peak time of withdrawal from the cell cycle (E15–17), for most sympathetic postganglionic neurons destined to innervate the targets investigated in the present study, precedes contact with the target.
If withdrawal from the cell cycle precedes contact with target tissues, it appears very likely that the functional identity of Group 1, Group 2, and Group 3 neurons is restricted prior to target contact. We cannot, however, rule out that the different types of neurons within Group 1 need to interact with their target tissues to specify their final phenotype, but they are differentiated from Group 2 and 3 neurons in the embryo prior to target contact. It is unclear what factors may be playing a role in determining identity in embryonic rat sympathetic postganglionic neurons, but a range of receptors to neurotrophins and other growth factors are expressed by embryonic rodent sympathetic neurons (Glebova and Ginty, 2005; Ernsberger, 2008, 2009).
There is no obvious pattern to the order in which the peak withdrawal from the cell cycle occurs for different functional classes of sympathetic. For example, vasomotor neurons are present in both Group 1 and in Group 3, with their time of peak withdrawal separated by over a day of gestation. Targets in the head (cutaneous vasomotor, muscle vasomotor, iridomotor, secretomotor, pilomotor) are also spread across all three groups. On the other hand, all classes of neuron-targeting structures associated with the skin (pilomotor, cutaneous vasomotor, and sudomotor) are born nearly simultaneously in Group 1.
One final consideration is that developing sympathetic neuroblasts continue to divide even after they start to express markers of a neuronal phenotype, including tyrosine hydroxylase (Rothman et al., 1978, 1980; Rohrer and Thoenen, 1987). There is also evidence that cell division can still occur even after axon extension (Wolf et al., 1999). While these properties make sympathetic neurons unusual among developing neurons, it is unclear how these properties might impact on the choice of target of an individual neuron. Dividing sympathetic neuroblasts express neuronal markers and catecholamine synthetic enzymes from E12 in a rat (Rothman et al., 1980), which is prior to the time that target choices described in the present study are being made. Dividing cells expressing markers of a neuronal phenotype and with axons have been reported on E 15.5–16.5 (Wolf et al., 1999) in rats. Wolf et al. (1999) labelled the neurons from a point 1 mm outside the superior cervical ganglion on the carotid nerves, so it is clear that the cells in question have an axon. However, they represent only a proportion of the total neurons/neuroblasts in the ganglion, and while they may have extended an axon into the carotid nerve, we have argued above that they are unlikely to be in contact with their final target.
The existence of different peak birth dates for different functional classes of rat sympathetic postganglionic neurons resembles that seen for other peripheral neurons of neural crest origin. The function of dorsal root ganglion neurons is related to their birth date (Lawson and Biscoe, 1979; Kitao et al., 1996), which seems to reflect the timing of their migration from the neural crest (for a discussion, see Marmigere and Ernfors, 2007). It is not known whether the timing of migration of different subtypes of sympathetic postganglionic precursor cells is correlated with their ultimate function. In the enteric nervous system, different chemical classes of neuron show variation in the peak rate of withdrawal from the cell cycle (Pham et al., 1991; Chalazonitis et al., 2008). In this case, the different chemical classes of neuron each include multiple functional classes of cell.
Differences in the Proportion of Neurons Retaining BrdU
The current study showed that the proportion of embryonic sympathetic neurons that retained BrdU differed between functional classes of neurons. One possible explanation is that the precursors of different functional classes of sympathetic neuron show differences in S-phase length during E13.5–E18.5, as this would alter the proportion of neurons labelling with BrdU. A less likely alternative is that the differences in the proportion of labelled neurons arise from differences in the handling of BrdU by the different functional classes of neuron. If this were true, it would also reflect differences between the precursors of different functional classes of neurons prior to target contact.
Interestingly, the two neuron types that were significantly different in the proportion of total neurons labelled with BrdU from all other classes of sympathetic neuron examined (and from each other) were both cholinergic (periosteal and sudomotor neurons). Sympathetic cholinergic sudomotor and periosteal neurons have a unique developmental history. They are phenotypically noradrenergic neurons until they contact their target tissues, at which time, under the influence of a neuropoietic cytokine released from the target tissue, they inactivate the genes associated with the noradrenergic phenotype and start expressing cholinergic markers (Landis and Keefe, 1983; Leblanc and Landis, 1986; Landis et al., 1988; Asmus et al., 2000, 2001).
It appears that many sympathetic postganglionic neurons in rodents are capable of responding to the presence of ectopic sweat glands and periosteum (Schotzinger and Landis, 1988; Asmus et al., 2000), which has supported the idea that the phenotype of rodent sympathetic postganglionic neurons is not determined until after target contact. However, based on our results, it now seems that while any sympathetic noradrenergic neuron may have the potential to become a cholinergic neuron under the influence of sweat glands or periosteum, normally a distinct subset of sympathetic neuron precursors project to these tissues and they show differences in the retention of BrdU from other sympathetic precursors prior to target contact, which occurs postnatally for sudomotor neurons (Guidry and Landis, 1998a) and around E18 for the periosteum (Asmus et al., 2000).
Our study has confirmed that rat sympathetic postganglionic neurons in the superior cervical and stellate ganglia withdraw from the cell cycle between E14.5 and E18.5, as previously reported by Landis and Damboise (1986). A previous report (Hendry, 1977) had suggested that some rat sympathetic postganglionic neuroblasts were still undergoing their final cell divisions at the time of birth. The discrepancy may lie in the reliance on cressyl-violet staining to identify neurons among the tritiated thymidine-labelled cells and it is possible that non-neuronal cells may have been being misidentified as neurons. Alternatively, Hendry (1977) may have detected late born neurons from functional populations not represented in our study.
Plug-mated Sprague Dawley rats were used, with E0.5 taken as the morning the plug was discovered. Each pregnant rat was given a single intraperitoneal injection of bromodeoxyuridine (BrdU, 100 mg/kg dissolved in 0.007N NaOH in 0.85% NaCl). A single BrdU injection was made on E12.5, E14.5, E15.5, E16.5, E17.5, E18.5, or E20.5 through a 2-mm incision made in the abdominal midline after the dam was anaesthetised with a mixture of ketamine and xylazine injected intramuscularly (i.m., 60 and 10 mg/kg, respectively). The incision through the body wall and skin was separately sutured and the animal allowed to recover and the pups to be born naturally. When the BrdU-labelled progeny were 28 days old, they were injected with the retrograde tracer, Fast Blue (2% in 10% dimethyl sulphoxide) in different targets on opposite sides of the animal in order to label specific populations of sympathetic postganglionic neurons in the superior cervical or stellate ganglia. For retrograde tracer injection, the animals were anaesthetised with a mixture of ketamine and xylazine i.m. (60 and 10 mg/kg, respectively) and Fast Blue was injected using a glass microelectrode. The pairs of targets injected were: left submandibular salivary gland and right masseter muscle or the skin of the forehead rostral to the left ear and the right anterior lobe of the interscapular brown fat. In separate animals, the anterior chamber of the left eye was also injected, without any other target being injected. The submandibular gland was exposed by a midline incision in the neck, the masseter muscle by a skin incision over the angle of the jaw, and the interscapular brown fat was exposed by a midline incision in the back. The anterior chamber of the eye was injected directly through a 30G needle hole in the lateral margin of the cornea, while the forehead skin was shaved prior to injection.
One week later, the animals were reanaesthetised with ketamine and xylazine i.m. (60 and 10 mg/kg, respectively) and perfused transcardially with fixative (2% formaldehyde, 15% saturated picric acid in 0.1M phosphate buffer) prior to removing the Fast Blue–labelled superior cervical and stellate ganglia. Ganglia were then post-fixed for a total of 2 hr in the same fixative, cryoprotected in 20% sucrose, and sectioned on a cryostat at 12 μm. Sections were then incubated with sheep anti-neuropeptide Y (NPY, AB1583, Chemicon International, Temecula, CA) raised against synthetic NPY conjugated to bovine thyroglobulin, at 1:500 for 24 hr, followed by donkey anti-sheep IgG labelled with Alexa 594 fluor (A-11016, Invitrogen, Carlsbad, CA) at 1:100 for 1 hr. A separate series of sections of stellate ganglia were stained with a rabbit anti-calcitonin gene-related peptide (CGRP) antiserum (CA-08-220, Genosys Biotechnologies Inc. Cambridge, UK) at 1:1,000 for 24 hr, followed by a donkey anti-rabbit IgG (711-075-152, Jackson Immunoresearch, West Grove, PA) at 1:150 for 1 hr.
Sections immunostained as described above were then processed to detect BrdU. Sections were incubated in 2M HCl acid for 30 min at room temperature, washed twice in 0.1M NaBO4 for 5 min, and then washed twice in PBS before incubating in mouse monoclonal anti-BrdU (A1-9452, Bioclone Australia, Marickville, Australia) at 1:100 for 24 hr, followed by a goat anti-mouse Alexa 488 (A-11029, Invitrogen) at 1:400 for 1 hr.
All antisera were diluted in hypertonic saline (1.7% NaCl containing 0.1% sodium azide and 1% Triton X-100). Primary and secondary antisera were washed off with 3 × 5–min washes in 0.01M phosphate buffered saline (PBS) and sections were mounted using fluorescence mounting medium (Dako Corporation, Carpinteria, CA).
Sections were analysed on a Zeiss Axioskop (Thornwood, NY) Axioskop microscope, equipped for fluorescence microscopy with filters capable of visualising Fast Blue, Texas Red/Alexa 594, and FITC/Alexa 488. Photographs were taken on a Zeiss Axiocam Mrm CCD camera, running on Zeiss Axiovision software.
Neurons labelled with Fast Blue from each target were analysed for their content of NPY and BrdU. Only sections that included the nucleus were analysed. In sections labelled with CGRP, all CGRP-immunoreactive neurons and all neurons surrounded by CGRP-immunoreactive pericellular baskets were assessed for BrdU labelling. A total of 72 animals labelled with BrdU were used for the retrograde labelling studies. At least three and up to five animals were used for each combination of age and target, except for the iris, brown fat, and forehead skin injections in E20.5 animals, where only two animals were used for each target.
For each class of neuron, curves representing the proportion of the total neuron projecting to each target labelled by an injection of BrdU on different days were constructed. A general linear model was used to fit separate cubic regression equations to each of the seven curves. Cubic regression equations gave a better fit to the data than did quadratic equations and the use of quartic equations did not improve the fit of the curves to the data. In order to determine which curves differed from which other curves, the location of the peak for each curve was calculated by finding the maximum turning point of the fitted cubic equation. The variability of the maximum turning point for each curve was estimated using a bootstrapping procedure, based on the residuals calculated by subtracting the fitted values from the observed values. These were then randomly resampled with replacement and added to the fitted values from the regression model to obtain new values of the response variable. This was repeated 1,000 times and the family of curves that resulted provided an estimate of the possible variation around the calculated position of the peak, which allowed the calculation of a standard error. The peak position and the estimated standard error were used to estimate the confidence intervals for the difference between each pair of peaks. If the confidence interval did not include zero, then the difference was deemed to be significant.
The proportions of the total number of cells labelled with BrdU across all ages injected for each class of neuron were compared using chi2 analysis. A significant difference within the contingency table was further analysed using a Tukey-type multiple comparison of proportions (Zar, 1984).
For all statistical comparisons, a difference was significant if P < 0.05.
This work was supported by a grant from the National Health and Medical Research Council of Australia. We thank Marnie Phillips of the Department of Mathematics and Statistics, University of Melbourne, for expert statistical advice.