Mouse subthalamic nucleus neurons with local axon collaterals

The neuronal population of the subthalamic nucleus (STN) has the ability to prolong incoming cortical excitation. This could result from intra‐STN feedback excitation. The combination of inducible genetic fate mapping techniques with in vitro targeted patch‐clamp recordings, allowed identifying a new type of STN neurons that possess a highly collateralized intrinsic axon. The time window of birth dates was found to be narrow (E10.5–E14.5) with very few STN neurons born at E10.5 or E14.5. The fate mapped E11.5–12.5 STN neuronal population included 20% of neurons with profuse axonal branching inside the nucleus and a dendritic arbor that differed from that of STN neurons without local axon collaterals. They had intrinsic electrophysiological properties and in particular, the ability to generate plateau potentials, similar to that of STN neurons without local axon collaterals and more generally to that of classically described STN neurons. This suggests that a subpopulation of STN neurons forms a local glutamatergic network, which together with plateau potentials, allow amplification of hyperdirect cortical inputs and synchronization of the STN neuronal population.

We hypothesized that early-born STN neurons would be good candidates for giving off axon collaterals inside the STN based on studies of early-generated hippocampal CA3 neurons which develop into functional hub cells (Marissal et al., 2012;Picardo et al., 2011).
To address this issue, we used inducible genetic fate mapping approaches and in vitro targeted patch-clamp recordings to determine the morpho-physiological features of fate mapped STN neurons. During development, glutamatergic neurons transiently express the transcription factor Neurogenin-2 (Ngn2) as they are becoming postmitotic. We have fate mapped glutamatergic neuron precursors using CreER, a tamoxifen-induced cre recombinase whose expression was placed under the control of the Ngn2 loci.
Glutamatergic progenitors express Ngn2 in accordance with their exit from mitosis, hence the timing of Ngn2 in a neuronal population provides a precise proxy indicator of their birthdate (Britz et al., 2006). We show that at least 20% of STN neurons born at embryonic days E11.5-12.5 displayed profuse axon collaterals with endings inside the STN suggesting the existence of a feedforward excitatory network inside the STN.

| Animals and fate mapping
All animal use protocols were performed under the guidelines of the French National Ethic Comity for Sciences and Health report, in agreement with the European Community Directive 86/609/EEC. All efforts were made to minimize pain and suffering and to reduce the number of animals used. To detect the date of birth of glutamatergic neurons of the STN (Rinvik and Ottersen 1993), we used the fate mapping strategy by taking advantage that glutamatergic neurons transiently express the transcription factor Ngn2 during development as they are becoming postmitotic. Eight-to ten-week-old, Swiss females were crossed with Ngn2CreER WT/Tg /AI14Tomato Tg/Tg males expressing the inducible CreER recombinase under the endogenous Ngn2 promotor and the fluorescent reporter TdTomato (AI14, Madisen et al., 2010).
We fed by gavaging (Feeding Needles, Fine Science Tools, Foster City, CA) pregnant females at embryonic days E10.5-16.5 post vaginal plug, with a tamoxifen solution (2 mg tamoxifen solution per 30 g body weight prepared at 10 mg/mL in corn oil, Sigma, St Louis, MO). For postnatal neuronal birth dates, we treated pups with an intraperitoneal injection of tamoxifen (0.33 mg/g body weight) at postnatal days (P) P1 and P7. Recombination of the reporter allele was achieved within 24 hr upon administration of tamoxifen, therefore, temporally labeling glutamatergic cells as they become postmitotic and providing a proxy indicator of their birthdate.

| Cell counting and immunocytochemistry
We anesthetized 8-week-old Ngn2CreER/AI14 mice (force feeding at E10.5-E16.5 or intraperitoneal injection at P1 or P7) (n 5 27) with a ketamine (100 mg mL 21 ) and xylazine (1 mg mL 21 ) solution (10 mL kg 21 i.p.) and perfused them transcardially with 4% paraformaldehyde in PB (1 mL g 21 ). Brains were postfixed overnight in fixative and then washed in PBS. We performed sagittal brain sections (70 lm) to count the number of Td-Tomato-positive neurons in the whole STN of each side. Brain sections from force feeding time E12.5 mice were routinely processed for immunofluorescence to calcium binding proteins. Briefly, after preincubation in 5% normal donkey serum, sections were first incubated overnight at room temperature under constant shaking in a solution (PBS plus 0.3% Trion-X100, PBST) containing goat-anti
Uncompensated access resistance values below 30 MOhms were considered acceptable and results were discarded if they changed by >20%. We analyzed their passive membrane properties in voltageclamp mode in response to a 10 mV hyperpolarizing voltage step (V H 5 260 mV). We evoked single action potentials (APs) in response to 5 ms depolarizing current pulses whose amplitude was minimally sufficient to reach spike threshold (from Vm 5 280 mV) and analyzed firing properties in response to intracellular sub-and supra-threshold 200 ms current pulses (20-300 pA) at 0.5 Hz. The maximum number of APs was analyzed as the maximum number of evoked APs independently of the injected current intensity.

| Neurobiotin-filled cells morphological analysis
After the recording session we identified recorded neurons as STN neurons by their location inside the STN and characterized their morphological properties, Slices were fixed overnight at 48C in Antigenfix, rinsed in PBS containing 0.3% Triton X-100 (PBST) and incubated overnight at room temperature in Alexa Fluor ® 488-streptavidin (1/1000 in PBST, Jackson Immunoresearch Lab. Inc., West Grove, PA, www.jacksonimmuno.com, RRID: AB_2337249). We performed post-hoc analysis using a confocal microscope (Leica SP5X, Wetzlar, Germany) and We chose this minimal length value of 300 mm based on the maximal length of the parent axon before its first bifurcation. Local axon collaterals of STN neurons, when present, bifurcated 2-15 times before 300 mm. In the quantification of local axonal terminals, we did not take into account axonal varicosities.

| Statistical analysis
We used Prism 6 (GraphPad Software Inc., La Jolla, CA) for statistical analysis. We established significance of all data using Mann-Whitney test except for Sholl analyses (Kolmogorov-Smirnov test). Proportions were compared with Chi-square (and Fisher's exact test). Data were expressed as mean 6 SEM. Significance threshold was set at p .05 for all statistical comparisons.

Comparative Neurology
In order to detect whether a subpopulation of fate mapped STN neurons had intra-STN axon collaterals, we labeled E11.5-12.5 Td-Tomato-positive STN neurons (n 5 115) with neurobiotin during patch-clamp recordings. We discarded neurons that were not clearly identifiable (n 5 19), and those which had no axon or an axon shorter than 300 mm (n 5 57) (see methods). The remaining 39 STN neurons had the classically described, simple, bifurcated, or trifurcated, long range projecting axons running up to the substantia nigra and/or globus pallidus. Among them, nearly 20% (n 5 7/39) had also intra-STN axon collaterals (Figure 3a,d). The number of axonal nodes of local axon collaterals ranged from 4 to 19 (11 6 2) (Figure 3c (Table 1).   Figure 4f).
The above results suggested that STN neurons that give off intranucleus axon collaterals have larger dendritic arbors. These collateralized STN neurons did not show preferential distribution inside the STN, either rostro-caudally or latero-medially (Figure 4g).
The two subpopulations of E11.5-12.5 STN neurons had similar mean membrane resistance and capacitance (Table 1). Mean properties of single action potentials (rheobase, threshold, peak amplitude, half-width duration, and AHP amplitude) and evoked or spontaneous firing were also similar for the two groups (Figure 5a-c and Table 1).
This is comparable to the birth dates identified in mice in the present study considering the longer duration of embryogenesis in rats (21 days instead of 19). STN neurons originate from the retromammillary region of prosomere four, which occupies the basal plate of the diencephalon (Puelles & Rubenstein, 2003). After exiting the cell cycle, prospective STN neurons migrate latero-dorsally, crossing over the descending tracts of the developing cerebral peduncle. They migrate over a 2-day period according to an outside-in pattern, with the earliest born neurons occupying the most rostral and dorsolateral regions of the STN, and later neurons residing caudal and ventromedial to earlier neurons (Altman & Bayer, 1979). The cytogenetic gradient identified in mice in the present study is in agreement with the above results in rats.

| Morphological-functional characteristics
All rodent STN neurons possess long-range axons, which bifurcate or trifurcate before running to internal and/or external pallidal segments (GPi in primates or entopeduncular nucleus (EP) in rodents, GPe in primates or GP in rodents) and SN. The situation is somewhat different in primates where the majority of STN axons have a single branch coursing rostrally to GPe and GPi. Therefore the primate STN neuronal population is not as homogenous as in rodents with regard to their projection sites (Sato, Parent, Levesque, & Parent, 2000). The present result also seriously challenges the homogeneity of rodent STN neurons since some of them also display profuse local axon collaterals but the others do not. Previous studies in rodents including our own have identified different proportions of STN neurons with short recurrent axon collaterals (Ammari et al., 2010;Chang et al., 1984;Hammond & Yelnik, 1983;Kita et al., 1983) whereas they have not yet been described in primates (Sato et al., 2000). The difference between the present study and previous ones is the extent of labeling of these collaterals and the identification of putative terminal boutons. Truncation of axon collaterals due to the slicing procedure in the present study may have resulted in underestimation of the number of retrieved boutons.
Previous reports described rat STN neurons as displaying five to eight long, sparsely spined dendrites arborizing mostly along the main axis of the nucleus (Afsharpour, 1985;Hammond & Yelnik, 1983).
Mouse fate mapped STN neurons had a mean of four dendritic trunks (2-9) and their dendrites were also sparsely spined. The ratio of the number of dendritic ends/trunks between collateralized and noncollateralized STN neurons was different suggesting that collateralized STN neurons had a more ramified dendritic tree, as previously reported in the rat (Kita et al., 1983). This was supported by the Sholl analysis showing a higher number of intersections. However, both groups of neurons had similar numbers of dendritic ends and trunks.
All intrinsic membrane properties tested at juvenile stage for fate mapped collateralized STN neurons were similar to that of fate mapped STN neurons without axon collaterals and to that of the juvenile STN neuronal population in general (Beurrier et al., 2000;Beurrier et al., 1999;Bevan, Magill, Hallworth, Bolam, & Wilson, 2002). Therefore, STN neurons with local axon collaterals are yet impossible to identify from their electrophysiological characteristics during patch-clamp recordings. Only post-hoc morphological studies can do so.

| Functional consequences
If local STN axon collaterals contact dendrites of other STN neurons (Chang et al., 1984), they are good candidates for sustaining a polysynaptic activity in the STN (Gillies & Willshaw, 1998). Indeed, a single stimulation of cortical afferents to the STN can evoke STN synchronization and widespread long lasting polysynaptic excitations in target nuclei and the mechanism of amplification resides in the STN itself (Magill, Sharott, Bevan, Brown, & Bolam, 2004); Ammari et al., 2010;Shen & Johnson, 2006). We propose that it combines the effect of the STN polysynaptic excitatory network and that of voltage dependent depolarizing currents of the STN neuronal membrane. Such amplification would underlie the broad suppressive effect (inhibition of action) implemented by the cortico-STN pathway, which then via GPi and SNr suppresses thalamocortical drive (reviewed in Wessel & Aron, 2017).