Dopaminergic control of postnatal mammalian spinal network function is predominantly inhibitory

Dopamine is well known to regulate movement through the differential control of direct and indirect pathways in the striatum that express D1 and D2 receptors respectively. The spinal cord also expresses all dopamine receptors however; how the specific receptors regulate spinal network output in mammals is poorly understood. We explore the receptor-specific mechanisms that underlie dopaminergic control of spinal network output of neonatal mice during changes in spinal network excitability. During spontaneous activity, which is a characteristic of developing spinal networks operating in a low excitability state, we found that dopamine is primarily inhibitory. We uncover an excitatory D1-mediated effect of dopamine on motoneurons and network output that also involves co-activation with D2 receptors. Critically, these excitatory actions require higher concentrations of dopamine; however, analysis of dopamine concentrations of neonates indicates that endogenous levels of spinal dopamine are low. Because endogenous levels of spinal dopamine are low, this excitatory dopaminergic pathway is likely physiologically-silent at this stage in development. In contrast, the inhibitory effect of dopamine, at low physiological concentrations is mediated by parallel activation of D2, D3, D4 and α2 receptors which is reproduced when endogenous dopamine levels are increased by blocking dopamine reuptake and metabolism. We provide evidence in support of dedicated spinal network components that are controlled by excitatory D1 and inhibitory D2 receptors that is reminiscent of the classic dopaminergic indirect and direct pathway within the striatum. These results indicate that network state is an important factor that dictates receptor-specific and therefore dose-dependent control of neuromodulators on spinal network output and advances our understanding of how neuromodulators regulate neural networks under dynamically changing excitability. Significance statement Monoaminergic neuromodulation of neural networks is dependent not only on target receptors but also on network state. We studied the concentration-dependent control of spinal networks of the neonatal mouse, in vitro, during a low excitability state characterized by spontaneous network activity. Spontaneous activity is an essential element for the development of networks. Under these conditions, we defined converging receptor and cellular mechanisms that contribute to the diverse, concentration-dependent control of spinal motor networks by dopamine, in vitro. These experiments advance understanding of how monoamines modulate neuronal networks under dynamically changing excitability conditions and provide evidence of dedicated D1 and D2 regulated network components in the spinal cord that are consistent with those reported in the striatum.


Introduction 1
Neuromodulators are critical for central nervous system function and diversify circuit outputs by altering 2 synaptic and intrinsic properties [1][2][3]. Dopamine is a monoamine neuromodulator that is well known for 3 action selection in vertebrates through the control of direct and indirect circuits of the basal ganglia that 4 express excitatory D1 and inhibitory D2 receptors respectively (for review, see [4] and [5][6][7] for 5 examples). Dopamine is also important for the regulation of spinal motor networks that control rhythmic 6 movements including but not limited to locomotion (for review, see [8]). In larval zebrafish, phasic and 7 tonic firing patterns of descending neurons that provide the primary source of spinal dopamine correlates 8 with locomotor episodes and quiescence, respectively [9]. These different firing patterns can impact the 9 cellular release of dopamine and as a consequence, the receptor subtypes it activates [10]. For example, 10 high levels of dopamine released during phasic cell firing activate lower affinity excitatory D1 receptors 11 and promote locomotor activity, and lower levels of dopamine released during tonic activity activate 12 higher affinity inhibitory D2 receptors and suppress motor output [11]. Similarly, dopamine has dose-13 dependent effects on locomotor circuits in precocial species with functional swim networks [12]. 14 15 We have recently demonstrated that neuromodulation of developing mammalian spinal circuits is state-16 dependent [13] which is consistent with work in invertebrates [14,15]. This is important in developing 17 spinal motor networks which produce a wide repertoire of patterned outputs at birth, including locomotor 18 activity, as a consequence of dynamically fluctuating network excitability. That being said, neonatal 19 rodents rarely produce coordinated bouts of locomotion and in vitro preparations of isolated spinal cord 20 require pharmacological or electrical stimulation to drive the network into a high excitability state to 21 produce fictive walking patterns. Instead, most of the movements observed in neonatal mice are ataxic. 22 (See Supplementary Video 1). These movements correlate with spontaneous network activity, which can 23 be observed in vitro [16][17][18][19]. Nevertheless, the vast majority of what we know about neuromodulation of 24 developing mammalian spinal networks has been derived from studies on fictive locomotor activities with 25 dopamine being predominantly excitatory when spinal networks are operating in this state [20][21][22][23][24][25][26][27][28][29][30][31]. 26 27 Based on our previous work, we hypothesized that receptor actions, and therefore concentration-28 dependent control of network output by dopamine, is linked to the underlying network excitability state 29 [13]. As a result, more complex receptor-dependent actions may have been masked during high 30 excitability states of fictive locomotion. We therefore examined how dopamine modulates spinal output at 31 the network and cellular level of neonatal mouse spinal cords in vitro during a low excitability state 32 characterized by spontaneous activity [16,17]. We found that the receptor-specific effects of dopamine 33 are fundamentally different during a low excitability state compared to what has been previously reported 34 during high excitability states. Specifically, during a low excitability state, dopamine has primarily 35 inhibitory effects on network output acting in parallel through activation of D2, D3, D4, and α2 receptors. 36 We uncover an excitatory effect of dopamine that is likely physiologically silent. This is because 37 endogenous dopamine levels in the spinal cord are low at this age and the excitatory effects, involving 38 coactivation of both D1 and D2 receptors, require higher concentrations of dopamine. We also found that 39 excitatory and inhibitory dopaminergic pathways act through the control of dedicated network 40 components. Specifically, the D1 pathway acts through excitation of motoneurons, likely recruiting 41 recurrent excitatory circuits [32], with the D2 pathway acting through hyperpolarization of multiple 42 ventral interneuron subclasses. Portions of these data were presented in abstract form [33,34]. 43 45 We focused on the modulation of perinatal spontaneous activity patterns to investigate how dopamine 46 modulates spinal network output during a physiologically relevant low excitability state. Using the same 47 experimental set-up as Sharples et al. [23], we recorded spontaneous motor activity with extracellular 48 suction electrodes from single ventral roots of the second or fifth lumbar (L2/L5) segments 49 simultaneously from two or four spinal cord preparations sharing the same chamber. Each preparation in 50 this series of experiments was naïve to dopamine exposure and received only a single dose of its 51 respective concentration. This configuration ensured consistent experimental conditions across several 52 To delineate receptor contributions to dopamine's bidirectional effects on endogenous spontaneous 60 activity in isolated spinal cord preparations, we used antagonists selective for the family of dopamine 61 receptors. At low concentrations of dopamine, we observed a negative response ratio ( Fig. 2A, Di), which 62 was due to a reduction in the number (Fig. 2Dii), but not the amplitude, of spontaneous episodes (Fig.  63 2Diii). In contrast to our hypotheses, the inhibitory effect of dopamine at low concentrations (10 µM) was Tukey post hoc, p = 0.98). Burst analysis revealed no difference in the number or amplitude of episodes 73 when antagonists were present, compared with baseline (Fig. 2Dii, Diii; two-way ANOVAs for number, 74 F(2,28) = 9.5, p < 0.0001; two-way ANOVA for amplitude: F(4,28) = 3.4, p = 0.023). Together these data 75

Dopaminergic modulation of spinal motor networks is dose-dependent and bidirectional
suggest that the inhibitory actions of dopamine depend on parallel action of D2, D3 and D4 receptors. 76 77 We considered the possibility that the inhibitory influence of low dopamine concentrations was partly due 78 to the activation of non-dopamine receptors. Previous work conducted in our laboratory showed that 79 dopamine inhibits cauda equina-evoked locomotion, partially via α2-adrenergic receptors [30]. The 80 remaining minor inhibitory effect of dopamine at 10 µM was blocked by antagonizing the α2-adrenergic suggesting that non-dopamine receptors also contribute to dopamine's inhibitory effects. 86 87

D1 -D2 receptor coactivation contributes to dopaminergic excitation of spinal network activity 88
Dopamine binding to D1 and D2 heteromers can lead to depolarization via an increase in intracellular 89 calcium levels, mediated by the enzyme phospholipase C (PLC) [35][36][37][38]. Thus, we first tested the role of 90 the lower affinity D1 receptor system and then examined whether the D2 receptor system has a 91 cooperative role in the control of spontaneous activity. With the addition of dopamine at high 92 concentrations (i.e., 50 & 100 µM), spontaneous activity patterns became rhythmic, often producing a 93 slow rhythm with episodes of high frequency rhythmic activity. Moreover, the presence of the D1-like 94 receptor antagonist, SCH-23390, reduced this effect ( Fig. 3; 100 µM dopamine with 10 µM SCH-23390; 95 n = 5; response ratio, F(5,35) = 11.4, p < 0.001); fast rhythm power, F(4,27) = 12.6, p < 0.001; slow rhythm 96 power, H(4) = 12.8, p = 0.013; Fig. 3  than excite responses. We tested this idea by administering D2-like antagonists (sulpiride + L745,870) and 103 found that the power of the fast rhythm elicited by dopamine at 100 µM was reduced ( Fig. 3 Aiii, Biii, 104 Cii; n = 4; F(4,27) = 12.6, p < 0.001) to the same extent as the D1-antagonist, with no effect on the power of To explore the interaction and co-activation profile of D1 and D2 receptors in isolated neonatal spinal 109 cords, we performed co-immunoprecipitation for D1 and D2 receptors and used agonists to activate both 110 receptor subtypes. After immunoprecipitating D2 receptors from neonatal spinal cord lysates, we used an 111 antibody to probe for D1 receptors. We detected D1 receptor protein within the D2 receptor post hoc: p = 0.02). We observed no difference in the amount of spontaneous network activity evoked 120 with co-application of a D2 agonist, compared with application of the D1 agonist alone, as indicated by the 121 response ratio (Fig. 4B, Cii; one-way ANOVA, F(3,29) = 12.0, p < 0.001; Tukey post hoc, p = 0.5). In 122 contrast, lower concentrations of the same agonists (10 µM) produced no effects (n = 8 for each 123 condition; DC potential, t(6) = 0.73, p = 0.24; response ratio, t(6) = 0.9, p = 0.19). Thus, consistent with 124 previous reports for striatal neurons [36], we found a dose-dependent effect of dopamine agonists wherein 125 co-applying high doses, but not low doses, of D1 and D2 receptor agonists, produced more robust 126 depolarization than a D1 agonist alone. 127

128
In addition to co-applying separate D1 and D2 agonists, we tested co-activating D1 and D2-like receptors 129 with the D1/D2 co-agonist SKF 83959 (50 µM) [38,39]. As predicted, the co-agonist elicited a more 130 robust depolarization of the ventral root DC potential, compared with the D1 agonist, when applied alone 131  Interestingly, the co-agonist also robustly facilitated superimposed spontaneous activity, as indicated by a 133 larger response ratio than co-application of the D1 and D2 agonists produced (

Low levels of endogenous spinal dopamine inhibit spontaneous activity 139
We next examined how the endogenous dopamine system regulates perinatal spinal network function.

D1 receptor activation increases motoneuron excitability by reducing afterhyperpolarization properties 161
In the next set of experiments, we were interested in determining the cellular mechanisms that mediate 162 dopamine's complex modulatory effects on spinal network output. As integrators of premotor network 163 activity that generate many of the rhythmic outputs of the spinal cord, motoneurons are ideally suited to 164 amplify spontaneous activity and respond to dopaminergic modulation. Given that they not only serve as 165 the final output for spinal networks, but they also participate in the generation of rhythmic activity [40],

Dopaminergic inhibition through D2 -receptor hyperpolarization of distributed populations of ventral 204 interneurons 205
We next set out to determine the cellular mechanisms that mediate the inhibitory effects of dopamine on 206 spinal network output with motoneurons as our first target. In contrast to our network recordings, low 207 concentrations of dopamine (1-10 µM; n = 20 cells across 14 animals; Table 1) and the D2 agonist 208 quinpirole (20 µM; n = 12 cells across seven animals; Table 1)  higher in responders than in non -responders (Table 2) and a greater proportion of responders were 237 localized to more medial regions of spinal slices (Fig. 7B), where putative commissural interneurons 238 reside [46]. 239 240 We next set out to determine the type of interneurons that were hyperpolarized by quinpirole. Given that 241 many of the responding cells were located medially, we next targeted descending commissural 242 interneurons (dCINs; n = 10 cells across five animals; Table 2) since this population can be identified 243 based on anatomical connectivity [47,48], display intrinsic burst properties [49] and are rhythmically-244 active during neurochemically-evoked fictive locomotion [50]. dCINs were retrogradely labelled with 245 tetramethylrhodamine-conjugated dextran amine (molecular weight (MW) 3000; Molecular Probes, Inc.) 246 inserted into the ventrolateral funiculus at the L4 segment (Fig. 7A). In contrast to our hypothesis, only 247 one dCIN responded with a sustained hyperpolarization and two were transiently hyperpolarized (Fig. 7B) 248 by quinpirole. Quinpirole did not alter any passive, spike, or repetitive firing properties of dCINs (n = 10; 249 While dCINs can be identified anatomically, they are heterogeneous with respect to their neurotransmitter 258 phenotype [51] and as a result have varying contributions to network activities [46,50,52]. We therefore 259 next targeted V3 interneurons which are exclusively glutamatergic, contribute to the stabilization of 260 locomotor-like rhythmicity and can be identified genetically based on the expression of the Sim1 261 transcription factor [53]. Heterogeneity with respect to location, morphology and electrophysiological 262 properties has also been reported in V3 interneurons [54,55] which may be accounted for in part by a 263 recently described hierarchical microcircuit whereby a medial population projects to a lateral population 264 which provide glutamatergic excitation of ipsilateral motoneurons. Both populations also project 265 commissurally and receive recurrent glutamatergic inputs from intra and intersegmental ipsilateral 266 motoneurons [32]. Given that dopamine inhibits ventral root-evoked locomotor activity, which may be 267 mediated by this circuit [56], through D2-receptor signaling [31], we hypothesized that V3 interneurons 268 may be a cellular locus for D2-mediated inhibition of spinal network activity. The magnitude of the response in the 5 cells that responded with a sustained hyperpolarization was 274 variable and approached, but did not reach, significance (Supp Figure 1; paired t-test: t(4)= 2.5, p = 0.06); 275 however, did reach significance when the cells that responded with a transient hyperpolarization were Dopamine is a monoamine neuromodulator that is important for the control of rhythmically active motor 283 circuits across phyla (reviewed by [8]) but is probably best known in vertebrates for the control of 284 dedicated circuits in the basal ganglia that control action selection (reviewed by [57] ). Work in small 285 circuits of invertebrates has established that circuit connectomes define the constraints on which networks 286 operate and that neuromodulators diversify outputs by altering intrinsic and synaptic properties of the 287 neurons that compose the circuit (for reviews see [1][2][3]. In line with this, the distribution of receptors 288 within circuits constrain the effect of neuromodulators on circuit output. For example, dopamine is 289 exclusively inhibitory in spinal circuits of Xenopus tadpoles prior to free-swimming stages [58] due to 290 expression of D2 but not D1 receptors. We show that dopamine has bidirectional concentration-dependent 291 effects on spinal network output in neonatal mice where all dopamine receptor types are expressed which 292 is consistent with what has been reported in tadpoles at free swimming-stages [12]. Our data highlights 293 that neuromodulator concentration is also important because receptors have varying ligand affinities 294 which underlie concentration-dependent actions of modulators. Although dopamine predominantly 295 inhibits spinal output in neonatal mice, similar to pre-free-swimming tadpoles, it is primarily due to the 296 concentration of spinal dopamine, not the distribution of receptors. Our previous work shows that 297 neuromodulation of mammalian spinal networks is dependent on network excitability state [13] which is 298 consistent with findings from invertebrates [14,15]. Our current work shows that receptor mechanisms 299 and concentration-dependent control of spinal network output is also state-dependent. This is important 300 because receptor expression, modulator concentration and network excitability are not fixed and fluctuate 301 dynamically [59,60]. Therefore, these three factors need to be considered if we wish to understand how 302 networks create diverse neuromodulator-dependent outputs ( Figure 8A). 303

A physiologically silent excitatory D1 pathway? 304
Previous work has shown predominantly excitatory D1-receptor mediated effects of dopamine on fictive 305 locomotion which is characteristic of the network operating in a high excitability state [13] -albeit in 306 these studies, higher concentrations of dopamine are necessary to elicit observable effects [20][21][22][23][24][25]. Here, 307 we show that endogenous levels of dopamine within the spinal cord of neonatal mice are low. Even 308

though HPLC suggests low concentrations of dopamine in the neonatal spinal cord, a critical question that 309
we addressed is what would happen when we manipulated endogenous dopamine? We accomplished this 310 by blocking dopamine reuptake and metabolism and found that the effects were not excitatory, but 311 inhibitory. Based on this, we suggest that although present, the excitatory D1-mediated pathway is 312 'physiologically silent' during early postnatal stages given that endogenous levels are not sufficient to 313 activate this pathway. A similar phenomenon has been reported for glutamatergic synapses in developing 314 circuits of the hippocampus which express NMDA but not AMPA receptors. As a result, glutamatergic 315 synapses fall 'physiologically silent' as the release of glutamate does not produce sufficient 316 depolarization of the postsynaptic membrane to remove the magnesium block from the pore of the 317 NMDA channel [61,62]. an increase in motor activity [63]. Similarly, the excitatory D1 system is more important for the control of 327 stepping movements whereas the inhibitory D2 system plays less of a role [22,64]. Instead, the inhibitory 328 D2 pathway may be more important in maintaining network quiescence during periods of immobility such 329 as has been inferred by the study of firing patterns in the zebrafish [9]. Our data demonstrates that in the 330 neonatal mouse, where spinal dopamine levels and network excitability are low, that inhibition of motor 331 output prevails (Figure 8Bii.). Thus, this points to the receptor-specific and therefore concentration-332 dependent control of modulators on network output being strongly influenced by the state of the network 333 on which they are acting. 334 335

Dedicated network components segregate excitatory and inhibitory control of spinal networks 336
Dedicated circuits regulated by non-overlapping populations of neurons that express D1 and D2 receptors 337 compose the direct and indirect pathways of the basal ganglia in vertebrates and have also been reported 338 in the superior colliculus of rodents [65]. Our work suggests dedicated network elements within the spinal 339 cord that are regulated by D1 and D2 pathways (Figure 8Biii.). Specifically, we found that D1 receptors 340 excite motoneurons through similar mechanisms that have been previously reported [44,45], but are not 341 affected by low concentrations of dopamine or D2 agonists. This points to the possibility of a dedicated 342 D1-dependent circuit that could underlie the generation of rhythmic activities elicited by high 343

concentrations of dopamine. Motoneurons compose key rhythm generating elements in invertebrate 344
circuits [66][67][68] and also participate in rhythm generation in vertebrates [69], including rodents 345 [32,40,56,70,71]. V3 interneurons are one subclass of genetically-defined spinal interneuron that are 346 important for the generation of rhythmic activities in mammalian spinal networks [53,55,72] and receive 347 recurrent excitatory collaterals from motoneurons in rodents [32]. Motoneurons in the rodent spinal cord 348 also form glutamatergic synaptic connections amongst each other [71] and activation of D1 receptors 349 could serve to synchronize motor pools. Previous results demonstrating D1-and not D2-mediated 350 increases in AMPA conductances on motoneurons [45] support this possibility. We cannot rule out the 351 possibility that there is a degree of overlap between D1 and D2 controlled network elements within the 352 spinal cord. While we did not examine the D1 control of ventral interneurons, we did find that D2 353 receptors hyperpolarize a subset of V3 interneurons. This population is therefore a potential cellular locus 354 where cooperative excitatory D1-D2 interactions that we report here could occur. D1 and D2 receptors have 355 been reported in the brain to become co-activated or form heterodimeric complexes that augment 356 neuronal excitability through PLC-dependent increases in intracellular calcium [35][36][37][38]. Our 357 pharmacological and immunoprecipitation data indicate that this may also occur in the neonatal mouse 358 spinal cord. Although this pathway may be physiologically silent during early postnatal development, 359 increasing dopamine concentrations at later stages may activate this pathway. 360

361
The inhibitory D2 pathway on the other hand does not appear to act through the modulation of 362 motoneuron intrinsic or synaptic properties [45] but instead hyperpolarizes a proportion of ventral 363 interneurons. Based on our responsiveness criteria, a majority of cells that responded to a D2 agonist with 364 sustained hyperpolarization of the membrane potential were localized more medially in lumbar slices. We 365 tested the hypothesis that a subpopulation of interneurons was D2 sensitive and tested both dCINs and 366 genetically identified V3 interneurons. Our hypothesis was not supported, and similar proportions of D2 367 sensitive neurons were found; however, the data suggest that there is heterogeneity in the responsiveness 368 to neuromodulators within a class of genetically defined interneurons and that D2 actions may not be 369 localized to a particular class of interneurons. One possibility is that neuromodulators elicit a robust effect 370 on network output through distributed control across multiple classes of genetically-defined interneurons 371 which would be consistent with findings that the locomotor rhythm generator is also distributed across 372 several classes of interneurons [73][74][75]. Our network data suggests that low concentrations of dopamine 373 inhibit network output by acting in parallel on D2, D3, D4 and ɑ2 receptors. We may therefore have 374 underestimated the cellular targets that underlie dopamine's robust inhibitory effect on the network given 375 that in this series of experiments we looked at the activation of D2 receptors alone. 376 377

Developmental considerations for the endogenous dopaminergic system 378
This robust inhibitory system may act as a brake on network activity during perinatal development when 379 chloride-mediated synaptic transmission still causes partial depolarization [76]. An inhibitory brake Signaling through [76] D2-like receptors may also play a role in driving the maturation of spinal 386 networks. In larval zebrafish, D4 receptors drive the maturation of spinal locomotor network organization 387 [77] and function leading to changes in locomotor behaviour [78]. Similar processes may also occur 388 perinatally in rodents, in that the preferential activation of the D2 receptor system may favour intracellular 389 signaling that results in network reorganization. Serotonin receptors have been found to shape network 390 function and inhibitory synaptic transmission during early postnatal days of rodents [79,80]. Dopamine 391 could, therefore, act analogously via the D2-system during perinatal development. 392 393

Conclusions 394
Here we present evidence for an inhibitory physiological role of dopamine in the regulation of developing 395 mammalian spinal networks. We also demonstrate an excitatory D1-mediated pathway that acts through 396 excitation of motoneurons, and possibly recurrent excitatory collaterals to CPG neurons, however given 397 that endogenous levels of dopamine are low, propose that this pathway is physiologically silent. These NaHCO3, 30 ᴅ-glucose; 310-315 mOsm.). We exposed the spinal cord with a ventral laminectomy and 414 isolated it by cutting the nerve roots that connected it to the vertebral column. The isolated spinal cord 415 was then transferred to a recording chamber, perfused with carbogenated aCSF, and placed ventral side 416 up. The bath temperature was gradually increased to 27°C [81]. We let the spinal cords stabilize for 1 417 hour before performing experiments. 418 419

Spinal cord slice preparation 420
Following isolation, spinal cords were transected above the tenth thoracic (T10) and below the first sacral 421 (S1) segments and transferred to a slicing chamber. Pre-warmed liquefied 20% gelatin was used to secure 422 cords to an agar (3%) block that was super-glued to the base of a cutting chamber and immersed in ice-423 cold, carbogenated, high-sucrose slicing aCSF (in mM, 25 NaCl, 188 sucrose, 1.9 KCl, 10 Mg SO4, 1.2 424 Na2HPO4, 26 NaHCO3; 25 D-Glucose; 340 mOsm). Using a vibratome (Leica, Bussloch, Germany) we 425 cut 250-µm-thick lumbar slices, collected and transferred them to a recovery chamber containing regular 426 carbogenated aCSF (see Tissue Preparation) heated to 32 o C for one hour, then maintained them at room 427 temperature for at least 30 minutes before transferring them to a recording chamber. 428 429

505
To reduce nonspecific binding, we first incubated lysates in anti-rabbit Ig agarose beads (Trueblot; 506 Rockland Inc., Limerick, PA) for 30 minutes, on ice and in the absence of primary antibody. We then 507 removed the supernatant and incubated the lysates on ice for 1 hour with rabbit antibody to D2 receptors 508 (1 μg per 100 μL, Millipore). Anti-rabbit Ig IP beads were added and samples were incubated overnight at 509 4°C with gentle agitation. Immunoprecipitates were washed with lysis buffer, heated in loading buffer 510 (350 mM Tris, 30% glycerol, 1.6% SDS, 1.2% bromophenol blue, 6% β-mercaptoethanol) to 95°C for 10 511 min, electrophoresed on a precast SDS gel (4-12% Tris HCl; BioRad, Hercules, CA), and transferred 512 onto a nitrocellulose membrane. After blocking, the membranes were incubated with guinea pig antibody All sections processed via immunohistochemistry were imaged on a Nikon A1R MP + microscope (Nikon, 537 Tokyo, Japan) operating in confocal mode with a 16× water-immersion objective lens (numerical aperture 538 [NA] = 0.8, working distance [WD] = 3 mm). Image acquisition used a z-step of 1 µm and averaged two 539 frames with a resolution of 2048 × 2048. Pixel dwell time was 2.5 ms and exposure settings were 540 maintained for all sections. We used NIS-Elements software (Nikon) for image acquisition and ImageJ to 541 perform maximum intensity projections of z-stacks. 542 543

High-performance liquid chromatography 544
Monoamine content of neonatal and adult spinal cords was measured using high-performance liquid 545 chromatography (HPLC). We dissected spinal cords from neonatal (P3, n = 11) C57BL/6 mice in aCSF as 546 described above and extracted adult spinal cords (P60, n = 17) with a pressure ejection method. Tissue 547 was then flash-frozen with liquid nitrogen, stored at −80ºC, and analyzed for biogenic amines with a 548 modified version of the Parent et al. HPLC method [76]. Tissue was homogenized in ice-cold 0.1 M 549 perchloric acid. We centrifuged the homogenate and used 10 μl of supernatant in the assay, employing an 550 Atlantis dC18 column (Waters, Milford, MA) and an electrochemical detector. 551

Data analysis 552
We determined the relative inhibitory or excitatory effects of dopamine and dopamine receptor agonists 553 on spontaneous motor network activity using methods similar to those in our previous work [23]: we 554 calculated a response ratio from single ventral root neurograms between the root mean square of 5 555 minutes of basal spontaneous activity and 5 minutes of activity recorded 20 minutes after adding the drug. 556 We subtracted 1 from the response ratio so that positive values reflect excitation and negative values 557 reflect inhibition. The response ratio was used as a high throughput assay to detect global changes in 558 network activity. Neurogram data were analyzed with Spike2 software. Bursts of spontaneous activity 559 were analyzed using Clampfit (Molecular Devices) to determine how episode number and amplitude 560 contributed to changes in response ratio. Spectral analyses were conducted using Spinalcore software [83]  interest that corresponded to the fast and slow rhythms were used as a measure of rhythm robustness. 570 Data were segmented into 30 s bins and averaged over 5-minute intervals for statistical analysis. We used 571 tools available in Spinalcore for all analyses of rhythmic motor activity [83], consistent with Sharples 572 and Whelan [13]. were analyzed as in our previous work [82,84], with the exception of repetitive firing analyses, which 576 examined the instantaneous firing rate for the first spike interval and steady-state firing separately. 577

Experimental design and statistical analysis 578
All experiments were repeated measures. We tested for differences in the magnitude of effects between 579 conditions with one-way ANOVAs, focusing on comparisons to time-matched vehicle controls. Two-way 580 ANOVAs compared baseline to multiple post-drug conditions. All effects surpassing a significance 581 threshold of p < 0.05 were further examined with post hoc analyses. We used Holm-Sidak post hoc tests 582 to compare all treatment conditions to the appropriate normalized time-matched vehicle control. Data that 583 violated assumptions of normality (Shapiro-Wilk test) or equal variance (Brown-Forsythe test) were 584 analyzed via nonparametric Mann-Whitney U (if two groups) or Kruskal-Wallis (if more than two 585 groups) tests. Table 1: Baseline motoneuron intrinsic properties. Passive, spike, and repetitive firing properties of motoneurons with dopamine or dopamine agonists did not differ at baseline in this series of experiments. Motoneuron identity was verified in a subset of experiments where cells were filled with fluorescein and verified post hoc for expression of choline acetyltransferase (ChAT); 100% of cells were ChAT + indicating that they were indeed motoneurons. Data are means ± SD and analyzed using one-way ANOVAs for each property. 0.5, 0.8 Table 2: Baseline ventral interneuron intrinsic properties Passive, spike, and repetitive firing properties of ventral interneurons that responded to quinpirole with sustained hyperpolarization of membrane potential (responders) were compared with those that did not respond (non-responders) and also retrogradely labelled descending commissural interneurons (dCINs). Responders had higher capacitance and holding current than non -responders and dCINs. Data are means ± SD; F and p values are reported from one-way ANOVAs for each property. Red values highlight variables where significant main effects were detected with p < 0.05. Superscript numbers reflect significant differences between respective conditions from post hoc analysis. (n=40 cells, 22 animals (P0-4)