|Home | About | Journals | Submit | Contact Us | Français|
Motor behaviors such as walking or withdrawing the limb from a painful stimulus rely upon integrative multimodal sensory circuitry to generate appropriate muscle activation patterns. Both the cellular components and the molecular mechanisms that instruct the assembly of the spinal sensorimotor system are poorly understood. Here we characterize the connectivity pattern of a sub-population of lamina V inhibitory sensory relay neurons marked during development by the nuclear matrix and DNA binding factor Satb2 (ISRSatb2). ISRSatb2 neurons receive inputs from multiple streams of sensory information and relay their outputs to motor command layers of the spinal cord. Deletion of the Satb2 transcription factor from ISRSatb2 neurons perturbs their cellular position, molecular profile, and pre- and post-synaptic connectivity. These alterations are accompanied by abnormal limb hyperflexion responses to mechanical and thermal stimuli, and during walking. Thus, Satb2 is a genetic determinant that mediates proper circuit development in a core sensory-to-motor spinal network.
The movements of animals are adapted to their environment by a highly integrative circuitry within the spinal cord that refines motor actions based on a variety of sensory cues (reviewed in Baldissera et al. 1981). Exteroceptive sensory information continuously streams into the central nervous system as an animal moves and includes detection of limb position, mechanical pressure and in the case of environmental threats, noxious pain. These sensory feedback pathways can rapidly and dynamically regulate motor actions via local spinal circuitry, even in the absence of supraspinal connections (Forssberg et al., 1975; Rossignol et al., 2006). Electrophysiological studies of sensorimotor circuits have found that both excitatory and inhibitory neurons within the dorsal spinal cord participate in the transformation of sensory cues into motor actions (Bannatyne et al., 2009; Cavallari et al., 1987; Schouenborg and Weng, 1994). Remarkable progress has been made using circuit mapping and genetics to identify neurons that gate and relay specific sensory modalities to higher brain centers (Bourane et al., 2015a; Bourane et al., 2015b; Braz et al., 2005; Duan et al., 2014; Fink et al., 2014), but less is known about the molecular and developmental features of spinal neurons that integrate different sensory streams for rapid modulation of the motor system.
It has long been recognized that multimodal sensory cues intersect within the medial deep dorsal horn of the spinal cord where direct-proprioceptive and indirect-nociceptive pathways converge (reviewed in Craig, 2003). Interestingly, circuit tracing has found that a significant fraction of spinal premotor neurons are likewise located in the medial deep dorsal horn (Levine et al., 2014; Stepien et al., 2010). Together these studies indicate that multiple types of sensory information including limb position and pain may target a somatotopically-organized sensorimotor circuit arrayed across lamina V that modulates motor activity (Schouenborg et al., 1995; Tripodi et al., 2011). Consistent with this view of sensorimotor integration, the ablation of broad groups of dorsal-spinal inhibitory neurons disrupts motor actions in response to sensory stimuli (Foster et al., 2015). Nevertheless, because the motor and sensory components of action are deeply intertwined (Rizzolatti and Sinigaglia, 2007), it has been difficult to study integrative sensorimotor circuitry without definitive cellular markers of the constituent components.
A subset of premotor cells in the deep dorsal horn can be identified by the expression of the transcription factor Satb2 (Levine et al., 2014). Satb2 is a DNA binding protein that interacts with nuclear matrix attachment regions and controls gene expression and chromatin remodeling. Functional studies of Satb2 have identified roles for this gene in musculoskeletal development, and shown this factor is necessary for establishing the proper identity and axon projections of callosal neurons (Alcamo et al., 2008; Britanova et al., 2008; Dobreva et al., 2006; Leone et al., 2014; Zhao et al., 2014). Interestingly, case studies of human mutations in the Satb2 gene have reported developmental delays in motor skill acquisition, coordination, and sensorimotor behavior (Balasubramanian et al., 2011; FitzPatrick et al., 2003; Van Buggenhout et al., 2005).
Here we describe the connectivity pattern of Satb2+ interneurons, and establish that the Satb2 gene is required for the proper development of spinal circuitry that mediates the transformation of sensory cues into motor behavior. Satb2+ neurons are predominantly inhibitory, receive monosynaptic input from proprioceptive neurons, and are activated by noxious stimuli via polysynaptic pathways. Satb2+ neurons form synaptic connections onto ventral locomotor-interneurons and motor neurons. Based on their neurotransmitter profile and connectivity we refer to the discrete population of Satb2+ cells in the dorsal horn as inhibitory sensory relay (ISRSatb2) neurons. When the Satb2 gene was mutated the cell position, molecular profile, synaptic inputs and synaptic outputs of ISRSatb2 neurons were markedly altered. Satb2 mutant mice display abnormal limb hyperflexion of the ankle during walking and hold the hindlimb in a prolonged flexed position following noxious thermal or mechanical stimulation of the paw. Our findings demonstrate that Satb2 is necessary for the connectivity of ISRSatb2 neurons, thereby revealing an intrinsic genetic program that contributes to the formation of a discrete subunit of intraspinal sensorimotor circuitry.
Both proprioceptive and nociceptive sensory neurons relay through the deep dorsal horn (lamina IV-VI) of the spinal cord, an area that was recently shown to contain a significant proportion of neurons that are directly upstream of motor neurons (Coulon et al., 2011; Levine et al., 2014; Stepien et al., 2010). We therefore sought to identify genes that are expressed in a specific group of neurons within this nexus of sensory-motor information. Satb1/2 was recently identified as a marker of a subset of premotor neurons located in spinal lamina V (Levine et al., 2014). We performed immunohistochemistry with a Satb2-specific antibody and found that, in contrast to many of the dorsal horn molecular markers (Lbx1, Lmx1b, Pax2, Tlx3, GlyT2) that label broad heterogeneous groups of neurons, Satb2 labels a discrete population of neurons in lamina V of the deep dorsal horn (Figure 1A, B).
Satb2 expression is first detectable at embryonic day e12 and continues through the first postnatal week (Figure 1B, S1). Satb2+ interneurons span the mediolateral extent of lamina V (Figure 1B), corresponding to a region where cells that are activated during the nociceptive withdrawal reflex (NWR) have been detected with electrophysiology (Schouenborg and Weng, 1994). We also observed transient expression of Satb2 in motor neurons of the medial and lateral motor columns as early as e10.5, which perdured in a subset of medial motor neurons into the first postnatal week (Figure S1). Satb2 expression is also detectable in the developing neocortex and specific regions of the musculoskeletal system (data not shown). These findings are consistent with reports of Satb2 expression in the nervous system and other tissues (Britanova et al., 2005; Dobreva et al., 2006; Szemes et al., 2006).
To extend previous studies and to determine whether Satb2 marks interneurons of known lineage, we performed colocalization experiments with antibodies for dorsal interneuron subtypes (Alaynick et al., 2011; Gross et al., 2002). Satb2 is expressed in the Lbx1+ lineage of dorsal interneurons (Britanova et al., 2005), a marker that is expressed broadly in the dorsal spinal cord including interneuron subtypes dI4–6 and dILA and dILB (Gross et al., 2002; Muller et al., 2002; Schubert et al., 2001). Satb2+ interneurons also co-express the LIM homeodomain markers Lim1/2 (Britanova et al., 2005), and a subset of Satb2+ interneurons coexpresses Ptf1a, a marker of dI4 and dIL lineages (Figure S2). However, when we performed colocalization experiments with markers that subdivide these broad lineage classes (Pax2, Tlx3, Brn3a, Lmx1b, Bhlhb5, or Isl1/2), we found that the Satb2+ interneuron population did not express additional markers of dI4, dI5, dIL, or dI6 lineages (Figure S2). These findings indicate that the classification of Satb2+ interneurons does not align with previously described populations of dorsal interneurons.
To gain genetic access to Satb2+ interneurons and define their circuit properties, we devised a temporally inducible Cre labeling strategy. To generate a Satb2:CreERT2 knock-in mouse line, CreERT2 was inserted into the ATG of the Satb2 locus (Figure 1C, and data not shown). Satb2:CreERT2 animals were crossed with the Cre-dependent reporter Rosa-CAG-LSL-TdTomato to generate Satb2:TdTomato animals. Although Satb2 is downregulated during the first postnatal week, tamoxifen administration at embryonic or early postnatal timepoints indelibly labels Satb2+ interneurons. Embryonic induction of Satb2:CreERT2 with tamoxifen at e12 labels spinal interneurons as well as a restricted population of motor neurons in the medial motor column (Figure 1D). Thus, TdTomato expression driven by our transgenic line accurately recapitulates Satb2 protein expression (compare Figure 1B to 1D, Figure S1).
Genetic labeling of Satb2+ interneurons facilitated further descriptive analyses of this population. To determine the neurotransmitter status of Satb2+ interneurons we performed double labeling experiments in which we examined the overlap of TdTomato expression with in situ hybridization probes for excitatory or inhibitory neurotransmitter markers. Only 3.8% of Satb2+ interneurons express the excitatory marker vGlut2, whereas the vast majority of cells express the inhibitory markers Gad65, Gad67, and/or GlyT2 (Figure 1E, F, S2). These experiments reveal that Satb2+ neurons represent a spatially restricted population of inhibitory dorsal interneurons arrayed along the mediolateral axis of the deep dorsal horn.
Tracing experiments have noted that peripheral sensory commands converge on spinal neurons in the deep dorsal horn of the spinal cord, representing the first point of intersection between the nociceptive and proprioceptive systems (Levine et al., 2014). Primary fibers of proprioceptive sensory neurons and second order neurons that are recruited during the hindlimb nociceptive NWR reside in overlapping domains within the deep dorsal horn (Craig, 2003). Although the approximate site of convergence for these two pathways overlaps with the position of Satb2+ interneurons (Figure 1A), their connectivity has not been described.
To test whether Satb2+ interneurons receive proprioceptive input, we used a genetic labeling strategy (Parvalbumin-Cre x Rosa-CAG-LSL-Syp-TdTomato-deltaNeo; abbreviated PV:Syn-Tom) in which the presynaptic terminals of proprioceptive fibers are labeled with a synaptophysin-TdTomato fusion protein (Levine et al., 2014). We identified TdTomato+ contacts on the cell bodies of Satb2+ interneurons (Figure 2A, B), indicating that Satb2+ interneurons receive direct input from proprioceptive fibers. We next used a second strategy to visualize the full cell morphology of Satb2+ interneurons. Immunohistochemistry for parvalbumin and vGlut1 revealed proprioceptive contacts along the soma and processes of Satb2:TdTomato+ spinal interneurons. We found that 97 +/− 2.6% of Satb2:TdTomato+ spinal interneurons received direct contacts from proprioceptive fibers (Figure S3, n=29 cells in 3 spinal cords), however we noted quantitative differences within the Satb2+ interneuron population, which are characterized in more detail below.
Nociceptive sensory fibers predominantly target superficial spinal laminae, but relay information through the deep dorsal horn via spinal interneurons. To explore whether Satb2+ interneurons are a site of convergence for multimodal sensory inputs, we performed double labeling experiments of proprioceptive fibers and nociceptive relay neurons in Satb2:TdTomato animals. Intramuscular injection of cholera toxin (CTB) into the tibialis anterior (TA) muscle was performed to label proprioceptive afferents of a muscle that is recruited following the NWR (Schouenborg, 2004). Capsaicin was then injected into the footpads of these animals, and c-Fos immunoreactivity was used to identify cells that are recruited in response to a painful stimulus (Figure 2C). As expected, c-Fos+ neurons were located in superficial laminae ipsilateral to the injection site, as well as in the deep dorsal horn. At spinal levels of peak c-Fos labeling, 32.3% of Satb2:TdTomato+ neurons coexpress c-Fos in response to capsaicin injection into the footpad (Figure 2D; 42 of 130 TdTomato+ cells, n=5 spinal cords), indicating that a significant fraction of Satb2+ interneurons are activated even in response to focal stimulation of nociceptive sensory neurons that innervate the foot. In addition, we identified Satb2:TdTomato+ interneurons that express c-Fos following noxious stimulation and receive vGlut1+ contacts from CTB-labeled proprioceptive afferents (Figure 2E). Thus, individual Satb2+ interneurons represent a convergence point for multimodal sensory input. Based on Satb2 expression, neurotransmitter identity, and sensory input, we refer to them as inhibitory sensory relay (ISRSatb2) neurons.
Sensory-motor circuit components are known to display a topographic organization along the mediolateral axis of the spinal cord (Arber, 2012; Levine et al., 2012). The convergence of synaptic inputs onto ISRSatb2 neurons and their distribution along the mediolateral axis prompted us to ask whether ISRSatb2 neurons are composed of a homogeneous population of neurons or whether there is diversity in their identity and connectivity. To test whether interneuron cell body location is a determinant of the amount of proprioceptive input, we quantified the number of PV:Synaptophysin-TdTomato+ contacts on neurotrace+ cell bodies of ISRSatb2 neurons, and mapped the number of contacts per cell along the mediolateral axis. These experiments revealed an inverse relationship between cell body distance from the midline and the amount of proprioceptive input: cells positioned closer to the midline had a greater number of proprioceptive contacts, while lateral cells had fewer contacts (Figure 2F). These findings indicate that mediolateral position is an organizational feature that correlates with proprioceptive sensory connectivity onto ISRSatb2 neurons.
Although interneurons are topographically organized along the mediolateral axis of the deep dorsal horn, genetic markers that subdivide these groups have not been described. In the cortex, Satb2 participates in an antagonistic genetic interaction with Ctip2 to subdivide cortical progenitors into functionally distinct pools (Alcamo et al., 2008; Britanova et al., 2008). To test whether Satb2 and Ctip2 have mutually exclusive expression in the spinal cord, we performed immunohistochemistry for both markers. Surprisingly, in the spinal cord Satb2 and Ctip2 are not mutually exclusive and define two subpopulations of ISRSatb2 cells; medial neurons are Satb2+/Ctip2+, while lateral neurons are Satb2+/Ctip2- (Figure 2G, H). This molecular code is detectable as early as e13.5 and is maintained into postnatal stages (data not shown). These experiments reveal that sensory connectivity and molecular profile subdivide ISRSatb2 neurons along the mediolateral axis of the spinal cord (Figure 2I).
There is a significant representation of premotor neurons in the deep dorsal horn (Coulon et al., 2011; Hongo et al., 1989; Levine et al., 2014; Puskar and Antal, 1997; Stepien et al., 2010), and neurons that target supraspinal motor control centers are also known to reside in dorsal laminae (Matsushita and Hosoya, 1979; Surmeier et al., 1988). To test whether ISRSatb2 neurons may serve as a cellular substrate for transmitting sensory cues to either ascending or ventral motor control centers, we used a combination of circuit tracing approaches.
First, we used an unbiased genetic labeling approach to identify synaptic targets of ISRSatb2 neurons. Satb2:CreERT2 x Rosa-CAG-LSL-Syp-TdTomato-deltaNeo (abbreviated Satb2:Syn-Tom) provided, cell-type specific expression of the Synaptophysin-TdTomato presynaptic fusion protein. Visualization of the overall distribution of synaptic terminals of ISRSatb2 neurons revealed a specific pattern of labeling across the ventral spinal cord. Syn-Tom+ terminals were detected throughout the ventral spinal cord, but enriched in a diagonal region covering lamina V-VII and IX (Figure 3A, B).
We compared the location of Satb2:Syn-Tom labeling with the presynaptic targets of other dorsal interneuron classes. In contrast to the specific pattern of labeling in ventral lamina that we observed in Satb2:Syn-Tom spinal cords, both Ptf1a: and Lmx1b:Syn-Tom expressing interneurons broadly target dorsal spinal laminae in which their cell bodies reside (Figure S3). The location and distribution of synaptic terminals from ISRSatb2 neurons indicates their connectivity patterns are distinct from those of other dorsal interneuron classes implicated in sensory transduction.
To test whether ISRSatb2 neurons target ventral cell types indiscriminately or with absolute specificity for synaptic partners, we performed labeling experiments for neuronal populations in the ventral spinal cord of Satb2:Syn-Tom animals. Viral labeling of motor neurons for tibialis anterior (TA, flexor), gastrocnemius (GS, extensor), and hamstring (H, flexor) muscles provided a clear visualization of specific motor neuron soma and processes (see Methods) and revealed contacts from ISRSatb2 neurons onto hindlimb motor pools (Figure 3D, and data not shown). We next used an independent strategy to confirm synaptic contacts onto motor neurons in which monosynaptic connections between spinal interneurons and motor neurons were identified by Rabies transsynaptic labeling (Stepien et al., 2010; Wickersham et al., 2007). These experiments confirmed that ISRSatb2 neurons target motor neurons (Figure 3C). Next we characterized the ventral interneuron populations that receive inputs from ISRSatb2 neurons. We found that many ventral (lamina VII) interneuron classes implicated in motor control receive input from ISRSatb2 neurons (Figure 3E–H; 82.1 +/− 6.2% of Chx10+ V2a, 66.8 +/− 1.8% Foxp2+ V1, and 69 +/− 9.4% of ChAT+ motor neurons).
To test whether ISRSatb2 neurons display quantitative differences in their connectivity, we determined the average number of ISRSatb2 synaptic terminals on each cell type. Chx10+ V2a interneurons receive numerous contacts on their soma (Figure 3E; 4.35 +/− 0.84 TdTomato+ contacts/Chx10+ neuron). Synaptophysin:TdTomato+ contacts were also identified on ChAT+ motor neurons (Figure 3F, 4.28+/−0.97 TdTomato+ contacts/ChAT+ neuron) as well as Foxp2+ V1 interneurons (Figure 3G, 1.83+/−0.14 TdTomato+ contacts/Foxp2+ neuron). Although contacts were less abundant, we also identified TdTomato+ contacts onto Evx1+ V0 interneurons, Calbindin+ ventral interneurons of mixed identity, and Bhlhb5+ ventral interneurons of mixed identity (Figure S3). We next compared the average number of Syn-Tom+ contacts onto various ventral interneuron populations. Chx10+ V2a interneurons receive significantly more contacts than Foxp2+ V1 interneurons (Figure 3I, p<0.05, Mann Whitney), suggesting that synaptic input from Satb2+ interneurons is weighted among various ventral interneuron populations. Syn-Tom+ contacts were not identified onto ISRSatb2 neurons (Figure 3H).
Next we examined whether Satb2+ interneurons have ascending projections that target supraspinal motor control centers. scAAV containing a Cre-dependent fluorescent reporter protein (scAAV-Flex-GFP) was injected into the lumbar spinal cord of neonatal Satb2-CreERT2 animals. GFP labeling was evident in a pattern consistent with that of Satb2:TdTomato labeling. GFP+ fibers were not detected at the pyramidal decussation nor found within the ventral brainstem where ascending fiber tracts normally reside (Figure S3). These findings suggest that the predominant targets of ISRSatb2 neurons are motor neurons and ventral interneurons that regulate motor activity within the spinal cord.
Given the established role of the Satb2 gene in developmental events that are critical for circuit formation, and the broad connectivity of ISRSatb2 neurons with ventral motor command neurons, we considered how loss of Satb2 gene function would impact motor activity. We first tested whether basic motor circuitry is intact in Satb2 null animals (Satb2lacZ/lacZ; Dobreva et al., 2006) using a fictive locomotor preparation that activates the central pattern generator (CPG) in the absence of sensory and descending brain input (Kiehn and Kjaerulff, 1996; Smith and Feldman, 1987). Patterned motor activity was normally coordinated in Satb2 null mutants between left-right and flexor-extensor pools (Figure 4A–C). Satb2 null animals also displayed normal cycle period length, cycle period variation, and burst duration (Figure S5 and data not shown), parameters that are altered when ventral interneuron function is perturbed (Alaynick et al., 2011; Goulding, 2009; Grillner and Jessell, 2009; Kiehn, 2016). These experiments indicate that the core components of ventral motor circuitry develop and function properly in the absence of Satb2.
To test whether targeted deletion of Satb2 from ISRSatb2 neurons impacts motor and sensory function in the behaving animal, we used a conditional Satb2flx allele. Lbx1:Cre was used as the deleter based on its coexpression with Satb2 in the spinal cord, and its exclusion from other Satb2+ cell types (Figure S2, S4, and data not shown). Use of Lbx1:Cre/Satb2flx/flx (Satb2-ISRKO) mutant animals allowed us to circumvent perinatal lethality in Satb2 null animals, allowing for behavioral analysis of sensory and motor function. We confirmed that Satb2-ISRKO animals lack Satb2 protein expression within spinal interneurons (Figure S4). Satb2-ISRKO mutant animals are born in normal numbers, have a normal lifespan, body weight and appearance at 8 and 20 weeks of age (data not shown). Behavioral tests that broadly assay motor function, including open field to test baseline motor activity and grip strength to test muscle strength, were normal in Satb2-ISRKO mutants (Figure S5). Satb2-ISRKO mutant animals also performed normally on the rotarod test measuring general motor coordination. Thus, Satb2 gene function appears to be dispensable in ISRSatb2 neurons for basic motor tasks.
Recent studies have shown that proprioceptive defects alter limb position and the timing of muscle recruitment during walking (Akay et al 2014). The dense proprioceptive input onto ISRSatb2 neurons prompted us to examine limb kinematics of Satb2-ISRKO animals. Satb2-ISRKO mutants had a visible disruption in limb position during the swing phase of the step cycle. In particular, the ankle joint was held in a hyperflexed position beginning at swing onset, resulting in a reduced ankle angle throughout early swing phase (Figure 4D–F, Figure S5).
Next, to test whether Satb2 is required for sensory-evoked behaviors, we examined motor responses of Satb2-ISRKO animals to noxious and non-noxious stimuli. To perform the Hargreaves thermal pain test, a high intensity light source was shined on the footpad of test animals, eliciting an isolated, reproducible nociceptive withdrawal response of the foot (Allen and Yaksh, 2004; Hargreaves et al., 1988). We performed this test to measure two key components of the pain response: latency to the withdrawal response, and the dynamics of the withdrawal behavior, including the initial flexion response and replacement of the limb following withdrawal. We observed a small but significant increase in hindlimb response latency in Satb2-ISRKO mutant animals when compared with control littermates (Figure 5A). Interestingly, Satb2-ISRKO animals maintained a flexed posture following thermal pain stimulation (Figure 5B, C; Videos S1 and S2), revealing an aberrant flexion behavior following deletion of Satb2 from ISRSatb2 neurons.
To test whether Satb2 expression in ISRSatb2 neurons is required to transduce other pain modalities, we performed the Von Frey test in Satb2-ISRKO animals to measure the response to mechanical stimulation. The threshold for producing the nociceptive withdrawal behavior following Von Frey fiber stimulation was normal in Satb2-ISRKO animals when compared with control littermates (Figure 5D). However, we observed a maintained flexion posture in Satb2-ISR animals following Von Frey stimulation (Figure 5E). In contrast, Satb2-ISRKO mutant animals have a normal withdrawal response to light touch (Figure 5F), and there are no differences in the withdrawn posture of the hindlimb following light brush stimulation (Figure 5G). Deletion of the Satb2 gene unmasks a genetic role for this transcription factor in ISRSatb2 neurons that leads to very selective defects in motor activity associated with hyperflexion while walking and retracting the limb from painful thermal and mechanical stimuli.
To understand how loss of the Satb2 gene contributes to the behavioral phenotype observed in Satb2-ISRKO animals, we monitored ISRSatb2 cell development using a lineage tracing strategy. Insertion of CreERT2 into the ATG of Satb2 generates a null allele (Figure S4), providing the opportunity to visually track the Satb2 null-interneurons in Satb2Cre/Cre:TdTomato animals (Satb2KO:Tom+). Satb2Cre/Cre (Satb2KO) animals phenocopy the null mouse line, in which there are defects in mandibular development and perinatal lethality (Dobreva et al., 2006; data not shown). We found that Satb2KO:Tom+ interneurons are generated in normal numbers (Figure 6A, B, and data not shown), however there were marked changes in cellular position along the mediolateral axis. ISRSatb2 (Satb2Cre/WT:TdTomato; abbreviated Satb2:Tom+) interneurons are normally organized in a tight band in the deep dorsal horn, spanning the mediolateral axis. In contrast, Satb2KO:Tom+ interneurons were shifted to a lateral position, and were more loosely organized with cells scattered dorsally (Figure 6A, B). The mean distance of Satb2:Tom+ neurons from the midline at lumbar levels was 201.6 +/− 4.7um, while Satb2:Tom+ neurons were located 301.5+/−4.4um from the midline (Figure 6C; p-value: <0.0001, Mann-Whitney), resulting in a final settling position at the lateral margin of the spinal cord. In addition to the main cluster of interneurons in the deep dorsal horn, we also observed a scattered population along the ventral midline in Satb2KO:TdTomato spinal cords. This population may represent a subset of neurons that were unable to exit the ventricular zone or that have an altered migration pattern (Figure 6B, arrowheads). These results indicate that Satb2 is necessary for establishing the medio-lateral position of ISRSatb2 neurons.
To compare the molecular profile of Satb2:Tom+ and Satb2KO:Tom+ spinal neurons, we performed double labeling experiments with immunohistochemistry or in situ hybridization. A panel of transcription factor markers for dorsal and ventral spinal interneuron populations were selected based on their coexpression within ISRSatb2 cells and the surrounding neuron populations (Figure S1; Gross et al 2002). Immunohistochemistry for these markers revealed a number of gene expression changes (Figure 6D–J, Figure S6, Table S1, and data not shown). Ptf1a and Lbx1, expressed by many dorsal interneuron populations, including subsets of ISRSatb2 neurons, were unchanged in response to loss of Satb2. In contrast, there were significant changes in the expression of Pax2 (Figure 6D, E) and Bhlhb5 (Figure S6), genes that regulate neuronal circuit assembly and inhibitory neurotransmitter identity, respectively (Brohl et al., 2008; Huang et al., 2008). In addition, Ctip2 expression is almost completely lost in the absence of Satb2 (Figure 6F, G).
The changes we observed in the molecular profile of Satb2KO:Tom+ neurons prompted us to test whether there were corresponding alterations in neurotransmitter identity, as this would likely alter the function of these cells within a circuit. To test this we performed in situ hybridization for neurotransmitter markers in Satb2:Tom+ and Satb2KO:Tom+ animals. Importantly, the ratio of inhibitory/excitatory status of Tom+ neurons is unchanged in the absence of Satb2 gene function (Figure 6H–J and Figure S6, Table S1). Consistent with the role of Satb2 in cortical neuron migration (Alcamo et al., 2008; Britanova et al., 2008), these results demonstrate that Satb2 regulates gene expression and cell position in ISRSatb2 spinal neurons.
Cell body positioning within the spinal cord is an important regulator of synaptic connectivity (Arber, 2012). To determine whether the mislocalized Satb2KO:Tom+ cells had alterations in proprioceptive input we performed immunohistochemistry for parvalbumin and vGlut1. In Satb2:Tom+ animals, a majority of ISRSatb2 neurons are located within the pathway intersected by proprioceptive fibers. In contrast, the percentage of Satb2KO:Tom+ cells receiving proprioceptive input at cervical and lumbar levels was greatly reduced (Figure 7A–E). Interestingly, the stereotypical trajectory and density of proprioceptive fibers in the spinal cord was unchanged despite the mislocalization of one of their target cell types in Satb2 mutants (Figure 7A, C, and data not shown).
To determine whether laterally shifted Satb2KOTom+ neurons might receive new inputs, we examined primary afferents of peptidergic nociceptors expressing CGRP that are known to target spinal neurons in lateral lamina V (Guo et al., 2011). Under normal conditions Satb2:Tom+ neurons are located medial to the CGRP+ fiber density (Figure 7F, G). The projection of CGRP+ fibers was unchanged in Satb2 mutants, however the lateral Satb2KO:Tom+ neurons ectopically overlap with the termination domain of CGRP+ fibers (Figure 7H–J).
To test whether Satb2 is necessary to regulate the connectivity of ISRSatb2 cells with postsynaptic motor-circuit targets, we compared the distribution of terminals in Satb2:Synaptophysin-TdTomato (Satb2:Syn-Tom+) and Satb2KO:Synaptophysin-TdTomato (Satb2KO:Syn-Tom+) animals. In Satb2:Syn-Tom+ animals, synaptic termini were distributed throughout the ventral spinal cord including contacts with V2a, V1 and motor neurons (Figure 8A, D; see also Figure 3C–G). In contrast, the overall distribution of terminals were less abundant and shifted to dorsal targets above the V2a cell domain in Satb2KO:Syn-Tom+ animals (Figure 8A–E). These results demonstrate that Satb2 gene function is required to ensure the proper connectivity between ISRSatb2 neurons and their postsynaptic partners within the ventral spinal cord that regulate motor output (Figure 8F).
Commands from different types of sensory neurons that monitor limb position and pain converge onto spinal neurons within the deep dorsal horn, which are thought to serve as a local hub for integrating multiple sensory modalities (Craig, 2003). The output of this local circuit is capable of directly modulating the motor system to produce rapid state-dependent behaviors such as reflexive withdrawal movements. The cellular composition and developmental mechanisms that give rise to this sensorimotor system within the spinal cord are not well understood, but have focused primarily on plasticity based learning mechanisms (Schouenborg, 2004). We found that a discrete subpopulation of lamina V inhibitory sensory relay neurons (ISRSatb2) require the transcription factor Satb2 to establish their proper input- and output-pattern of connectivity. Deletion of the Satb2 gene from ISRSatb2 neurons does not disrupt most types of motor activity. Rather, we observed specific behavioral abnormalities in which hyperflexion defects were apparent during specific phases of walking and in response to sensory stimuli. Our studies demonstrate that Satb2 controls the development of a discrete component of the sensorimotor system.
Within the deep dorsal horn sensorimotor neurons have diverse properties including wide dynamic range responses and distinct patterns of inputs from cutaneous and proprioceptive afferents (Harrison and Jankowska, 1985; Hongo et al., 1989). Furthermore, both inhibitory and excitatory sensorimotor interneurons are intermingled and some project locally while others project to supraspinal targets. Using Satb2 as a marker we found that ISRSatb2 neurons express inhibitory neurotransmitters and are arrayed along the mediolateral axis of lamina V. They receive input from proprioceptive and nociceptive pathways and target multiple cellular components of the ventral motor network, but appear to lack supraspinal projections. Thus, ISRSatb2 neurons represent a specific subset of cells within the sensory-to-motor pathway (Figure 8G).
Mutation of Satb2 does not perturb the function of the core circuitry associated with central pattern generation. Rather, Satb2 mutation disrupts both the location and connectivity of ISRSatb2 neurons: they reposition to the lateral region of lamina V, they have a greatly diminished number of synaptic inputs from proprioceptive fibers, and they lack outputs to the ventral spinal cord where motor actions are controlled (Figure 8G). Interestingly, the ablation of all proprioceptive sensory neurons leads to severe locomotor deficits including rotarod defects (Ilieva et al., 2008; Oliveira Fernandes and Tourtellotte, 2015), whereas Satb2 mutant mice had normal rotarod function. Taken together, these findings suggest that the remaining proprioceptive connections in Satb2 mutants are sufficient for many motor tasks.
We found that Satb2 expression in spinal interneurons is required for specific motor and sensory-evoked behaviors including proper limb position during runway walking, and limb flexion-withdrawal following thermal and mechanical stimulation. Studies in the cat suggest that muscle afferents help to set the timing of limb flexion and the transitions from swing to stance (Hiebert et al., 1996; Stecina et al., 2005). Perhaps the hyperflexion phenotypes observed in Satb2 mutants are caused by a defect in how proprioceptors control the timing of limb flexion in specific behavioral contexts such as when sensory cues activate a limb flexion response.
Satb2 mutants display an increased latency to withdrawal from thermal stimulation, revealing a breakdown in the transformation of sensory input into motor output (Figure 8G). The aberrant hindlimb hyperflexion that was observed following thermal and mechanical stimulation was not produced by light touch stimulation, indicating that Satb2 is required for the execution of a specific set of sensory-motor behaviors. Although it is unclear whether the behavioral changes observed in Satb2 mutants represent a loss- or aberrant gain-of-function of ISRSatb2 neurons, our findings demonstrate that Satb2 is necessary to ensure the proper development of a subcomponent of the sensorimotor circuitry whose function is only apparent during specific behavioral contexts. Importantly, the defects that arise from the targeted deletion of Satb2 from ISRSatb2 neurons functionally-reveals a cellular pathway that transforms multi-modal sensory stimuli into motor action.
We report an altered molecular profile within ISRSatb2 neurons in response to loss of the Satb2 gene. Transcription factors such as Lbx1 and Ptf1a that specify progenitor domains within dorsal interneurons were unchanged in Satb2 null interneurons. We instead observed changes in the expression of genes that regulate subsequent phases of neural development. For instance, Bhlhb5, a gene that is involved in neuronal circuit assembly (Ross et al., 2012), is upregulated in a subset of Satb2 null interneurons that are positioned near the central canal. Although Satb2 null interneurons retain their inhibitory neurotransmitter identity, Pax2 expression is significantly upregulated. Pax2 is a downstream target of Lbx1 and Ptf1a, and is known to regulate glycinergic cell fate as well as the expression of neuropeptides used for inhibitory neurotransmission (Brohl et al., 2008; Cheng et al., 2004; Gross et al., 2002; Huang et al., 2008). Although not extensively investigated in this study, the genetic changes we observed raise the possibility that Satb2 mutant ISRSatb2 neurons may have an abnormal composition of neurotransmitters and/or neuropeptides.
Satb2 is considered a direct repressor of Ctip2 in cortical neurons, however recent studies have identified a population of Satb2/Ctip2 coexpressing cortical neurons (Harb et al., 2016). Likewise, we found that medial ISRSatb2 neurons coexpress Ctip2, whereas lateral ISRSatb2 cells lack Ctip2. Unlike the cortex, when Satb2 is mutated Ctip2 expression is not upregulated in ISRSatb2 neurons, instead Ctip2 expression is lost. Our findings reveal two critical aspects of gene regulatory networks in the spinal cord. First, ISRSatb2 neurons lack the antagonistic gene interaction that segregates most Satb2+ and Ctip2+ cortical neurons (Alcamo et al., 2008; Britanova et al., 2008). Second, the Ctip2 status of ISRSatb2 neurons reveals a genetic heterogeneity among ISRSatb2 neurons along the mediolateral axis of lamina V. The loss of Satb2 appears to disrupt this cellular heterogeneity. Future studies will be important for understanding whether Ctip2 is a stringent indicator of synaptic input and output patterns within the ISRSatb2 neuron population.
Cell position is an important organizational feature within the diverse neuronal types that occupy the deep dorsal horn. For example, a subset of spinal interneurons in lamina V, termed “reflex encoders”, are topographically organized such that cellular position along the mediolateral axis correlates with patterns of synaptic input and output (Schouenborg et al., 1995). In addition, lamina V premotor neurons display a medial-extensor and lateral-flexor bias (Tripodi et al., 2011). We found that the medial ISRSatb2 neurons receive more proprioceptive inputs than the lateral cells. Thus, the subdivision of ISRSatb2 neurons based on their mediolateral Ctip2 status represents a molecular correlate that aligns with the orderly mapping relationships between pre- and/or post-synaptic targets. While past studies have focused on activity-based mechanisms during a critical window around birth (Schouenborg, 2004; Tripodi et al., 2011), our findings demonstrate that a complementary mechanism is also involved whereby Satb2 is an essential genetic regulator of the cellular connectome for spinal sensorimotor circuitry.
Full methods can be found in the Supplemental Experimental Procedures
We gratefully acknowledge the generosity and advice of M. Goulding, L. Bachmann, C. Farrokhi, S. Bourane, and L. Garcia-Campmany; Q. Ma, R. Johnson, and C. Birchmeier for providing reagents, Y. Dayn for help with generation of the Satb2-CreERT2 transgenic mouse line. K. Lettieri, and C. Heller provided technical support and advice. K.L.H was supported as a National Science Foundation Graduate Research Fellow and the Chapman Foundation. A.J.L was supported by George E. Hewitt Foundation for Medical Research and Christopher and Dana Reeve Foundation. C.A.H. was supported by a US National research Service Award Fellowship from US National Institutes of Health NINDS. M.H was supported by the Timken-Sturgis Foundation and the Japanese Ministry of Education, Culture, Sports, Science, and Technology Long-Term Student Support Program. S.L.P is supported as a Howard Hughes Medical Institute Investigator and as a Benjamin H. Lewis chair in neuroscience. This research was supported by funding from the Howard Hughes Medical Institute, the Marshall Foundation and the Sol Goldman Charitable Trust.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Author ContributionsK.L.H and S.L.P designed the study and wrote the manuscript. K.L.H and A.J.L designed and carried out the experiments with help from C.A.H and M.H. J.M.M and M.G assisted with behavioral testing design and execution. S.P.D conducted statistical analysis. R.G. provided the Satb2lacz/lacZ and Satb2flx/flx mouse lines. Y.K and T.K.S provided the Satb2 antibody.