Sexual dimorphism of spinal neural circuits controlling the mouse external urethral sphincter with and without spinal cord injury

Abstract

Spinal cord injury (SCI) disrupts coordination between the bladder and the external urinary sphincter (EUS) leading to transient or permanent voiding impairment, which is more severe in males. Male vs. female differences in spinal circuits related to the EUS as well as post-SCI rewiring are essential for understanding of sex/gender specific impairments and possible recovery mechanisms. To quantitatively assess differences between EUS circuits in males vs. females and in spinal intact (SI) vs. SCI animals, we retrogradely traced and counted EUS-related neurons. In transgenic ChAT-GFP mice motoneurons (MNs), interneurons (INs) and propriospinal neurons (PPNs) were retrogradely trans-synaptically traced with PRV614-RFP injected to EUS. EUS-MNs in dorsolateral nucleus (DLN) were separated from other GFP⁺ motoneurons by tracing them with Fluorogold (FG). We found two morphologically distinct cell types in DLN: FG⁺ spindle-shaped bipolar (SB-MNs) and FG⁻ rounded multipolar (RM-MNs) cholinergic cells. Number of MNs of both types in males was twice as large as in females. SCI caused partial loss of MNs in all spinal nuclei. After SCI, males showed a 4-fold rise in number of RFP-labeled cells in retro-dorsolateral nucleus (RDLN) innervating hind limbs. This suggests (a) an existence of direct synaptic interactions between spinal nuclei and (b) post-SCI increase of non-specific inputs to EUS-MNs from other motor nuclei. Number of INs and PPNs deferred between males and females: in SI males the numbers of INs and PPNs were ~10 times larger than in SI females. SCI caused two-fold decrease of INs and PPNs in males, but not in females.

Graphical Abstract

Spinal neuronal pool innervating external urethral sphincter (EUS) in female mice has less motoneurons and interneurons compared to males. Spinal cord injury caused reduction of EUS-related neuronal pool in males but not in females.

INTRODUCTION

Spinal cord injury (SCI) is a debilitating condition that leads to loss of motor functions, including impairment of voluntary voiding due to interrupted communication between spinal neural circuits and the supraspinal levels of the central nervous system [1, 2]. Autonomic neural circuits of the lower urinary tract (LUT) are located in lumbar-sacral segments of the spinal cord, so that SCI interrupts the flow of descending commands from the brainstem to these circuits. In spinal intact (SI) animals tonic and/or phasic activity of smooth muscles of the bladder (BL) and the external urethral sphincter (EUS) is highly coordinated to guarantee continence and timely release of urine [3–5]. Voiding is initiated by an increased bladder volume and firing of bladder afferents, which triggers the micturition reflex when BL contracts at a threshold magnitude of intravesical pressure. When the previously quiescent smooth muscle of the BL contracts, the striated muscle of the EUS relaxes allowing removal of urine through the opened urethra. During urine storage the urethra is closed by tonic contraction of the striated EUS muscle while the BL is quiescent. During voiding the EUS relaxes and the smooth muscle of BL contracts to expel urine through opened urethra. This reciprocal relationship between BL and EUS is controlled by visceral reflex circuits, which are under voluntary control. Experimental studies have shown that the micturition reflex is mediated by a spinobulbospinal pathway passing through a subcortical coordination center (the pontine micturition center – PMC) [2, 6, 7]. The PMC is in turn modulated by higher centers in the cerebral cortex that are involved in the voluntary control of micturition [3]. SCI at cervical or thoracic levels breaks communication between spinal and supraspinal compartments, thereby disrupting pathways providing voluntary and reflex voiding as well as the normal reflex that coordinates bladder and sphincter functions. The lack of supraspinal BL/EUS coordination immediately after SCI and during some following period results in persistent spasmatic contraction of the EUS regardless of the degree of BL filling. This condition is called detrusor-sphincter dyssynergia (DSD), which can lead to BL rapture if not treated and to development of LUT pathologies if becomes a chronic condition [4, 8, 9]. However, experiments in rats with a complete spinal cord transection at the thoracic level have shown that despite a total disconnection between spinal and supraspinal circuits the reflex voiding can recover in a few weeks. It was assumed that recovery is due to re-wiring of the pre-existing LUT-related lumbar networks or local axonal sprouting and recruitment of other spinal neurons to restore BL/EUS coordination [10]. However, LUT motoneurons do not receive direct commands from the PMC or sensory input from the bladder. Instead, these commands and inputs are modulated though a cascade of local and propriospinal interneurons [11–15]. Thus, well-tuned ensembles of spinal interneurons play a critical role in coordination of the EUS and BL muscles allowing efficient elimination of accumulated urine, although it is often incomplete even in SI animals [68]. To decode changes in neural networks after SCI and to reveal what is necessary for at least partial restoration of the LUT function, one needs to know what happens with certain types of neurons. For example, how their connections have been modified, whether their relative numbers are the same as in SI, whether the size of the circuit increased due to sprouting or decreased due to apoptosis, etc.

Another important factor is the sex, because certain body features, including the LUT organs’ structure, often display evolutionary justifiable differences subserving sex-specific functions, for example, place marking by the urine scent. Therefore, not only external and internal LUT-related organs differ in males vs. females, but also the neural control making these organs functional should be different.

Spinal motoneurons innervating the EUS are located in Onuf’s nucleus in the ventral horn as it was described by Bronislaw Onuf-Onufrowicz in 1899 [16]. However, details of the autonomic spinal circuits vary among species. In the golden hamster, cat, chimpanzee and human motoneurons of the EUS and the external anal sphincter (EAS) are located close to each other within Onuf’s nucleus [16–19]. In rats and mice Onuf’s nucleus is split into two distant clusters: dorsolateral nucleus (DLN) and dorsomedial nucleus (DMN). The former is located in the ventrolateral corner of the ventral horn and contains EUS motoneurons (EUS-MNS), whereas the latter is close to the crest of the ventral column of white matter and contains EAS motoneurons (EAS-MNs) [20, 21]. Most studies of EUS/EAS-MNs in mammals have been done in cats and rats [22–26], but it is not possible in these animals to use techniques employing genetically modified specific cell types in order to uncover details of circuit organization and dynamic post-SCI changes at the molecular level. Therefore, in recent years, SCI studies have been conducted in mice to utilize these methods [27–33]. In this paper, we used transgenic ChAT-GFP mice expressing green fluorescent protein in cholinergic neurons to visualize all available motoneurons and reveal their fate after SCI. In addition, we retrogradely traced EUS-MNs with Fluorogold to separate them from non-EUS neurons of the DLN. The viral retrograde trans-synaptic tracer PRV614-RFP injected into the EUS of mice first labeled EUS-MNs, and then propagated further labeling neurons presynaptic to the MNs. This allowed us to reveal synaptic coupling between motoneurons of different nuclei. Besides, PRV tracing of other cells presynaptic to EUS-MNs allowed us to quantitatively estimate and compare size and distribution of EUS-related interneuronal populations in males and females. Furthermore, we revealed post-SCI changes in these populations. Schematically spinal neural circuits for LUT control comprise two main compartments in L6-S1 and L3-L4, and one auxiliary compartment in L1-L2 (Fig.3A). Motoneurons are the last in the chain of command for signals arriving from the supraspinal compartments involved in cortical and subcortical control. Their activity is synaptically modulated through intraspinal interneuronal circuits [12]. The majority of local L6-S1 EUS-related interneurons (EUS-INs) are located in the dorsal commissure (DCM) near the central canal [11, 34, 35]. The EUS-MNs in the DLN are the first to be labeled when retrogradely traced with PRV-RFP from the EUS. A subpopulation of cells presynaptic to them creates a pool of the first order or primary EUS-interneurons (pINs), whereas a subpopulation of cells presynaptic to the pINs includes the second order EUS-interneurons, consisting of a subset of secondary interneurons (sINs). In choosing the optimal post-inoculation survival time we focused on obtaining the maximal number of RFP-labeled EUS-INs and PPNs in the first wave of RFP expression. At this time both pINs and sINs were fluorescently labeled showing a gradient of staining from bright (pINs) to moderate or weak (sINs). Unfortunately, by this time most of EUS-MNs were destroyed by PRV, and we had to use other techniques to assess their numbers.

Fig.3.

A – Schematics illustrating the current view on spinal circuits controlling the lower urinary tract. Neurons related to EUS are in green and neurons related to smooth muscles of urethra are in blue. Neurons relevant to the bladder are in red and black. EUS-MNs of the DLN (= Onuf’s nucleus) (large green triangles) receive excitatory inputs from local DCM-INs (small green triangles) in L6-S1 segments. They were also suggested to have an auxiliary excitatory input from propriospinal neurons of the LSCC located in L3-L4 segments (dotted green line). A source of inhibitory input to EUS-MNs is still unknown. B – Cross-sections of urethra in male (upper photo) and female (bottom photo) adult mice. Note a considerable sex difference in the urethral diameters and overall thickness of the muscle wall. C – The fixed spinal cord was transected at the L4-L5 border (middle dotted line), then the rostral part was curtailed above L1 (upper dotted line) and the caudal part was curtailed at or below S2 (lower dotted line). Blue arrows indicate the direction of cutting serial sections beginning from the L4-L5 border. Abbreviations: bPPGN – bladder parasympathetic preganglionic neurons; DCM – dorsal commissure; EUS – external urethral sphincter; EUS-MN – motoneuron projecting to the EUS; LSCC – lumbar spinal coordinating center; SM – smooth muscle; SPGN – sympathetic preganglionic neuron; uPPGN – urethra parasympathetic preganglionic neuron. Scale bar in B is 500 μm.

MATERIALS AND METHODS

Animals

A colony of ChAT-EGFP mice was generated by breeding a homozygous ChAT-EGFP males (obtained from Jackson Labs, RRID: IMSR_JAX:007902) with wild-type C57BL/6 (RRID: IMSR_JAX:000664) females or heterozygous ChAT-EGFP females from the colony. Thus, offspring always were EGFP-positive, although some might be homozygous and some heterozygous. For experiments we used mice of ages between 3 and 6 months selected in a random manner. No substantial difference was found in visually assessed fluorescence expression within the experimental pool of the same sex and same (SI or SCI) status. Microscopic images were digitally photographed with LAS-X software using standardized setting parameters such as exposure time, gain etc. Therefore, the data were merged in four groups accordingly: SI males, SI females, SCI males and SCI females.

Spinal cord injury

The goal of the study was to reveal chronic changes in the EUS-related neuronal populations in lumbar-sacral segments after total interruption of communication with the supraspinal compartments. Therefore, we made complete transection of the spinal cord at the level of T9-T10. Transection was made under isoflurane anesthesia, then animals were kept in the animal facility in accordance with all post-operative requirements. Postoperative care included subcutaneous injections of analgesic ketamine (5mg/kg) and antibiotic Polyflex (100 mg/kg) for 5 days. In both males and females, the micturition reflex was suppressed and voiding was blocked for 1–2 weeks. During this time manual compression of the lower abdomen was performed twice a day to release accumulated urine. It should be mentioned that manual urine elimination in females was much easier to conduct and it took less time than in males. Post-SCI recovery of spinally mediated voiding in mice takes 2–4 weeks for females and 2–6 weeks for males [36]). Therefore, females after 4 weeks and males after 6 weeks after SCI were taken for viral tracing experiments. A few separate male and female mice were also used for measurements of the urethra cross-sections (Fig.3B) to correlate the robustness of the EUS with an anticipated number of neurons subserving its function.

Viral tracing

Retrograde trans-synaptic viral tracing of the spinal motoneurons innervating the EUS and interneurons presynaptic to motoneurons was made with pseudorabies virus PRV614 encoding red fluorescent protein (RFP) (Princeton University, NIH Virus Center grant P40 OD010996). Using the microinjector Nanoliter 2010 (WPI, Sarasota, FL) and a G34 needle on a flexible tubing we made multiple injections of PRV614 in the proximal and middle thirds of the EUS in each animal. The criterion for selecting an optimal dose and post-inoculation survival time was obtaining the best fluorescent labeling of the first order spinal interneurons immediately presynaptic to EUS-MNs with the minimal labeling of higher order interneurons involved in the EUS control. In the initial series of experiments, we injected 5 μL of PRV614 in each animal and set 4 days as survival time, which yielded good tracing of L6-S1 and L3-L4 EUS-related interneurons in the rat [13] and in our preliminary studies in mice where we used another batch of the same virus. However, with this approach we did not find any virus-infected interneurons in SI female mice; in SI males the number of interneurons was much lower than in earlier pilot experiments (probably, due to a different batch of the virus). Therefore, we doubled the volume of PRV614 to 10 μL per animal. This resulted in an increase of the number of labeled interneurons in males, but not in females. We assumed that in females spread and expression of the virus takes longer due to poor EUS innervation and lesser amount of PRV reaching the spinal cord. Therefore, for females the survival time was prolonged to 5 days. It should be noted, that with the doubled PRV614 dose the majority of infected EUS-MNs died leaving fluorescent puncta of debris at the site of DLN. To determine whether the number of cholinergic motoneurons in the DLN is changed after SCI, we counted GFP-positive cells in the DLN of SI and SCI animals. To check how many of GFP⁺ MNs of the DLN can be retrogradely traced with the given protocol, we made a set of FluoroGold (FG) (CAS 223769-64-0) injections in SI mice (2% FG, 2 μL per animal, 10 days survival time). FG is a retrograde tracer which, unlike PRV, cannot cross a synaptic cleft or penetrate adjacent cells through gap-junctions. Therefore, FG labels only the first order neurons, i.e. motoneurons innervating the EUS.

Histology

In the end of the survival period, animals were anesthetized with isoflurane and intracardially transfused with 4% paraformaldehyde in 0.1 M phosphate buffer. Spinal cords were removed, fragments containing (a) L5-S1 and (b) L4-L2 were dissected as shown in Fig.3C, post-fixed overnight and placed in 25% sucrose for 2–3 days. Each of L5-S1 and L4-L2 fragments was cut on a cryostat to 60–70 50μm thick sections starting from the L4/L5 border, i.e. serial sections of L5-S1 were cut in the caudal direction and sections of L4-L2 were cut in the rostral direction. Thus, a set of serial sections spanned ~3 mm in each direction covering the areas of interest. No amplification of GFP or RFP with antibodies was made. Sections were placed on slides in the order of cutting, air dried, embedded in Fluoroshield (Sigma-Aldrich) and coverslipped. Then areas of interest were photographed in each section using 10x objective on the Leica DM 5000 B microscope (RRID: SCR_000011) and LAS-X software (RRID: SCR_013673) (Leica Microsystems). Fluorescent neurons were visually identified and counted. To compare robustness of muscle layers in the EUS, fragments of mid-urethra were dissected, cut to 50μm sections and coverslipped. Autofluorescence of the urethra cross-sections was sufficiently bright to collect representative images.

The EUS-MNs are the first order neurons labeled retrogradely. Presynaptic to them, are the second order neurons representing primary (pINs) or 1^st order interneurons, which are labeled later than MNs with a time lag necessary for PRV614 to cross the synaptic cleft, reach the soma of an interneuron and produce a sufficient amount of RFP. Due to the next trans-synaptic time lag the 3^rd order neurons presynaptic to pINs (i.e. secondary (sINs) or 2^nd order interneurons), will be labeled even later, etc. Thus, higher order cells express weaker if identifiable fluorescence. However, by the time when expression of RFP in spinal interneurons reaches saturation, their postsynaptic cells may already be destroyed by the virus. Usually, neurons of no more than one order could be reliably counted assuming a condition with a similar rate of axonal transport and similar axonal length. Therefore, we could not determine the number of PRV-traced EUS-MNs because most of them were destroyed by the time of highest number of RFP-labeled interneurons. Instead, to quantify EUS-MNs, we used two other approaches: counting the number of DLN cells having GFP-fluorescence and tracing DLN cells with FluoroGold injected into the EUS.

In this study, we focused on analysis of two interneuronal populations of the first order immediately presynaptic to EUS-MNs: (a) L6-S1 local interneurons in the dorsal commissure (DCM-INs) and (b) L3-L4 short axon propriospinal interneurons in the lumbar spinal coordinating center (LSCC-INs). Since retrograde trans-synaptic labeling progresses in an incremental fashion, the intensity of staining helps to roughly discriminate pINs and sINs: when pINs are brightly fluorescent, sINs are distinguishable, but faint [13]. We assumed that post-SCI rewiring might recruit neurons, which were previously foreign to EUS function. Therefore, to assess the maximal number of INs involved to the EUS control in this article we did not separate traced interneurons by intensity of staining and counted all labeled cells.

Cell counts in histological sections were performed in a blind fashion. Photo images with fluorescent neurons were number-coded prior to cell counts, and a person who performed counting did not know to what experimental group the images belonged.

Cell counting

The key to evaluating correctness of obtained data is using unbiased or minimally biased counting methods [69]. There are a few conventionally accepted methods of cells counts in serial sections reviewed in [70]. They include (a) total profile numbers, when counts are done with gaps between analyzed sections (e.g. in every 5^th or 10^th section), so that the end result is approximated by multiplying a total sum by a corresponding coefficient (i.e. k=5 or k=10 in this example); (b) area or volume density and ratio instead of estimating profile numbers; (c) section separation, which is a variant of (a) in an assumption that analyzed sections are separated by a minimal distance greater than diameter of a largest cell; (d) central profile counts, where only cells with prominent nuclei are counted; (e) small objects in thick sections, which typically refer to synaptic boutons. We did not use any of these approaches because in mice LUT-related clustered neuronal assemblies are located within 2–3 mm along the spinal cord, and arbitrary gaps between 50μm thick sections might result in larger errors than counting each profile in each successive section. Furthermore, EUS motoneurons in DLN tended to pack in tight irregular groups, so that analyzing even every 2^nd section might lead to either under- or overestimation of total cell numbers. To minimize this risk and to assess the probability of double counts in our study, we took advantage of using digitized fluorescent microscopic images, which could be visualized in any desired pseudo-color. If neurons in the image of one section are colored in green, whereas neurons in an adjacent section are colored in red, the precise superimposition of sections will reveal overlaps of profiles in yellow. Fig.1 illustrates the principle of the method. If transverse spinal cord sections were separated in the plane indicated as thick dotted line (Fig.1 Aa), some cells can be separated into two parts: one in the upper and one in the lower section (Fig.1 Aa, ##1–5). However, in sections substantially thicker than a maximal soma diameter the majority of neurons will remain intact within one or the other section (Fig.1 A, ## 6–9). After changing the green pseudo-color in the lower section to red (Fig.1 Ab) and merging images (Fig.1 A c,d), the combined profiles will show either complete overlap (Fig.1 Ad, ##1–5), or partial overlap (Fig.1 Ad, ##1–5). A complete overlap revealed by a yellow spot within either green or red larger profile will be indicative of a cell divided into parts. Therefore, such two-color profile is to be counted only once. On the other hand, partial overlap of profiles showing all three colors with differently oriented red and green extensions is indicative of two different cells belonging to different sections. Coinciding outlines and individual peculiarities of adjacent sections can serve as good reference for precise superimposition of sections’ images (Fig.1 B). Precise overlay of images from adjacent sections (one of which was converted from green to red) was performed using LAS-X (RRID: SCR_013673) and ImageJ (RRID: SCR_003070) imaging software. This approach gave us an estimate of a complete overlap probability between profiles belonging adjacent sections as 1%−2% (see Results, Table 1), which is well within the typical 5% confidence level. We did not use immunocytochemical amplification of a reporter signal with antibodies. The recorded fluorescence was intrinsic to neurons, i.e. it was due to intracellularly synthetized fluorophore encoded either by the GFP-transgene (in ChAT-GFP cells), or by RFP-gene delivered with RNA of PRV614. Therefore, intensity of fluorescence varied depending on a cell type, its metabolic rate, functional state at the moment of fixation with 4% PFA, or on the rate of trans-synaptic PRV propagation and a number of viral particles infected a given neuron. In this work we intended to reveal the presence and location of different neurons with no consideration of their physiological state (provided they look healthy) and/or expression of transgenes. Frequently observed partial overlap of profiles having similar or different degree of labeling within the same section could be readily discriminated by size, shape, orientation and/or intensity of staining (Fig.2). A complete inseparable overlap of profiles in a section seems to have a low probability similar to that assessed for adjacent sections. The criteria for counting were: a fluorescent profile should be well distinguishable with distinct outline, not swollen or shrank, not disintegrated by the virus, i.e. a cell must be visually healthy. Since SI animals were perfused with PFA without prior interventions potentially causing pathological changes, a typical shape, size and fluorescence intensity of a certain cell type were taken as traits of healthy neurons to compare them with that of SCI animals. Thus, we counted all healthy-looking profiles of cells expressing a reporter gene regardless of cell size and intensity of fluorescence.

Fig.1.

A - An algorithm of double counting minimization. Fluorescent cell profiles obtained from two adjacent sections may overlap either completely or partially. Most probably the former implies a single cell cut to parts, whereas the latter corresponds to different cells. Schematic examples on the left (#1 to #5) cells separated to parts by cutting in a transverse plane; on the right (#6 to #9) non-cut cells localized either in one or the other section. Vertical grey lines indicate limits of corresponding projections onto transverse plane. a – white dotted line designates a border line between i-th (above the line) and (i+1)-th (below the line) sections; b – pseudo-color of (i+1)-th image changed; a and b depict cells in the longitudinal (e.g. of completely (left) or partially overlapped (right) projections; profiles in #9 display a total overlap of despite they belong to different cells, thus resulting in underestimation of non-cut cell number. B – Schematic transverse sections of the spinal cord. a – image of i-th section with its original pseudo-color (green). b – image of an adjacent (i+1)-th section with a pseudo-color changed to red. In c images a and b are merged. Three cell pairs on the right in the dorsolateral nucleus (DLN) and a pair in the retro-dorsolateral nucleus (RDLN) reveal partial overlap of profiles (yellow) along with non-overlapping segments (red and green). However, one pair of cells in DLN (the most lateral) shows a total overlap. This co-localization of profiles is ambiguous and does not allow to make a unanimous conclusion. Thus, with high probability partially overlapping profiles belong to a pair (or more) cells with somata localized completely within one of adjacent sections, whereas a total overlap predominantly indicates a single cell cut to two parts separated into adjacent sections.

Table 1.

Assessment of potential double counts using pseudo-color-based approach through N=25 pairs of adjacent sections.

Cell type	Total number of MNs in 50 sections	Partial overlap in 25 pairs of sections	Partial overlap per section (%)	Complete overlap in 25 pairs of sections	Complete overlap per section (%)
SB-MNs	342	73	42.7%	3	1.8%
RM-MNs	180	23	25.5%	1	1.1%
RDLN-MNs	557	86	30.9%	3	1.1%

Fig.2.

In pairs of adjacent initially green-colored sections the pseudo-color of one of sections was converted from green to red. Superimposition of green and red sections shows partial overlaps of green and red visualized as yellow. Although in both (A and C) pairs of sections neurons in the DLN are tightly packed, superimposition reveals multiple, but only partial overlap of differently labeled profiles. Differently oriented red and green fragments extending from a common yellow spot indicate presence of two independent (red and green) cells, each entirely belonging to its section. RLDN neurons in A also demonstrate a few partial overlaps, whereas superimposed RDLNs in C do not show yellow spots. Dotted frames in A and C indicate areas zoomed in B and D. CC – central canal, DLN – dorsolateral nucleus, RDLN – retro-dorsolateral nucleus, VMN – ventromedial nucleus. Yellow lines designate a section’s outline, location of the CC and the border between grey and white matter. All of them were used as references to achieve correct superimposition of adjacent sections.

Ethical considerations

All animal experiments were conducted in accordance with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) and National Institutes of Health (NIH) guidelines, and approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh (Protocol approval D16-00118).

Statistics

Statistical analysis was performed using OriginPro 2022 (OriginLab) (RRID: SCR_014212) software package. Numbers of cells per section for each type of neurons in each group were accumulated sequentially starting from the pre-defined reference levels: (a) L4/L5 border for interneurons and (b) the first appearance of SB-MNs for ChAT-GFP cholinergic neurons. Averaging through each rostro-caudal step (Δ=50 μm) resulted in a rostro-caudal distribution of a given neuron type in each experimental group. The data were represented as M±SE and analyzed for differences between groups by one-way ANOVA with P<0.05 significance level.

RESULTS

Testing of pseudo-color-based approach to avoid double-counting.

In pairs of adjacent sections from ChAT-GFP male and female mice we tested probability of complete vs. partial overlap of differently colored profiles, as it was suggested in METHODS. In caudal DLN neurons displayed a denser packing compared to more rostral DLN or other nuclei. Nevertheless, even there the majority of superimposed profiles showed only partial overlap with distinctive red and green fragments differentiating two separate cells (Fig.2). In Fig.2A not only DLN-MNs show partial overlaps, but also some RDLN-MNs. In Fig.2C RDLN-MNs are more dispersed and do not show overlaps. To assess the probability of double counting we scrutinized 25 pairs of superimposed adjacent sections, where one of sections in a pair changed its pseudo-color from green to red. It was found that estimated probability for complete overlap of two profiles was at a level of 1%−2% for motoneurons of the L6-S1 ventral horn (Table 1). Therefore, we assume that our counts through serial sections are correct with more than 95% level of significance.

Sexual dimorphism of urethra

The size of a motoneuron population in the spinal cord should correlate with the number of motor units in a relevant muscle. In turn, the number of motor units should correlate with the size of the muscle. Therefore, assessment of the male vs. female difference in volumes of the EUS is the first approximation to quantify LUT sexual dimorphism. We compared diameter and thickness of the urethra muscles in transverse sections. The urethra of an adult male had external and internal diameter of ~2.2 mm and ~1 mm (Fig.3B top). In the adult female corresponding dimensions were ~0.75 and ~0.35 mm (Fig.3B bottom). Thickness of the EUS striated muscle (determined as a layer of circularly oriented fibers in cross-sections) in a male was ~0.44 mm and in a female ~0.04 mm. This observation implies that the population of motoneurons innervating the urethra in the male mouse is substantially larger than that in the female mouse. To verify this assumption, we counted all cholinergic GFP⁺ neurons in Onuf’s nucleus in ChAT-GFP male and female mice.

Unidentified Motoneurons in L6-S1 segments

The number of DLN cholinergic neurons expressing GFP was counted in the spinal cord of intact (SI) and spinal cord injured (SCI) male and female mice. In transverse sections Onuf’s nucleus contained two morphologically distinct types of neurons. Neurons located in close proximity to the border between grey and white matter (g/w border) tended to spread their dendrites along this border frequently invading deep into lateral and ventral funiculi, reaching the ventral column medially and projecting along or within the lateral funiculus up to RDLN. These cells often had a predominantly elongated or spindle-like shape and bipolar primary dendrites. They typically grouped in “flattened” tight clusters stretched along the g/w border (Fig.4 D, E). The average long and short dimensions in the transverse plane were 31.6±0.6 μm and 13.9±0.4 μm respectively, with the ratio of 2.4±0.06 (M±SE, Table 2). Neurons of another type which were located deeper in the grey matter, were mostly larger in size with rounded somata and multipolar dendrites spreading randomly, except into the white matter. Their long and short dimensions were estimated as 34.9±0.8 μm and 26.3±0.5 μm with the ratio 1.34±0.03 (Table 2). The somata of spindle shaped bipolar cells usually touched each other or overlapped within groups in transverse 50 μm thick sections, whereas rounded multipolar cells were rather distant from each other, but tended to be distributed within a circle-like area (Fig.3 B, C). For convenience, we named border-linked spindle-shaped bipolar cell SB-MNs and rounded multipolar cells RM-MNs. Both types of MNs were spread along the rostro-caudal axis demonstrating a partial overlap but also shifted relative to each other along this axis. It should be emphasized that all analyzed sections were transverse and rostro-caudal shapes/dimensions of profiles were not studied here. Despite a different plane of view might add to cells’ morphological characteristics, we will keep SB-MN and RM-MNs terminology for simplicity in this paper. The most caudal part of DLN contained only SB-MNs (Fig.4E), the middle part contained both SB-MNs and RM-MNs (Fig.4C), and the most rostral part contained only RM-MNs (Fig.4B). In the rostral direction the circle-like structure of RM-MNs became less organized and merged with the RDLN (Fig.4A). To determine the spatial profile for each of these two neuron types we constructed longitudinal distributions in SI animals (Fig.5). Criteria for separation of neurons into two types included: (a) border vs. grey matter location, (b) border-linked vs. random orientation of main dendrites, (c) bipolarity vs. multipolarity, (d) medio-laterally spindle-shaped vs. rounded/uncertain shape of the soma. We started counting GFP⁺ cholinergic DLN neurons in L5-S1 segments from the caudal end, because close to L5/L6 border a distinctive gap between DLN and RDLN disappeared (Fig.4A), so that the rostral end of DLN could not be clearly identified. Thus, GFP-fluorescent DLN cells were counted between 3000 μm and 1000 μm from the L4-L5 border, whereas more caudally than 3000 μm the DLN ended. No correction for double counting was made. Longitudinal distributions of SB-MNs and RM-MNs partly overlapped but had different shapes and mode positions, implying that within the DLN motoneurons of distinct morphology are separated into two subnuclei. The shape of longitudinal distribution of the SB-subnucleus had a prominent mode at ~2000 mm and was similar in males and females, although in females the peak of the distribution was almost three times lower than in males (Fig.5 A, C). Distributions of RM-MNs were also similar in males and females showing an overlap with the SB-subnucleus between 1300 μm and 2300 μm. Total averaged number of RM-MNs in a male (174 ± 7) was significantly greater than in a female (86 ± 13). Total number of SB-MNs in a male (116 ± 8) was also significantly larger than in a female (59 ± 4) (Fig.5 B, D).

Fig.4.

Consecutive GFP-fluorescent images of cholinergic neurons in spinal cord transverse sections for a female (upper row) and a male (lower row). Only the right sides of cross-sections are shown. Due to uncertainty of where the DLN begins on the rostral end, the sections on the right side of the figure start from the first appearance of GFP-positive neurons in the DLN (panel E) at the caudal end, i.e. moving right-to-left on this figure (green arrows in caudo-rostral direction). The most caudal part of DLN contains mostly spindle-shaped bipolar cholinergic neurons (SB-MNs, panels E and D) located at the grey and white matter spreading their dendrites along the grey-white matter border ventral to the RDLN (as in SB-MNs on the rightest images). The white arrow in D (lower row) is directed to a cluster of SB-MNs in the male spinal cord; similar clusters are visible in E and in D-E of the female spinal cord. More rostrally SB-MNs were complemented by larger rounded multipolar cholinergic neurons (RM-MNs) located in the adjacent grey matter but not intermingling with SB-MNs (panels C and B). Further in the rostral direction SB-MNs practically disappeared and the DLN became a circle containing only RM-MNs. The white arrow in B (lower row) is directed to a circle-shaped nucleus containing RM-MNs in the male spinal cord; in the female this nucleus is less robust and contains less neurons. Eventually, the circle-shaped cluster of RM-MNs merged with the RDLN (panel A). Scale bar is 500 μm. CC – central canal (yellow circle).

Table 2.

Neuronal dimensions and length/width ratios in the DLN and RDLN

Dimension	N total	Mean (μm)	SD (μm)	SE (μm)
SB-N width	100	13.932	3.77903	0.3779
SB-N length	100	31.6305	5.9336	0.59336
RM-N width	100	26.325	4.62707	0.46271
RM-N length	100	34.938	7.68309	0.76831
RDLN-N width	100	23.544	4.73257	0.47326
RDLN-N length	100	32.4945	6.6808	0.66808
SB-N length/width ratio	100	2.3728	0.5857	0.05857
RM-N length/width ratio	100	1.34	0.28379	0.02838
RDLN-N length/width ratio	100	1.40777	0.30085	0.03009

Fig.5.

Two types of morphologically discriminated cholinergic neurons in the DLN of females (upper row) and males (lower row). A and C – rostro-caudal distributions of GFP-positive cells in female’s and male’s DLN correspondingly. Counts in the left and right DLN were merged. RM-MNs predominate in the rostral part showing rather monotonic decay towards the caudal end. SB-MNs have the maximal density around the middle of the DLN. B and D - In either sex, numbers of RM-MNs were significantly larger than the numbers of SB-MNs. Data are represented as M±SE.

EUS Motoneurons in L6-S1 segments

To test whether both types of neurons project to the EUS equally, we traced EUS MNs in SI ChAT-GFP male (n=3) and female (n=4) mice with a retrograde tracer FluoroGold (FG) to label only the output first order cells, i.e. only spinal motoneurons projecting to the EUS. FG fluorescence was detected exclusively in neurons within 0–50 μm from the g/w border. In transverse sections most of FG⁺ cells spread primary dendrites along or parallel to the border, whereas their branches often invaded deep into the ventral or lateral funiculi (Fig.6 A, B, C, E, F). The spatial distribution of FG-labeled cells resembled the distribution of SB-MNs, indicating that (1) these neurons represent the majority of EUS-MNs and (2) RM-MNs typically do not project to the EUS but innervate some other muscle(s). In male mice, the number of SB-MNs filled with FG strongly prevailed over FG⁺ RM-MNs with an approximate ratio 10:1 (140/15, 105/9, 80/5). The average number of FG⁺ SB-MNs in a male (108.7±17.4, n=3, M±SE) was significantly larger than the average number of FG⁺ RM-MNs (9.6±5.0, n=3, M±SE) (Fig.7). In a female, the number of FG⁺ SB-MNs was also greater than number of RM-MNs (48.5±8.2 and 1.5±0.95, n=4, M±SE) (Fig.7). Importantly, those RM-MNs, which were traced with FG, directed their main dendrites mostly parallel to g/w border, although some branches could spread into the grey matter (Fig.6F). On the contrary, GFP-positive, but FG-negative RM-MNs, which occupied most of the circle-shaped nucleus, did not have a preferable orientation of primary dendrites. The ventral edge of this circle (RM-subnucleus) merged with the dorsal edge of the flattened clusters of SB-MNs and/or FG⁺ cells at the g/w border (SB-subnucleus), so that there was no distinctive gap for anatomical separation between two subnuclei. Taking this into account, the number of FG⁺ RM-MNs in our counts might be overestimated with concomitant underestimation of FG⁺ SB-MNs, because some FG⁺ neurons (i.e. definitely projecting to the EUS) were classified as RM if they were located at a distance from the g/w border and were multipolar, although their main dendrites were oriented parallel to g/w border. Thus, the spatial profile of distribution and morphology of the FG⁺ DLN neurons coincides with distribution and morphology of SB-MNs. However, the above characteristics of SB-MN pool does not resemble that of RM-MNs. Taken together, it confirms the assumption that RM-MNs in the circle-like part of Onuf’s nucleus do not project to the EUS. Instead, they may innervate the ishiocavernosus muscle (IC), which is well developed in male, but not in female mice. In females the total number of FG-traced cells was expectedly lower than in males. Importantly, in females FG⁺ RM-MNs were exceedingly rare (with SB/RM ratio 35/0, 71/2, 38/4, 50/0 in four SI females), i.e. DLN consisted almost exclusively of SB-MNs. Since in either sex the majority of cells traced with FG from the EUS were classified as SB-MNs, we could compare numbers of FG⁺ SB-MNs and GFP⁺ SB-MNs to assess the percentage of EUS-MNs revealed by retrograde tracing. In males, GFP⁺/FG⁺ ratio was ~1.0 (108±17 and 109±21, correspondingly). In females GFP⁺/FG⁺ ratio was ~0.8 (48±8 and 59±4). Very similar numbers of FG⁺ and GFP⁺ MNs suggest that a tracer injected in multiple sites labeled most of spinal EUS-MNs represented predominantly by SB-MNs.

Fig.6.

EUS-MNs traced with Fluorogold: A, B, C in females, D, E, F in males. Green and blue images are merged, therefore FG-filled neurons are white here. Note long GFP- labeled distal dendrites of FG-traced neurons; green arrows indicate visible branch endings within the funiculus or in the vicinity of the g/w border. A and D show the first appearance of the SB-MNs in the rostral end of DLN. B and E show tightly packed clusters of SB-MNs. C – Multiple dendritic branches of FG-traced SB-MN end near the periphery of the ventrolateral funiculus, whereas the longest branch climbed within the lateral funiculus to the level of the RDLN. F – a rare case when FG-traced SB-MN was far from the g/w border among RM-MNs. Scale bars are 500 μm. The outline of the section and the g/w border are indicated by dotted lines. CC – central canal. DMN – dorsomedial nucleus. DLN – dorsolateral nucleus. RDLN – retro-dorsolateral nucleus. VN – ventral nucleus. Scale bars are 500 μm.

Fig.7.

Predominance of SB-MNs among neurons traced with FG. Both in males and females, RM-MNs constitute <10% of cholinergic neurons traced from the EUS. Here SBN stands for SB-MNs, RMN stands for RM-MNs.

Changes in Motoneurons in L6-S1 segments after Spinal Cord Injury

To determine whether the motoneuron populations are changed in SCI animals, we performed the same cell counts in ChAT-GFP SCI male mice as we did for SI males. These counts have shown similarity in numbers of RM-MNs and SB-MNs per DLN (148.8±13.8 and 106.6±2.1 in SCI vs.174.8±7.3 and 116 ± 8 in SI males) (Fig.8B). Rostro-caudal distributions of DLN-MNs were also similar to that in SI animals (Fig.8A). However, in SCI mice, many GFP-fluorescent cells in all nuclei, especially in the RDLN, were shadowed by dark dead cell bodies and some looked swollen with poor expression of GFP (Fig.8C).

Fig.8.

A - In SCI male mice, distributions of BB-MNs and RM-MNs remain the same as in SI mice. B - numbers of BB-MNs and RM-MNs do not differ from those in SI animals. However, some neurons are deformed and/or screened by bodies of dead cells (C).

Short intraspinal axonal collaterals of motoneurons are not unusual in locomotor circuits. Axons of motoneurons can synapse with neighboring motoneurons of the same nucleus, and with inhibitory neurons like Renshaw cells or glutamatergic excitatory cells taking part in locomotion [37]. We have noticed earlier in SI rats and now in SI mice that PRV tracing from the EUS occasionally results in delayed labeling of motoneurons in the DMN or in RDLN. We found that in SI male mice ~10 GFP⁺ cells in the DMN and in RDLN were labeled with RFP (11.75 ± 2.98 and 13.58 ± 2.27) (Fig.9, SI and plot). At 6 weeks post-SCI, the number of RFP-labeled DMN-MNs (13.37 ± 2.38) did not change, but in the RDLN this number increased 4-fold (55.62 ± 15.17) (Fig.9, SCI and plot). This non-specific labeling was most likely due to trans-synaptic virus propagation because it was delayed relative to EUS-MNs labeling and FG never labeled neurons in the DMN or RDLN. In none of the cases did we see DMN or RDLN neurons destroyed by PRV infection, whereas most of EUS-MNs on the 4^th day were already dead leaving fluorescent debris at the DLN location. The latter fact implies that DMN-MNs and RDLN-MNs can synapse with some motoneurons in the DLN despite large distances between nuclei. Axons originating from RDLN or DMN can synapse DLN-MNs en passant, because they enter the ventral root crossing the DLN (Fig.10).

Fig.9.

In SI animals, GFP⁺ neurons in DMN and RDLN can be trans-synaptically infected by PRV614; RM-MNs in the DLN also can be infected by PRV (yellow = green+red) (left photo). This indicates a possibility of cross-synapsing and information exchange among motoneurons within the same nucleus and between MNs innervating different muscles. SCI induced a 4-fold increase of PRV-labeling in the RDLN, whereas in the DMN the number of infected cells per nucleus was the same as in SI animals (right photo and the plot).

Fig.10.

PRV could retrogradely infect MNs in the RDLN through their axons crossing proximal dendrites of EUS-MNs and synapsing them en passant. On the insets (left inset belongs to A, right inset belongs to B). Note the bundles of RDLN axons passing through the DLN into the ventral root. White arrowheads show the width of axonal bundles within grey matter and the following entrance to the ventral root.

Spinal interneurons of the EUS circuit

Spinal interneurons (INs) related to EUS were traced with PRV by the 4^th day in males, but in females tracing took longer (5 days). Traced INs grouped in the central canal (CC) area: in L6-S1 they were above the CC, and in L3-L4 they were on sides and above the CC (Figs 11 and 12). Importantly, not all fluorescent elements represented neurons. Since the soma size of interneurons was within 10–20 μm, all puncta smaller than 10 μm were considered as stained synaptic boutons or debris from dead cells killed by PRV, so they were not counted. Notably, the average visually assessed INs’ soma size in females was smaller than in males (Fig.11 and Fig.12). Besides, we frequently encountered necrotic foci (e.g., in Fig.12D) in SCI animals. These foci were not related to PRV infection: they appeared randomly in the grey and white matter. In some sections there may be up to 3–5 necrotic spots often with a hollow center and brightly fluorescent tissue around (not shown). It seems that they resulted from ischemia after SCI.

Fig.11.

Interneurons located near the central canal of the male spinal cord traced by PRV614-RFP injected into the EUS. On the left – EUS-related INs in the DCM of L6; on the right – EUS-related INs in the LSCC of L3-L4. A and B – spinal intact animals; C and D – spinal injured animals. Dotted lines indicate borders between the neuropil and white matter; dc – dorsal column, vc – ventral column. The central canal is outlined by yellow. Two DMN-MNs trans-synaptically labeled with RFP are visible at the bottom of the section in A. The scale common for all frames is 500 μm. The yellow dot on the right from the scale has a diameter ~13.5 μm which is close to a minimal size of INs. Red puncta of <10 μm are stained synaptic boutons or debris of dead traced INs.

Fig.12.

Interneurons located near the central canal of the female spinal cord traced by PRV614-RFP injected into the EUS. The composition of the figures and the labelling are the same as in Fig.11. The scale common for all frames is 500 μm. The bright red spot in D is the area of necrosis unrelated to PRV labeling.

The total number of INs in L6-S1 DCM and L3-L4 LSCC differed strikingly between sexes. The DCM of a SI male contained 265.6±62 INs (n=7) with a prominent rise of distribution curve within 1000–2500 mm range, whereas in the DCM of SI females the total number of labeled INs was significantly lower, being 21.4±8.7 per animal (n=10) and showing a flat rostro-caudal distribution (Fig.13, upper row). In chronic SCI mice the total number of DCM-INs was reduced two-fold both in males and females, showing 120.4±23.6 DCM-INs per male and 11.6±5.6 per female (Fig.13, lower row). After SCI the male/female ratio for DCM-INs remained practically the same as for SI animals and was highly significant (p<0.01), showing 12:1 for SI and 11:1 for SCI mice. For males the reduction of DCM-INs after SCI was significant (p<0.05) (Fig.13, left bar plot), but for females it was not (Fig.13, right bar plot).

Fig.13.

Upper row - Rostro-caudal distributions of DCM-INs in SI male (left) and female (right) mice traced by PRV614 injected into the EUS. Lower row – Similar distribution of PRV614 labeled DCM-INs in chronic SCI mice. RFP-labeled neurons were counted from L4-L5 border in caudal direction. The bar plots between distributions in each row demonstrate significant prevalence of DCM-INs number in males (blue) over that in females (pink). The bar plot on the left indicates a significant reduction of DCM-INs population after SCI (light blue) in males. The bar plot on the right shows some, but not significant loss of DCM-INs in females (light pink).

The population of L3-L4 LSCC neurons traced from the EUS in SI mice was significantly more numerous in males compared to females: 125±14.5 (n=7) and 16.1±3 (n=10) per animal correspondingly. In SI males the caudo-rostral distribution revealed the highest density of LSCC-INs per section within L4 and the caudal half of L3 (Fig.14, upper left). In SI females the distribution was practically flat without an obvious mode (Fig.14, upper right). After SCI the number of LSCC neurons reduced 5-fold in males but did not change in females, with 65.5±22.1 (n=6) and 14±4.2 (n=5) cells per animal respectively. This difference was significant, although not as striking as in SI mice. Thus, the post-SCI populations of LSCC-INs did not differ significantly between sexes. The profile of distributions in SCI males became substantially flattened compared to control, and in SCI females it did not change (Fig.14, lower row).

Fig.14.

Upper row - Distributions of PRV614 labeled LSCC-INs in males (left) and females (right) differ in SI animals. In SI males, most labeled cells were found within the second millimeter rostral to the L4/L5 border. In SI females, the distribution was flat without significant deviations of the envelope. Lower row – After spinal cord injury (SCI) both males and females exhibit essentially flat distributions without obvious maxima, although, in males, there is a tendency for an increase in the caudal direction. The bar plots between distributions of SI animals demonstrate a significant prevalence of LSCC-INs in males (blue) over that in SI females (pink). The bar plot in the lower row indicates the absence of significant differences between males and females in total numbers of LSCC-INs after SCI. The bar plot on the left indicates a significant loss of LSCC-INs after SCI in males (light blue). The bar plot on the right shows no change in LSCC-INs number in females (light pink).

DISCUSSION

Sexually dimorphic DLN is the commonly acknowledged location of EUS-MNs in rats and mice. In addition to EUS-MNs, the DLN in males contains motoneurons innervating musculus ishiocavernosus (IC), which is absent in females [21]. Another sexually dimorphic bilateral cluster of motoneurons is the DMN. It contains motoneurons of the external anal sphincter (EAS) and motoneurons of the musculus bulbospongiosus (or bulbocavernosus); the latter is present in males, but not in females [21]. Absence of m. ishiocavernosus and m. bulbospongiosus in female rats corresponds to lower number of motoneurons in DLN and DMN as compared to males [38]. McKenna and Nadelhaft [21] also found, that in female rats EUS motoneurons accounted for almost all DLN neurons, and anal sphincter motoneurons accounted for almost all DMN neurons. They found no difference between males and females in the number of MNs innervating EUS and EAS, but in the male rat motoneurons related to EUS are located more laterally and neurons related to IC are located more medially within the DLN. Reconstructions of traced DLN neurons in rats showed that they are strictly bipolar extending dendrites along the g/w border and into the white matter [39]. In our experiments in mice we also identified EUS-MNs as bipolar cells with long dendrites extending along the g/w border. However, sexual dimorphism in mice is expressed differently compared to rats. First, the number of EUS-MNs in females is twice as low as in males. Second, we did not see a mediolateral subdivision to EUS and IC subnuclei. Instead, flattened clusters of bipolar EUS-MNs on the g/w border first occurred most caudally alone, then further in the rostral direction they were distributed adjacent to the circle-like sub-nucleus consisting of large rounded multipolar cells. These multipolar cells were positioned dorsally from SB-MNs and are assumed to be IC-MNs, although this assumption must be verified by tracing from the IC in future studies. If it is correct, one would expect very low or zero percentage of RM-MNs in the DLN of females compared to males. However, Fig.5 shows similar SB-to-RM proportions in males and females with similar spatial distributions. The similarity of SB-to-RM proportions in males and females is a puzzling finding, which should be clarified in the future.

Our data reveal two previously underestimated facts in the mouse DLN anatomy. First, the number of DLN MNs in females is smaller not only because the male DLN contains a set of IC-MNs, but also because a set of purely EUS-related MNs is twice as small in females compared to males. Considering differences in the diameter and thickness of the urethra (Fig.3C), it is feasible to assume that an apparent number of motor units in the female EUS is smaller, so the corresponding number of EUS-MNs must be smaller than in the male. There may be alternative explanations of the observed sexual difference in DLN-MNs’ numbers. For example, lesser number of DLN-MNs in females may be due a larger size of a female’s individual motor unit involving more muscle fibers innervated by a single motoneuron. However, in this case either thickness of the EUS would be comparable in males vs. females (or even larger in females), or striated muscle fibers in females must be much thinner to occupy smaller volume by a larger number of fibers and be in accordance with the observed anatomical difference (see Fig.3C). Both alternatives seem to be less likely than the suggested “the less motor units - the less motoneurons” relations, but they should be certainly tested in future studies. The second underestimated feature of the DLN anatomy is an overwhelming localization of distal SB-MNs’ dendrites within white matter, where they span through entire thickness of the lateral and ventral funiculi and reach the ventral column, indicate that the EUS-MNs collect most of their afferent inputs from the ventrolateral funiculus, but not from within the central grey matter of L6-S1. The latter fact corroborates our previous finding that propriospinal neurons send their axons from L3-L4 down to L6-S1 levels within the ventral column [40], i.e. they can synapse SB-MNs either in the ventral column or in the ventral funiculus. At the same time, this fact raises a question: what is the spatial arrangement of the DCM-INs’ axons through which they can be traced? One possibility is that DCM-INs also send axons to the ventral column and further to the ventral funiculus to synapse SB-MNs dendrites. Other possibilities are: (a) their axons reach some infrequent dendritic branches of PRV-infected RM-MNs, (b) reach SB-MNs through the neuropil, (c) send their axons to the lateral funiculus, or (d) INs’s axons synapse dendrites of PRV-infected DMN- and/or RDLN-MNs. The latter is a feasible option as DMN-MN dendrites extend to the lamina X, and a compact bundle of RDLN-MN dendrites approaches DCM through the lamina VII (observed but not shown in figures). To test which option is correct, a separate experiment is needed in the future. There is also a consequence of the assumed separation between the EUS- and IC-related subnuclei based on characteristic dendritic morphology. Namely, spatial distributions of SB- and RM-MN types show the longitudinal shift of one population relative to the other, thus supporting the notion of structurally and functionally different neuronal pools. However, the suggested correlation between morphology of RM-MNs and their connection with the IC is hypothetical and needs testing in a separate experiment. Since we could not identify a distinctive end of the rostral circle-like part of the DLN, there is a possibility that close to its rostral end the IC-related part of DLN merges with or converts into another nucleus innervating some other muscles.

Activity of all motoneurons is under complex control of interneuronal populations either local or distributed along the spinal cord [41]. Local L6-S1 interneurons involved in the EUS spinal circuit can be trans-synaptically traced in the dorsal commissure (DCM) of rats [13, 35]. In addition, a population of propriospinal interneurons presynaptic to EUS-MNs was found in L3-L4 segments in the vicinity of the central canal [13, 40]. This population was named the lumbar spinal coordinating center (LSCC), because in the rat spinal cord transection at or below L4 prevented restoration of voiding after several post-SCI weeks, whereas transections above L3 did not [10]. Therefore, a hypothesized intraspinal post-SCI rewiring involving propriospinal neurons was suggested to compensate for the lack of supraspinal signaling in chronic SCI animals [12]. An enhancement of LSCC circuit due to sprouting and recruiting new neurons would explain partial post-SCI recovery of the micturition reflex. To test this hypothesis, we compared the sizes of DCM and LSCC interneuronal populations traced with PRV614 encoding RFP.

Numbers of EUS-INs in male vs. female do not correlate with numbers of output motoneurons. Female mice have ~10 times less DCM- and LSCC-INs than males, whereas they have only 2 times less EUS-MNs (= SB-MNs). This discrepancy may be due to sexually different patterns of urethra muscle contraction, which implicate different network requirements. In males voiding is accompanied by EUS-electromyography (EUS-EMG) busting and rhythmic 5–7 Hz EUS contractions actively pumping out portions of urine. In females voiding occurs during a series of smooth waves of urethra relaxation allowing passive release of urine driven by intravesical pressure [67] Thus, difference in mechanisms of voiding may explain the observed fact of much easier manual emptying of the bladder in females than in males before reflex voiding recovers after SCI (personal observation). Restoration of reflex voiding in females also occurs earlier than in males [36] which is consistent with the notion of functional dimorphism as a consequence of anatomical dimorphism. Furthermore, insignificant change in numbers of female DCM- and LSCC-INs after SCI indicates that these interneurons, in contrast to those in males, neither play a big role in micturition block during first post-SCI weeks, nor they play a role in reflex voiding restoration.

It seems plausible that circuits involved in control of different muscles may interact through random or evolutionary balanced synaptic contacts. Therefore, we need to consider existence of the non-dimorphic RDLN located bilaterally near the middle of the lateral funiculus and the dimorphic DMN located near the crest of the ventral column. RDLN-MNs innervate hind limb muscles using ~50% of motor fibers in the pudendal nerve [42, 43]. We did not count GFP⁺ RDLN cells, but visual assessment of RDLN size (see Fig.4) suggests an approximate equality of the neuronal content in males and females, i.e. corroborates non-dimorphic nature of the RDLN. On the contrary, GFP⁺ neurons in the DMN of males were encountered more frequently and were more numerous by visual assessment than in females, thus corroborating dimorphism of the DMN. In our previous experiments injections of PRV into the EUS of rats labeled not only DLN-MNs and DCM-INs, but sometimes a few DMN-MNs (unpublished data). In mice we observed the same phenomenon. Moreover, we saw PRV-labeling of RDLN-MNs in SI and SCI animals. After the optimal (for revealing second order neurons) post-inoculation survival during 4 days GFP⁺ MNs in DMN and RDLN double-stained with RFP never showed signs of decay or destruction, whereas in the same section many, if not all, DLN-MNs could be partly or totally decomposed by PRV. This indicates that in our tracing protocol DMN-MNs and RDLN-MNs are the second order cells and can be PRV-labeled through DLN-MNs. In principle, secondary labeling might happen either by invasion of PRV across synaptic cleft or through gap-junctions between infected and non-infected cells. Intranuclear gap-junctions between motoneurons in DLN, DMN and RDLN have been documented [44, 45]. Furthermore, gap-junctions in DLN and DMN were found to connect sexually dimorphic with non-dimorphic motoneurons within these nuclei [46]. In the latter work connexin36-containing gap-junctions were found on somata and dendrites of >94% of motoneurons within each nucleus. Since dendrites spreading along the white/grey matter border from both DLN and DMN meet and overlap, they might also establish gap-junctional connections making possible dye propagation. However, gap-junctions provide only exchange with ions and small molecules of dyes like Lucifer yellow [47]. Large particles or molecules like viruses and proteins are too large to penetrate the gap-junctional pore. Thus, the only remaining option is the retrograde transport of the virus through synaptic connections from DLN-MNs dendrites to axons or axonal collaterals of DMN- or RDLN-MNs.

The unexpected selective enhancement of PRV retrograde transport through the synaptic cleft from DLN to RDLN after SCI may be due to either more active virus uptake by presynaptic boutons of RDLN-MNs or to sprouting and an increased number of synaptic contacts. Synaptic connections between motoneurons of different nuclei do exist. For example, a regular retrograde tracing of DMN neurons results in only ipsilateral labeling, whereas trans-synaptic retrograde tracing labels DMN-MNs bilaterally [34].In the latter case this can provide an additional non-specific excitatory tone to EUS-MNs causing spasm of the EUS. The DLN is positioned just at the entrance of all efferent fibers into the ventral root. Fluorescently labeled axonal bundles exiting RDLN and converging into the ventral root through DLN are good candidates to synapse with DLN-MN an-passant or via short axonal collaterals. Axons of DMN-MNs also have to cross the DLN to enter the ventral root, and they also can synapse with DLN-MNs. Auto-correlation analysis of motor unit discharges in the EUS and EAS showed that their motoneurons do not share a common drive, because a rhythmic component in EUS-EMG has a twice higher preferred frequency compared to that in the EAS [48]. However, cross-correlation of two simultaneously recorded EUS and EAS single-unit trains revealed a possibility of functional connection between driving motoneurons. The constant time lag unequivocally indicated monosynaptic coupling where the EAS-MN’s action potential was 1 ms ahead of the EUS-MN spike [48]. This finding suggests that EUS-MNs in the DLN can be synaptically excited not only by adjacent motoneurons or by interneurons, but also by inputs from other motor nuclei. It opens a possibility that tonic activity of the EUS can be modulated by non-specific motoneurons making it partially correlated with tonic activity of other sexually dimorphic and/or non-dimorphic muscles. Enhancement of the RDLN-to-DLN synaptic input after SCI suggests either (a) upregulation of non-specific synapses compensating for the disappearance of previously available specific inputs or (b) pathological increase of previously dormant non-specific excitatory inputs. In both cases enhancement of RDLN’s modulatory action on EUS could make it a noticeable player in the post-SCI LUT function although further studies are needed to clarify this point.

Apart from synaptic connections with local L6-S1 interneurons, the EUS-MNs can receive inputs from other segments or from supraspinal structures. If the distant connections come through the ventrolateral funiculus, then we should assume that they synapse long EUS-MNs’ dendrites en passant within white matter and go further to other synaptic targets. In the cat ascending short propriospinal fibers are derived mainly from cells in the intermediate zone of lumbar segments. The bulk of these fibers is derived from the medial and central parts of ipsilateral lamina VII and then converges to the ventral funiculus or ventral part of the lateral funiculus [49]. Descending short propriospinal fibers from neurons in the intermediate zone also connect to their targets within lateral or ventral funiculi [50]. Axons of propriospinal interneurons of L3-L4 LSCC project to L6-S1 through the ventral column where they ramify within the ventral funiculus and the nearby grey matter [12, 40]. Serotoninergic axons descending from the brainstem within the lateral funiculus also occur more densely in caudal segments and are most prominent at the sacral level. In lower lumbar and sacral segments, serotonin fibers are dispersed throughout the ventral and lateral funiculi. Large motoneurons, especially within the lumbar enlargement, are surrounded by fine networks of serotonin fibers and terminals [51]. Since retrograde trans-synaptic tracing did not reveal other interneurons presynaptic to EUS-MNs except for DCM- and LSCC-INs, we should consider a few additional options to explain positioning of EUS-MNs dendrites within white matter. First, there is a possibility that DCM-INs send their axons through the lateral funiculus. Second, there is another intraspinal origin of excitatory or modulatory inputs to EUS-MNs arriving in the lateral/ventrolateral funiculus. Third, a bulk of synaptic inputs arrive to EUS-MNs from supraspinal structures. Fourth, combinations of the above options.

To understand the process of functional restoration or irreversible loss of function after SCI, one should take into account all interconnected processes occurring in the spinal cord after injury. Chronic SCI induces multiple structural, metabolic and functional pathologic changes in the injury microenvironment and in distant areas below injury (reviewed in [52]). Neuronal damages involve neural inflammation, synaptic disfunction, neuronal degeneration, cell death, demyelination, etc.[53]. The primary SCI caused by the initial trauma results in ischemia, oxidative damage, edema, and glutamate excitotoxicity in the spinal cord; physiologically it associated with hyperflexion, hyperextension and axial loading [54, 55]. This process initiates a secondary injury cascade, which starts just a few hours post-injury and may continue for more than 6 months, leading to additional cell death and spinal cord damage. A robust and complex inflammatory cascade is known to be a prominent component of the secondary injury following SCI [56–58]. Inflammation on the one hand causes neuronal loss or functional silencing, but on the other hand it transiently increases neuronal excitability [59]. Early administration of anti-inflammatory drugs alleviates consequences of SCI and facilitates functional recovery [60–63] [64]. After a compression-induced SCI without breaks in the vascular system or white-matter axonal pathways, a wave of secondary damages develops due to lasting inflammation below the site of injury [56, 65, 66]. In the SCI model with complete spinal cord transection all anterior and posterior spinal veins and arteries are cross-cut leaving intact only auxiliary segmental blood vessels below the site of transection. This should give rise to even more dramatic impairments including cell death, because the fragment below the transection suffers from deprivation of nutrients and oxygen, and from accumulation of toxic waste.

In our experiments the spinal cord transection was complete, which means development of trophic deprivation, dramatic reduction of oxygenation and possible penetration of toxic substances in the tissue below transection. At 4–6 weeks after SCI we observed multiple foci of tissue dystrophy and necrosis, while some nearby neurons preserved normal shape and apparently could be functionally active (as in Fig.12D). Thus, on top of neuronal loss below transection, necrotic sites could interrupt some pre-existing pathways as well as communication in local networks. The bladder-related interneuronal pool most probably also suffers massive loss. Reduced interneuronal networks are more likely to lose the ability to coordinate different executive (i.e. motoneuronal) compartments. Therefore, in a possible but oversimplified scenario of post-SCI restoration of voiding in males, the leading component of recovery may be a gradual decay of non-specific inflammation-driven EUS-MNs’ excitation. After this excitation decreases, the remaining interneuronal pool becomes capable to take MNs’ activity under control and make it coordinated with bladder contractions. It seems that available afferent signals from the bladder become sufficient to initiate generation of EUS bursting by the local circuit. In females a decay of this non-specific EUS-MNs’ excitation may be sufficient by its own as females need only relaxation of the EUS to perform voiding. This scenario may be a simplification of the real chain of events, but in order to dismiss or confirm it, other experiments involving intracellular recording from EUS-MNs in SI and SCI conditions are necessary.

Abbreviations:

BL

bladder

DCM

dorsal commissure

DLN

dorsolateral nucleus

DMN

dorsomedial nucleus

DSD

detrusor-sphincter dyssynergia

EAS

external anal sphincter

EUS

external urethral sphincter

EUS-MNs

EUS related motoneurons

INs

local to L6-S1 EUS-related interneurons

pINs

primary interneurons

PMC

pontine micturition center

PPNs

propriospinal EUS-related neurons

PRV

pseudorabies virus

L3-L4

a fragment of the spinal cord consisting of 3^rd and 4^th lumbar segments

L6-S1

a fragment of the spinal cord consisting of 6^th lumbar and 1^st sacral segments

LSCC

lumbar spinal coordinating center

LUT

lower urinary tract

RDLN

retro-dorsolateral nucleus

RM-MNs

rounded multipolar motoneurons

SB-MNs

spindle-shaped bipolar motoneurons

SCI

spinal cord injury

SI

spinal intact

sINs

secondary interneurons

RFP

red fluorescent protein

VMN

ventromedial nucleus