Disproportionate Loss of Excitatory Inputs to Smaller Phrenic Motor Neurons Following Cervical Spinal Hemisection

Abstract

Excitatory glutamatergic input mediating inspiratory drive to phrenic motor neurons (PhMNs) emanates primarily from the ipsilateral ventrolateral medulla. Unilateral C₂ hemisection (C2SH) disrupts this excitatory input resulting in cessation of inspiratory-related diaphragm muscle (DIAm) activity. In contrast, after C2SH, higher force, non-ventilatory DIAm activity persists. Inspiratory behaviors, require recruitment of only smaller PhMNs, whereas with more forceful expulsive/straining behaviors, larger PhMNs are recruited. Accordingly, we hypothesize that C2SH primarily disrupts glutamatergic synaptic inputs to smaller PhMNs, whereas glutamatergic synaptic inputs to larger PhMNs are preserved. We examined changes in glutamatergic presynaptic input onto retrogradely labelled PhMNs using immunohistochemistry for VGLUT1 and VGLUT2. We found that 7 days after C2SH there was an ~60% reduction in glutamatergic inputs to smaller PhMNs compared to ~35% reduction at larger PhMNs. These results are consistent with a more pronounced impact of C2SH on inspiratory behaviors of the DIAm, and the preservation of higher force behaviors after C2SH. These results indicate that the source of glutamatergic synaptic input to PhMNs varies depending on motor neuron size and reflects different functional control -perhaps separate central pattern generator and premotor circuits. For smaller PhMNs, the central pattern generator for inspiration is located in the pre-Bötzinger complex and premotor neurons in the ventrolateral medulla, sending predominantly ipsilateral projections via the dorsolateral funiculus. C2SH disrupts this glutamatergic input. For larger PhMNs, a large proportion of excitatory inputs appear to exist below the C₂ level or from contralateral regions of the brainstem and spinal cord.

Introduction

A motor unit comprises a motor neuron and the muscle fibers that it innervates (Liddell & Sherrington, 1925). For mixed muscles like the diaphragm (DIAm), an orderly recruitment of different motor unit types is particularly important to establish an appropriate motor output (Sieck, 1988; Sieck & Fournier, 1989; Sieck et al., 1989). It is well established that motor unit recruitment depends on motor neuron size (the ‘size principle’) such that smaller motor neurons are recruited before larger motor neurons (Henneman, 1957; Henneman et al., 1965; Clamann & Henneman, 1976; Henneman & Mendell, 1981). The size of motor neurons is matched to the contractile and fatigue properties of muscle fibers comprising motor units. Accordingly, smaller motor neurons innervate muscle fibers that are fatigue resistant and generate lower forces compared to motor neurons that are recruited later, which innervate muscle fibers that are fatigable but generate greater forces (Sieck et al., 1985; Sieck & Fournier; Kernell, 2006).

As in other skeletal muscles, motor units in the DIAm can be classified into four types based on their contractile and fatigue properties (Fournier & Sieck, 1988). Only the recruitment of smaller phrenic motor neurons (PhMNs) comprising slow (type S) and fast fatigue resistant (type FR) motor units is required to generate the trans-diaphragmatic pressures (P_di) necessary during ventilatory (breathing) behaviors (Fournier & Sieck, 1988; Sieck, 1988; Sieck & Fournier, 1989; Sieck, 1994; Mantilla et al., 2010; Mantilla & Sieck, 2011; Sieck et al., 2013; Seven et al., 2014). In contrast, higher force expulsive DIAm behaviors require recruitment of larger PhMNs comprising fast fatigue intermediate (type FInt) and fast fatigable (type FF) motor units.

The descending glutamatergic excitatory input to PhMNs responsible for “inspiratory drive” and inspiratory-related DIAm activity is primarily ipsilateral emanating from the rostral ventral respiratory group (rVRG) (bulbospinal pathway) (McCrimmon et al., 1989; Lipski et al., 1994,; McCrimmon et al., 1995). However, there is a smaller but significant contralateral input to PhMNs that, although latent and ineffective in eliciting inspiratory-related activation of PhMNs immediately after C2SH, appears to strengthen with time and underlies spontaneous recovery of inspiratory-related DIAm activity (Goshgarian et al., 1991; Moreno et al., 1992; Miyata et al., 1995; Fuller et al., 2006; Fuller et al., 2008; Mantilla et al., 2013a; Mantilla et al., 2013b; Gransee et al., 2017). Importantly, we recently found that DIAm activity during higher force sighs and inspiratory efforts against an occluded airway is relatively unaffected by C2SH (Martinez-Galvez et al., 2016; Hernandez-Torres et al., 2017). These higher force activities of the DIAm would require recruitment of larger PhMNs. Based on this observation, we hypothesize that C2SH predominantly disrupts ipsilateral excitatory glutamatergic synaptic input to smaller PhMNs that is responsible for inspiratory drive during breathing.

A previous electron microscopy (EM) study reported a loss of glutamatergic synapses at motor neurons in the cervical spinal cord (presumed PhMNs) at four days post-C2SH (Tai & Goshgarian, 1996). However, this study did not examine differences in glutamatergic synapses on motor neurons of varying size, and the EM technique employed was extremely limited in its sample size. In a recent study, we designed a novel confocal imaging procedure to systematically evaluate glutamatergic terminal distribution (using VGLUT immunoreactivity) on a large number of retrogradely labelled PhMNs (Issa et al., 2010; Rana et al., 2019b). In the present study, we employed this technique to assess size dependent loss of VGLUT terminals at PhMNs at seven days post-C2SH (SH7D). We hypothesized that C2SH would result in pronounced loss of glutamatergic inputs to smaller PhMNs as compared to larger PhMNs ipsilateral to the injury.

Materials and Methods

Ethical Approval

All procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85–23, revised 2011) and the policies and regulations set out in the editorial in The Journal of Physiology and Experimental Physiology by Grundy (Grundy, 2015). All procedures were also approved by the Institutional Animal Care and Use Committee at the Mayo Clinic (protocol no. A3105–17). All steps were taken to minimize animals’ pain and suffering.

Experimental Animals

A total of 13 adult (12 weeks) male Sprague-Dawley rats obtained from Envigo (Hsd:Sprague Dawley® SD, Indianapolis, IN) were assigned to either control, uninjured (n=6) or SH7D (n=7) groups. Based on previous studies in which we found no differences in PhMN number, somal surface areas, VGLUT terminal distribution or DIAm activation across behaviors between female and male rats (Fogarty et al., 2018; Khurram et al., 2018; Rana et al., 2019b), sex was not considered as a biological variable in the present experimental design. Animals were housed individually in cages under a 12-h light/dark cycle with ad libitum access to food and water. Animals were anesthetized with ketamine (90 mg/kg) and xylazine (10 mg/kg) via intramuscular injection for all experimental procedures.

PhMN Labelling and Tissue Collection

As previously described, PhMNs were retrogradely labelled by injecting cholera toxin subunit B (CTB) solution (Cat#104, List Biological, Campbell, CA) bilaterally into the DIAm (Prakash et al., 1993; Prakash et al., 2000; Rana et al., 2019b). Briefly, following induction of anesthesia, a midline laparotomy was performed such that the inferior surface of the DIAm was exposed. Using a 10 μl 20-gauge Hamilton syringe (Cat # 7731–10, Hamilton Company, Reno, NV), three 5 μl injections of CTB (0.2%) were made at different sites equally spread across the dorsal to ventral surface of the muscle to ensure adequate labelling of the entire motor neuron pool. Thereafter, muscle and skin layers were sutured using 3–0 Vicryl (polyglactin 910). Carprofen (50 mg/ml) was administered ad libitum in the drinking water and started 48 h prior to surgery. Animals also received Penicillin G (25,000 U/kg) intra-muscularly and parenteral dextrose and saline injections sub-cutaneously following surgery. Animals were maintained on a heating pad until they were awake and alert.

During the same surgery for CTB injection, EMG electrodes were placed in the DIAm (see below). These surgeries occurred four days prior to the C2SH surgery (see below) to allow ample time for labelling of the PhMN pool. Animals in the control group received CTB injections 11 days prior to the terminal experiment to match the time course of CTB injection in SH7D group.

Assessment of DIAm EMG Activity

Chronic DIAm EMG activity was recorded using a pair of bilateral wire electrodes inserted into the mid-costal region of the DIAm and externalized for chronic use, as previously described (Miyata et al., 1995; Mantilla et al., 2011; Mantilla et al., 2013a; Gransee et al., 2015; Rana et al., 2017). Briefly, a midline laparotomy was performed to expose the inferior surface of the DIAm. A pair of insulated stainless-steel fine wire (AS631, Cooner Wire Inc., Chatsworth, CA) electrodes with ~3 mm of insulation exposed at the recording tip were placed into each side of the mid-costal DIAm with an inter-electrode distance of ~3 mm. The electrode pair was secured in place using a surgical knot at the abdominal surface of the muscle, looped within the abdomen, tunneled sub-cutaneously, externalized and secured to the dorsum of the animal. Animals were allowed to recover for three days to avoid possible laparotomy-induced inhibition of neural activation of the DIAm, which peaks ~6–8 h after laparotomy and returns to baseline by ~48 h (Ford et al., 1983; Dureuil et al., 1986).

For the SH7D group, baseline inspiratory-related DIAm EMG activity was recorded at 1 day prior to the injury, and further recordings were performed at 3 days postoperatively to confirm the absence of inspiratory-related DIAm EMG and at the terminal experiment at 7 days post C2SH (SH7D group). For every session, the EMG signal for each pair of DIAm electrodes was differentially amplified (gain: 2000x) and band-pass filtered (20–1000Hz) using an analog amplifier (Model 2124, DATA Inc.). The signal was digitized using a data acquisition board (National Instruments, Austin, TX) with sampling rate of 2 kHz and recorded using a custommade program (LabView 8.2, National Instrument). The root mean square (RMS) of DIAm EMG was calculated using a 100-ms window, and the peak RMS (RMS_peak) was determined during different ventilatory and non-ventilatory DIAm behaviors.

DIAm EMG activity was recorded during eupneic breathing (room air), during breathing stimulated by exposure to a hypoxic-hypercapneic (10% O₂, 5% CO₂) gas mixture, and during sustained airway occlusion, as in previous studies (Mantilla et al., 2010; Mantilla & Sieck, 2011; Rana et al., 2017). Animals were allowed to recover for 10 min between hypoxic-hypercapnic exposure and airway occlusion. The RMS_peak of DIAm EMG was averaged across a 120-s period of eupnea, during the last 60-s of a 5-min period of hypoxic-hypercapneic exposure, and for the last 10 s of a 45-s period of airway occlusion.

Spinal Hemisection

Details of the C2SH surgical procedure have been previously published (Miyata et al., 1995; Prakash et al., 1999; Mantilla et al., 2007; Gransee et al., 2013, 2015; Gransee et al., 2017). At the start of the procedure, animals were checked for depth of anesthesia using a toe pinch, and a change in heart rate, whisker twitch and hindlimb withdrawal. An appropriate level of anesthesia was verified continuously during the surgical procedure and animal was re-dosed with a ¼ dose of initial ketamine/xylazine dose if needed. Under sterile conditions, a bilateral dorsal C₂ laminectomy was performed and the right half of the spinal cord was sectioned lateral to the dorsal fissure, thus transecting only the ventral and lateral funiculi, while preserving the dorsal funiculus on the side of C2SH. Previous studies by our group have evaluated the extent of histological damage following unilateral C2SH (Miyata et al., 1995; Sieck & Mantilla, 2009; Mantilla et al., 2014). Muscle and skin layers were sutured using 3–0 Vicryl (polyglactin 910). Carprofen (50 mg/ml) was administered ad libitum in the drinking water and started 48 hours prior to surgery. A single Buprenorphine sustained release (0.6 mg/kg) bolus was administered immediately post-surgery. Animals also received Penicillin G (25,000 U/kg) intra-muscularly and parenteral dextrose and saline injections sub-cutaneously. Animals were maintained on a heating pad until alert and awake. Animals were monitored on a daily basis for signs of distress, dehydration, and weight loss, with appropriate veterinary care given as needed.

As in previous studies, DIAm EMG was recorded prior to C2SH surgery and at 3D after C2SH to confirm completeness of C2SH by the absence of ipsilateral inspiratory-related DIAm EMG activity during eupnea in anesthetized animals (Miyata et al., 1995; Zhan et al., 1997; Prakash et al., 1999; Gransee et al., 2013; Mantilla et al., 2013a; Mantilla et al., 2013b; Gransee et al., 2015; Martinez-Galvez et al., 2016). The C2SH injury consistently results in damage to the ventral and lateral columns at C2SH resulting in absence of ipsilateral inspiratory-related DIAm EMG activity at three days after C2SH. At the terminal experiment, animals were deeply anesthetized, euthanized by exsanguination and transcardially perfused with 4% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS, pH 7.4). The spinal cord was resected from C₂ to T₁, post-fixed in 4% paraformaldehyde for 24 h and transferred to 30% sucrose in 0.1 M PBS (pH 7.4) for three days at 4°C. Spinal cords were subsequently embedded in cryomolds (VWR, Radnor, PA), sectioned longitudinally at 70 μm and stored in 0.1M PB solution in 24-well plates until processing.

Immunohistochemistry

Immunohistochemical detection of synaptic input at PhMNs was based on previously reported techniques (Issa et al., 2010; Rana et al., 2019b). Briefly, spinal cord sections were blocked for 1 h in 10% normal donkey serum in TBS containing 0.3% Triton X-100 (TBS-Tx), followed by incubation overnight (16–24 h) with the appropriate primary antibodies in 5% normal donkey serum TBS-Tx. An anti-CTB (Goat polyclonal, Cat # 703, Lot # 7032A9, 1:5000; List Biologicals) was used to label PhMNs. An anti-vesicular glutamate transporter 1 (VGLUT-1) antibody (Mouse monoclonal IgG2a, 135311; Synaptic Systems, Goettingen, Germany; 1:400) and anti-vesicular glutamate transporter 2 (VGLUT-2) antibody (Mouse monoclonal IgG1, MAB5504, Lot # 2817471; Millipore, Billerica, MA; 1:400) were used to identify glutamatergic synapses at PhMNs due to the predominant VGLUT-1 and VGLUT-2 expression at glutamatergic synapses around motor neurons (Oliveira et al., 2003; Herzog et al., 2004; Issa et al., 2010; Rana et al., 2019b). After being washed three times in TBS (15 min each), sections were incubated with donkey Alexa Fluor 488 anti- goat (Cat # 705–545-003) and Alexa Fluor 594 anti-mouse secondary antibodies (Cat # 715–585-150, IgG; Jackson Immuno Research Laboratories, Inc. West Grove, PA) for 3h. Tissue sections were then washed three times in TBS before being mounted with ProLong Gold Antifade Mountant (Thermo Fisher Scientific, Waltham, MA) and cover-slipped. For all primary/secondary antibody pairs, additional studies not including the primary antibody (blank) or alternately using Alexa Fluor conjugated secondary antibodies were conducted to confirm the specificity of immunostaining. Blank sections were processed in parallel for all immunohistochemical reactions and animals.

Confocal Imaging

Labelled PhMNs in the cervical spinal cord sections were visualized usng an Olympus FV2000 laser confocal microscope (Olympus Life Sciences Solutions, Waltham, MA) equipped with argon (488 nm), green HeNe (543 nm) and red HeNe (633 nm) lasers.). Three-dimensional imaging techniques have been previously reported (Prakash et al., 1993, 1994; Sieck et al., 1999; Prakash et al., 2000; Zhan et al., 2000; Issa et al., 2010; Fogarty et al., 2018; Rana et al., 2019b). All images were acquired at 12-bit resolution in an array of 1024×1024 pixels using a 60x Plan Apo (NA 1.40, WD 0.15 mm) oil immersion objective lens with a step size of 0.5 μm (voxel dimensions: 0.207 × 0.207 × 0. 5 μm). Sequential two-channel imaging was performed using a dichroic mirror beam splitter that allows transmission of 555–615 nm (reflects 480–555 and 615800 nm) and appropriate band pass emission filters (495–535 nm and 575–630 nm long pass for Alexa 488 and 594 respectively). Since image analysis is highly dependent on image quality, several steps were taken to ensure images were of optimal quality. To reduce cross-talk, laser illumination was done sequentially for the imaging of each optical slice. Confocal aperture was set optimally for the 60x lens using the empirically-calculated point spread function to set the step size, as previously reported (Prakash et al., 1993, 1994; Sieck et al., 1999). Laser intensity and photomultiplier settings were adjusted to maintain black level of the blank sections (no primary antibody) to less than 10% of the dynamic range and to prevent saturation. All PhMNs were visualized in the rostro caudal direction using Fluoview software version 5.0 (Olympus). Only motor neurons with visible nuclei that were not cut during spinal cord sectioning were sampled for synaptic distribution analyses. Registration was verified empirically using multiwavelength micro-bead calibration techniques similar to previous reports (Prakash et al., 1993, 1994; Sieck et al., 1999; Prakash et al., 2000; Zhan et al., 2000).

Quantitative Analyses

A total of 480 PhMNs (n=240 and n=240 PhMNs from 6 animals in control and SH7D group each) were analyzed from both ipsilateral and contralateral sides (20 PhMNs each side per animal) of the PhMN pool. Confocal image stacks were analyzed using NIS Elements AR Imaging Software Ver 5.00 (Nikon Instruments Inc., Melville, NY). The procedures for PhMN somal volume measurements have been described previously (Prakash et al., 1993, 1994; Prakash et al., 2000; Issa et al., 2010; Fogarty et al., 2018; Rana et al., 2019b). Somal surface areas of PhMNs were estimated by measuring major and minor diameters and assuming a prolate spheroid shape.

To create a 2.5 μm shell around CTB labelled PhMNs and then perform semi-automated processing of confocal stacks to isolate VGLUT terminals within this shell, we employed a previously described macro program developed on the NIS elements platform (Rana et al., 2019b). Briefly, an edge detection-based threshold algorithm was used to construct a 2.5 μm shell around PhMNs and VGLUT positive terminals were isolated within this shell. Images were processed using the deconvolution module in the NIS-Elements AR software package (Nikon). A blind deconvolution algorithm (Point Scan Confocal, 3 iterations) was used in which the most probable combination of point spread function is estimated for the given image and obviates the need to obtain a point spread function for our optical system for each image acquisition session. Following deconvolution, the three-dimensional object measurement plugin in the NIS-Elements AR software package was utilized to identify VGLUT terminals at the PhMN. Based on previously reported size of synapses at cervical motor neurons (putative PhMNs) in studies utilizing EM (Goshgarian & Rafols, 1984; Ellenberger et al., 1990; Tai & Goshgarian, 1996) VGLUT terminals smaller than 1 μm³ were excluded and the maximum size of a synapse was set to be 6 μm³. VGLUT terminals larger than 6 μm³ were counted as multiple terminals (e.g., >612 μm³ counted as 2 terminals; >12–18 μm³ counted as 3 terminals, etc.). VGLUT Terminal density was determined by normalizing counts to the shell volume of each motor neuron. Total number of VGLUT terminals at the soma were estimated using VGLUT terminal density and estimated surface area for that PhMN.

Images were converted to 8-bit resolution in NIS Elements for presentation only. No thresholding or post-imaging processing was applied for any images. Each channel was pseudocolored by changing the color gamut (RGB). Only brightness and contrast levels were adjusted linearly if needed to facilitate presentation of multiple colors.

Statistical Analysis

All statistical evaluations were performed using JMP statistical software (version 14.0, SAS Institute Inc., Cary, NC). The study statistical design was powered to consider both the within and across subject variance. Power analysis considering within subject variance in VGLUT terminals was used in order to detect a 15% difference at p<0.05 across PhMNs of varying somal surface areas. Accordingly, we stereologically sampled 20 PhMNs per side per animal, which provided more than sufficient power to test our hypothesis that there is a greater loss of VGLUT terminals at smaller compared to larger PhMNs following C2SH. We also powered the study to consider across subject variance in order to detect a 15% difference at p<0.05 across animals in control vs. C2SH group (n=6 animals per group). For measurements repeated for the same animal (e.g., VGLUT terminal density and total number of estimated VGLUT terminals), our model included PhMN somal surface area, group and side as model variables. Variance across animals within a group was considered as a random effect. An additional model was generated after grouping PhMNs into tertiles according size for each animal. All experimental data are presented as means ± standard deviation (SD), unless otherwise specified. Using the mixed linear model, statistical significance was established at the 0.05 level and adjusted for any violation of the assumption of sphericity in repeated measures using the Greenhouse-Geisser correction. When appropriate, post hoc analyses were conducted using Tukey-Kramer honestly significant difference (HSD).

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Results

All SH animals had successful implantation of DIAm electrodes for monitoring of EMG activity (Fig. 1). As in previous studies, DIAm EMG was recorded during C2SH surgery and at three days post C2SH (SH3D) to confirm completeness of C2SH by the absence of ipsilateral inspiratory-related DIAm EMG activity during eupnea in anesthetized animals (Gransee et al., 2013; Mantilla et al., 2013a; Mantilla et al., 2013b; Gransee et al., 2015; Martinez-Galvez et al., 2016). Based on the presence of ipsilateral inspiratory-related DIAm EMG activity during eupnea at SH3D, one animal was excluded from further analysis and terminated resulting in a final n=6 for this group. Representative DIAm EMG tracings are shown in Figure 2. Compared to control (n=6) rats, all SH7D (n=6) animals displayed altered movements of the rib cage and abdomen resulting from DIAm hemiparalysis (Miyata et al., 1995; Prakash et al., 1999).

Figure 1. Experimental Timeline.

Experimental timeline of study. In the 7DSH group, in-dwelling EMG electrodes and intramuscular injection of CTB was performed 4 days prior to C2SH injury. Pre-injury EMG baseline was collected 1 day prior to C2SH injury. At SH3D, completeness of C2SH injury was verified by recording EMG activity during eupneic breathing. At the terminal timepoint, SH7D, DIAm EMG activity was recorded, followed by collection of spinal tissue for histology. In the control group, CTB was injected into the DIAm 11 days prior to the terminal timepoint. At the terminal timepoint, spinal tissue was collected for histology.

Figure 2. Representative EMG recordings.

Representative ipsilateral DIAm EMG recordings and corresponding root-mean-square (RMS) calculations during eupnea, hypoxia-hypercapnia (10% O₂, 5% CO₂), spontaneous deep breath (sigh), and airway occlusion from a SH7D animal before and after injury. Chronic DIAm EMG recordings were obtained over a 10-day period for pre- and post-injury (SH7D) comparisons. Note that ipsilateral inspiratory-related DIAm EMG activity disappeared post-injury during eupnea and hypoxia-hypercapnia, while EMG activity persisted during airway occlusion and sighs.

Retrograde Labelling of PhMNs

Bilateral injection of CTB into the DIAm resulted in robust retrograde labelling of PhMNs from C₃ to C₅ bilaterally (Fig 3A). Consistent with previous reports, PhMNs were clustered and oriented in a rostro-caudal direction in the ventral horn of the cervical spinal cord. Near complete labelling of PhMN somata and most primary dendrites was evident; however, labelling of more distal dendrites was inconsistent. There was no apparent bias in rostral or caudal CTB labelling of PhMNs. Additionally, there were no discernable rostral to caudal differences in the size of PhMNs.

Figure 3. Retrogradely-labelled PhMN pool and PhMN somal surface area.

A. Representative image of retrogradely labelled PhMNs on both ipsilateral and contralateral sides of the cervical spinal cord for a SH7D animal. Images were obtained from maximum intensity projections of 70 μm thick longitudinal spinal cord sections. Note disruption of white and grey matter regions of spinal cord in rostral section of the right spinal cord (depicted by *). Scale bar represents 1 mm. B. The somal surface areas of retrogradely labelled PhMNs were separated into tertiles for each animal (different color symbols represent values for n=6 rats per control and SH7D groups with 20 PhMNs sampled per side per animal – total of 240 PhMNs per group). Box plots of PhMN somal surface area represent variability across PhMNs within an animal. For statistical analysis, a mixed linear model with tertiles based on PhMN somal surface area and group as model variables and variance across animals within a group considered as a random effect was employed

Somal surface areas of PhMNs were calculated based on measurements of the long and short axis at the mid-nuclear section, assuming a prolate spheroid shape (Prakash et al., 1993, 1994). The size and distribution of sampled PhMNs (n=480 PhMNs from 12 animals) were consistent with previous reports and reflected the overall PhMN pool (Prakash et al., 1994; Prakash et al., 2000; Fogarty et al., 2018; Rana et al., 2019b). There were no statistically significant differences in PhMN somal surface areas across animals (one-way ANOVA; F_(11,479)=1.07; p=0.38), with mean somal surface area in each animal ranging from 2,882 to 3,267 μm². Table 1 summarizes PhMN size parameters across groups (control, SH7D) and side (ipsilateral, contralateral).

Table 1. PhMN somal dimensions.

Dimensions of PhMN soma did not differ between control and SH7D groups (n = 120 PhMNs per group, 6 animals in control and SH7D group each). Comparisons used a mixed linear model with animal as a random effect (see materials and methods for details). Data shown are mean ± SD. There was no significant effect of group or side on any dimension (p > 0.12).

Group	Side	Long axis	Short axis	PhMN Volume (x1000 μm3)	PhMN Surface Area (x1000 μm2)
Control	Contralateral	44.69 ± 10.01	24.65 ± 4.39	14.58 ± 6.02	3.02 ± 0.80
Control	Ipsilateral	45.52 ± 11.76	25.02 ± 4.74	15.36 ± 6.42	3.12 ± 0.88
SH7D	Contralateral	43.43 ± 10.73	25.65 ± 4.37	15.53 ± 6.63	3.11 ± 0.91
SH7D	Ipsilateral	44.23 ± 11.76	24.70 ± 5.22	14.64 ± 6.83	3.01 ± 0.92

PhMNs were grouped into tertiles based on somal surface area for each animal in order to segregate the assessment of pre-synaptic glutamatergic innervation (Fig 3B). The tertile grouping is expected to reflect proportions of the different motor unit types that would be recruited to accomplish a range of motor behaviors (Sieck & Fournier, 1989; Mantilla et al., 2010; Khurram et al., 2019). Based on recruitment models from our group, PhMNs in the lower tertile would comprise motor units that are activated during lower force ventilatory behaviors of the DIAm, followed by the middle tertile of PhMNs that are activated to accomplish higher force ventilatory behaviors such as deep breaths (sighs) or breathing efforts against an occluded airway. The upper tertile of PhMN somal surface areas likely comprise DIAm motor units that are only recruited to accomplish near maximal forces associated with expulsive/ straining behaviors (cough, sneeze and defecation). The mean PhMN somal surface areas for the tertiles were: Lower: 2,153 ± 406 μm² (n=161), Middle: 2,996 ± 316 μm² (n=158) and Upper: 4,044 ± 496 μm² (n=161). There was no effect of group (F_(1,10.1)=0.35; p=0.57) and side (F_(1,460)=0.73; p=0.39) on PhMN somal surface area, or interaction between group*side (F_(1,460)=0.55; p=0.46) and group*side*PhMN tertile (F_(1,462.7)=0.11; p=0.89).

VGLUT Terminals at PhMNs Following C2SH

VGLUT immunoreactivity and distribution was consistent with previous reports in the spinal cord (Alvarez et al., 2004; Herzog et al., 2004; Issa et al., 2010; Rana et al., 2019b). VGLUT positive terminals had bouton like morphology and were present around soma and dendrites of retrogradely labelled PhMNs in high abundance. VGLUT terminals were subsequently isolated within a 2.5 μm distance from the PhMN somal membrane (Fig. 4). Morphometrics of VGLUT terminals are represented in Table 2. These three-dimensional measurements are consistent with the size of synaptic terminals previously identified using EM techniques (Goshgarian & Rafols, 1984; Ellenberger et al., 1990; Tai & Goshgarian, 1996). Reduction in VGLUT terminals in SH7D group likely resulted in reduced clustering of terminals following injury. Therefore, we compared changes in morphometrics of individually identified VGLUT terminals that fell within the range of 1μm³ and 6 μm³ that would reduce bias for clustered terminals in the control group. VGLUT terminal volume and EqDiameter was compared across animal group and side (Fig. 5). For VGLUT terminal volume, there was a significant interaction between group and side (F_1,29125=23.64; p<0.001) and a significant effect of group (F_1,10.95=5.78; p=0.035) and side (F_1,29125=13.31; p<0.001). Isolated VGLUT terminals in the ipsilateral SH7D group were ~5% smaller as compared to ipsilateral control group. For VGLUT terminal EqDiameter, there was a significant interaction between group and side (F_1,29202=21.30; p<0.001) and a significant effect of side (F_1,29202=10.07; p=0.002). However, there was no effect of group on VGLUT EqDiameter (F_1,10.76=4.77; p=0.052).

Figure 4. Three-dimensional of VGLUT terminals isolated at labelled PhMNs.

A. VGLUT immunoreactivity (red pseudo-colored) surrounding a retrogradely labelled (intramuscular injection of CTB) PhMN (grey) in a rat spinal cord. B. Using a three-dimensional workflow described in detail in methods section, VGLUT terminals (red) were isolated within a 2.5 μm wide shell around the somal surface of PhMNs labelled with CTB.

Table 2. VGLUT terminal morphometry in control and SH7D animals.

VGLUT terminal dimensions are summarized for control and SDH animals (n=6 rats per group and 20 PhMNs per side, per animal). A total of 9,710 ipsilateral and 10,074 contralateral VGLUT terminals were analyzed in control rats, and a total of 6,295 ipsilateral and 10,235 contralateral in SH7D animals. Equivalent diameter (EqDia; diameter of a sphere with the same volume for each identified VGLUT terminal); Elongation (ratio of MaxFeret and MinFeret); Sphericity (ratio of surface area of the VGLUT object and that of a sphere with the same volume). Data shown are mean ± SD. Comparisons used a mixed linear model with animal as a random effect

Group	Side	Volume (μm3)	EqDia (μm)	Major Axis (μm)	Minor Axis(μm)	Sphericity	Elongation
Control	Contralateral	4.40 ± 6.03*	1.84 ± 0.57*	3.66 ± 1.89	1.88 ± 0.90	0.61 ± 0.11*	2.51 ± 1.03
Control	Ipsilateral	4.08 ± 5.23	1.81 ± 0.54	3.57 ± 1.76	1.87 ± 0.87	0.61 ± 0.11	2.45 ± 0.93
SH7D	Contralateral	4.35 ± 6.26	1.84 ± 0.57	3.67 ± 1.92	1.87 ± 0.90	0.61 ± 0.10	2.52 ± 0.97
SH7D	Ipsilateral	3.19 ± 3.30*†‡	1.71 ± 0.43*†‡	3.26 ± 1.45*†‡	1.69 ± 0.65*†‡	0.63 ± 0.09‡	2.48 ± 0.94

^*,significantly different from the ipsilateral control group

^†,different from the contralateral control group

^‡different from the contralateral SH7D group; post hoc Tukey-Kramer HSD, p<0.05.

Figure 5. VGLUT terminal volume following C2SH.

A. Quantile and violin plots (probability density) of the volume of identified VGLUT terminals with volume <6 μm³ (defined as a single synapse) at labelled PhMNs from control and SH7D animals (n=6 rats per group with 20 PhMNs sampled per side, per animal – total of 240 PhMNs per group). All identified VGLUT terminals at PhMNs were analyzed (Ipsilateral: Control 8,080 and SH7D 5,595 VGLUT terminals; Contralateral: Control 8,179 and SH7D 8,372 VGLUT terminals). Note the reduction in the volume of individual VGLUT terminals defined as a single synapse on the ipsilateral side in SH7D animals. This likely reflects a reduction in gross VGLUT terminal size in addition to loss of individual VGLUT terminals ipsilateral to injury at SH7D. (*significantly different from the ipsilateral control group; post hoc Tukey-Kramer HSD, p<0.05). For statistical analysis, a mixed linear model with tertiles based on side and group as model variables and variance across animals within a group considered as a random effect was employed

Disproportionate Loss in VGLUT Terminal Density Across PhMNs Following C2SH

VGLUT terminals isolated within a 2.5μm distance from the PhMN somal membrane were quantified bilaterally in control (n=6 animals (120 PhMNs ipsilateral, 120 PhMNs contralateral) and SH7D animals (n=6 animals (120 PhMNs ipsilateral, 120 PhMNs contralateral)). Representative three-dimensional reconstructed images of VGLUT terminals around ipsilateral small and large PhMNs in control and SH7D groups are depicted in Fig 6. Similar to a previous study by our group, there was an effect of PhMN somal surface area on VGLUT terminal density when grouped based on somal surface area (F_(2,232.1)=3.60; p=0.03). Post hoc analysis revealed a ~10% higher VGLUT terminal density at smaller PhMNs compared to larger PhMNs in uninjured rats.

Figure 6. Representative distribution of VGLUT terminals at ipsilateral small and large PhMNs.

VGLUT terminal (red pseudo-colored) were isolated within a 2.5 μm shell around labelled PhMNs (grey) from control and SH7D rats. Note that for a representative smaller PhMN (in the lower tertile of somal surface areas) in a SH7D rat, there were ~60% fewer VGLUT terminals as compared to a representative smaller PhMN from a control animal. In contrast, the number of VGLUT terminals at a larger PhMN (in the upper tertile of somal surface areas) in a SH7D animal was reduced to a much lesser extent (~35% fewer) as compared to a representative control.

VGLUT terminal densities were subsequently compared across somal surface areas of PhMNs and between ipsilateral and contralateral sides using a mixed linear model, with group as a fixed effect and animal as a random effect. Ipsilateral to injury, there was a significant effect of PhMN somal surface area (F_(1,228.6)=13.42; p<0.001) and group (F_(1,10.02)=168.31; p<0.001) on VGLUT terminal density (Fig. 7A). There was also a significant interaction between PhMN somal surface area and group (F_(1,228.6)=40.55; p<0.001). VGLUT terminal density was also compared across PhMN tertiles based on somal surface area (Fig. 7B). There was a significant effect of PhMN tertile (F_(2,225.6)=7.71; p<0.001) and group (F_(1,10.07)=153.89; p<0.001) on VGLUT terminal density, as well as a significant interaction between PhMN tertile and group (F_(2,225.6)=17.44; p<0.001). Post hoc analysis revealed statistically significant pairwise differences in the VGLUT terminal density between SH7D and control groups. There was a greater loss in VGLUT terminal density at PhMNs in the lower (~58%) and middle (~53%) tertiles as compared to the upper tertile (~35%).

Figure 7. VGLUT terminal density across ipsilateral PhMNs.

A. Scatterplot of VGLUT terminal density vs. PhMN somal surface area on the ipsilateral side in control (blue) and SH7D (red) groups (n=6 rats per group and 20 PhMNs per animal). VGLUT terminal density is negatively correlated with PhMN somal surface area (r²=0.11; 0.075 – 1.7 e⁻³ * PhMN Somal Surface Area), in the control group and positively correlated with PhMN somal surface area (r² = 0.49; 0.016 + 6.3 e⁻³ * PhMN Somal Surface Area) in SH7D group. For statistical analysis, a mixed linear model with PhMN somal surface area and group as model variables and variance across animals within a group considered as a random effect was employed. Post hoc analysis revealed statistically significant pairwise differences in the VGLUT terminal density between SH7D and control groups. B. Mean (± 95% CI) of VGLUT terminal density across individual PhMNs grouped by tertile of somal surface area for each animal. For statistical analysis, a mixed linear model with tertiles based on PhMN somal surface area and group as model variables and variance across animals within a group considered as a random effect was employed (*, significantly different from same tertile in control group; †, different from lower and middle tertile in SH7D group; post hoc Tukey-Kramer HSD, p<0.05).

Contralateral to C2SH, there was a significant effect of PhMN somal surface area (F_(1,228.9)=18.82; p<0.001) and group (F_(1,9.94)=10.31; p=0.01) on VGLUT terminal density, however, there was no interaction between PhMN somal surface area and group (p=0.55) (Fig. 8A). VGLUT terminal density was also compared across PhMN tertiles based on somal surface area (Fig. 8B). There was a significant effect of PhMN tertile (F_(2,225.9)=10.92; p<0.001) and group (F_(1,10.11)=11.11; p = 0.008) on VGLUT terminal density, but no interaction between PhMN tertile and group (p = 0.45).

Figure 8. VGLUT terminal density across contralateral PhMNs.

A. Scatterplot of VGLUT terminal density vs. PhMN somal surface area on the contralateral side in control (blue) and SH7D (red) groups (n=6 rats per group and 20 PhMNs per animal). VGLUT terminal density is negatively correlated with PhMN somal surface area in control (r²=0.11; 0.081 – 2.8 e⁻³ * PhMN Somal Surface Area), and in SH7D (r² = 0.30; 0.074 – 4.0 e³ * PhMN Somal Surface Area) groups. For statistical analysis, a mixed linear model with PhMN somal surface area and group as model variables and variance across animals within a group considered as a random effect was employed. B. Mean (black diamond, ± 95% CI) of VGLUT terminal density across individual PhMNs grouped by tertile of somal surface area for each animal. For statistical analysis, a mixed linear model with tertiles based on PhMN somal surface area and group as model variables and variance across animals within a group considered as a random effect was employed.

Number of VGLUT Terminals Vary Across PhMNs Based on Size

Table 3 summarizes the loss in VGLUT terminal numbers by PhMN tertile in SH7D group as compared to respective PhMN tertile in control group. Ipsilateral to C2SH, there was a significant effect of PhMN tertile (F_(2,225.8)=196.26; p<0.001) and group (F_(1,10.09)=82.05; p<0.001) on VGLUT terminal number, but no interaction between PhMN tertile and group (p=0.16) (Fig. 9). Post hoc analysis revealed an ~50% reduction in VGLUT terminal number ipsilateral to injury in the SH7D group as compared to the control group. Contralateral to injury, there was a significant effect of PhMN tertile (F_(2,225.3)=119.14; p<0.001) and group (F_(1,10.08)=6.56; p=0.03) on VGLUT terminal number, but no interaction between PhMN tertile and group (p=0.32). Post hoc analysis revealed an ~18% reduction in VGLUT terminal across PhMNs on the contralateral side to injury in SH7D group as compared to control group.

Table 3. Percent loss of VGLUT terminals at SH7D by PhMN tertile.

Percent loss in the number of VGLUT terminals at PhMNs in SH7D relative to control animals (n = 120 PhMNs per group, 6 animals in control and SH7D group each). PhMNs analyzed by ipsilateral vs. contralateral side to injury and grouped by tertiles of somal surface area.

Side	Lower	Middle	Upper
Ipsilateral	58.57%	54.20%	34.85%
Contralateral	15.86%	14.91%	17.45%

Figure 9. Total number of VGLUT terminals across PhMNs.

A. Quantile plot of total number of VGLUT terminals across ipsilateral PhMNs grouped by tertile of somal surface area for control (blue) and SH7D (red) groups (n=6 rats per group and 20 PhMNs per animal). B. Quantile plot of total number of VGLUT terminals across contralateral PhMNs grouped by tertile of somal surface area for control (blue) and SH7D (red) (n=6 rats per group and 20 PhMNs per animal).

Discussion

The present study examined changes in excitatory glutamatergic innervation at PhMNs of varying size following C2SH that disrupts descending excitatory bulbospinal inspiratory-related drive. Our results show that glutamatergic inputs onto smaller PhMNs were markedly reduced after C2SH at 7D (~60%), consistent with removal of predominant ipsilateral excitatory input mediating inspiratory-related drive, as reflected by the silencing of inspiratory-related ipsilateral DIAm EMG activity. In contrast, after C2SH, glutamatergic input at larger ipsilateral PhMNs was reduced to a much lesser extent (~35%), consistent with the relative sparing of higher force behaviors of the DIAm after C2SH. Our data suggest that there is a dichotomy in the distribution of inspiratory-related descending excitatory glutamatergic input to small vs. large PhMNs that reflects their differential recruitment in lower force vs. higher force inspiratory and nonventilatory (e.g., coughing, sneezing) behaviors.

PhMN Morphology

Robust labelling of PhMNs was observed from C₃ to C₅ segments of the rat spinal cord, consistent with previous reports (Goshgarian & Rafols, 1981; Furicchia & Goshgarian, 1987; Prakash et al., 1993; Prakash et al., 2000; Issa et al., 2010; Mantilla et al., 2018). PhMN somal surface areas displayed a unimodal distribution consistent with previous reports from our group (Prakash et al., 2000; Mantilla et al., 2018; Rana et al., 2019a; Rana et al., 2019b). Motor neurons innervating other limb skeletal muscles typically display a bimodal distribution reflecting a population of smaller cell bodies associated with gamma motor neurons and a second population of larger cell bodies associated with alpha motor neurons (Burke et al., 1977). However, in the rat DIAm there are few if any muscle spindles, and thus there are few if any smaller gamma motor neurons, consistent with the absence of a distinct bimodal distribution in PhMN size (Barstad et al., 1965; Holt et al., 1991; Sieck & Gransee, 2012). Also consistent with previous studies from our lab (Prakash et al., 2000; Mantilla et al., 2018), we observed that some PhMN had dendrites traversing the midline to the contralateral side. The importance of bilateral dendritic branching of PhMNs remains unclear.

Recently, we reported PhMN morphological adaptations in response to C2SH as reflected by a marked reduction in PhMN somal volume at 14 days post C2SH (Prakash et al., 2000; Mantilla et al., 2018). We suggested that this adaptation at 14 days post C2SH is consistent with an increase in intrinsic size-dependent PhMN excitability. In the current study, we did not observe an overall reduction in PhMN somal dimensions at 7 days after C2SH. It is plausible that there is a time course to such morphological changes, and that gross structural adaptations are not apparent at SH7D.

Loss in VGLUT Terminals following C2SH

Little is known about the ultrastructural changes in excitatory synapses at PhMNs following an incomplete spinal cord injury that leaves PhMNs partially denervated. However, it is likely that by disrupting descending premotor neuron axons, there is stripping of synapses at PhMNs. In the present study, three-dimensional analysis revealed a marked reduction in VGLUT terminal density at PhMNs following C2SH. Previously, Goshgarian et al. assessed changes in glutamatergic synapses at putative PhMNs using electron microscopy (Tai & Goshgarian, 1996). Although EM studies are considerably limited by their sample size, the study by Tai and Goshgarian reported an ~40% loss in excitatory glutamate terminals at putative PhMNs at 30 days post C2SH. In contrast, we observed an ~50% loss in VGLUT terminal density at ipsilateral PhMNs at SH7D. Following axotomy, the distal segment of an axon undergoes Wallerian degeneration that typically occurs in the first 24–48 h (Mack et al., 2001). This includes synaptic disruption that ensues after the initial degenerative period and is completed by ~72–96 h post injury (Wang et al., 2012). Thus, by 7 days post C2SH, there should be a complete loss of VGLUT terminals associated with the disrupted descending excitatory inputs. The apparent reduced extent of VGLUT terminal loss at putative PhMNs by 30 days post C2SH reported in the study by Tai and Goshgarian (Tai & Goshgarian, 1996) most likely reflects the inclusion of cervical motor neurons other than PhMNs in their sample. In addition, it is possible that there was some glutamatergic re-innervation of the PhMN pool by 30 days post C2SH.

Changes in synaptic architecture at putative PhMNs have been reported at 30 days post C2SH and includes an increased number of double synapses, an increase in the number of active synaptic zones and an increase in the mean length of active zones at glutamatergic synapses (Tai & Goshgarian, 1996). In the present study, we observed that for PhMNs ipsilateral to C2SH there was an ~6% reduction in VGLUT terminal EqDiameter and a significant increase in VGLUT terminal sphericity. These acute changes in the morphometry of VGLUT terminals after C2SH would be indicative of a loss of larger synapses with the remaining population of VGLUT terminals being smaller and rounder in shape.

PhMN Size Dependent Loss of VGLUT Terminals

Previously, we found that C2SH results in the silencing of lower force inspiratory-related DIAm EMG activity, while DIAm EMG activity associated with higher force inspiratory (e.g., sighs and efforts against an occluded airway) and non-ventilatory (e.g., cough and sneeze) behaviors were spared (Martinez-Galvez et al., 2016; Hernandez-Torres et al., 2017). Based on the size principle for motor unit recruitment, lower force inspiratory-related DIAm activity is accomplished by the recruitment of smaller PhMNs (Fournier & Sieck, 1988; Sieck, 1988; Sieck & Fournier, 1989; Sieck, 1994; Mantilla et al., 2010; Mantilla & Sieck, 2011; Sieck et al., 2013; Seven et al., 2014). In agreement, we observed that after C2SH, there was a greater decrease in VGLUT terminal density at smaller PhMNs. When PhMNs were grouped in tertiles of somal surface areas that would reflect their corresponding size-dependent recruitment during different ventilatory and non-ventilatory behaviors, we observed a ~60% reduction in VGLUT innervation at PhMNs in lower tertile. These smaller PhMNs would be recruited to accomplish lower force inspiratory efforts of the DIAm that are silenced by C2SH. In contrast, C2SH resulted in only an ~35% reduction in VGLUT terminal density at larger PhMNs in the upper tertile of somal surface areas, consistent with the sparing of higher force behaviors of the DIAm. These results indicate that in addition to size-dependent differences in the intrinsic properties of PhMNs (i.e., the size principle), size dependent differences in excitatory glutamatergic input to PhMNs may contribute to appropriate DIAm motor unit recruitment. Inspiratory-related drive via bulbospinal premotor input is primarily ipsilateral and directed to smaller PhMNs.

Source of Glutamatergic Innervation of Phrenic Motor Neurons

The origin and organization of excitatory synaptic inputs to PhMNs has been previously mapped using a number of approaches including, but not limited to, neuroanatomical studies with tracers, lesions, antidromic stimulations as well as correlational studies using intra- and extracellular recordings from PhMNs and afferent input sources (Lee & Fuller, 2011; Ghali, 2017). Unfortunately, none of these approaches distinguished the source of inputs to PhMNs of varying size. Monosynaptic descending excitatory inputs from the rostral ventral respiratory group (rVRG) constitute the most prominent bulbospinal input to PhMNs (Ellenberger & Feldman, 1988; Ellenberger et al., 1990; Dobbins & Feldman, 1994; Lipski et al., 1994; Tian & Duffin, 1996a; Tian & Duffin, 1996b), and it is this rVRG input that was most likely disrupted by C2SH. Our results do not support the general view that the rVRG input comprises a broadcast system that sends distributed projections to the entire PhMN pool. Unfortunately, in previous studies exploring bulbospinal inputs from the rVRG, PhMN size and DIAm motor unit type were not considered. However, it is likely that the rVRG sends ipsilateral input that is broadly distributed to smaller PhMNs to achieve an orderly size-dependent recruitment of DIAm motor units during inspiratory efforts (Fournier & Sieck, 1988; Sieck, 1988; Sieck & Fournier, 1989; Sieck, 1994; Mantilla et al., 2010; Mantilla & Sieck, 2011; Sieck et al., 2013; Seven et al., 2014).

In addition to excitatory inputs from the rVRG, a complex network of interneurons throughout the spinal cord send excitatory inputs to PhMNs and can modulate DIAm activity (see reviews (Jensen et al., 2019; Sunshine et al., 2020)). There exist anatomical projections from interneurons in upper cervical segments to PhMNs (Dobbins & Feldman, 1994; Lane et al., 2008a; Lane et al., 2008b), constituting a premotor interneuron population that does not mediate inspiratory drive but may affect PhMN excitability (Lu et al., 2004; Gonzalez-Rothi et al., 2017; Streeter et al., 2017). Following spinal cord injury, an increase in the excitatory input from these spinal interneurons may be important in recovery (Jensen et al., 2019). However, size-dependent differences in the excitatory input of spinal interneurons to PhMNs has not been examined. Interestingly, in zebra fish, it has been suggested that different populations of spinal V2a interneurons exhibit differential synaptic connectivity to ‘slow’ or ‘fast’ motor neurons in order to facilitate an additional layer of motor control for varying locomotor speeds (Ampatzis et al., 2014; Song et al., 2018). Such selectivity in the respiratory system remains elusive and would be a worth investigating.

A population of ascending propriospinal excitatory inputs from intercostal muscle proprioceptors to PhMNs also exists, and these excitatory inputs comprise the afferent limb of the intercostal to phrenic reflex (Decima et al., 1969; Remmers, 1973; Lane et al., 2008b). These propriospinal inputs are likely to be widely distributed to smaller PhMNs, since intercostal afferent stimulation induces size dependent inspiratory-related activation (or inactivation) of PhMNs (Decima et al., 1969). Together, these results indicate that ascending propriospinal inputs generally affect the overall excitability of PhMNs, thereby enhancing or inhibiting inspiratory-related bulbospinal drive. The involvement of propriospinal input in modulating excitability of larger PhMNs involved in higher force motor behaviors of the DIAm is unclear.

Organization of Descending Bulbospinal Pathways

The predominant ipsilateral nature of descending inspiratory excitatory drive to PhMNs is supported by the cessation of DIAm EMG activity during eupnea following C2SH (Miyata et al., 1995; Vinit et al., 2006; Gransee et al., 2013; Mantilla et al., 2014; Gransee et al., 2017). However, higher force DIAm behaviors such as deep sighs and inspiratory efforts against an occluded airway are relatively unimpaired following a C2SH (Martinez-Galvez et al., 2016; Hernandez-Torres et al., 2017). These changes in DIAm activity are consistent with the observation that C2SH disproportionately eliminates glutamatergic input to smaller PhMNs while larger PhMNs are far less impacted.

Another important aspect to consider is the contribution of contralateral bulbpospinal input to PhMN activation. Porter (Porter, 1895) provided the first clear demonstration that an extant contralateral descending excitatory input to PhMNs could be activated following C2SH, thereby restoring inspiratory-related DIAm activity. In this study, Porter performed a contralateral phrenicotomy in dogs after C2SH, and showed that inspiratory-related DIAm activity reappeared. This reactivation of the DIAm, which is termed the crossed-phrenic phenomenon is thought to be mediated by activation of extant contralateral inputs to the PhMNs that were previously sub-threshold for inspiratory-related recruitment. Subsequently, it has been shown that any augmentation of inspiratory drive allows for the unmasking of latent contralateral input to PhMNs after C2SH and reactivation of inspiratory-related DIAm activity. Many anatomical studies have characterized the laterality of descending bulbospinal projections to PhMNs using anatomical tracing techniques. (Feldman et al., 1985; Ellenberger et al., 1990; Goshgarian et al., 1991; Boulenguez et al., 2007; Vinit et al., 2007). Consistent with the crossed-phrenic phenomenon and the contribution of a contralateral descending inspiratory-related input to PhMNs, we observed that DIAm EMG activity during higher force inspiratory efforts (e.g., during sighs and efforts against an occluded airway) persists after C2SH (Martinez-Galvez et al., 2016; Hernandez-Torres et al., 2017). It is likely that these higher force inspiratory-related efforts reflect descending excitatory input via spared contralateral inputs. The current study shows an ~60% overall reduction in presynaptic terminal density at smaller ipsilateral PhMNs and an ~16% reduction on the contralateral side. This provides substantial evidence that C2SH results in a significant loss of ipsilateral inspiratory-drive to smaller PhMNs resulting in the cessation of eupneic DIAm activity. However, there is sufficient sparing of contralateral descending inspiratory-drive input to smaller PhMNs to promote recovery of breathing especially if PhMN excitability is enhanced. It is likely that ~25% (at smaller PhMNs) to ~50% (at larger PhMNs) of excitatory glutamatergic input to PhMNs derives from sources other than the rVRG (e.g., cervical or thoracic premotor neurons). This premotor excitatory input does not provide inspiratory-related drive but does modulate PhMN excitability. It appears that this noninspiratory related excitatory input is preferentially distributed to larger PhMNs.

Figure 10. Conceptual Framework.

The primary excitatory drive to the phrenic pool is ipsilateral. In the current model, we propose a distributed neural drive to small vs. large PhMNs comprising low threshold type S and FR motor units vs. high threshold type FInt and FF motor units respectively. Following a C2SH there is a ~60% reduction in glutamatergic innervation to small and medium sized PhMNs (thick line), suggesting a loss in predominant excitatory innervation from the brainstem. Contrastingly, there is a ~35% loss in glutamatergic innervation at larger PhMNs (thin line), suggesting ipsilateral innervation from supra-spinal pre-motor centers is not the predominant source of excitatory inputs. In addition, there is a ~15% loss in glutamatergic innervation across PhMNs, regardless of size on the contralateral side of the spinal cord. Alternate sources of glutamatergic innervation to PhMNs depicted in the figure include excitatory inputs from spinal interneurons (black) which could be ipsilateral or bilateral in nature and present above and below the phrenic pool. It is likely that ~25% (at smaller PhMNs) to ~50% (at larger PhMNs) of excitatory glutamatergic input to PhMNs derives from sources other than the rVRG (e.g., cervical or thoracic interneurons, intercostal to phrenic reflex, abdominal propriospinal inputs).

Key Points.

• Motor units, comprising a motor neuron and the muscle fiber it innervates, are activated in an orderly fashion to provide varying amounts of force.

• A unilateral C₂ spinal hemisection (C2SH) disrupts predominant excitatory input from medulla, causing cessation of inspiratory-related diaphragm muscle activity, whereas higher force, non-ventilatory diaphragm activity persists.

• In this study, we evidence a disproportionately larger loss of excitatory glutamatergic innervation to small phrenic motor neurons following C2SH, as compared to large phrenic motor neurons ipsilateral to injury.

• Our data suggest that there is a dichotomy in the distribution of inspiratory-related descending excitatory glutamatergic input to small vs. large PhMNs that reflects their differential recruitment.

Abbreviations:

VGLUT

Vesicular Glutamate Transporter

C2SH

C₂ spinal hemisection

SH7D

Seven days post-C2SH

DIAm

Diaphragm muscle

PhMN

Phrenic motor neuron

CTB

Cholera toxin subunit β

Biography

Dr. Sabhya Rana received her B.S. from the University of California, Irvine in 2011 and her PhD in Neuroscience from the Mayo Clinic under the mentorship of Dr. Carlos B. Mantilla and Dr. Gary C. Sieck (2018). Her thesis projects focused on the neuromotor control of breathing, impact of mid-cervical spinal cord injuries on diaphragm muscle function and mechanisms of neuroplasticity at the motor neuron level. Dr. Rana is now a postdoctoral fellow with Dr. David D. Fuller at the University of Florida where she is studying the role of ampakines in restoring respiratory function following cervical spinal cord injury.