AAV2-BDNF promotes respiratory axon plasticity and recovery of diaphragm function following spinal cord injury

Abstract

More than half of spinal cord injury (SCI) cases occur in the cervical region, leading to respiratory dysfunction due to damaged neural circuitry that controls critically important muscles such as the diaphragm. The C3-C5 spinal cord is the location of phrenic motor neurons (PhMNs) that are responsible for diaphragm activation; PhMNs receive bulbospinal excitatory drive predominately from supraspinal neurons of the rostral ventral respiratory group (rVRG). Cervical SCI results in rVRG axon damage, PhMN denervation, and consequent partial-to-complete paralysis of hemidiaphragm. In a rat model of C2 hemisection SCI, we expressed the axon guidance molecule, brain-derived neurotrophic factor (BDNF), selectively at the location of PhMNs (ipsilateral to lesion) to promote directed growth of rVRG axons toward PhMN targets by performing intraspinal injections of adeno-associated virus serotype 2 (AAV2)-BDNF vector. AAV2-BDNF promoted significant functional diaphragm recovery, as assessed by in vivo electromyography. Within the PhMN pool ipsilateral to injury, AAV2-BDNF robustly increased sprouting of both spared contralateral-originating rVRG axons and serotonergic fibers. Furthermore, AAV2-BDNF significantly increased numbers of putative monosynaptic connections between PhMNs and these sprouting rVRG and serotonergic axons. These findings show that targeting circuit plasticity mechanisms involving the enhancement of synaptic inputs from spared axon populations is a powerful strategy for restoring respiratory function post-SCI.—Charsar, B. A., Brinton, M. A., Locke, K., Chen, A. Y., Ghosh, B., Urban, M. W., Komaravolu, S., Krishnamurthy, K., Smit, R., Pasinelli, P., Wright, M. C., Smith, G. M., Lepore, A. C. AAV2-BDNF promotes respiratory axon plasticity and recovery of diaphragm function following spinal cord injury.

Trauma to the spinal cord can be neurologically devastating because it causes neuronal death and compromise in supporting glial function as well as disruption of neural circuits via axonal damage. Resulting functional deficits include sensory impairment, autonomic dysregulation, and motor compromise (1, 2). These deficits can be particularly dire when spinal cord injury (SCI) occurs in the cervical region because of damage to respiratory neural circuitry, including the rostral ventral respiratory group (rVRG)–phrenic motor neuron (PhMN) pathway that is responsible for diaphragm muscle control during rhythmic inspiratory breathing (3, 4). More than half of all human SCI cases occur in the cervical spinal cord region, where a critical component of the rVRG-PhMN-diaphragm respiratory neural circuit is located (5).

PhMNs, located along spinal cord levels C3–C5, receive projections from the ventral respiratory column (VRC), a collection of medullary nuclei responsible for breathing pattern generation (4, 6). The predominate excitatory bulbospinal input to PhMNs is from neurons located in the rVRG, a nucleus within the VRC and the prominent driver of diaphragm activation (7, 8). The diaphragm is a major inspiratory muscle; its ability to maintain proper inspiratory contraction is a critical component of independent ventilation (3). Cervical SCI can damage PhMNs and the descending rVRG axons, leading to respiratory dysfunction. Clinical implications for repairing this critically important neural circuit include restoration of independent breathing, improvements in quality of life, and abrogation of risks and morbidity associated with mechanical ventilation.

Disruption of the rVRG-PhMN-diaphragm circuit and failure to restore proper connections can lead to respiratory compromise or failure, requiring permanent respiratory maintenance on artificial ventilation (6, 9). Several barriers to axon growth and synaptic reconnection exist after CNS damage that are either cell-intrinsic or -extrinsic to injured and spared neurons. These include, but are not limited to, low intrinsic axon growth capacity of adult CNS neurons and the inhibition of axon growth mediated by various astrocyte- and myelin-associated inhibitory factors (e.g., chondroitin sulfate proteoglycans, Nogo, myelin-associated glycoproteins), fibroblasts, and pericytes (10–17). Importantly, axon guidance that occurs during nervous system development is largely absent after adult CNS injury (18); this lack of guidance hinders recovery by limiting appropriate axon-neuron targeting. Therefore, in addition to promoting regeneration of damaged axons and sprouting of spared fibers, providing a mechanism to sculpt this axon plasticity into meaningful connections is crucial for recovery.

To achieve targeted growth of relevant respiratory axon populations toward PhMNs following cervical SCI, we used the axon guidance properties of brain-derived neurotrophic factor (BDNF) in the rat model of C2 hemisection SCI by performing intraspinal injections of an adeno-associated virus serotype 2 (AAV2)-BDNF vector focally at the location of denervated PhMNs. Importantly, we used this experimental manipulation to study neuroanatomical modes of rVRG-PhMN-diaphragm circuit plasticity that may be capable of driving recovery of respiratory function following cervical SCI.

MATERIALS AND METHODS

Experimental design and statistical analysis

Experimental design

A total of 37 rats were used in this study. Twenty-one rats were randomly assigned to either AAV2–green fluorescent protein (GFP) or AAV2-BDNF treatment groups before receiving a C2 hemisection, and 13 animals were used as uninjured/intact controls. These 34 rats were divided into 3 groups to assess diaphragm muscle depolarization via electromyography (EMG): 1) intact (n = 13), 2) C2 hemisection and AAV2-GFP intraspinal injections (n = 11), and 3) C2 hemisection and AAV2-BDNF intraspinal injection (n = 10). A subset of these 34 rats were further divided into 4 groups for histologic analysis: 1) right rVRG injection, AAV2-BDNF treatment (n = 3), 2) right rVRG injection, AAV2-GFP treatment (n = 3), 3) left rVRG injection, AAV2-BDNF treatment (n = 3), and 4) left rVRG injection, AAV2-GFP treatment (n = 3). Three rats were used for tissue extraction and ELISA after receiving injections with AAV2-BDNF. EMG, compound muscle action potential (CMAP), grip strength, and all histologic analyses were carried out in a fashion in which the experimenter was blinded to the group identify of each sample.

We histologically assessed all animals for anatomically correct targeting of rVRG and efficient transduction of rVRG neurons following intramedullary AAV2-mCherry injection, as well as for completeness of C2 hemisection. Any rats with incomplete hemisection were excluded from all analyses, and any animal with mistargeted AAV2-mCherry injection or lack of transduction were excluded from rVRG axon growth analyses. We authenticated relevant regents to ensure that they performed similarly across experiments and to validate the resulting data. We generated the AAV2-mCherry, AAV2-GFP, and AAV2-BDNF vectors in-house; whenever we used a new batch of the vector, we verified that the virus performed equivalently from batch to batch. We have provided Research Resource Identification Initiative (RRID) numbers for all relevant reagents (i.e., antibodies and computer programs) throughout the Materials and Methods section.

Statistical analysis

Results are expressed as means ± sem. Statistical significance between 2 groups was analyzed by an unpaired Student’s t test, between 3 or more groups by repeated measures ANOVA as appropriate, and multiple comparisons used post hoc test (Fisher’s least significant difference test). Statistics were computed with Prism 5 (RRID: SCR_002798; GraphPad Software, La Jolla, CA, USA) and Sigma Plot 12 (RRID: SCR_003210; Systat Software, Chicago, IL, USA) software. All tests were conducted as 2-tailed. A value of P < 0.05 was considered as statistically significant.

Animals and surgical procedures

All procedures were carried out in compliance with the Guide for the Care and Use of Laboratory Animals [National Institutes of Health (NIH), Bethesda, MD, USA], the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines, the Society for Neuroscience’s Policies on the Use of Animals and Humans in Neuroscience Research, and the Thomas Jefferson University Institutional Animal Care and Use Committee (IACUC). All adult female Sprague-Dawley rats were purchased from Taconic Biosciences (Rensselaer, NY, USA) and housed in a temperature-, humidity-, and light-controlled animal facility and provided food and water ad libitum.

C2 hemisection

Adult female Sprague-Dawley rats (250–300 g) were anesthetized with intraperitoneal injection of ketamine (100 mg/kg), xylazine (5 mg/kg), and acepromazine (2 mg/kg) (KXA). When corneal reflexes and response to painful stimuli (i.e., toe pinch) were absent, the dorsal skin and underlying muscle layers of the rats were incised along the midline from the occipital bone to the caudal tip of the spinous process of C2. The dorsal muscle layers were retracted to expose the spinous process of C2. Following laminectomy above the C2 spinal cord level, the dura was cut parallel to the rostral-caudal axis of the spinal cord. Rats were subjected to a single hemisection injury that extended from midline to the lateral edge of the spinal cord using a sterile scalpel #11 blade (Feather Surgical Blade, VWR, Radnor, PA, USA) rostral to the C3 dorsal rootlets. A piece of biobrane (Biobrane-L) was placed over the injury site to prevent scarring and interference with subsequent intraspinal AAV2 injection surgery. Individual muscle layers were sutured (4-0 Sofsilk; Covidien, Dublin, Ireland), and the overlying skin was closed with sterile wound clips (Braintree Scientific, Braintree, MA, USA). Animals were monitored for pain until they were euthanized and were treated accordingly with buprenorphine (0.5 mg/kg). Food was left ad libitum on the cage floor, and subcutaneous injections of Lactated Ringer’s were given until rats recovered upright mobility.

Intraspinal injection

Rats were anesthetized as previously described. The dorsal skin and underlying muscle layers were then incised along the midline between the spinous processes of C2 and T1 to expose the cervical laminae and overlying muscles. The dorsal muscle layers were retracted, and the paravertebral muscles overlying C3–C5 were removed. Rats were subjected to unilateral laminectomy on the right side above the C3-C5 spinal cord levels. To target the ventral horn, a 33-gauge needle on a microsyringe was lowered 1.7 mm ventral from the dorsal surface of the spinal cord at the rostral tip of the C5 dorsal root ganglion rootlets (Hamilton, Reno, NV, USA). Three minutes after needle placement, 2 μl of AAV2-BDNF or AAV2-GFP vector was delivered to the spinal cord over 5 min, controlled by a UltraMicroPump and Micro4 Microsyringe Pump Controller (World Precision Instruments, Sarasota, FL, USA). After injection, the needle was left in place for 5 min before being slowly removed. This was then repeated at C4 and C3. In a similar manner, 2 injections of 0.5 μl of FluroGold (Fluorochrome, Denver, CO, USA) were delivered to the spinal cord over 2 min, with the exception that injection sites were restricted to half the distance between C4/C5 rootlets and C3/C4 rootlets. Overlying muscle layers and skin were then sutured and clipped as previously described.

rVRG injection

Three weeks before they were euthanized, rats were anesthetized as described above, and the dorsal skin and underlying muscle layers were incised along the midline from the occipital bone to the caudal tip of the spinous processes of C2. The dorsal muscle layers were retracted to expose the occipital bone. The C1 ligament was cut and the occipital bone removed on either side. AAV2-mCherry-P2A–wheat germ agglutinin (WGA) was loaded into a 33G syringe mounted on a micromanipulator (Hamilton). Beginning from the obex, a stereotaxic injector was used to move the needle 2 mm lateral, 0.4 mm rostral, and 2.4 mm ventral (Kopf Instruments, Tujunga, CA, USA). After needle placement into the medulla, the needle was left to rest for 3 min before injecting 300 nl of AAV2-mCherry-P2A-WGA over 45 s. After injection, the needle was left in place for 3 min. Individual muscle layers were sutured, and the overlying skin was closed with sterile wound clips. Rats received postoperative care as previously described.

Intrapleural injection

To identify PhMNs within the ventral horn of cervical spinal cord, 20 μl of cholera toxin subunit B (CTB; 2.5 µg/µl) (List Biological Laboratories, Campbell, CA, USA), a retrograde neuronal tracer, was delivered into the intrapleural space 3 d prior to euthanization, as previously described (19). Intrapleural injections were performed under isofluorane anesthesia (2.0% in 100% oxygen at 2 L/min). An 18-gauge needle was inserted 3 mm through the intercostal space between the sixth and seventh ribs (intersection of the anterior axillary line and xiphoid process). Rats were removed from isofluorane and allowed to recover in home cages. Animals were monitored for pain postprocedure and treated as needed.

EMG

All EMG recordings were terminal experiments. At 8 wk postinjury (6.5 wk posttreatment with AAV2-BDNF-GFP), animals were anesthetized with KXA, as described in C2 hemisection. Following a laparotomy, bipolar electrodes spaced 3 mm apart were inserted into the ventral, medial, or dorsal regions of the right hemidiaphragm, as previously described (20–26). Activity was recorded and averaged during eupneic breathing. EMG signal was amplified, filtered through a band-pass filter (50–3000 Hz), and integrated using LabChart 7 software (RRID: SCR_001620; ADInstruments, Sydney, NSW, Australia). Discharge duration and integrated peak amplitude were averaged over a 1-min sample period.

Diaphragm CMAPs

Animals were anesthetized with isoflurane. Supramaximal stimuli (0.5 ms duration; 6 mV amplitude) were delivered through needle electrodes placed 0.5 cm apart along the right and left phrenic nerves, as previously described. A ground needle electrode was placed in the tail; a reference electrode was placed subcutaneously in the right abdominal region. CMAP response was recorded via a surface strip along the costal margin of the right hemidiaphragm. Peak-to-peak CMAP amplitude from 5 stimulations was recorded to assess intra-animal variability and confirm reproducibility. Recordings were made using an ADI Powerlab 8/30 stimulator and BioAMP amplifier (ADInstruments) followed by computer-assisted data analysis (Scope v.3.5.6, RRID: SCR_001620; ADInstruments).

Grip strength test

Forelimb muscle grip strengths were determined using a Grip Strength Meter (DFIS-2 Series Digital Force Gauge; Columbus Instruments, Columbus, OH, USA). Grip strength testing was performed by allowing the animals to grasp a thin bar attached to the force gauge using either forelimb individually. This was followed by pulling the animal away from the gauge until the forelimb released the bar. This provides a value for the force of maximal grip strength. The force measurements were recorded in 3 trials, and the means were used in analyses. Grip strength measurements were collected once per week, beginning 1 wk prior to injury.

Viruses, reagents, and other procedures

Viruses

Viruses used included AAV2-pAM/chicken β actin–BDNF-WPRE-bGHpA (AAV2-BDNF) expressing human BDNF under the chicken β actin promoter and AAV2-eGFP (AAV2-GFP), injected at 1 × 10¹² genome copy (GC)/ml and 2.5 × 10¹¹ GC/ml, respectively. AAV2-BDNF and AAV2-GFP were used in the treatment and control groups, respectively, after C2 hemisection and were intraspinally delivered 10 d postinjury. AAV2-mCherry-P2A-WGA was used at 2.1 × 10¹³ GC/ml for rVRG injections to selectively label descending rVRG axons.

GFP fluorescence histogram

All images for analysis were taken in 30-μm sections using a Zeiss upright epifluorescence microscope (Carl Zeiss, Oberkochen, Germany) at ×10 and stitched together using Metamorph software (RRID:SCR_002368; Molecular Devices, San Jose, CA, USA). GFP + area was selected using the threshold tool and the area measured in 500-μm bins beginning at the rostral edge of the PhMN pool, moving rostrally through the lesion border and caudally 6 mm.

Protein extraction

Fresh spinal cord from level C3–C5 in rats treated with AAV2-BDNF was isolated and subdissected into ipsilateral (side of AAV2-BDNF injection) and contralateral halves. The tissue was then flash frozen in liquid nitrogen. The frozen tissue was fragmented with mortar and pestle on dry ice. After briefly thawing on ice, 300 μl of RIPA buffer (0.05 M Tris-HCl, 1% IGEPAL CA-630, 0.67 M NaCl, 1 mM EDTA, 0.5% sodium deoxycholate, 0.1% SDS, 1:100 phosphatase inhibitor cocktails 2 and 3, complete mini EDTA-free protease inhibitor cocktail) (Santa Cruz, Dallas, TX, USA) was added, and a micropestle was used for further disintegration. All samples were kept on ice. Samples were vortexed 3 times for 2 s each, with breaks on ice. Samples were incubated for 20 min and then spun down in a tabletop microcentrifuge at 4°C for 10 min at 13,000 rpm. Supernatant was collected, flash frozen, and stored at −80°C.

ELISA

Human/mouse BDNF DuoSet ELISA Kit (R&D Systems, Minneapolis, MN, USA) was used to determine protein concentration in ipsilaterally and contralaterally isolated spinal cord halves. After reaction termination, plates were read with a microplate reader (Molecular Devices, San Jose, CA, USA) at 450 nm.

Neuromuscular junction morphology analysis

Right hemidiaphragms were collected after the terminal EMG experiment, stained and imaged for neuromuscular junctions (NMJs), as previously described (27). Briefly, animals were euthanized with KXA overdose and the right hemidiaphragm was removed and then pinned flat on a silicon dish, fixed in 4% paraformaldehyde (PFA) and washed. Superficial fascia was removed prior to staining. Diaphragms were incubated in 0.1 M glycine, α-bungarotoxin conjugated to Alexa Fluor 555 (1:200; Thermo Fisher Scientific, Waltham, MA, USA) to label postsynaptic nicotinic acetylcholine receptors and then washed before they were incubated in ice-cold methanol. After washing, diaphragms were blocked at room temperature in 2% bovine serum albumin and 0.2% Triton X-100 in PBS, then incubated overnight at 4°C in blocking solution with anti–synaptic vesicle protein 2 (1:10, RRID: AB_2315387; Developmental Studies Hybridoma Bank, Iowa City, IA, USA) as a presynaptic vesicle marker and anti-SMI-312 (1:1000, RRID: AB_2314906; Covance, Princeton, NJ, USA) as a neurofilament marker. After washing, diaphragms were incubated in blocking solution with FITC anti-mouse IgG secondary (1:100; Jackson ImmunoResearch Laboratories, West Grove, PA, USA) and then washed. Diaphragms were mounted with Vectashield mounting medium (Vector Laboratories, Burlingame, CA, USA), cover-slipped, and stored at −20°C until they were imaged on a FluoView FV1000 confocal microscope (RRID: SCR_014215; Olympus, Center Valley, PA, USA). Images were used to analyze the phenotype of individual NMJs: intact, partially denervated or completely denervated, multiple axons, and terminal sprouting.

Tissue processing for histologic analysis

At the time of euthanization, rats were anesthetized with KXA and transcardially perfused with 0.9% normal saline solution, followed by 4% PFA. Spinal cords were immediately dissected and stored at 4°C in PFA for 1 d, 0.1 M phosphate buffer for 1 d, and then 30% sucrose solution for at least 3 d. Spinal cord tissue was embedded in a freezing medium, sectioned serially in either the transverse or sagittal plane at 30-μm thickness, collected on charged glass slides, and stored at −20°C until analysis.

Immunohistochemistry

Frozen sections were dried at room temperature in the dark before washing with Tris-buffered saline, incubated in blocking solution (0.3% Triton X-100, 10% normal donkey serum, in 1× PBS), and labeled with primary antibodies in blocking solution overnight at 4°C (0.3% Triton X-100, 10% normal donkey serum, in 1× PBS). After washing, sections were incubated with secondary antibodies in blocking solution (0.3% Triton X-100, 5% normal donkey serum, in 1× PBS). Sections were washed and dried overnight in the dark, dipped in dH₂0, and coverslipped with fluorescence-compatible mounting medium (FluorSave; MilliporeSigma, Burlington, MA, USA). Multiple primary antibodies were used in combination, including rabbit-anti-dsRed (1:500, Living Colors DsRed pAb, RRID: AB_10013483; Takara Bio USA, Mountain View, CA, USA) to enhance mCherry fluorescence, goat-anti-CTB (1:10,000; List Biological Laboratories), mouse-anti–vesicular glutamate transporter 2 (VGlut2) (1:250, RRID: AB_2187552; MilliporeSigma), rabbit-anti–5-hydroxytryptamine (5-HT; 1:15,000, RRID: AB_572263; ImmunoStar, Hudson, WI, USA), and mouse-anti-synaptophysin (Syn; 1:500, RRID: AB_2198854; Cedarlane Laboratories, Burlington, ON, Canada). All secondary antibodies were used at 1:200 dilution and included: Alexa Fluor 405–conjugated donkey anti-goat, Alexa Fluor 488–conjugated donkey anti-mouse, Alexa Fluor 594–conjugated donkey anti-mouse, Alexa Fluor 647–conjugated donkey anti-goat, Alexa Fluor 647–conjugated donkey anti-rabbit, Alexa Fluor 647–conjugated donkey anti-mouse (Jackson ImmunoResearch Laboratories), and Alexa Fluor 594–conjugated donkey anti-rabbit (Abcam, Cambridge, MA, USA).

Imaging

All mosaic scans were taken using an upright fluorescence microscope (Axiovert 200M; Carl Zeiss) and MetaMorph Software (Nashville, TN, USA) at either ×10 or 20, as indicated. Images for axon and synaptic quantification were taken using a confocal microscope (Leica TCS SP8; Leica Microsystems, Wetzlar, Germany) and LAS X software (RRID: SCR_013673; Leica Microsystems) at either ×20 or 40, as indicated. Images were quantified using ImageJ (RRID: SCR_003070; NIH).

Estimation of PhMN counts

All images for analysis were taken from 30-μm sections using a Zeiss upright epifluorescence microscope at ×10 and stitched together using MetaMorph software. All sections containing CTB + PhMNs per animal were assessed. The 2-dimensional counting method was used to determine the estimated cell count of PhMNs (28, 29). For quantification, the average nuclear size was 10.68 μm, and the average section thickness was 30 μm. Therefore, a correction factor of 0.74 was used based on the following equation: N = n(T/(T + D)), where N is the cell count estimate, n is the number of nuclear profiles, T is the section thickness, and D is the mean profile diameter (28). The mean nuclear profile diameter was calculated by measuring the longest diameter of each visible nucleus and taking the average. All CTB + PhMNs with a visible nucleus in 30-μm sagittal sections were counted for either the right or left PhMN pool and in either AAV2-GFP or AAV2-BDNF–treated rats. With this counting method, we likely underestimated the PhMN cell count, as shown by previous PhMN counts (19).

PhMN soma size

All images for analysis were taken in 30-μm sections using a Zeiss upright epifluorescence microscope at ×10 and stitched together using MetaMorph software. Soma diameter was determined by the widest horizontally separated points for each PhMN. PhMNs were determined with the same criteria as for estimated cell counts (i.e., the appearance of a visible nucleus). Twenty-five CTB + PhMN soma diameters were measured and averaged per rat.

Axon regeneration analysis

All images for rVRG axon analysis were obtained at ×20 in sagittal 30-μm sections using a Zeiss upright epifluorescence microscope and stitched together using MetaMorph software. An axon profile was defined as a discrete dsRed + line that was >5 μm in length. The number of axon profiles was determined in 1 section per animal in the lesion site and in areas 500-μm rostral and 500-μm caudal to lesion-intact borders.

Axon sprouting analysis

All images for rVRG and 5-HT + axon analyses were taken in transverse 30-μm sections using a Leica confocal microscope at ×40 and 20 magnification, respectively, using the same laser power and gain parameters. All dsRed + axons were selected using the threshold tool and the total area of axons was measured in a 164-μm diameter around the PhMNs. The total area of dsRed + axons was averaged across 3 sections per animal. The 5-HT + axons were similarly selected and measured at increasing distances from the center of a PhMN cluster using a modified Sholl analysis (i.e., increasing radii at 100-μm intervals) and averaged across 3 sections per animal. For the analysis of rVRG sprouting, putative synaptic connections between sprouting rVRG axons and PhMNs, 5-HT axon sprouting, and putative synaptic connections between 5-HT axons and PhMNs, we performed quantification of all of these effects at spinal cord level C4. As we observed (in the AAV2-BDNF group) significant recovery of EMG amplitude at ventral, medial, and dorsal diaphragm subregions and evidence of increased density of both 5-HT axons and contralateral rVRG-originating mCherry + axons along the entire length of the PhMN pool (i.e., C3–C5) ipsilateral to injury, we focused our detailed quantification of rVRG axon and 5-HT axon sprouting and putative rVRG-PhMN and 5-HT–PhMN synaptic connections at only a single spinal cord level (i.e., the C4 spinal cord). We choose the C4 location because it is at the rostral-caudal center of the phrenic nucleus.

Putative synaptic connection analysis

All images for analysis were taken in transverse 30-μm transverse sections using a Leica confocal microscope at ×40. Single Z-plane images were used to count the total number of colocalized synaptic markers (either VGlut2⁺/dsRed + or Syn1⁺/5-HT⁺) presynaptic to CTB + PhMN somata. The total number of double-labeled puncta directly presynaptic to CTB + PhMNs was tallied for 2 to 4 sections per rat and then averaged.

Data availability

All data for studies are available from the corresponding author upon request.

RESULTS

Intraspinal AAV2-BDNF injection focally enhanced BDNF expression within the PhMN pool

Supraspinal respiratory neurons located in the rVRG nucleus predominately stimulate PhMNs of the C3–C5 spinal cord to activate the diaphragm during inspiration (7, 8). Hemisection SCI at C2 causes selective removal of this descending excitatory drive to the ipsilateral PhMN pool and consequent hemidiaphragm paralysis. We evaluated a novel viral vector–based approach to restore diaphragm function after SCI by targeting expression of the neurotrophic factor, BDNF, to the locations of denervated PhMNs. Ten days after a right C2 hemisection in adult female rats, we delivered an AAV2 vector expressing BDNF by intraspinal injections to C3, C4, and C5 ventral horn ipsilateral to the injury (Fig. 1A).

Figure 1.

Intraspinal AAV2-BDNF injection focally enhanced BDNF expression within the PhMN pool. A) Illustration of intraspinal delivery of AAV2 to the ventral horn of spinal cord levels C3–C5 10 d after C2 hemisection. B) GFP reporter expression from AAV2 vector in the ventral horn on the side of injection. Scale bar, 250 μm. C–E) Sagittal section of rat spinal cord 8 wk postinjury with GFP reporter and PhMN pool labeled with CTB. E) GFP spread was restricted mostly to the ventral horn. Scale bar, 1 mm. The most rostral point of the PhMN pool was located 2.03 mm (sd ± 0.26 mm, n = 3) from the caudal lesion border. F) Histogram of GFP + area of fluorescence in 500-μm bins throughout the length of the spinal cord in relation to the most rostral point of the PhMN pool (n = 3). G) TrkB receptor expression in rVRG neurons that were retrogradely labeled from the ipsilateral C3–C5 ventral horn with FluroGold (F.Gold). Scale bar, 50 μm. H) Bar graph illustrating BDNF protein concentrations in ipsilateral and contralateral spinal cord halves 2 wk after AAV2-BDNF injection (n = 3). Unpaired Student’s t test; t = 4.351. *P = 0.0121.

AAV2 delivery resulted in efficient and selective transduction of cells with neuronal morphology in the ventral horn on the side of injection, as shown by GFP reporter expression (Fig. 1B). Transduction of neuronal somata was restricted to the ventral horn (Fig. 1B), extended continuously throughout the rostral-caudal length of the C3–C5 PhMN pool (Fig. 1C–F), and was noted up to the caudal lesion-intact border (Fig. 1C–F) but was absent within the lesion site (Fig. 1F).

Using immunohistochemistry on brainstem sections, we first established that the receptor for BDNF, tropomyosin receptor kinase B (TrkB), is expressed by rVRG neurons. We injected the retrograde tracer, FluoroGold, into C3–C5 ventral horn (i.e., location of rVRG axon terminals) and found that FluoroGold + rVRG neurons expressed TrkB (Fig. 1G), suggesting that rVRG axons will be responsive to BDNF chemotactic signals generated by the AAV2-BDNF injections.

To promote plasticity within a defined region of the spinal cord (i.e., the C3–C5 ventral horn on the side of the hemisection), we aimed to spatially limit BDNF up-regulation. To determine BDNF protein concentration after intraspinal delivery of AAV2-BDNF, we conducted ELISA of extracted protein from entire C3–C5 hemicord ipsilateral and contralateral to AAV2 injections. Compared with the contralateral cord, AAV2-BDNF increased BDNF protein expression at the site of injection in ipsilateral spinal levels C3–C5 (Fig. 1H).

AAV2-BDNF delivery to the PhMN pool promoted recovery of diaphragm function

After establishing our method of BDNF delivery, we determined whether AAV2-BDNF injection could promote diaphragm recovery after C2 hemisection SCI. EMG measures muscle depolarization to assess respiratory neural circuit drive and diaphragm activation in vivo during inspiratory breathing. Hemidiaphragm innervation by the PhMN pool is divided into 3 diaphragm subregions: ventral, medial, and dorsal. The ventral portion of the hemidiaphragm is largely innervated by PhMNs located in the C3 spinal cord, whereas medial hemidiaphragm is innervated by C4 and dorsal by C5 (30). Therefore, we assessed EMG recordings separately at these 3 locations for each hemidiaphragm ipsilateral to the lesion 8 wk after C2 hemisection (20–26).

Focal overexpression of BDNF promoted significant recovery of rhythmic inspiratory diaphragm activation 8 wk after injury (Fig. 2A–D). Rats treated with AAV2-GFP control showed minimal to no spontaneous recovery of diaphragm activation, whereas treatment with AAV2-BDNF increased diaphragm activity in all 3 subregions, with no difference in recovery across them (Fig. 2B, C). Although diaphragmatic EMG recordings documented substantial recovery in AAV2-BDNF–treated rats, EMG amplitudes were not returned to the levels of uninjured controls (Fig. 2A, D). These data reveal our ability to promote diaphragm recovery with a novel, anatomically targeted approach of AAV2-BDNF injection.

Figure 2.

Focal AAV2-BDNF injection promoted recovery of diaphragm function. A) Diaphragmatic EMG recording of ipsilateral hemidiaphragm in intact rats. B, C) Diaphragmatic EMG of ipsilateral hemidiaphragm 8 wk postinjury in AAV2-GFP–treated (B) and AAV2-BDNF–treated (C) rats. Top: raw traces; bottom: integrated signals. D) Quantification of EMG integrated signal for intact (n = 13), injured plus AAV2-GFP (n = 11), and injured plus AAV2-BDNF (n = 10) rats. Repeated measures 2-way ANOVA (post hoc Fisher’s least significant difference test): treatment, F(2,31) = 21.21, P < 0.0001; intact vs. AAV2-GFP, P < 0.0001, t = 6.44; intact vs. AAV2-BDNF, P = 0.0008, t = 3.725; AAV2-GFP vs. AAV2-BDNF, P = 0.0199, t = 2.456; region, F(2,62) = 0.572, P = 0.567.E, F) CMAP recordings of ipsilateral hemidiaphragm after supramaximal phrenic nerve stimulation in AAV2-GFP– treated (E) and AAV2-BDNF–treated (F) rats 8 wk post-SCI (n = 5/group). G) Quantification of CMAP amplitude. Ns, not significant. Unpaired Student’s t test, P = 0.0528, t = 2.271, df = 8. A subset of rats in each injury group was assessed weekly for grip strength, beginning 1 wk prior to injury (baseline = preinjury). H) No difference was found in recovery between AAV2-GFP (n = 3) and AAV2-BDNF–treated (n = 4) C2 hemisection rats. Repeated measures 2-way ANOVA: treatment, F(1,5) = 0.1646, P = 0.702. *P < 0.05; ***P < 0.001; ****P < 0.0001.

We conducted additional analyses of diaphragm activity, including burst frequency and percent of maximum amplitude, to determine whether AAV2-BDNF treatment exerted effects on additional EMG measurements. We found no significant differences in the number of inspiratory bursts per minute among uninjured rats and injured rats treated with either AAV2-GFP or AAV2-BDNF (uninjured: 52.1 bursts/min; hemisection plus AAV2-GFP: 51.8 bursts/min; hemisection plus AAV2-BDNF: 57.7 bursts/min; 1-way ANOVA; F = 0.5090, P = 0.6072). We also analyzed the amplitude of inspiratory bursts in uninjured and hemisection rats compared with the maximum diaphragm activity (i.e., sigh) in the same rat (31). There were no significant differences across groups in the percentage of maximum amplitude during eupneic breathing (uninjured: 47.2%; hemisection plus AAV2-GFP: 42.2%; hemisection plus AAV2-BDNF: 36.6%; 1-way ANOVA; F = 0.5111, P = 0.6099).

AAV2-BDNF did not alter functional innervation of the diaphragm

BDNF may be driving diaphragm recovery by stimulating rVRG-PhMN-diaphragm circuit plasticity either centrally within the cervical spinal cord or peripherally at the NMJ, the critical synapse between phrenic motor axons and diaphragm muscle. To assess functional innervation of the diaphragm by PhMNs independent of bulbospinal drive, we performed evoked nerve conduction studies by recording the CMAP from the ipsilateral hemidiaphragm after supramaximal stimulation of the phrenic nerve (20–26). AAV serotypes can be anterogradely transported along axons, including along the phrenic nerve (32), raising the possibility that AAV2-BDNF could be stimulating diaphragm recovery by altering NMJ structural connectivity and function. We found no difference in peak CMAP amplitude between AAV2-BDNF and AAV2-GFP on the side of injury (Fig. 2E–G).

AAV2-BDNF injection at C3–C5 did not affect forelimb motor function

BDNF may also exert effects on other circuits in the region of overexpression. Although we were unable to assess all possible behavioral effects of AAV2-BDNF injection, we chose to focus on one modality that is relevant to cervical spinal cord anatomy. The motor neurons that innervate distal forelimb muscles are located from C6–T1 and are therefore in close proximity to targeted areas of AAV2-BDNF injection. Additionally, descending motor pathways that control forelimb motor function (e.g., red nucleus/rubrospinal tract, corticospinal tract) pass through the C3–C5 spinal cord to reach caudal targets. These motor pathways and their motor neuron innervation may be affected by BDNF (33). To evaluate potential effects on forelimb motor function, we assessed forelimb grip strength weekly in injured rats after AAV2-BDNF or AAV2-GFP delivery. There was no difference in motor recovery between AAV2-GFP or AAV2-BDNF rats after C2 hemisection (Fig. 2H). This suggests that AAV2-BDNF injection had a relatively targeted functional effect on diaphragm recovery, supporting the utility of restricting BDNF up-regulation to a specific region of interest (i.e., within the PhMN pool).

AAV2-BDNF did not alter diaphragm NMJ innervation, PhMN survival, or PhMN somal morphology

In addition to CMAP recordings, we assessed NMJ morphologies to determine the effects of AAV2-BDNF on PhMN innervation of the diaphragm (23–27). As with EMGs, we examined NMJs at all 3 subregions of the hemidiaphragm (ventral, medial, dorsal) 8 wk after injury. Using confocal analysis of individual NMJs throughout the muscle, we counted the total number of NMJs and assessed the proportion of several morphologic phenotypes. We found no differences in the percentage of intact, partially denervated or completely denervated NMJs between rats treated with AAV2-GFP (Fig. 3A) and AAV2-BDNF (Fig. 3B), with >98% of all NMJs remaining completely intact at all 3 subregions (Fig. 3C–E). We also observed no differences between GFP control and BDNF groups in the number of NMJs with terminal sprouting (Fig. 3C–E).

Figure 3.

AAV2-BDNF injection did not alter diaphragm NMJ innervation, PhMN survival, or PhMN somal morphology. Synaptic vesicle protein 2 and SMI-312 as markers for presynaptic terminals and neurofilament, respectively, were used to identify the presynaptic component of the diaphragm NMJ. A, B) α-Bungarotoxin was used to label nicotinic acetylcholine receptors to identify postsynaptic receptors of the diaphragm NMJ in hemidiaphragms of AAV2-GFP–treated (n = 4) (A) and AAV2-BDNF–treated (n = 5) (B) rats. Scale bar, 30 μm. Arrowheads (B) denote an example of an NMJ that is partially denervated (arrowheads point to the portion of this NMJ that is denervated: no overlap of green and red labeling). NMJs were analyzed to determine the percentage of intact, partially denervated (P. Denv), and fully denervated (Denv) NMJs, in addition to the percentage of terminal sprouting and multiple axons. C–E) Quantification of ipsilateral NMJ morphologies showed no difference in all 3 regions of the hemidiaphragm between AAV2-GFP and AAV2-BDNF rats. Repeated measures 2-way ANOVA: intact, F(1,7) = 0.298, P = 0.602; partially denervated, F(1,7) = 0.269, P = 0.620; fully denervated, F(1,7) = 0.777, P = 0.407; thin preterminal staining (T.S.), F(1,7) = 1.597, P = 0.247; NMJ innervated by multiple axons (M.A.), F(1,7) = 1.263, P = 0.298. F, G) CTB-labeled PhMNs ipsilateral to hemisection in injured rats with AAV2-GFP (F) or AAV2-BDNF (G). H) Quantification of estimated ipsilateral and contralateral PhMN cell counts in AAV2-GFP–treated (n = 8) and AAV2-BDNF–treated (n = 6) rats. Two-way ANOVA: treatment P = 0.2062, F(1,10) = 1.828; side P = 0.523, F(1,10) = 0.4385. I) Quantification of CTB + PhMN soma size. Two-way ANOVA: treatment P = 0.4604, F(1,10) = 0.5893; side P = 0.511, F(1,10) = 0.4642. Ns, not significant.

BDNF has a known role in neuronal survival and excitability (34–37). AAV2-BDNF may therefore directly impact PhMN properties, a key link between rVRG input and the diaphragm. Using 2-dimensional stereology techniques, we counted PhMN somata both ipsilateral and contralateral to the C2 hemisection in AAV2-GFP and AAV2-BDNF animals (Fig. 3F, G). To selectively label PhMN cell bodies within the cervical spinal cord, we performed intrapleural injection of the retrograde neuronal tracer, CTB. We found no difference in estimated CTB + PhMN counts between AAV2-GFP and AAV2-BDNF–treated rats in either the ipsilateral or contralateral PhMN pool (Fig. 3H). Together, our PhMN counting as well as CMAP and NMJ results suggest that AAV2-BDNF did not promote functional diaphragm recovery by altering diaphragm innervation by PhMNs.

Though we did not assess PhMN intrinsic properties using electrophysiological techniques, we did examine PhMN size given that somal size is important in determining neuronal excitability and that SCI can induce changes in somal characteristics such as via atrophy (38). Again, using CTB to selectively label PhMNs, we found no differences in average PhMN somal size between AAV2-GFP and AAV2-BDNF rats on the side of injury or between ipsilateral and contralateral sides (Fig. 3I). These data suggest that AAV2-BDNF did not alter the morphologic properties that contribute to PhMN intrinsic excitability (though rigorous electrophysiological analyses of PhMN properties are necessary to fully address this issue) or that are indicative of neuronal health.

Thus far, our results show continuous viral transduction throughout the length of the PhMN pool, in addition to transduction that was restricted to the ventral horn of spinal cord levels C3–C5 (i.e., location of PhMNs) on the side of the C2 hemisection SCI. This resulted in BDNF up-regulation around the PhMNs on the lesioned side of the spinal cord, resulting in a robust restoration of diaphragm function 8 wk after cervical SCI. We found no associated effects of AAV2-BDNF on PhMN survival or somal size as well as no peripheral changes to functional or morphologic innervation of the diaphragm by PhMNs using CMAP and NMJ morphology analyses. Taken together, these data suggest a central mechanism of recovery, including possible changes to PhMN innervation, by descending excitatory and modulatory axon populations. Thus, we sought to assess the effects on different forms of targeted axon growth and synaptic connection with PhMNs, including rhythmic excitatory rVRG axons and modulatory serotonergic axons, to determine whether plasticity of these circuits within the spinal cord may account for diaphragm recovery induced by AAV2-BDNF delivery.

AAV2-BDNF did not promote rVRG axon regeneration

To explore possible modes of axon plasticity underlying diaphragm recovery, we examined neuronal populations providing rhythmic excitatory input to PhMNs. There are 2 primary populations: axons of neurons residing in the ipsilateral rVRG and axons from neurons located in the contralateral rVRG. C2 hemisection is a powerful SCI paradigm for studying modes of axon regrowth in the context of respiratory dysfunction because it results in complete axotomy of ipsilaterally descending rVRG axons and sparing of rVRG axons descending in the contralateral spinal cord (26). We can label these descending excitatory rVRG axons with a system previously developed by our group using an AAV2 vector that selectively transduces rVRG neurons to express the anterograde tracer, mCherry (26, 39). Therefore, by delivering the AAV2-mCherry vector to either the ipsilateral or contralateral rVRG nucleus, we can trace the descending projections of rVRG neurons as well as their synaptic connectivity with PhMNs and other neuronal targets (Fig. 4A). In transverse sections 500-μm caudal to the caudal border of the C2 hemisection lesion site, we show mCherry + puncta of descending rVRG axons in the ventral white matter on the uninjured side of the spinal cord but no mCherry + puncta on the side of injury, indicating a complete hemisection (Fig. 4B, B″).

Figure 4.

AAV2-BDNF injection did not promote rVRG axon regeneration. A) Schematic of AAV2-mCherry injection for labeling of ipsilateral-originating rVRG axons. B) Schematic of transverse spinal cord. B′, B″) Representative images of transverse spinal cord 500-μm caudal-to-caudal lesion border demonstrate hemisection completeness. mCherry + puncta can be found in ventral white matter of the contralateral uninjured side of the cervical spinal cord after rVRG labeling with AAV2-mCherry (B′), whereas no mCherry + axons are present on the side of injury (B″). Scale bar, 100 μm. C) mCherry-labeled axons from the ipsilateral rVRG 8 wk postinjury in AAV2-BDNF–treated animal. Scale bar, 500 μm (C, C′). Average lesion width was 430 μm. D) Quantification of number of mCherry + axon profiles in AAV2-GFP (n = 3) and AAV2-BDNF (n = 3) rats. Two-way ANOVA (post hoc Fisher’s least significant difference test); treatment P = 0.4841, F(1,4) = 0.5934. MN, motor neuron.

We first assessed whether focal AAV2-BDNF injection promoted diaphragm recovery via reconstruction of the damaged rVRG-PhMN circuit (i.e., regeneration of axons originating in the ipsilateral rVRG and reconnection of these regrowing axons with denervated PhMNs on the same side as the lesion). To do so, we assessed ipsilaterally descending rVRG axons by injecting AAV2-mCherry into the right rVRG (i.e., the side of injury). Eight weeks after hemisection, we histologically analyzed sagittal sections of cervical spinal cord for growth of mCherry + axons into and through the lesion site. We found that in all rats, mCherry-labeled rVRG axons extended into the lesion but did not grow further caudally through the caudal lesion-intact interface or into the caudal intact spinal cord where the PhMNs are located (Fig. 4C–D). These data suggest that regeneration of injured rVRG axons was not a plasticity mechanism responsible for functional diaphragm recovery induced by AAV2-BDNF.

AAV2-BDNF promoted significant sprouting of spared rVRG axons locally within the PhMN pool

We next assessed the effect of BDNF overexpression on spared respiratory circuitry, in particular, descending bulbospinal axons originating in the contralateral (left) rVRG (Fig. 5A). AAV2-mCherry injection into the left rVRG resulted in labeling of the intact rVRG-PhMN circuit on the uninjured side ipsilateral to brainstem injection (Fig. 5B, C). Following injection of AAV2-mCherry into the contralateral rVRG, we observed a region of mCherry + axon (using anti-dsRed antibody) concentration packed densely at the location of PhMNs on the side of injury in AAV2-BDNF–treated rats (Fig. 5D–E′). In transverse spinal cord sections at 8 wk posthemisection, we found that AAV2-BDNF (Fig. 5E, E′) promoted a robust increase in the density of mCherry + rVRG axons around CTB + PhMN clusters within the ventral horn on the side of injury compared with AAV2-GFP controls (Fig. 5D). This increase in mCherry + rVRG axons within the PhMN pool was increased by ∼2.5-fold in AAV2-BDNF–treated animals vs. AAV2-GFP (Fig. 5F), suggesting that rVRG axon sprouting spared by the injury is possibly a mechanism underlying—at least in part—the functional recovery induced by AAV2-BDNF.

Figure 5.

AAV2-BDNF injection promoted significant sprouting of spared rVRG axons locally within the PhMN pool. A) Schematic of AAV2-mCherry injection for labeling of spared contralateral-originating rVRG axons. B, C) Transverse image of mCherry + axons around CTB + PhMNs in the contralateral cervical spinal cord after injection of AAV2-mCherry into the left rVRG and right C2 hemisection. Scale bar, 50 μm. D–E′) Confocal images of mCherry + rVRG axon bundles around denervated CTB + PhMNs (right side) 8 wk after C2 hemisection in AAV2-GFP (D, D′) and AAV2-BDNF (E, E′) rats. Scale bar, 100 μm. F) Quantification of area of mCherry + axons in a 164-μm diameter radius around the CTB PhMNs on the side of hemisection (n = 3/group). Unpaired Student’s t test; P = 0.02, t = 3.654, df = 4. MN, motor neuron. *P < 0.05.

AAV2- BDNF significantly enhanced putative excitatory synaptic input to PhMNs from spared rVRG axons

To determine whether the increased sprouting of spared rVRG axons resulted in enhanced excitatory bulbospinal input to PhMNs, we analyzed putative glutamatergic synaptic connections between CTB-labeled PhMNs (located ipsilateral to the hemisection) and mCherry + rVRG axons originating in the contralateral medulla (Fig. 6). We conducted triple immunolabeling for mCherry + rVRG axons (using anti-dsRed antibody), CTB + PhMNs, and the excitatory presynaptic terminal marker VGlut2. Using single z-section confocal analysis, we counted numbers of overlapping VGlut2⁺/mCherry + puncta directly presynaptic to cell bodies of CTB + PhMNs (Fig. 6A, B). AAV2-BDNF increased putative excitatory synaptic connections per ipsilateral PhMN from spared rVRG axons 3-fold (Fig. 6D), indicating that the increase in rVRG axon sprouting around these PhMNs was associated with an enhancement of putative monosynaptic input. Together, our findings suggest this plasticity of spared rVRG axons may have contributed to the diaphragm recovery promoted by AAV2-BDNF.

Figure 6.

AAV2-BDNF injection significantly enhanced putative excitatory synaptic input to PhMNs from spared rVRG axons. A, B) Representative confocal images with orthogonal projections of CTB/dsRed/VGlut2 triple labeling on the side of hemisection in AAV2-GFP (A) and AAV2-BDNF (B) rats. Scale bar, 20 μm. Yellow arrows indicate colocalization of VGlut2⁺/mCherry + puncta directly presynaptic to CTB-labeled PhMNs. C) Location of PhMNs in ventral horn for synaptic connection analysis. D) Quantification of number of VGlut2⁺/mCherry + presynaptic puncta per CTB + PhMN (n = 3 rats/group). Unpaired Student’s t test; P = 0.0043, t = 5.837, df = 4. *P < 0.01.

AAV2-BDNF promoted sprouting of modulatory serotonergic axons locally within the PhMN pool

In addition to the rVRG-PhMN pathway, other axonal input relevant to diaphragm recovery may have also been affected by BDNF. There are additional axon populations that contribute to diaphragm activation that do not originate in the VRC, a particularly important one being descending serotonergic (5-HT) axons originating in the raphe (40). 5-HT axons can affect PhMN excitability by modulating, for example, synaptic input from rVRG axons (41, 42). We performed immunohistochemistry to detect 5-HT axons and determined the density of 5-HT + axons in concentric circles of 200-, 300-, and 400-μm diameters around CTB-labeled PhMNs on the side of the hemisection. Compared with the AAV2-GFP controls (Fig. 7A), the AAV2-BDNF–treated rats (Fig. 7B) displayed enhanced growth of serotonergic axons, as shown by an increase in total 5-HT axon area at all 3 distances around these CTB-labeled PhMNs. When compared with the contralateral ventral horn (Fig. 7C), 5HT + axon area was only increased in AAV2-BDNF rats on the side of AAV2 injection (Fig. 7D). This increase in modulatory serotonergic axon input may have augmented the rhythmic excitatory drive from spared contralateral rVRG axons that was also increased by AAV2-BDNF treatment.

Figure 7.

AAV2-BDNF injection significantly enhanced putative synaptic input to PhMNs from sprouting serotonergic axons. A, B) Total area of 5-HT + axons was determined on the side of C2 hemisection using a modified Sholl analysis from the center of PhMN clusters in AAV2-GFP (A) and AAV2-BDNF (B) rats. Scale bars, 200 μm. C, D) Quantification of 5-HT + axon area in the left/contralateral ventral horn (C) and right/ipsilateral ventral horn (D). Right ventral horn repeated measures 2-way ANOVA (post hoc Fisher’s least significant difference test): treatment P = 0.0107, F(1,4) = 20.38, at +200 μm P = 0.0031, at +300 μm P = 0.0005, at +400 μm P = 0.0020; distance P = 0.1514, F(2,8) = 2.412. Left ventral horn repeated measures 2-way ANOVA (post hoc Fisher’s least significant difference test): treatment P = 0.6042, F(1,4) = 0.3156, at +200 μm P = 0.8017; at +300 μm P = 0.4677; at +400 μm P = 0.7718. E, F) Representative confocal images with orthogonal projections of CTB/Syn1/5-HT triple labeling on the side of hemisection in AAV2-GFP (E) and AAV2-BDNF (F) rats. Scale bar, 20 μm. G) Quantification of Syn1⁺/5-HT + double-labeled puncta onto CTB + PhMNs (n = 3 rats/group). Unpaired Student’s t test; P = 0.0019, t = 7.274, df = 4. Ns, not significant. **P < 0.01; ***P < 0.001.

AAV2-BDNF significantly enhanced putative synaptic input to PhMNs from sprouting serotonergic axons

Lastly, we determined whether the enhanced serotonergic axon sprouting effect was associated with an increase in synaptic density onto PhMNs, which would support a role for 5-HT + modulatory axons in recovery by augmenting the excitatory drive from sprouting spared rVRG axons. We conducted triple immunolabeling for 5-HT + serotonergic axons, CTB + PhMNs, and the pan-presynaptic terminal marker Syn1. Using a similar single z-section confocal analysis as with the mCherry/VGlut2/CTB analysis, we identified overlapping Syn1⁺/5-HT + puncta directly presynaptic to CTB + around PhMNs on the side of injury (Fig. 7E, F). AAV2-BDNF significantly increased the density of putative synaptic connections onto ipsilateral PhMNs from sprouting 5-HT + axons (Fig. 7G).

DISCUSSION

Anatomically targeted BDNF expression

Although previous work has shown that BDNF can promote partial recovery of diaphragm function after cervical SCI, most of these studies delivered exogenous BDNF in a relatively nonanatomically targeted manner, including bathing whole cervical spinal cord with BDNF or releasing BDNF over an extended area from an engineered hydrogel (25, 31, 43–45). Unlike this previous work, we restricted BDNF expression to a specific region of interest, the PhMN pool on the side of injury. We used a viral vector–based approach to deliver a focal source of BDNF to the location of denervated PhMNs to promote targeted growth of injured and spared rVRG axon populations, and, importantly, to sculpt this plasticity into synaptic connections to restore rVRG-PhMN-diaphragm circuit function (46). Using this delivery system, we promoted significant functional diaphragm improvement beyond the minimal spontaneous recovery found with AAV2-GFP controls. We determined the diaphragm recovery resulting from this BDNF delivery approach 8 wk after injury to establish the longer-term effect on synaptic formation and circuit connectivity as opposed to only transient plasticity.

By anatomically restricting AAV2-BDNF delivery, we were able to promote plasticity primarily within ipsilateral C3–C5 ventral horn. Along these lines, we found no effect of AAV2-BDNF on forelimb motor recovery, indicating limited influence on motor circuits innervating distal forelimb muscles located from C6–T1, sites caudal to injections (33, 47). We expect that this anatomically targeted BDNF expression strategy would result in limited off-target effects, such as altered pain neurotransmission that is modulated by BDNF signaling (48, 49), though additional work is necessary to confirm this prediction.

Regeneration vs. sprouting of rVRG axons

We found no regeneration through the lesion of axons originating in the rVRG ipsilateral to hemisection. Although ipsilaterally descending rVRG axons grew into the lesion, they did not extend into the intact caudal spinal cord, suggesting that this mode of plasticity (i.e., regeneration) likely did not contribute to the observed diaphragm recovery. Repulsive cues from the intact caudal spinal cord or stabilizing signals from within the lesion may be acting on these axons to retain them in the lesion center. For example, axons may be attracted to NG2 + cell processes, confining growth to the lesion site (50).

Interestingly, we observed growth of injured rVRG axons into the lesion in both the AAV2-GFP and AAV2-BDNF conditions, with no difference between the 2 groups. This regeneration of injured rVRG axons into the injury—particular in the AAV2-GFP condition—differs from other control conditions in the C2 hemisection model in our prior (26) and unpublished work. In hemisection-only conditions, we consistently do not observe rVRG axon regeneration into the injury site. As we found similar rVRG regeneration with both AAV2-GFP and AAV2-BDNF, these findings suggest that some aspect of the experimental manipulation, such as the intraspinal injection procedure itself, stimulated the regrowth response, possibly by altering the local microenvironment encountered by the injured rVRG axons in and around the lesion site.

By focusing on only regeneration of damaged axons, substrates for promoting recovery are limited. rVRG neurons primarily project to ipsilateral PhMNs but also have projections to contralateral PhMNs (51, 52). It is believed that these contralaterally projecting axons contribute to the crossed phrenic pathway, a set of contralateral rVRG axons that become activated to maintain ventilation after C2 hemisection and contralateral phrenicotomy (53–56). The C2 hemisection largely does not disrupt the contralateral rVRG-PhMN circuit; therefore, these spared axons are promising substrates for driving recovery (51, 52, 55, 57). We assessed whether rVRG axons originating from the contralateral rVRG may contribute to diaphragm recovery. We found a dense concentration of contralateral-originating rVRG axons surrounding denervated PhMNs on the side of injury; this concentration was significantly increased after BDNF delivery, suggesting a role in functional improvement induced by AAV2-BDNF. Although there was a significant increase in this rVRG axon sprouting around the PhMNs after AAV2-BDNF injection, the overall density of axons was still lower than the intact circuit, which may explain why EMG amplitudes did not recover fully to levels observed in uninjured rats.

Our results with PhMN cell body counting, CMAP amplitude recordings and NMJ morphologic analysis suggest that AAV2-BDNF did not promote functional diaphragm recovery by altering diaphragm innervation by PhMNs. These finding are not unexpected given that the C2 hemisection was located rostral to the PhMN pool and therefore did not directly injure any PhMN cell bodies and result in muscle denervation. Nevertheless, these data are important in that they confirm that the hemisection did not induce loss of PhMNs despite this distance, and the findings further support the notion that plasticity mechanism involving synaptic input to PhMNs within the spinal cord were possibly responsible for—at least in part—diaphragm recovery in response to AAV2-BDNF.

Plasticity of synaptic input to PhMNs

In addition to stimulating a robust rVRG axon sprouting response, focal AAV2-BDNF injection increased putative monosynaptic connections onto PhMNs. In addition to this increased synaptic innervation, BDNF could also be altering the strength of these new (and/or previously established) synapses. BDNF has a well-characterized role in altering synaptic function, including at rVRG-PhMN synapses (45). Thus, BDNF may strengthen new and/or basal rVRG input to PhMNs, possibly by increasing glutamatergic release from rVRG axon terminals or by altering glutamate receptor recruitment to rVRG-PhMN synapses (58, 59).

Importantly, we are likely underestimating the number of rVRG-PhMN synapses by failing to assess connections on dendritic trees given that retrograde CTB only labels PhMN somata and proximal dendrites. In addition, we are not taking into account potentially important rVRG axon input onto prephrenic interneurons (i.e., polysynaptic circuits), which we can assess in future work using trans-synaptic labeling methods (60–63).

In AAV2-BDNF rats, we also observed a significant and anatomically localized increase in serotonergic axon sprouting within the PhMN pool that was accompanied by an increase in putative serotonergic synapses onto PhMNs. Serotonergic innervation of PhMNs is responsible for modulating excitatory glutamatergic rVRG input (41, 64, 65). These data suggest that this enhanced serotonergic innervation of PhMNs on the side of injury acted in concert with the increased rhythmic excitatory input from the contralateral rVRG to drive diaphragm EMG recovery. In future work, we could examine specific 5-HT receptor subtypes to distinguish their various contributions to functional improvement in response to AAV2-BDNF. For example, 5-HT_2A, 5-HT_2/1C, or 5-HT₇ receptors may positively modulate rVRG input, whereas other receptor subtypes may dampen it (41, 64, 65).

Acute intermittent hypoxia (AIH) has been used as a means of neurorehabilition to endogenously increase BDNF expression and promote diaphragm recovery after high-cervical SCI via a form of spinal plasticity known as long-term phrenic facilitation (40). AIH enhances PhMN activation by increasing de novo BDNF synthesis in a serotonin-dependent manner (40), suggesting that AAV2-BDNF may be promoting recovery directly by increased BDNF synthesis and possibly bypassing the need for serotonergic changes. BDNF up-regulation by AAV2 intraspinal injections may also be promoting recovery via a feed-forward mechanism by increasing serotonergic input onto PhMNs.

Our analyses demonstrated enhanced putative rVRG and serotonergic synaptic input to PhMNs with AAV2-BDNF delivery using anatomic methods; however, we do not know whether these synapses are actually functional, though the robust recovery we observe using EMG recordings support the notion that functional connectivity of these circuits is being enhanced. Moving forward, it will be important to use methods, such as electrophysiological recordings of PhMNs, to explore this important issue. Furthermore, we employed in vivo EMG recordings to assess diaphragm activation and function of the rVRG-PhMN-diaphragm circuit; however, it will be critical to extend this analysis to approaches such as whole-body plethysmography and blood gas measurements to examine the overall functional effects on ventilatory behavior, including under conditions of low-O₂ or high-CO₂ respiratory challenge.

Comparison with other BDNF delivery approaches that target diaphragm function after SCI

Several different approaches have been used to deliver BDNF to the PhMN pool in models of cervical SCI. Important work from Mantilla et al. (43) has shown that intrathecal BDNF delivery to the general location of the PhMN pool using a catheter directed at the C4 spinal cord promotes recovery of ipsilateral hemidiaphragm function following C2 hemisection (31). Gransee et al. has also shown that intraspinal transplantation (directly into the C2 spinal cord) of mesenchymal stem cells engineered to express BDNF promotes partial recovery of ipsilateral hemidiaphragm function after C2 hemisection (44). It is difficult to directly compare the degree of EMG recovery between our current work and these previous studies because there are important differences in experimental design, such as timing and/or duration of BDNF delivery, time postinjury of EMG recordings, and anesthesia used. Nevertheless, the recovery we observe with AAV2-BDNF under eupnic conditions is in line with the diaphragm EMG amplitude observed in these previous experiments; recovery of EMG amplitude in all of these studies (including the current work) in response to BDNF delivery reached at least 50% of preinjury or uninjured controls. In addition, the majority (and, in some studies, all) of BDNF-treated animals in these previous studies showed significant hemidiaphragm recovery, which is similar to our current findings. Based on the available data, there does not appear to be a particular BDNF delivery approach that results in more robust therapeutic effects on diaphragm function in the cervical hemisection model, which suggests that future work may be necessary to compare these strategies under similar experimental conditions, test diaphragm function in response to various respiratory challenges, and evaluate additional outcome measures beyond just EMG recordings (e.g., blood gas measurements).

We recently employed an engineered hydrogel to deliver BDNF in a nontargeted manner by placing the hydrogel on the dorsal surface of the injured spinal cord after C4/C5 contusion SCI (25). In addition to functional recovery as evidenced by enhanced EMG and CMAP amplitudes, we observed robust effects of BDNF hydrogel on 5-HT axon growth and monosynaptic innervation of PhMNs by these serotonergic fibers on the side of injury; however, we did not assess these same effects in the contralateral PhMN pool. Given that this hydrogel strategy released BDNF not only on the ipsilateral side, we may have also promoted axon plasticity on the uninjured side of the spinal cord. In contrast, in the current work, our anatomically targeted AAV2-BDNF injection approach resulted in 5-HT axon sprouting only at the site of injection and not in the nontargeted contralateral PhMN pool, demonstrating the utility of this strategy for restricting the location of plasticity vs. more diffuse delivery paradigms such as intrathecal injection and hydrogel-mediated release.

As previously discussed, AIH promotes ipsilateral hemidiaphragm recovery in lateralized cervical SCI by modulating excitatory neurotransmission in PhMNs in a serotonin- and BDNF-dependent manner (40). It is possible that AIH may only be promoting plasticity in receptor signaling involving synaptic connections with PhMNs that are spared by the injury, whereas our AAV2-BDNF approach stimulated robust focal sprouting of both rVRG and 5-HT axons within the phrenic nucleus and consequent generation of additional monosynaptic inputs of 5-HT and rVRG axons onto PhMNs. AIH may also promote sprouting of these important axon populations within the PhMN pool given that AIH involves BDNF synthesis, but this has not yet been evaluated. It is also possible that the level of BDNF expression produced by AIH is significantly less than that delivered using this viral vector–based strategy, raising the question of whether AIH is capable of stimulating a similar robust axon growth response observed as compared with AAV2-BDNF.

CONCLUSIONS

The essential function provided by the rVRG-PhMN-diaphragm circuit makes it a central focus for SCI intervention. Furthermore, cervical spinal cord is the most common location of traumatic SCI in the clinical population, leading to debilitating respiratory compromise that includes the need for mechanical ventilation, increased risk of upper respiratory infections, and reduced cough reflex (5, 6, 9). Because of the obstacles present in the adult CNS that limit or prevent axon regeneration, achieving regrowth of damaged rVRG axons and successful reconnection with their original PhMN targets is challenging following cervical SCI. There is thus a critical need to explore alternative modes of axon plasticity to develop therapeutic interventions, particularly those involving respiratory pathways spared by the injury.

In this study, we found that focally increasing BDNF expression at the location of PhMNs promoted significant restoration of diaphragm function after C2 hemisection SCI. By selectively labeling rVRG neurons, we observed that this recovery was accompanied by both a robust sprouting of spared rVRG axons locally within the PhMN pool on the side of injury and a significant increase in excitatory synaptic connections between these rVRG fibers and PhMNs. This excitatory input was likely also enhanced by an increase in serotonergic axon sprouting within the PhMN pool as well as an increase in synaptic input of these 5-HT axons directly onto PhMNs. Collectively, we uncovered various modes of neuroanatomical plasticity involving excitatory rVRG and modulatory serotonergic input within the cervical spinal cord which may have—including possibly in a synergistic manner—promoted PhMN activation and diaphragm recovery following SCI (Fig. 8). Furthermore, these data suggest that targeting circuit plasticity mechanisms involving the enhancement of synaptic inputs from spared axon populations is a powerful strategy for restoring diaphragm function after SCI.

Figure 8.

Model of axon plasticity after focal AAV2-BDNF injection. 5-HT axons modulate rhythmic excitatory input onto PhMNs from glutamatergic rVRG axons. A, B) Diagram depicting different forms of input onto PhMNs in AAV2-GFP–treated (A) and AAV2-BDNF–treated (B) rats after a C2 hemisection. C, D) Image of an individual PhMN with rVRG and serotonergic monosynaptic connections. E, F) Effect on diaphragm EMG amplitude after changes in synaptic input to PhMNs.

Glossary

5-HT

5-hydroxytryptamine

AAV2

adeno-associated virus serotype 2

AIH

acute intermittent hypoxia

BDNF

brain-derived neurotrophic factor

CMAP

compound muscle action potential

CTB

cholera toxin subunit B

EMG

electromyography

GC

genome copy

GFP

green fluorescent protein

KXA

ketamine-xylazine-acepromazine

NMJ

neuromuscular junction

PFA

paraformaldehyde

PhMN

phrenic motor neuron

RRID

Research Resource Identification Initiative

rVRG

rostral ventral respiratory group

SCI

spinal cord injury

Syn

synaptophysin

TrkB

tropomyosin receptor kinase B

VGlut2

vesicular glutamate transporter 2

VRC

ventral respiratory column

WGA

wheat germ agglutinin

AUTHOR CONTRIBUTIONS

B. A. Charsar designed and performed experiments, analyzed data, generated figures, and wrote the manuscript; M. A. Brinton, K. Locke, A. Y. Chen, B. Ghosh, M. W. Urban, S. Komaravolu, K. Krishnamurthy, R. Smit, and M. C. Wright performed experiments, analyzed data, helped generate figures, and commented on the manuscript; P. Pasinelli and G. M. Smith contributed key experimental reagents and commented on the manuscript; and A. C. Lepore designed experiments, contributed to writing the manuscript, and had final approval of the manuscript.