The crossed phrenic phenomenon.

The cervical spine is the most common site of traumatic vertebral column injuries. Respiratory insufficiency constitutes a significant proportion of the morbidity burden and is the most common cause of mortality in these patients. In seeking to enhance our capacity to treat specifically the respiratory dysfunction following spinal cord injury, investigators have studied the "crossed phrenic phenomenon", wherein contraction of a hemidiaphragm paralyzed by a complete hemisection of the ipsilateral cervical spinal cord above the phrenic nucleus can be induced by respiratory stressors and recovers spontaneously over time. Strengthening of latent contralateral projections to the phrenic nucleus and sprouting of new descending axons have been proposed as mechanisms contributing to the observed recovery. We have recently demonstrated recovery of spontaneous crossed phrenic activity occurring over minutes to hours in C1-hemisected unanesthetized decerebrate rats. The specific neurochemical and molecular pathways underlying crossed phrenic activity following injury require further clarification. A thorough understanding of these is necessary in order to develop targeted therapies for respiratory neurorehabilitation following spinal trauma. Animal studies provide preliminary evidence for the utility of neuropharmacological manipulation of serotonergic and adenosinergic pathways, nerve grafts, olfactory ensheathing cells, intraspinal microstimulation and a possible role for dorsal rhizotomy in recovering phrenic activity following spinal cord injury.

Introduction

Respiratory insufficiency is a common complication in patients sustaining spinal cord injury (SCI; Winslow and Rozovsky, 2003). This accounts for a significant proportion of the morbidity and mortality burden in patients with cervical SCI (DiMarco, 2005) and may (Oo et al., 1999) or may not result in ventilator dependency. The higher the level of injury, the more pronounced the degree of ventilatory dysfunction (Winslow and Rozovsky, 2003). Patients rendered ventilator-dependent by cervical SCI may benefit from phrenic nerve stimulation or pacing of respiratory (diaphragm, intercostal) muscles (DiMarco, 2009). The programming of such devices may benefit from determining the time-frequency pattern of individual phrenic motoneurons (PhMNs), which we have investigated in a previous study (Marchenko et al., 2012).

Functional consequences of hemidiaphragmatic paralysis include decreases in ventilation, perfusion, and ventilation/perfusion matching in the lower lobe of the lung ipsilateral to paresis. Small decreases in PO2 and vital capacity in the supine position along with a minimal increase in the alveolar-arterial oxygen difference may also occur (Arobrelius et al., 1975; Ridyard et al., 1976; Clague et al., 1979; Easton et al., 1983). Unilateral diaphragm paralysis is generally minimally symptomatic, but patients may exhibit mild compensatory tachypnea and dyspnea upon exertion (Easton et al., 1983). Patients with bilateral diaphragm paralysis, in contrast, typically are severely tachypneic and orthopneic, exhibiting dyspnea at rest and atelectasis (Gibson et al., 1989) and a respiratory acidosis that is more marked during sleep (Newsom et al., 1976).

The development of animal models of respiratory dysfunction following high cervical SCI has been instructive in elucidating underlying mechanisms contributing to recovery of previously silent neural respiratory outputs. As will be discussed, the C2-hemisected rat has dominated this exciting field and investigations have revealed mechanisms underlying spontaneous recovery following complete loss of ipsilateral medullophrenic drive. C2 hemisection (HSx) of the spinal cord silences the ipsilateral phrenic nerve (PhN) and hemidiaphragm (Goshgarian, 1979; O’Hara and Goshgarian, 1991), as well as the ipsilateral intercostal motoneuron pools (Dougherty et al., 2012b; Beth Zimmer et al., 2015). The crossed phrenic phenomenon (CPP), strictly defined, involves re-emergence of PhN or hemidiaphragm activity ipsilateral to a C2 HSx in response to respiratory stressors. Crossed phrenic activity, broadly defined, involves any re-emergence of PhN or hemidaphragm activity ipsilateral to injury: this may occur spontaneously (acutely or chronically) or in response to respiratory stressors (CPP). These observations demonstrate the potential for the phrenic motor neuron pool ipsilateral to hemisection to be driven by pathways in the contralateral hemicord crossing at cervical spinal levels and may be exploited therapeutically via a variety of means.

We will first discuss the historical context for our understanding of crossed phrenic activity, the models used to study the CPP, and changes in respiratory (mechanical and blood gas) parameters following high cervical SCI. We will then critically compare and evaluate the different electrophysiologic methods used to measure respiratory neural output (phrenic electroneurography and diaphragm electromyography (EMG)). We will then discuss neuroanatomical, neurochemical, and molecular mechanisms underlying crossed phrenic recovery, then conclude by discussing the potential for regenerative therapeutic approaches to be used in respiratory neurorehabilitation following cervical SCI.

The CPP: Historical Perspectives

In a well-designed set of investigations by Porter (1895), the supraspinal source for primary respiratory drive was confirmed and the CPP was systematically characterized. A major goal of Porter's studies was to disprove a spinal origin of respiration proposed by Brown-Séquard (1858), Langendorff (1880), and Wertheimer (1886), who reported observing breathing following spinal transection (Ghali and Marchenko, 2016a). Those advocating for the existence of a spinal respiratory central pattern generator proposed that hemisective or transective spinal cord injury transiently interrupts respiration by spinal shock inhibition. Porter challenged these findings: “irregular contractions have been seen in the diaphragm after its isolation from the cord by the section of the phrenic nerves; and that rhythmical contractions of this muscle have occurred from the stimulation of the phrenic nerve by the current of action of the contracting heart. To these sources of error I would add one previously undescribed, namely, the changes of intrathoracic pressure caused by the contractions of the trapezius and sterno-cleido-mastoideus muscles….I shall show….that long inhibition of respiration from section of the upper cervical cord does not exist.”

Brown-Sequard (1869), and previously Vulpian (1866), demonstrated that a HSx performed between C1–C4 occasionally does not affect breathing on either side and bilateral diaphragm contractions were later reported following C1/C2 HSx by Bert (1970). In two separate studies by Knoll (1885, 1888), respiration continued after HSx at the calamus scriptorius. The interpretation of these results is of course confounded by the subjectiveness of determining whether diaphragmatic/chest movement is active or passive. Porter goes on: “The phrenic nuclei on the cut side are not inhibited. On the contrary, they remain completely functional, for they begin again to send out normal impulses the moment the phrenic nerve of the opposite, uninjured side is cut”.

In his first experiment investigating crossed phrenic activity, Porter (1895) performed a left HSx at the C2 level on a morphine-sedated dog. “On cutting the right phrenic nerve just above the first rib, the right side of the diaphragm ceased to contract, while the left side, which had hitherto been passive, contracted strongly.” He repeated the same study on rabbits and confirmed his previous results. Thus, the CPP proper is defined as such: the appearance of phrenic or diaphragmatic activity ipsilateral to a supra-phreninuclear HSx of the cervical spinal cord following a phrenicotomy contralateral to the same in a spontaneously breathing animal. The latter distinction is significant in that phrenicotomy may induce CPP via 1) increased respiratory drive consequent to decreased tidal volume or 2) the severing of phrenic afferents that may inhibit contralateral PhMN output. The definition of the CPP has been extended to include induction by a variety of respiratory stressors, including hypercapnia, hypoxia, and asphyxia.

In contrast to current understanding of CPP, Porter (1895) did not hypothesize the existence of a bulbophrenic projection coursing contralateral to bulbar origin. Instead, he hypothesized PhMN dendrites crossing the midline receive inputs from descending projections in the contralateral lateral funiculus. He suggested that these contacts are insufficient by number and strength under resting conditions but are induced to activity following phrenicotomy on account of ‘redirection of phrenic impulses’: “The section of one phrenic nerve interrupts the ordinary respiratory path of the same side, and a greater portion, perhaps the whole descending impulse of that side passes through the crossed dendrites into the phrenic cells of the opposite side. The impulse reaching them is now strong enough to call forth contractions in the half of the diaphragm innervated by them.”

In 1951, Lewis and Brookhart investigated CPP in nembutal-anesthetized artificially-ventilated cats, recording diaphragm EMGs. Following a high cervical HSx between C1/C2, minute ventilation (VE, mL/min) was altered before and after phrenic nerve transection or cold block. No breathing was found at VE of 600 mL. With progressive decreases in VE, tonic, followed by phasic, respiratory activity, appeared on the intact and injured sides, with the injured side's threshold higher than the intact side. At 240 mL, both hemidiaphragms exhibited phasic activity. Additionally, intravenous injection of potassium cyanide, a potent peripheral chemoreceptor stimulant (Franchini and Krieger, 1993) with VE set at the apneic threshold, induced diaphragmatic activity bilaterally. Lastly, phrenicotomy or phrenic nerve cold block under conditions of artificial ventilation did not induce CPP, however, and it was concluded that increased respiratory drive is the mechanism underlying crossed phrenic activity.

Animal Models of Respiratory Function and Recovery following SCI

Overview

The C2-hemisected animal is the classic model used to study respiratory recovery following SCI (and the principal model described throughout this review; Goshgarian, 1979; see Additional file for comprehensive listing of studies). Typically, C2 HSx paralyzes the ipsilateral half of the diaphragm. However, as is rarely the case that human spinal cord injuries involve precise hemisections of the spinal cord (Gauthier et al., 2002; Li et al., 2003), other models for cervical SCI, with variable and/or incomplete loss of respiratory function (Baussart et al., 2006), may more accurately reflect SCI in patients. To this end, several high cervical SCI animal models have been developed.

High- and mid-cervical contusion injury models of high cervical SCI

El Bohy et al. (1998) developed a lateralized C2 contusion injury model in the rat by applying an impactor to the lateral aspect of the hemicord. Terminal experiments were performed 5 weeks following injury under chloral hydrate anesthesia. Rats with a C2 contusion used approximately 78% and 55% of respiratory reserve at rest, as determined by expressing PhN amplitude during eupnea as a percentage of that during asphyxia, ipsi- and contralateral to HSx, respectively; the latter value was not significantly different from spinal-intact animals.

Baussart et al. (2006) also utilized a C2 contusion model. Diaphragm activity recorded from spontaneously-breathing pentobarbital-anesthetized rats at one week (as % of contralateral control) was 27.4% and not significantly different in the subset of animals tested later at 1 month. Regrettably, there is no mention of activity recorded immediately following injury. In terminal experiments on pentobarbital-anesthetized artificially-ventilated rats, PhN activity ipsilateral to HSx was 38.2% of the contralateral; in a qualitative sub-study on these animals (n = 2), a contralateral right C1 HSx eliminated ipsilateral PhN activity and reduced recovery in the HSx-ipsilateral PhN by approximately 50– 75%, demonstrating the involvement of contralateral projections in recovery of phrenic output.

A modified version of the cervical contusion model, the dual injury, hemi-contusion post-contralateral hemisection, was developed by Awad et al. (2013). In this model, spared phrenic activity ipsilateral to a hemi-contusion is abolished by contralateral hemisection. The authors posit this model may be more clinically relevant for investigating cervical SCI, wherein diaphragm activity is more directly compromised. In order to create models more representative of human cervical SCI, other investigators have used contusions of the mid-cervical cord, which compromise white matter and damaged PhMNs simultaneously (Choi et al., 2005; Golder et al., 2011; Lane et al., 2012). Contusions involving C5 (+/– C4) resulted in acute tachypnea with subacute resolution, whereas higher contusions (C3/C4) did not change respiratory pattern. Importantly, although contusion and incomplete HSx injuries exhibit high variability in anatomical extent and physiological deficits even under the most controlled conditions, they are more representative of human cervical SCI and may prove useful in testing therapies that have been shown to improve respiratory recovery in the C2-hemisected animal.

Far lateral-restricted model of high cervical SCI

Vinit and colleagues (2006, 2008) developed a far lateral C2 HSx model for cervical SCI with the purpose of sparing the ventromedial funiculus to serve as a potential pathway for recovery. This group has shown variable PhN and diaphragm recovery (see below for comparison of 2006 vs. 2008 study) in terminal experiments conducted on anesthetized rats. Further studies are required to more thoroughly characterize recovery patterns in this model under different conditions.

The lateralized contusion injury model (El Bohy et al., 1998; Baussart et al., 2006) is advantageous in that it more closely resembles spinal cord injuries occurring in patients (than precise HSx without impact and cold block without injury). By preserving parenchymal medial continuity, this model allows the testing of more therapeutic approaches for neurorehabilitation following high cervical SCI. (i.e., attempts at regeneration through the ipsilateral hemicord).

Cold block model of high cervical SCI

Castro-Moure and Goshgarian (1996) developed a technique using cold block of the ventrolateral funiculus at C2 in the rat. Much of the innovation comes from the surgical approach, especially with regards to avoiding injury to the bilaterally-paired rostrocaudally-oriented large epidural venous complex in this area. Using recirculation of cooled ethyl alcohol through a cooling probe in contact with a thermo-paste placed on the spinal cord (to avoid direct contact), surface temperature of the spinal cord was reduced to 7°C. This resulted in a complete silencing of ipsilateral diaphragm EMG activity and crossed phrenic activity occurred in response to asphyxia. Diaphragm EMG activity was restored following reversal of the cold block. The cold block model is not ideal as an injury model per se, but it is useful in dissecting out the effects of interrupted descending drive from parenchymal damage proper.

The C1-hemisected unanesthetized decerebrate rat

We have recently shown progressive acute recovery of spontaneous crossed phrenic activity occurring over 2 hours in the C1-hemisected unanesthetized decerebrate adult rat (; Ghali and Marchenko, 2015). This new model offers several advantages, by avoiding the neural-suppressant effects of anesthesia () and sparing the C1–2 pre-phrenic interneuron group. Our findings suggest that crossed phrenic activity is state-dependent and/or that C1–2 pre-phrenic interneurons play a role in respiratory recovery following SCI. Anesthesia potently suppresses neural activity, including the respiratory network, insofar that administration of isoflurane anesthesia essentially eliminated acutely recovered spontaneous crossed phrenic activity in C1-hemisected decerebrate rats.

Dynamic changes in phrenic motor output following hemisection.

Following readministration of isoflurane, there is a suppression of phrenic nerve (PhN) amplitude. Hemisection performed under isoflurane silences the ipsilateral PhN. Over the ensuing recovery period, spontaneous crossed phrenic activity returns and there is a marked increase in central respiratory drive. Phrenic nerve amplitude is shown in millivolts (mV). Timescale bar is shown in lower right hander corner. Modified with permission from Figure 3 of Ghali and Marchenko, 2015. HSx: Hemisection; PhL: left phrenic nerve; PhR: right phrenic nerve.

Figure 3

Bulbophrenic network organization.

Color traces under phrenic neurograms indicate phase of activity (i.e., inspiratory, expiratory [post-I and E2], or tonically discharging units) of indicated excitatory (+) and inhibitory (–) synapses. The retrotrapezoid nucleus is a central chemoreceptor region which stimulated by low pH and provides tonic excitation to bulbar respiratory nuclei. The parafacial respiratory group and Bötzinger and pre-Bötzinger complexes generate the respiratory rhythm and control the output of pre-motor neurons in the ventral and dorsal respiratory groups. Pre-motor neurons in rVRG, in turn, PhMNs during inspiration, monosynaptically (pathways 1, 5, 6, and 7) and polysynaptically (pathways 2, 3, and 4) via C1–2 pre-phrenic interneurons (pathways 2 and 3) and possibly via local phrenic interneurons (pathway 4). Medullophrenic rVRG units project both ipsi- (pathways 1, 2, 4, and 7) and contralaterally (pathways 5 and 6; decussating at medullary or spinal levels) in lateral and ventromedial funiculi. A subset of medullophrenic rVRG units (crossed-and-recrossed) may decussate in the medulla and subsequently in the spinal cord to ultimately supply ipsilateral PhMNs (pathway 7). Inhibitory control of PhMNs is provided by descending medullophrenic BötC units (pathway 8) as well as local phrenic interneurons (pathway 9), which may relay the former as well as contribute to tonic inhibition of PhMNs independent of extrinsic drive. C1–2 pre-phrenic interneurons are concentrated in the C1 and rostral half of C2 spinal segments; C1 hemisection would thus spare the majority of C1–2 pre-phrenic interneurons, while a C2 hemisection would dissociate these units from caudally-related ipsilateral PhMNs. Not shown are putative contralateral monosynaptic bulbophrenic projections decussating between above the classic C2 hemisection and below our C1 hemisection site. Pathways: 1, ipsilateral monosynaptic medullophrenic projection; 2, ipsilateral polysynaptic medullophrenic projection via C1–2 pre-PhINs; 3, collaterals to C1–2 pre-PhINs from rVRG pre-MNs; 4, ipsilateral polysynaptic medullphrenic pathway via PhINs; 5, contralateral monosynaptic spinal-decussating medullophrenic pathway; 6, contralateral monosynaptic medullary-decussating medullophrenic projection; 7, crossed-and-recrossed monosynaptic medullary- and spinal-decussating medullophrenic projection; 8, descending ipsilateral medullophrenic inhibitory projection; 9, intraspinal inhibitory interneurons. Modified with permission from Figure 10 of Ghali and Marchenko, 2016a. BötC: Bötzinger complex; C1–2 pre-PhINs: C1–2 pre-phrenic interneurons; cVRG: caudal ventral respiratory group; HSx: hemisection; pFRG: parafacial respiratory group; PhL: left phrenic nerve; PhNucl: phrenic nucleus; PhR: right phrenic nerve; pre-BötC: pre-Bötzinger complex; RTN: retrotrapezoid nucleus; rVRG: rostral ventral respiratory group.

Anesthesia potently suppresses crossed phrenic activity.

Two hours following a hemisection, isoflurane is readministered to the animal. There is a potent suppression of phrenic bursting bilaterally, with discharge of the phrenic nerve (PhN) ipsilateral to HSx (PhL) nearly eliminated by anesthesia. Phrenic nerve amplitude is shown in millivolts (mV). Timescale bar is shown in lower right hander corner. Modified with permission from Figure 6 of Ghali and Marchenko, 2015. HSx: Hemisection; PhL: left phrenic nerve; PhR: right phrenic nerve.

The C4-hemisected anesthetized rat

Lee et al. (2014) compared and contrasted the effects of C2 versus C4 hemisection in an effort to produce an anatomically precise lesion of the mid-cervical cord, which is the most common site of human cervical SCI, and distinguish the effects of white matter from combined white and gray matter injury seen in mid-cervical contusion models (Choi et al., 2005; Golder et al., 2011; Lane et al., 2012). They found mid-cervical hemisection to be associated with improved tidal volume in comparison to high-cervical hemisection at rest but not under hypercapnic conditions and a diminished response to asphyxia after the mid-cervical lesion compared with the high-cervical one.

Sex differences in phrenic recovery following cervical SCI

Possible sex difference in CPP are suggested by the finding that increases in PhN amplitude ipsilateral to C2 HSx in response to hypoxia and asphyxia, as well as increases in tidal volume in response to hypercapnia, are greater in female rats than male and ovariectomized female counterparts (Doperalski et al., 2008). Examples of well-characterized hormonal-based differences in respiratory neural networks include the involvement of testosterone aromatization in long-term facilitation of phrenic and hypoglossal discharges in male rats in response to intermittent hypoxia (Zabka et al., 2006). Thus, sex differences may account for experimental variability and should be taken into account when making cross-study comparisons. This is clinically significant insofar that female patients exhibit better neurological recovery compared to male patients following SCI (Sipski et al., 2004), though this remains debated (Greenwald et al., 2001; Chan et al., 2013).

Changes in Respiratory Parameters and Arterial Blood Gases following Upper Cervical SCI

Goshgarian et al. (1986) were the first to systematically investigate changes in respiratory parameters and arterial blood gases following high cervical SCI. Twenty-four hours following C2 HSx, increases in respiratory rate, PO2, and arterial pH were observed, along with a non-significant decrease in PaCO2 and increase in HCO3– in unanesthetized rats (Goshgarian et al., 1986). One to two months following C2 HSx, spontaneously-breathing urethane-anesthetized rats exhibit increased respiratory frequency, diminished tidal volume, and respiratory alkalosis relative to control (Golder et al., 2001a, b), similar to breathing in humans following SCI (Estenne and De Troyer, 1987; Loveridge and Dubo, 1990; Loveridge et al., 1992). Recovery of minute ventilation and tidal volume occurs progressively over weeks to months following C2 HSx (Fuller et al., 2008; Dougherty et al., 2012a) and phrenicotomy reduces tidal volume when performed 2 or 8 weeks, but not 1–3 days, following injury, providing evidence for phrenic-mediated functional recovery (Dougherty et al., 2012a).

Functionally, upper cervical SCI results in an acute reduction in tidal volume compensated for increased respiratory frequency in both awake (e.g., Fuller et al., 2006) and spontaneously breathing anesthetized (e.g., Golder et al., 2001b) rats, resulting in unchanged minute ventilation (e.g., Fuller et al., 2006). It has been suggested that the increases in respiratory frequency compensating for diminished tidal volume capacity results from a vagal-dependent mechanism (Golder et al., 2001b), although chemoreceptor-drive likely plays an equally important role. Vagal-dependent pathways relaying pulmonary stretch receptor information may also tonically inhibit crossed phrenic activity, as the incidence of spontaneously-active CPP at 2 and 8 weeks was increased by intra-investigational bilateral vagotomy in artificially-ventilated urethane-anesthetized rats (Lee et al., 2010). It should be noted that the central respiratory rhythm synchronizes to the ventilator in vagus-intact animals (e.g., Ghali, 2015; Ghali and Marchenko, 2016b); vagotomy results in decoupling of the animal's central respiratory rhythm from the ventilator and decreased baseline respiratory frequency.

Tachypnea was observed 2 months following C2 HSx in anesthetized, vagotomized, artificially-ventilated rats (Golder et al., 2001a). Bilateral vagotomy in controls and upper C2-hemisected animals effected a reduction of respiratory frequency and augmentation of tidal volume that resulted in no statistical differences of these respiratory parameters between both groups (Golder et al., 2001b, 2003). This suggests that information regarding altered pulmonary mechanics is relayed centrally via vagal afferents to induce tachypneic hypopnea.

Curiously, vagotomy reduced minute ventilation but decreased PaCO2, suggesting a significant increase in alveolar ventilation occurring via reduction in deadspace by improved ventilation/perfusion matching in spinal-injured animals (Golder et al., 2001b). Reduced PaCO2 following vagotomy may also theoretically reflect decreased metabolic production as a consequence of gastrointestinal hypomotility (Macleod, 1974). Other changes in peripheral vagal modulation of respiratory function include diminution (Tsai and Lee, 2014) and attenuated recovery (Lee and Chang, 2014) of the pulmonary component of the Bezold-Jarisch reflex in C2-hemisected animals, which may permit recovery of the phrenic motor system as well as contribute to protection of minute ventilation following SCI (Lee, 2016).

The frequency response to hypercapnia in C2-hemisected unanesthetized rats is not different from spinal-intact animals, but some have shown that tidal volume is blunted and fails to improve with time (Fuller et al., 2006) and crossed phrenic activity remains delayed in onset in response to hypercapnia compared to contralateral PhN (Fuller et al., 2008). Dougherty et al. (2012a) demonstrated a progressive recovery in tidal volume responses to hypercapnia, but, at 2 months post-C2 HSx, this failed to approach the magnitude of those by spinal-intact animals. Analogously to responses to hypercapnia, those to hypoxia improve in C2-hemisected animals over the same time frame but do not normalize (Fuller et al., 2008).

Electrophysiologic Determination of Recovery Following High Cervical SCI

A major short-coming of studies investigating CPP is the high variability of results between and within different preparation types. This was first noted by Lewis and Brookhart (1951): “no group of experiments dealing with this subject has been productive of uniform results; inter- and intraspecies variations in the degree of respiratory hemiparesis produced by the spinal hemisection and in the amount of crossing produced by experimental manipulations are reported in every study.” One possible reason may involve inconsistencies in the electrophysiological recording of neural and diaphragmatic respiratory output.

Diaphragm electromyography assessment of crossed phrenic recovery: inter-study inconsistency

The crossed phrenic phenomenon occasionally occurs spontaneously in the subacute recovery period in the days to weeks after SCI in rats (Golder and Mitchell, 2005; Fuller et al., 2006). However, diaphragmatic recovery as recorded by EMG does not necessarily parallel that seen in phrenic neurograms. Some investigators show that spontaneous CPP recovery as expressed in diaphragm EMG is negligible (Goshgarian, 1981; Vinit et al., 2006, 2008) while others demonstrate significant recovery (Nantwi et al., 1999; Alilain and Goshgarian, 2008). Studies in guinea pigs have also yielded variable results when using diaphragm EMG to assess recovery, with demonstration of earliest CPP inducibility as early as 3.5 hours (Goshgarian and Guth, 1977) and as late as 2–7 months (Guth, 1976) following C2 HSx. Diaphragm EMG activity, however, can be consistently elicited in spontaneously-breathing animals by a contralateral phrenic transection (Vinit et al., 2006, 2008) or other respiratory stressor.

In spontaneously-breathing chloral hydrate-anesthetized rats, spontaneous recovery of CPP in diaphragm EMG occurred as early as 6 weeks following C2 Hsx, progressively increasing up to 16 weeks, but no recovery was observed at 4 weeks (Nantwi et al., 1999). The incidence of spontaneous recovery in vagotomized, artificially-ventilated, anesthetized rats was shown to be 25% at 1 month and 73% at 2 months (Golder et al., 2001a). Corroborating these findings, far-lateral C2 HSx (HSxL) (but neither medial nor dorsolateral section) silenced both the phrenic electroneurogram and diaphragmatic electromyogram in pentobarbital/morphine-anesthetized rats (with paracetamol/codeine on board; Vinit et al., 2006) acutely. However, spontaneous recovery occurred at 3 months in PhN but not diaphragm, in which CPP was activated only in response to contralateral phrenicotomy. In one study, HSx-ipsilateral diaphragm EMG remained silent 2 weeks following a C2 HSx sparing the dorsal funiculus, even in response to combined hypercapnia and hypoxia (5% CO2, 10% O2), in sponatenously-breathing ketamine/xylazine-anesthetized rats (Mantilla et al., 2007). These results underscore the better sensitivity of phrenic electroneurography compared to diaphragm EMG for detecting recovery of the phrenic motor system, especially in anesthetized animals (Sapru and Krieger, 1979).

In a later study, Vinit et al. (2008) showed spontaneous activity in ipsilateral PhN and/or diaphragm (in contradistinction to their 2006 study) within one week after injury which was greater at 3 months in the same animal model (pentobarbital/morphine-anesthetized rats with C2 HSxL). That spontaneous diaphragm activity was found as early as one week following injury in the 2008 study (Vinit et al., 2008), but not at 3 months in the 2006 study (Vinit et al., 2006) in the exact same animal model implicates inter-experimental variation (e.g., thoroughness of and avoidance of tissue distortion during HSx), inter-animal variability (e.g., hemodynamics, arterial blood gases, temperature stability), and/or spinal shock as important and over-looked confounders in interpreting the HSx literature.

Diaphragm electromyography assessment of crossed phrenic recovery: sources of confounding

It should be noted that since many studies do not quantify EMG activity, variability in threshold used to determine presence or absence of muscle unit firing may lead to inconsistency in the reporting of results leading to different conclusions regarding whether or not diaphragmatic recovery was obtained (Vinit and Kastner, 2009). One should also note the possibility of picking up contralateral diaphragm-synchronous motion artifact and reporting the same as recovery. The use of anesthesia, by state-dependently suppressing phasic and/or tonic descending bulbospinal drive, may also result in overestimation of the time-to-recovery of spontaneous CPP. Failure to observe recovery in diaphragm EMG after C2 HSx in the subacute period of recovery (Vinit et al., 2006; Mantilla et al., 2007) could also represent poor sensitivity as a consequence of the few motor units sampled by such recordings and may be addressed by recording from multiple diaphragm sites.

As a final consideration, – an unlikely scenario but nevertheless interesting and deserving of special mention – the Vulpian-Heidenhain-Sherrington phenomenon, first described in the 19th century, involves slow muscular contraction in the absence of neuromuscular innervation in response to activation of cholinergic fibers supplying associated vasculature (Kusurkar, 2004) and may underlie diaphragmatic contractions that have been infrequently seen following bilateral phrenicotomy (Porter, 1895). In theory, this phenomenon could account for false-positive determination of post-HSx recovery as assessed by diaphragm EMG, but would not confound results obtained by phrenic electroneurography.

Phrenic electroneurography assessment of crossed phrenic recovery: sources of confounding

The use of contralateral PhN amplitude as a control for PhN activity on the hemisected side with a uniform gain (10,000) applied to recordings bilaterally (e.g., Polentes et al., 2004) may not be truly representative of left-right differences, as the vagaries of PhN microdissection (i.e., desheathed vs. partial or complete desheating) and electrophysiologic recordings (i.e., exact distance between bipolar electrodes) rarely yield neurogram amplitude that are precisely left-right equivalent even under the most optimal conditions. Thus, ideally, phrenic neurograms should be recorded bilaterally, normalized to an equivalent voltage, and monitored continuously before and after HSx. This also permits the use of pre-HSx activity as control, leading to more consistent reporting of results. An alternative that has been used by investigators whose investigations necessitate foregoing pre- and post-HSx recordings of PhN activity is use of the homolateral PhN of spinal-intact animals as control (e.g., Nantwi et al., 2003a) in conjunction with a large sample size.

Neuroanatomical Basis for Recovery of Crossed Phrenic Activity

Adult animals

Since ipsilateral PhN activity becomes silent after acute C2 HSx in guinea pigs (Guth, 1976), mice (Minor et al., 2006) and rats (Goshgarian, 1979; Ling et al., 1994), it has been proposed that CPP does not contribute to PhN output at rest, but can generally be elicited in response to respiratory stressors such as contralateral phrenicotomy (Lewis and Brookhart, 1951), asphyxia (e.g., O’Hara and Goshgarian 1991; Yu and Goshgarian, 1993; Gould and Goshgarian, 1999; Golder et al., 2003; Doperalski et al., 2008), and chronic intermittent hypoxia (Fuller et al., 2003). In fact, asphyxia applied several minutes after C2 HSx has been shown to recover PhN activity ipsilateral to injury (Porter, 1895; Rosenbleuth and Ortiz, 1936).

Despite extensive structural and functional characterization of bulbospinal pathways in several animal models, to what extent ipsi- versus contralateral and phasic versus tonic pathways are active under uninjured versus injured and unstressed versus stressed conditions remains to be clarified () (Ghali, 2017). CPP is thought to be mediated by descending bulbospinal pathways coursing in the lateral funiculus contralateral to the injury (Goshgarian, 2003) and crossing within the spinal cord at the level of the phrenic nucleus (Goshgarian et al., 1991; Moreno et al., 1992) in the ventral white commissure to supply PhMNs in the contralateral ventral horn. Alternatively, these axons may not actually cross in the ventral white, but may make synaptic contact on contralateral PhMN dendrites extending past midline (Prakash et al., 2000; Boulenguez et al., 2007), as first proposed by Porter (1895).

Additionally, the ventromedial funiculus may also contribute to CPP by providing excitatory drive, as CPP elicited by a contralateral phrenicotomy in mice was greater if the ventromedial funiculus was spared (Minor et al., 2006). In another study, crossed phrenic activity occurred in all control rats where integrity of the ventral funiculus was preserved, but in none wherein HSx was complete (Li et al., 2003). Moreover, during terminal experiments in pentobarbital/morphine-anesthetized rats (Vinit et al., 2008), a complete C2 HSx performed contralateral to the initial HSxL eliminated spontaneous activity ipsilateral to the original side of injury in animals studied 7 days, but not 3 months, after initial C2 HSxL. Thus, subacute and chronic recovery may involve different pathways; in subacute recovery, contralateral bulbophrenic tracts descending in the lateral and ventral funiculi are utilized whereas, in delayed recovery, the ipsilateral ventral funiculus is involved.

Neonatal animals

Overview

The commencement of CPP immediately following C2 HSx is fundamentally different from CPP commencing several hours or days later, in that acute CPP is consistent with spontaneously active pathways whereas CPP occurring later may occur as a consequence of intracellular changes taking place as early as two to four hours following injury (Goshgarian et al., 1989; Sperry and Goshgarian, 1993; Castro-Moure and Gosharian, 1997; Hadley et al., 1999a). Evidence suggests that crossed phrenic activity may be spontaneously active following HSx in younger animals and this pathway may enter into functional latency with maturation, although the use of anesthesia precludes definitive conclusions.

In vitro studies

Prior to our study in C1-hemisected unanesthetized decerebrate rats (Ghali and Marchenko, 2015), spontaneous crossed phrenic activity occurring acutely following injury was only shown in neonatal animals (see below), initially in the in vitro brainstem-spinal cord en bloc preparation (pontomedullary and T8 transection; cervical and thoracic dorsal rhizotomy). Recovery ipsilateral to injury was approximately 20% of ipsilateral activity pre-injury (peak amplitude), although the raw recordings show that integrated burst area may be significantly less. Importantly, the respiratory discharge was irregular, exhibiting decrementing firing dynamics suggestive of gasping, consistent with the fact that this preparation is pontomedullary transected (St. John and Paton, 2004).

In vivo studies

Studies conducted in vivo in neonates reveal a rapid entrance into latency of crossed phrenic activity occurring well before the juvenile period. In spontaneously-breathing anesthetized rats, crossed activity is manifest throughout the entire diaphragm in early neonatal (P2–P4) animals, becoming restricted to the ventral diaphragm in late neonatal (P7–P28) animals, and completely abolished in P35 animals (Huang and Goshgarian, 2009a). Retrograde diaphragm labeling reveals midline-crossing PhMN dendrites in P2 animals which become diminished in P7 and P28 rats, and are not observed by P35, suggesting that retraction of midline-crossing PhMN dendrites are responsible for maturational changes in crossed phrenic activity (Prakash et al., 2000; Huang and Goshgarian, 2009c).

Juvenile and young adult animals

Remarkably, more than two decades prior to demonstrated spontaneously-active crossed phrenic activity in neonates, Goshgarian (1979) showed that there is a longer delay to elicit CPP following C2 HSx in juvenile rats (5–9 weeks old) compared to older animals (26 weeks old), suggesting a latent pathway contributing to contralateral phrenic activity that becomes functional over maturation. Another previous study by this lab (Yu and Goshgarian, 1993) investigated age differences in inducible CPP, noting a greater amplitude in older than younger animals but a similar overall incidence in the two groups. In our experience, we observe diminution, but persistence of phrenic nerve activity ipsilateral to a C1 hemisection with gradual recovery in the artificially-perfused in situ preparation of the decerebrate juvenile rat (unpublished data).

Mechanisms underlying prominent spontaneous crossed phrenic activity in younger animals

Spontaneous crossed phrenic activity in neonatal animals is likely attributed in large part to midline-crossing PhMN dendrites (Allan and Greer, 1997; Song et al., 2000), which represent a higher proportion of the PhMN pool in younger animals (Prakash et al., 2000), perhaps facilitated by removal of tonic inhibitory influence of the pons on respiration (Hilaire et al., 1989) in the en bloc preparation (St. John and Paton, 2004). Midline-decussating PhMNs may become retracted with age. Additionally, spontaneous crossed phrenic activity may be more prominent in neonatal animals as a consequence of higher input resistance of smaller PhMNs, necessitating smaller amounts of current to reach action potential threshold.

To reconcile the apparent contradiction of spontaneously-active crossed phrenic activity in neonates and more facile induction of crossed phrenic activity in older than younger animals that have passed the neonatal period, one may argue that weak but consistently active crossed activity mediated by PhMN dendrites crossing the midline in the neonate diminishes rapidly with growth and age-related increases in projections result in a stronger but latent inducible crossed phrenic drive.

Cellular and Neurochemical Adaptations Underlying Recovery of Crossed Phrenic Activity

Acute cellular adaptations following SCI

Structural adaptations

It is rather noteworthy that within 2–4 hours following C2 hemisection, the neural respiratory network within the spinal cord undergoes significant ultrastructural changes (Goshgarian et al., 1989; Sperry and Goshgarian, 1993; Castro-Moure and Gosharian, 1997; Hadley et al., 1999a). These include, but are not limited to, glial process retraction along with a significant increase in dendro-dendritic oppositions and double (and multiple) synapses. Under normal conditions, glial processes are interposed between and separate the cell body and primary dendrities; retraction of glial processes may effect the unmasking of dendro-dendritic appositions, which may improve coupling between and amongst PhMNs, amplifying bulbophrenic phasic and tonic synaptic inputs. Analogous ultrastructural adaptations occur in rats sacrificed 4 hours following hemi-cold block at the C2 level, indicating acute plasticity is a consequence of loss of descending drive rather than representing a non-specific effect of trauma (Castro-Moure and Gosharian, 1997).

Acute morphologic changes in the phrenic nucleus induced by a C2 HSx are attenuated by pre-treatment with an inhibitor of tryptophan hydroxylase, para-chlorophenylalanine (Hadley et al., 1999a), and the same treatment reduces asphyxia-inducibility of CPP acutely following C2 HSx (Hadley et al., 1999b). This suggests an important role for serotonergic signaling in respiratory cellular neuroplasticity. Acute recovery of inducible CPP occurring over several hours following HSx may parallel morphological adaptations, as asphyxia-induced crossed phrenic activity progressively increases over 1–6 hours following a C2 HSx, but does not improve significantly at 12 and 24 hours post-injury (O’Hara and Goshgarian, 1991). Mechanisms operant during the hyperacute/acute post-injury period may differ fundamentally from those occurring subacutely and chronically and understanding these differences may reveal new treatment targets to “protect” the respiratory network in the spinal cord, in addition to the current protocol of intravenous methylprednisolone aimed at protecting neural parenchyma in general.

Electrophysiological adaptation

A study by El-Bohy and Goshgarian (1999) provides some insight into the single cell firing properties of crossed pathways. They performed individual PhN axon recordings ipsilateral to C2 HSx along with intact HSx-contralateral PhN recordings and induced crossed phrenic activity by asphyxia 2–3 hours following HSx. Responses included increased single unit firing frequency as well as recruitment of previously silent PhMNs, with a greater increase in late-onset inspiratory (I) than early-onset I units. Theophylline-induced crossed phrenic activity was shown in one set of investigations to elicit predominantly late-I activity in PhN discharge (described by the investigators as “asynchrony”), further corroborating the hypothesis that PhMNs/diaphragm motor units underlying CPP are recruited from silence in an orderly manner (Nantwi et al., 1996), presumably in accordance with the size principle of Henneman (1965a, b). Moreover, hypoxia-induced short-term potentiation appears to preferentially affect late-recruited inspiratory PhMNs (Lee et al., 2015).

Recovery of predominantly late-I units was observed by our group (; Ghali and Marchenko, 2015) and delayed burst onset in recovering PhMNs ipsilateral to a C2 hemisection was similarly observed by Lee and colleagues (Lee et al., 2013; Lee, 2016). Late-onset I units tend to fire with lower frequencies while early-onset I units are comprised by PhMNs underlying synchronous medium- and high-frequency oscillations in PhN spectra (Marchenko et al., 2012). Uncovering the recruitment order and activation patterns of specific subpopulations of the PhMN pool following HSx in response spontaneously and in response to respiratory stressors would greatly further current understanding of laterality of bulbophrenic inputs controlling PhMNs with different firing patterns and overall respiratory network organization. This knowledge, in turn, could be exploited to optimize the design of phrenic and diaphragm pacing devices (i.e., time-frequency pattern of stimulation).

Recovery of late-inspiratory activity ipsilateral to hemisection.

Following a C1 HSx, late-inspiratory activity is often observed to recover initially. Phrenic nerve amplitude is shown in millivolts (mV). Timescale bar is shown in lower right hander corner. Modified with permission from of Ghali and Marchenko, 2015. HSx: Hemisection; PhL: left phrenic nerve; PhR: right phrenic nerve; sec: second.

Role of neurochemical pathways in crossed phrenic activity

Zimmer and Goshgarian (2007a) have shown that CPP is tonically inhibited by a mechanism utilizing GABAAergic pathways, demonstrating that fast inhibitory synaptic transmission plays a critical role in CPP, as it contributes to phrenic pattern formation in normal uninjured conditions (Marchenko and Rogers, 2009; Marchenko et al., 2015). Additionally, other investigators have demonstrated the involvement of serotonergic signaling in CPP (Ling et al., 1994; Zhou et al., 2001). Thus, a variety of neurochemical pathways appear to interact to regulate phrenic output at rest as well as modulate respiratory plasticity following injury. For example, it has been shown that serotonergic transmission is involved in long-term potentiation in PhMNs induced by stimulation of pre-motor neurons (Fuller et al., 2002) and administration of a 5-HT1A agonist increases PhN output (Zimmer and Goshgarian, 2006). Moreover, 5-HT1A agonists have been shown to hyperpolarize membrane potential of dorsal horn neurons (Hains et al., 2003b), which, in turn, have been shown to inhibit PhMNs by a GABAA-dependent, but glycine-independent, mechanism (Zimmer and Goshgarian, 2007a). Clearly, therefore, it is the interaction of and intersection between neurochemical pathways which underlies spontaneous CPP as well as acute, subacute, and chronic recovery of neural respiratory function.

Role of fast inhibitory transmission

Evidence for the role of fast inhibitory synaptic transmission (i.e., GABAA- and glycinergic signaling) following SCI comes from a study by Zimmer and Goshgarian (2007a). In vagotomized, paralyzed, artificially-ventilated, urethane-anesthetized rats, GABAA- or glycinergic antagonists were topically applied to the dorsal surface of the cervical spinal cord 1 week following C2 HSx. The method of application was as follows: laminectomy and durotomy inclusive of C3–7 were performed and drugs pipetted across this length during bilateral PhN recordings. Bicuculline and gabazine (GABAAergic receptor antagonists), but not strychnine (glycinergic receptor antagonist), induced CPP, and the effects of the former were reversed by application of muscimol. This suggests that CPP is tonically inhibited via GABAA- but not glycinergic mechanisms. Elevation of the animal's head and unchanged respiratory frequency argue against the proposition that these drugs affected supraspinal respiratory centers, although it is not inconceivable that the activity of upper cervical inspiratory neurons at C1/C2 (Lipski et al., 1993; Tian and Duffin, 1996a, b) was somehow modified.

GABAAergic inhibition of CPP may derive from supraspinal regions, local intraspinal, or both. Irrespective of the origin, gradual down-regulation of GABAAergic inhibition over time may account for the observation that CPP becomes spontaneous (i.e., in the absence of respiratory stress) eventually (Nantwi et al., 1999). Other than GABAAergic mechanisms, serotonergic signaling may be implicated in inhibitory control of phrenic output. For example, it has been shown that activation of 5-HT1A receptors increases PhN activity on the side of a C2 HSx (Zimmer and Goshgarian, 2006) and dorsal rhizotomy can activate a latent CPP (Fuller et al., 2002), suggesting PhN afferents activate dorsal horn neurons (DHNs), which consequently tonically inhibit CPP.

Role of serotonergic mechanisms

Intraventricular administration of the serotonergic-selective neurotoxin 5,7-dihydroxytryptamine has been shown to reduce the incidence of spontaneous CPP recovery 2 months following C2 HSx in vagotomized artificially-ventilated anesthetized rats (Golder et al., 2001a) and blockade of 5-HT2A receptors prevents acute intermittent hypoxia-induced long term facilitation in PhN activity ipsilateral to a C2 HSx in the same preparation (Golder and Mitchell, 2005). Conversely, intravenous administration of 5-hydroxytryptophan (Zhou et al., 2000) or the 5-HT2A/2C agonist DOI (Zhou et al., 2001) one day following HSX recovered ipsilateral PhN activity. Activation of 5-HT2 receptors has been shown to enhance theophylline-induced crossed phrenic activity (Basura et al., 2002), providing evidence for interaction among pathways. The usefulness of these findings is challenged by a study showing enhancement of hemidiaphragm EMG ipsilateral to HSx, but not tidal volume, in response to systemic serotonergic stimulation (Hsu and Lee, 2015). The therapeutic utility of manipulating specific serotonergic signaling pathways in respiratory neurorehabilitation following SCI requires further investigations.

Role of adenosinergic signaling

A role for adenosinergic signaling in recovery of crossed phrenic activity has been extensively investigated and holds promise as a therapeutic target (Bascom et al., 2005). Nantwi et al. (1996) demonstrated that intravenous administration of theophylline during recording induced crossed phrenic activity in chloral hydrate-anesthetized rats one day following C2 HSx. Crossed phrenic activity could not be induced by theophylline if the animal was pre-treated with an adenosine receptor agonist (L-PIA), nor in response to a methylxanthine lacking adenosine antagonistic properties (enprofylline; Nantwi et al., 1998b; Nantwi and Goshgarian, 2001) nor one that is peripherally selective lacking central activity (8-(p-sulfophenyl)theophylline; Nantwi and Goshgarian, 2001). Later, it was shown blockade of adenosine A1, but not A2, receptors following HSx which is specificially responsible for inducing the crossed phrenic phenomenon. These investigations establish theophylline as promising in the treatment of respiratory dysfunction following SCI and implicate central adenosine A1 receptor antagonism as an important mechanism underlying observed crossed phrenic recovery in response to methylxanthines. Efficacy of methylxanthines in respiratory recovery has also been shown in animal models of cervical contusion (Hoy and Alilain, 2015). Since blockade of adenosine 2A receptors has been shown to improve long-term facilitation in diaphragm in response to chronic intermittent hypoxia (Navarette-Opazo et al., 2014), adenosinergic pathways involved respiratory neuroplasticity and recovery following SCI are complex and a more thorough understanding of the same will require further investigations.

As adenosine A1 receptors are Gi-protein coupled, their blockade results in increased activity of adenylate cyclase, greater cAMP synthesis, and consequent protein kinase A activity and pharmacological manipulation of these pathways should elicit similar effects to those observed with methylxanthines. This, indeed, is the case, as crossed phrenic activity has been induced in rats following C2 HSx by treatment with the cAMP analogue 8-Br-cAMP (Kajana and Goshgarian, 2008a), the phosphodiesterase (PDE) inhibitors pentoxyfylline (Kajana and Goshgarian, 2008b) and rolipram (Kajana and Goshgarian, 2008b, 2009), and the adenosine A1 receptor antagonist DPCPX (Nantwi and Goshgarian, 2002; Kajana and Goshgarian, 2008b).

Dose-dependent efficacy for use of theophylline during subacute/chronic recovery has been shown in rats treated with theophylline for 3 – 30 days after a C2 HSx. In these animals, crossed phrenic activity was evident at post-treatment recording times but absent in untreated animals (Nantwi et al., 1998b, 2003a) and persisted following discontinuation of treatment (in some cases improving with increasing duration of post-treatment discontinuation, see Nantwi et al., 2003a). However, theophylline treatment for 4 months following HSx suppressed or did not affect crossed phrenic recovery that had occurred spontaneously (Nantwi et al., 2003b), suggesting a critical period wherein this treatment is effective and after which may be harmful.

The systemic administration of adenosine antagonists may have untoward and undesired consequences. To this end, Minic et al. (2017) innovatively investigated the administration of the adenosine antagonist DPCPX in a nanoconjugate-bound form directly into the diaphragm following C2 hemisective injury and found it to effect increased tidal volume and minute ventilation and diminished increases in respiratory rate associated with high cervical cord injury. Such a therapy has the benefit of producing the desired clinical response without exposing the patient to systemic adverse effects of adenosine antagonists.

Modulation of respiratory recovery following SCI by concomitant pre-phrenicotomy

Animals with combined C2 HSx and ipsilateral phrenicotomy studied 2 months later did not exhibit CPP spontaneously although crossed output was inducible by hypoxia (Golder et al., 2003), in contrast to HSx-only animals, 12 of 15 of which had spontaneous phrenic activity ipsilateral to injury. This may be attributable to axotomy induced molecular and cytoarchitectural alterations in PhMN somata intraspinally (Gould and Goshgarian, 1997; Liou and Goshgarian, 1994, 1997) and altered firing properties of phrenic axons (Titmus and Faber, 1990; Liou and Goshgarian 1994). This process can loosely be considered as the transynaptic variant of double crush syndrome (Upton and McComas, 1973), whereby a single lesion activates plasticity mechanisms but a preceding second lesion significantly compromises recovery.

Gould and Goshgarian (1999) suggest that a microglial, but not astrocytic, proliferative response occurs in response to phrenicotomy. They showed that intracisternal treatment prior to HSx with the mitotic inhibitor cytarabine reduces microglial proliferation and diminished phrenicotomy-related attenuation of recovery following SCI. Microglial proliferation may be induced by the axotomy of the phrenic nerve and subsequent activation of these cells may include features such as increased cytokine signaling capacity which may underlie morphological alterations at the phrenic pre-MN-PhMN synapse (Giulian and Baker, 1985; Barron et al., 1990; Giulian et al., 1994). This has important implications for multi-trauma patients, who may sustain cervical SCI in conjunction with peripheral nerve (e.g., phrenic) damage.

Role of afferents in recovery of crossed phrenic activity

Lewis and Brookhart (1951) demonstrated that CPP is elicited by phrenicotomy as a consequence of increased central respiratory drive as opposed to the transection of fibers proper, which argues somewhat against a role for phrenic afferents in CPP inducibility. However, PhN stimulation was shown to result in inhibition of 40% of the contralateral PhMN pool (as determined by intracellular recording) and this effect was abolished by sectioning dorsal roots of the stimulated PhN (Gill and Kuno, 1963) and unaffected by complete spinalization above the phrenic nucleus. These, among other, findings led investigators studying high cervical hemisection to hypothesize that loss of afferent inhibition may additionally play a role in phrenicotomy-induced CPP (Goshgarian, 1981). An anatomical basis for crossed afferent inhibition has been shown in mammals, where afferent fibers terminate in the contralateral ventral (Escolar, 1948; Sprague, 1958; Kerr, 1961; Sprague and Hongchien, 1964) and dorsal (Culberson et al, 1979) horns.

Several investigators have suggested a role for phrenic afferents in CPP recovery (Hadley et al., 1999a, b), relaying ventrally via dorsal horn neurons. Notably, and in contrast to other muscle groups, there exist a paucity of muscle spindles in the diaphragm (Corda et al., 1965; Von Euler, 1968), yet PhN stimulation results in increased neurogram amplitude (Butler et al., 2003), suggesting that the few afferents that do exist in the phrenic system are capable of producing marked changes in the motor output. It has been suggested that sensory neurons in the PhN may subsume a greater role after injury than normal conditions and may contribute to recovery following cervical SCI (Corda et al., 1965).

Dorsal rhizotomy in the subacute recovery period following C2 HSx elicits CPP (Goshgarian, 1981), suggesting afferents may play a predominantly inhibitory role in phrenic recovery. Additionally, following a far-lateral C2 HSx, PhN transection eliminates previously recovered activity; this does not occur in cervical dorsal root-gangliectomized animals, supporting the hypothesis that the effect is due to afferent as opposed to motoneuron axotomy (Vinit et al., 2007). Since axoaxonic synapses occur in the phrenic nucleus (Goshgarian and Rafols, 1984), presynaptic inhibition is another candidate mechanism by which phrenic afferents may inhibit CPP or phrenic output in general. Understanding afferent regulation of crossed phrenic activity may lead to the development of medical therapies targeting specific neurochemical pathways or the use of direct surgical intervention, such as selective C3–6 dorsal rhizotomy, to improve respiratory function following SCI.

Experimental studies

Three to twenty-eight days following C2 HSx, terminal experiments were performed on chloral hydrate-anesthetized C2-hemisected rats. Dorsal rhizotomy of the cervical spinal cord (C3–6; a few additionally at C7–8) on the intact side (contralateral to HSx) elicited CPP (diaphragm EMG) ipsilateral to HSx and sometimes augmented activity ipsilateral to rhizotomy (Goshgarian, 1981). In the animals in which additional C7–8 dorsal rhizotomization was performed, CPP was further enhanced. These findings physiologically demonstrate that phrenic nucleus receive contralateral inhibition from afferents at and below phrenic nucleus levels. The role of phrenic primary afferents in inhibiting CPP was later revisited by Zimmer and Goshgarian (2006) who performed C2 HSx in vagotomized, paralyzed, artificially-ventilated, anesthetized rats. Following variable postoperative periods (1 day, 2 days, 1 week, 16 weeks), the 5-HT1A agonist 8-OH-DPAT was topically applied to the dorsal spinal cord surface or administered systemically, resulting in activation of CPP.

It is instructive to comment on a series of investigations by Fuller and colleagues (2002). One week following cervical dorsal rhizotomy (C3–6), vagotomized, paralyzed, artificially-ventilated, urethane-anesthetized rats were hemisected at C2. Relative to cervical afferent-intact animals, dorsally-rhizotomized rats exhibited greater PhN amplitude in response to contralateral ventrolateral funiculus stimulated rostral to HSx level. Responses in rhizotomized animals were unaffected by intravenous administration of the serotonin antagonist methysergide, arguing against a role for 5-HT mediated inhibition. However, since the animals were phrenicotomized (for PhN recordings), what was actually being investigated was the role of non-phrenic cervical afferents, which may tonically inhibit PhMN output via serotonin-independent mechanisms.

Mechanisms underlying phrenic afferent inhibition of crossed phrenic activity

There are several candidate pathways mediating motor inhibition by phrenic afferents: primary afferent-ascending relay-descending inhibition, via synapses on cervical respiratory-modulated neurons, via direct synapse on PhMNs, and via non-respiratory-modulated DHNs (Iscoe and Duffin, 1996). The latter two are the most viable options, as phrenic afferents in close approximation to PhMNs (Song et al., 1999) have been demonstrated and the activity of a subset of non-respiratory-modulated DHNs nevertheless responds to PhN stimulation (Cleland and Getting, 1993; Iscoe and Duffin, 1996). Descending inhibition of dorsal horn neurons has been well demonstrated in locomotor systems. For example, multireceptive DHNs in the lumbar spinal cord become more excitable following a T13 HSx (Hains et al., 2003b) and fire more frequently in response to a rostrally-placed cold block (Wall, 1967; Brown, 1971; Handwerker et al., 1975), in part as a result of the elimination of descending serotonergic inhibition (Yaksh and Wilson, 1979; Hains et al., 2002).

Following SCI, 5-HT levels are significantly reduced in the spinal cord (Hains et al., 2002; Golder and Mitchell, 2005) and DHNs become hyperexcitable (Hains et al., 2003a, b). Since DHNs are known to be inhibited by descending serotonergic (5-HT1A) signaling (Thor et al., 1993), it is possible that 8-OH-DPAT mediates its effect by silencing of DHNs and facilitation of PhMN activity. Thus, loss of descending serotonergic inhibition of serotonergic inhibitory DHNs may account for reduced responsiveness (i.e., excessive hyperpolarization) of PhMNs caudal to a HSx to contralateral bulbophrenic inputs. Descending serotonergic inputs, originating from raphe magnus, are bilateral and have intraspinal decussations through lamina X (Skagerberg and Bjorklund, 1985) and bilateral monosynaptic supply to PhMNs from raphe structures has been demonstrated by Dobbins and Feldman (1994), suggesting some sparing following HSx and possible involvement in acute and/or chronic respiratory recovery.

Long-Term Respiratory Recovery following SCI

Functional considerations

Long-term respiratory recovery has been demonstrated in humans, with diaphragmatic recovery occurring in approximately one-third of patients ipsilateral to SCI in one series (Oo et al., 1999) and some individuals achieving ventilator independence (DiMarco, 2005). Although spontaneous CPP may be observed a few days after C2 HSx in animal models, a progressive increase in PhN output is observed over long time scales in rats (Nantwi et al., 1999; Fuller et al., 2006; Vinit et al., 2006, 2008); recovery after 1 week is approximately 25% and after 3 months is 40% of control (Vinit et al., 2008). Despite progressive recovery of the PhN ipsilateral to C2 HSx over the months following injury (Nantwi et al., 1999), this is not correlated with increased resting tidal volume (Fuller et al., 2006), but phrenicotomy ipsilateral to the side of injury 2 months post-C2 HSx reduces tidal volume suggesting improved PhN output account for functional recovery (Golder et al., 2003). Additionally, increased vital capacity is demonstrated by the finding that chronically-injured animals exhibit larger tidal volumes under respiratory stressors (i.e., hypercapnia, hypoxia, asphyxia; Golder et al., 2003) compared to acutely-injured animals.

As normoxemic normocapnia typically characterizes arterial blood gases following C2 HSx (Goshgarian et al., 1986; Golder and Mitchell, 2005; Doperalski and Fuller, 2006; Fuller et al., 2006), the progressive increase in PhN output observed during chronic recovery is not the consequence of increased chemo-drive, although the increased respiratory frequency may be sustained by, in addition to vagal afferent feedback mechanisms, minor alterations in arterial pH from decreased tidal volume which is adequately compensated by the tachypnea. Although animal models of high cervical SCI remain normoxic, chronic (Fuller et al., 2003) and acute (Golder and Mitchell, 2005; Doperalski and Fuller, 2006; Lee et al., 2015), intermittent hypoxia does enhance recovery of PhN activity ipsilateral to C2 HSx. This is significant insofar that patients may develop hypoxemia following trauma due to a variety of causes (i.e., pneumonia, atelectasis, pulmonary embolism, respiratory insufficiency), which may play a role in long-term respiratory neuroplasticity contributing to functional recovery (Oo et al., 1999).

Neuroanatomical substrate for long-term recovery of crossed phrenic activity

A fundamental question is whether phrenic recovery following SCI reflects activation of pre-existing pathways or a regenerative process, or both. In chronic recovery of a far-lateral C2-sectioned animal, PhN recovery is eliminated by damage to the contralateral medial spinal cord, but unaffected by damage to the contralateral lateral funiculus (Vinit et al., 2008), consistent with a hypothesis whereby long-term recovery involves pathways projecting in the ventromedial funiculus and not requiring the lateral funiculus. Despite the observation that PhN output increases time-dependently, no such correlation is observed for PhN activity elicited from electrical stimulation of contralateral lateral funiculus (location of CPP; Fuller et al., 2006) and in far-lateral C2-sectioned animals, PhN recovery occurred only if the ventromedial funiculus was preserved (Li et al., 2003), providing further evidence for CPP-independence of long-term recovery. The latter finding goes a step further to suggest that the acute/subacute recovery observed with activation of latent CPP not only is not responsible for long-term recovery but in some manner becomes silenced. Another study showed that at 1 month following C2 HSx, respiratory frequency was lower and tidal volume higher in ventromedially-spared animals compared to those in which the ventromedial funiculus was lesioned (Fuller et al., 2009). Vinit et al. (2006, 2007) suggest that bulbophrenic projections in the ventromedial funiculus are not used normally and are too weak to be involved in acute/subacute recovery, being recruited later by the induction of long-term respiratory plasticity.

Sprouting of ventral corticospinal tract (CST) axons into ventral horns has been demonstrated following damage to the dorsal CST via a dorsal funiculus-limited HSx in rats (Weidener et al., 2002) and mice (Steward et al., 2008), demonstrating that the animal may utilize pathways other than the native projections to achieve functional recovery. This has also been shown for the respiratory system in rats following a C2 HSx and transplantation with olfactory ensheathing cells at the lesion site (Li et al., 2003; see below).

There are several proposed mechanisms for increased functionality of the normally dormant ventromedial bulbophrenic synapses onto PhMNs in chronic injury models. Goshgarian (2003) suggests that nerve terminals already synapsing onto PhMNs become more strongly coupled to the latter. Evidence for this is provided by a study by Alilain and Goshgarian (2008) demonstrating an increase in the expression of medullary NR2A and GluR1 after SCI, suggesting the activation of mechanisms involved in strengthening of glutamatergic synapses. Moreover, both serotonin and 5-HT2A receptors are found in greater abundance following C2 HSx (Golder and Mitchell, 2005; Fuller et al., 2006). Additionally, axons that normally supply adjacent non-phrenic motoneurons may form collaterals onto PhMNs or propriospinal interneurons may be recruited into the phrenic network (Vinit and Kastner, 2009).

Molecular mechanisms underlying long-term recovery of crossed phrenic activity

Molecular basis for phrenic motor system neuroplasticity

Increased excitability of PhMNs ipsilateral to hemisection is thought to underlie long-term recovery of crossed phrenic activity. This is believed to be mediated, in part, by changes in glutamate receptor subtype expression. It is well known that recovery of crossed phrenic activity is associated with upregulation of NMDA receptor subunits and a downregulation of AMPA receptor subunits (Alilain and Goshgarian, 2008). Specifically, chronic recovery following C2 HSx is characterized by increased expression of NR2A and GluR1 and decreased expression of GluR2 in the phrenic nucleus, consistent with enhanced NMDA and non-NMDA glutamatergic signaling (Alilain and Goshgarian, 2008) within the PhMN pool. Glutamate receptor expression is in turn governed by tropomyosin-related kinase receptor subtype B (TrkB), which has been shown to be critically involved in phrenic recovery following high cervical spinal cord injury during eupnea (Mantilla et al., 2013, 2014; Martínez-Gálvez et al., 2016; Gill et al., 2016; Gransee et al., 2013, 2017; Hernandez-Torres et al., 2017) and other respiratory behaviors (Hernandez-Torres et al., 2017). These changes act in concert to render PhMNs more excitable and receptive to descending inputs.

Transfection of PhMNs with adenoviral-associated vector-delivered TrkB increases the probability of phrenic recovery following SCI and effects a corresponding large increase in the expression of NMDA receptor subunit 1. Hemisection proper diminishes the expression of the AMPA receptor subunit 1, which is expressed to a greater degree in animals with greater degrees of recovery (Gransee et al., 2013, 2017). Intrathecal administration of BDNF to stimulate TrkB signaling was shown to enhance phrenic recovery following high cervical hemisective cord injury and blockade of the same using TrkB-Fc soluble fusion protein or siRNA targeting TrkB was shown to prevent respiratory recovery (Mantilla et al., 2013). Whether this recovery was a consequence of phrenic versus supraspinal effects remains to be clarified. Importantly, expression of glutamatergic signaling machinery can be targeted using therapeutic administration of brain-derived neurotrophic factor (BDNF) to stimulate the TrkB pathway.

Following axonal transection, increased expression of GAP-43 and c-Jun, which promotes axonal sprouting (Raivich et al., 2004). Increased expression of GAP-43 in the ventral horn caudal to C2 HSx, but a decrease in entire ipsilateral C1–6 hemicord, suggests that increased capacity for functional plasticity is specific for MN pools (Vinit et al., 2009). In contradistinction, upregulation of GAP-43 in the contralateral cervical hemicord is observed, suggesting that recovery of bulbophrenic input involves de novo axonogenesis via primarily crossed pathways (Vinit et al., 2009). Moreover, it has been shown that regenerative potential is inversely proportional to the distance between the site of injury and corresponding soma (Jenkins et al., 1993). Following SCI at cervical levels, bulbospinal and rubrospinal neuron somata increase the expression of genes associated with regeneration (Jenkins et al., 1993; Houle et al., 1998; Vinit et al., 2005, 2009), while corticospinal tract somata do not (Mason et al., 2003), since they are further from the site of injury. Specifically, c-Jun increases in rubrospinal neurons following HSx at C3 (Jenkins et al., 1993; Houle et al., 1998) but not T10 (Jenkins et al., 1993) and increased expression of genes implicated in regenerative processes occur in layer V pyramidal neurons of CST following axotomy within the cortex but not at the cervical spinal cord (Mason et al., 2003).

Involvement of nitric oxide in respiratory neuroplasticity was demonstrated by Capková et al. (2011), wherein staining for neuronal nitric oxide synthase in terminals near PhMNs and b1 subunit of guanylyl cyclase in phrenic motoneurons was significantly reduced contralateral to a C2–3 hemisection and nearly abolished ipsilateral to the cord injury. Further studies are necessary to elucidate the role of nitric oxidergic signaling in respiratory neuroplasticity following SCI, an enhanced understanding of which could hold significant therapeutic promise, given the wide availability of pharmacological compounds acting on this pathway.

Molecular basis for supraspinal neuroplasticity

In afferent-intact spontaneously-breathing C2-hemisected animals, respiratory rate is increased to compensate for the decrease in tidal volume (Goshgarian et al., 1986; Fuller et al., 2006; Dougherty et al., 2012a); however, in vagotomized, anesthetized, paralyzed, and artificially-ventilated rats (Golder et al., 2001a), as well as the in vitro brainstem-spinal cord en bloc preparation of the neonatal rat (Zimmer and Goshgarian, 2007b), respiratory frequency is lower following C2 HSx, demonstrating that central drive is reduced. Moreover, in the in vitro neonatal rat preparation, the response to hypercapnia is paradoxically characterized by shorter burst duration and smaller burst area than control. The observed increase in respiratory rate following SCI in vivo in vagus-intact animals and decrease in respiratory rate in vivo in vagotomized animals and in vitro are likely consequences of supraspinal plasticity secondary to alterations in pulmonary stretch receptor, chemoreceptor, and spinobulbar pathway activity.

Adaptations of supraspinal respiratory circuitry following high cervical SCI (Buttry and Goshgarian, 2014; Ghali and Marchenko, 2015) likely underlie changes in central drive and increased phrenic/hemidiaphragm activity contralateral to hemisection (Ghali and Marchenko, 2015; Lee and Hsu, 2017). In seeking to uncover molecular mechanisms underlying the observed plasticity, Zimmer and Goshgarian (2007b) quantified changes in neurotransmitter receptor subunit levels two days following a C2 HSx in the in vitro brainstem-spinal cord en bloc preparation of P0/P1 neonatal rats. Medullary levels of glutamate decarboxylase (65 kDa isoform; GAD65) and NR2B (NMDA receptor subunits) were increased ipsilateral to HSx and of A1 (adenosine receptor subunits) and neurokinin-1 (NK1) contralateral to HSx. Levels of NR1, phospho-NR2B/totalNR2B, GluR2, and NK1 were decreased ipsilateral to HSx and levels of phospho-NR1, phospho-NR1/totalNR1, phospho-NR2, phospho-NR2/totalNR2, and GluR2 were decreased contralateral to HSx.

The aforementioned changes in neurotransmitter receptor subunits may be non-specific, reflecting glutamate excitotoxicity (Bittigau and Ikonomidou, 1997; Liu et al., 1999) at the site of HSx or axotomy-induced cell death. However, the general trend of an increase in subunits associated with inhibitory transmission and a decrease in those involved in excitatory transmission suggests coordinated change and is consistent with the central bradypnea reported in the in vitro neonatal rat preparation and vagotomized anesthetized animal in vivo. The supraspinal increase in NK1 contralateral and decrease ipsilateral to injury are noteworthy, since NK1-positive cells in the ventral medulla have been implicated in chemoreception in conscious rats (Nattie and Li, 2006) and their selective obliteration in the pre-Bötzinger complex, the kernel of respiratory rhythmogenesis (Smith et al., 1991), causes respiratory dysrhythmia (Gray et al., 1999, 2001). Thus, SCI-induced molecular changes are expressed in respiratory network dynamics and manifest as changes in breathing pattern.

Role of interneurons in respiratory plasticity following high cervical hemisection

It has been suggested that following injury, spinal circuitry undergoes such significant reorganization to the degree that its microanatomy and neurocytoarchitecture are fundamentally altered (Dimitrijevic et al., 1997). One of the mechanisms utilized in recovery of PhN output ipsilateral to a C2 HSx, contusion, or analogous injury may involve the recruitment and incorporation of interneurons (Lane et al., 2009b; Sandhu et al., 2009) as shown in other motor systems (Jankowska, 2001; Bareyre et al., 2004; Courtine et al., 2008, 2009).

In contrast to spinal-intact animals, where pseudorabies virus transynaptic retrograde tracing fails to label the C1–2 respiratory-related interneuron pool, following C2 HSx, retrograde tracing of the diaphragm contralateral to injury demonstrates extensive bilateral labeling of pre-phrenic interneurons at the C1–2 spinal levels (Lane et al., 2009a). Recruitment of these cells subsequent to hemisection and their coaptation by the phrenic motor network may be a consequence of development of polysynaptic projections to the phrenic nucleus. Whether these units are one and the same as the C1–2 respiratory-related units previously described and shown to receive supraspinal input (Lipski et al., 1993), as described in the rat (Lipski et al., 1994) and cat (Hoskin and Duffin, 1987a, b; Mateika and Duffin, 1989), remains to be determined.

Diaphragmatic changes following cervical SCI

Plasticity of the neural respiratory network following SCI involves not only neural elements subserving respiration but also effector muscles. Thus, therapeutic interventions aimed at improving respiratory function via neurorehabilitation should take into consideration concomitant changes occurring in the target muscle over time, which may lead to changes in the recruitment order and firing patterns of motoneurons and motor units (Henneman et al., 1965a, b; Webber and Pleschka, 1976; Marchenko et al., 2012).

The diaphragm contains all muscle fiber classes (Fournier and Sieck, 1988; Sieck et al., 1989, 1995), with resting respiration principally utilizing MHCslow and MHC2A fibers, while MHC2X and MHC2B fibers are reserved for strenuous exertion (Sieck et al., 1995; Geiger et al., 2000). The diaphragm actively contracts (i.e., duty cycle) during approximately one-third of each respiratory cycle (Kong and Berger, 1986). In contrast to appendicular and axial musculature, the diaphragm does not bear weight. Thus, changes in the diaphragm following SCI reflect primarily altered myoneural interactions as opposed to loss of gravitational loading (Rowley et al., 2005).

Following C2 HSx, the diaphragm atrophies (Gill et al., 2014). Neuromuscular junction (NMJ) microanatomy (planar area, nerve terminal and motor end plate length) of myosin heavy chain (MHC)slow and MHC2A fibers does not change significantly, while an increase in planar area occurs at MHC2X and MHC2B NMJs, perhaps as an adaptation to enhance synaptic efficiency (Prakash et al., 1999). In fact, neuromuscular transmission failure in an isolated phrenic nerve-diaphragm preparation taken from C2-hemisected animals two weeks following injury was less than that observed in spinal-intact animals, possibly reflecting increased synaptic overlap at MHCfast (2A, 2X, 2B) fibers (Prakash et al., 1999).

The relatively unaffected MHC content per half-sarcomere (MHC-hs) following SCI, representative of the number of actomyosin cross-bridges that can form in parallel, suggests that the capacity for maximal tetanic force generation is at the most minimally affected, which, when measured directly, exhibits a decrease of only ~10% in hemisected animals (Miyata et al., 1995). MHC-hs is also roughly correlated with Ca2+-activated actomyosin ATPase activity and the latter has also been shown to be relatively unchanged following SCI, with only a small decrease observed in MHC2B fibers (Zhan et al., 1997).

Respiratory and locomotor recovery: similarities and differences

Differences between the pre-existing pathways subserving respiration and locomotion are reflected in differential mechanisms underlying recovery following injury (Majczyński and Sławińska, 2007; Ghali and Marchenko, 2015; Ghali, 2017). For example, the corticospinal tract relays to lumbosacral MNs polysynaptically via interposed pre-motor propriospinal interneurons (Courtine et al., 2008). As a result, locomotor recovery is characterized mainly by the outgrowth of collaterals (Fouad et al., 2001; Weidner et al., 2001) and specific interruption of the dorsal corticospinal tract will result in sprouting of collaterals from the ventromedial corticospinal tract (Weidner et al., 2001; Steward et al., 2008). In contrast, as the bulbospinal respiratory network, especially the bulbophrenic network, is characterized by direct monosynaptic bulbospinal projections (Dobbins and Feldman, 1994; Kastner and Gauthier, 2008; Vinit and Kastner, 2009), a potential substrate for recovery is the recruitment and incorporation of extra-motor propriospinal interneurons into descending pathways for use as relays and/or amplifiers (Vinit and Kastner, 2009). However, the few pre-phrenic interneurons (Lipski et al., 1993, 1994; Tian and Duffin, 1996a, b; Lu et al., 2004; Lane et al., 2008) that do exist, and which possibly function in a minor polysynaptic pathway, remain unchanged in number 2 weeks after C2 HSx (Lane et al., 2008), arguing against this as a primary mechanism facilitating long-term respiratory recovery.

Regenerative and Non-Regenerative Therapeutic Approaches

Respiratory muscle training regimens and phrenic nerve pacing are useful therapeutic modalities (Winslow and Rozovsky, 2003), but they are limited in the degree to which they can achieve lasting improvement of intrinsic function. While many studies in animal models of respiratory dysfunction following SCI have emphasized basic physiology and underlying neurochemical mechanisms, a few studies have looked at the applicability and utility of regenerative neurobiology, which may be combined with traditional treatments, to achieve functional recovery of the respiratory system. Recently, techniques in regenerative neurobiology, including implantation of olfactory ensheathing cells across an SCI site, nerve grafts, as well as intraspinal microstimulation, have been applied toward enhancing respiratory recovery in spinal cord-injured animals and offer promising potential for patients suffering ventilator-dependency following spinal trauma.

Nerve grafting

Nerve grafts provide the central nervous system circuits with the microenvironmental milieu of the peripheral nervous system to encourage regrowth of axons and regeneration (Houle, 1991). Gauthier et al. (2002) implanted the proximal end of a nerve graft in the ventrolateral medulla (control graft implanted in dorsolateral medulla) 3 months before performing a HSx at C3 followed by implantation of the distal nerve graft in C4 in phrenic nucleus. PhN activity was elicited by stimulation of grafts implanted in the medulla ventrolaterally, but not those implanted dorsolaterally, suggesting a specific innervation of the phrenic nucleus by inspiratory pre-motor neurons. The use of medullo-phreninuclear bridges to achieve respiratory recovery is a promising approach in patients with extensive SCI significantly compromising descending respiratory drive.

Efficacy of nerve grafting in improving PhN recovery has also been demonstrated (Gauthier et al., 2006). The proximal cut end of the left recurrent laryngeal nerve was grafted to the distal cut end of the PhN and exhibited recovery five months later, at which time diaphragm EMG was recorded following complete C2 spinal transection or right phrenicotomy. Moreover, in grafted animals, respiratory compensation was adequate to sustain survival following right phrenicotomy, but not in non-grafted controls. Thus, the use of peripheral nerves as donor grafts, combined with advances in growth factor neurobiology, may be a therapeutically-effective modality for treating patients with spinal cord or PhN injury (Houle, 1991).

Olfactory ensheathing cells

Olfactory ensheathing cells (OECs) have attracted interest on account of their propensity to encourage regeneration in the central nervous system (Li et al., 1998). Following an upper cervical HSx (approach through atlanto-occipital membrane without laminectomy), animals were transplanted with OECs and neural respiratory recovery compared to controls (Li et al., 2003). Terminal experiments were performed 2 months later on rats under paralyzed artificially-ventilated, gallamine-paralyzed, and anesthetized (unspecified drug) conditions. Ipsilateral to HSx, spontaneous PhN activity in control animals occurred only if there was sparing of the ventral funiculus and no PhN recovery was observed during asphyxia in rats with a histologically-confirmed complete HSx. In 19 of 24 treated animals, PhN activity recovered under spontaneous breathing conditions. Curiously, this recovery only persisted in 17 animals following paralysis with gallamine and asphyxia, suggesting the unspecified anesthetic type and depth confounded the results. The authors found neural repair of descending tracts in the ventral funiculus correlated with PhN recovery in treated animals; whether this represents regeneration of axotomized fibers (Li et al., 1997, 1998; Ramón-Cueto et al., 1998) or de novo axonogenesis (Thallmair et al., 1998).

Polentes et al. (2004) performed a similar experiment to Li and colleagues (1998), wherein OECs were transplanted into the lesion site in C2-hemisected rats, with a recovery period was 3–6 (rather than 2) months. Subsequently, PhN and diaphragm activity ipsilateral to C2 HSx were 80.7 and 73% of the side contralateral to injury, respectively, in terminal experiments conducted under spontaneously-breathing pentobarbital-anesthetized conditions. Subsequently, a C1 HSx performed contralateral to initial C2 HSx eliminated, reduced, or did not change recovery in the phrenic neurogram, suggesting variable contributions of regenerative processes on the ipsilateral side, crossed pathways, and possible effects of spinal shock. Subsequent mid-sagittal section of the cervical spinal cord (C1–6) did not change PhN activity ipsilateral to HSx, arguing against the role of crossed and re-crossed pathways. Moreover, stimulation ipsilateral and rostral to the original C2 HSx elicited PhN responses before, and weaker responses after, contralateral C1 HSx in transplanted (6/8), but not in control, animals. These findings suggest that therapeutic intervention may shift regenerative processes from natural mechanisms of recovery toward pathways that would otherwise be unused.

Evidence for the utility of other neural stem cells in promoting similar recovery derives from the work of Sandhu et al. (2017). Neural progenitor cells derived from the subventricular zone transplanted intraspinally below the C2 lesion site in adult rats enhanced phrenic recovery. This recovery correlated with the proportion of surviving transplanted cells. They localized principally to the white matter and a subset of them was found to have differentiated into glia.

Intraspinal microstimulation

Use of intraspinal microstimulation has been extensively investigated in the context of locomotor recovery following spinal cord injury, but no such study had explored the utility of this intervention in respiratory recovery following SCI until recently. Mercier et al. (2017) found that genioglossus-triggered intraspinal microstimulation at the ventral aspect of spinal segment C4 in rats effectively enhanced phrenic recovery acutely and subacutely following a rostrally-related C2 hemisection of the cord. These data suggest this approach to be promising in promoting respiratory recovery following SCI.

Conclusion and Perspectives

The present-day conceptual model underlying crossed phrenic activity still remains nebulous. The crossed phrenic phenomenon currently is understood as a latent pathway that can be recruited under conditions of moderate to severe respiratory stress. However, one may propose the more general conclusion that CPP is “state-dependent.” Redefining crossed phrenic activity in this manner, then, suggests that it is not necessarily the loss of ipsilateral phasic control from the rostral ventral respiratory group which fully accounts for silence in PhN activity (and hemidiaphragmatic paresis) ipsilateral to C2 HSx, that is unrecoverable by the contralateral projection acutely, but rather, the loss of descending tonic excitatory support of the membrane potential. This would be the case if 1) tonic drive is primarily ipsilateral and/or 2) contralateral tonic drive is polysynaptic, whereas ipsilateral tonic drive is monosynaptic, which would make the former more state-dependently sensitive to suppression by anesthesia, hypocapnia, or lung stretch receptors in experimental investigations. In this regard, it becomes critical to note that many studies investigating the C2-hemisected rat model of acute SCI do so under conditions of anesthesia, which may confound results by diminishing the already reduced tonic drive to the ipsilateral phrenic nucleus, leading to inaccurate inferences about the native configuration of the system.

Current management of SCI involves standard resuscitation, intensive care, and/or surgical measures and thus, has remained rudimentary at best, failing to parallel the tremendous leaps made in the fields of regenerative neurobiology and network neuroscience. The study of CPP following high cervical spinal cord hemisection in the rat has significantly furthered our understanding of the respiratory network's capacity to adapt following injury. It has also served as an impetus for the development of novel therapeutic approaches aimed at recovering respiratory function in patients sustaining SCI, including neuropharmacological agonists and antagonists targeting serotonin and/or adenosine receptors. Additional strategies which appear promising include use of brief periods of mild hypoxia to induce long-term facilitation in respiratory networks and dorsal rhizotomy to disinhibit crossed phrenic activity. Regenerative approaches have also been used successfully in experimental models to enhance respiratory recovery following SCI, including nerve grafts to promote axonal regeneration (into PhN or from the medulla into the phrenic nucleus), implantation of OECs at the injury site to encourage axonal growth, as well as intraspinal microstimulation. These strategies require extensive further investigation before they can become candidate for human clinical trials, but nevertheless hold tremendous promise as therapeutic options for respiratory neurorehabilitation following SCI.

Additional file:

Crossed phrenic phenomenon in rats as assessed by diaphragm electromyography.

Crossed phrenic phenomenon in rats as assessed by diaphragm electromyography

Click here for additional data file.

Crossed phrenic phenomenon in rats as assessed by phrenic electroneurography.

Crossed phrenic phenomenon in rats as assessed by phrenic electroneurography

Click here for additional data file.