Secondary Logo

Journal Logo


Synergy between Acute Intermittent Hypoxia and Task-Specific Training

Welch, Joseph F.1; Sutor, Tommy W.1; Vose, Alicia K.1; Perim, Raphael R.1; Fox, Emily J.1,2; Mitchell, Gordon S.1

Author Information
Exercise and Sport Sciences Reviews: July 2020 - Volume 48 - Issue 3 - p 125-132
doi: 10.1249/JES.0000000000000222
  • Free
  • Journal Club


  • After spinal cord injury, combined acute intermittent hypoxia (AIH) and task-specific training (TST) enhance motor function beyond that expected from either treatment alone.
  • Mechanisms underlying this synergistic therapeutic strategy are unknown.
  • We propose a hypothetical working model including both neural network and cellular elements to explain synergy between AIH and TST.
  • Our model is intended to spur needed mechanistic studies and accelerate progress toward clinical application of AIH as a plasticity primer in neurorehabilitation.


Plasticity is a fundamental property of neural systems, defined as a persistent change in neural system morphology and function based on prior experience. Experiences initiating neuroplasticity include synaptic activity, exercise/physical activity, environmental challenges such as altitude sojourn, and the onset of neurological disease or injury.

Spinal cord injury (SCI) impairs motor function, leaving many patients with chronic disability. Rehabilitation interventions seek to harness activity-dependent plasticity to partially restore function after neurological insult; unfortunately, functional gains are limited. Combinatorial therapies may enhance plasticity and functional improvements. One emerging method of augmenting the impact of task-specific exercise training (TST) is therapeutic (low-dose) acute intermittent hypoxia (AIH) (1). In this brief review, we consider the knowns and unknowns regarding the disproportionate benefits of combined AIH-TST, and propose a working model to explain the synergy (to produce a combined effect greater than the sum of individual treatment effects) between these treatments.

Concepts driving our model are largely derived from decades of basic science research on the phrenic motor system and other neural systems such as the hippocampus. Whether the same mechanisms pertain to other motor systems or to humans is unknown, representing key areas for future investigation. In rodents, brief (≤5 min) episodic (3–15 exposures per day) periods of moderate hypoxia (inspired oxygen: 9%–15%) increase phrenic nerve activity for more than 90 min post-AIH (2) — an effect known as phrenic long-term facilitation (LTF). AIH should not be confused with the chronic intermittent hypoxia associated with sleep-disordered breathing (15, 60-s hypoxic episodes, more than five episodes per hour; approximately 8 h·d−1; many days to years).

Intermittent electrical stimulation of axons from chemoafferent neurons activates brainstem neural networks that quickly (and reversibly) increase breathing (i.e., chemoreflex), and triggers slower mechanisms of spinal synaptic plasticity. Specifically, serotonergic neurons in the medullary raphe nuclei are activated by carotid chemoafferent neurons, initiating and orchestrating phrenic motor plasticity (3). Raphe neurons project to spinal motor nuclei where they release serotonin on or near alpha-motor neurons. Serotonin type 2 receptors (5-HT2) then activate intracellular signaling cascades that ultimately strengthen synapses between brainstem premotor and spinal alpha-motor neurons in an activity-independent manner (4). These neural network and intracellular mechanisms of motor plasticity underlie the therapeutic potential of AIH. Indeed, single or repeated AIH presentations improve breathing in rodent models of cervical SCI (5).

After decades of work investigating AIH-induced plasticity in the phrenic motor system, we came to realize that AIH effects are not unique to phrenic motor neurons. In fact, AIH influences diverse motor neuron pools, including hypoglossal respiratory (6), and locomotor limb motor neurons (1). For example, repetitive AIH improves both forelimb and respiratory function in rats with cervical SCI (7). Similar cellular mechanisms are thought to underlie plasticity in each of the relevant motor pools. However, forelimb improvements are minimal with AIH or TST presented alone. Significant functional benefits are observed only when AIH is delivered before TST on each training day (8). Because combined AIH plus TST produces an effect greater than the sum of individual treatments, there is synergy in their impact on motor plasticity.

Human AIH trials after SCI are underway. To date, nine studies have been published demonstrating efficacy in leg/ankle strength (9), hand function (10), locomotion (11), dynamic balance (12), and breathing (13). Moreover, combined AIH preceding over-ground walking practice (i.e., TST) improves walking function more than the sum of their individual effects when presented alone (11). Thus, the most effective rehabilitation strategy may incorporate AIH as a preconditioning stimulus (or plasticity primer) to amplify the benefits from TST (8,11).

Although functional improvements with combined AIH-TST are impressive (e.g., more than twofold greater increase in walking endurance with daily AIH and walking practice versus daily AIH alone [37% vs 17%] [11]), little evidence exists concerning mechanisms giving rise to their synergy. Understanding such mechanisms in multiple motor systems is of considerable importance because it may enable 1) optimization of AIH-TST therapy; 2) development of new treatments for other neuromuscular disorders that compromise movement (e.g., amyotrophic lateral sclerosis, Pompe disease, multiple sclerosis); and 3) conceptual translation to other forms of neuromodulation or plasticity, such as closed-loop vagal nerve, epidural, or transcutaneous spinal cord stimulation. In this review, we present a hypothetical model of motor plasticity with the aim of guiding new research, enabling a greater understanding of the synergistic relation between AIH and TST. Before we present our model, we provide additional background on core concepts guiding model development.


Over the past 30 years, considerable progress has been made toward a comprehensive understanding of cellular mechanisms giving rise to AIH-induced phrenic motor plasticity. This work inspired, and continues to guide, translation of basic science discoveries into human clinical trials, including investigations of nonrespiratory motor systems. Human studies demonstrating improved limb function after AIH (9) were directly influenced by earlier studies in rodent models that shaped our current knowledge (14).

Neural Network Mechanisms

During hypoxia, peripheral chemoreceptors located at the carotid artery bifurcation activate chemoafferent neurons that synapse in the brainstem nucleus of the solitary tract (nucleus tractus solitarii [NTS]). Second-order NTS neurons project to contralateral respiratory neurons of the ventral respiratory column, including rhythm generating neurons of the pre-Bötzinger complex and premotor neurons of the ventral respiratory group. Projections from NTS (direct) or ventral respiratory group (indirect) neurons activate midline serotonergic neurons of the caudal raphe nuclei. Serotonergic neurons in the raphe nuclei project broadly, including the spinal cord where they release serotonin in the phrenic motor nucleus. Cervical spinal 5-HT2 receptor activation is both necessary (15) and sufficient (16) for moderate AIH-induced phrenic LTF. Although serotonin type 7 (5-HT7) receptor activation elicits similar phrenic motor facilitation (17), its cellular mechanism is completely distinct. Overall, moderate AIH-induced phrenic LTF requires caudal raphe neuron activity, serotonin release, and 5-HT2 receptor activation in the phrenic motor nucleus versus tissue hypoxia per se.

In contrast to moderate AIH, the same AIH protocol consisting of severe hypoxic episodes (arterial oxygen pressure <30 mm Hg) elicits a phenotypically similar but distinct form of phrenic LTF that requires spinal tissue hypoxia and adenosine accumulation (18), presumably from glial adenosine release. In this situation, adenosine 2A (A2A) receptor activation on phrenic motor neurons initiates a distinct intracellular cascade, leading to phrenic motor facilitation (18). Although the practical utility of severe AIH protocols in humans is limited because of safety concerns, it is important to recognize that subthreshold A2A receptor activation competes with and undermines the serotonin-induced mechanism necessary for functional improvement.

Intracellular Mechanisms

At least two opposing phrenic LTF mechanisms arise from AIH protocols that vary in severity and duration of hypoxic episodes. Both mechanisms operate within phrenic motor neurons but differ in their reliance upon specific serotonin and adenosine receptors that trigger completely distinct intracellular signaling cascades. These signaling cascades are named for the G proteins canonically coupled to the initiating receptor, the Q (Gq) and S (Gs) pathways to phrenic motor facilitation, respectively (14).

With moderate AIH, the serotonin-dependent Q-pathway dominates phrenic LTF. Serotonin activates 5-HT2 receptors, initiating a signaling cascade that includes extracellular signal-regulated kinase/mitogen-activated protein kinase (ERK/MAPK) activity (19), de novo brain-derived neurotrophic factor (BDNF) protein synthesis, activation of the high-affinity BDNF receptor TrkB within phrenic motor neurons (20), and downstream signaling via the protein kinase C isoform, PKCθ. We postulate that PKCθ phosphorylates NMDA and AMPA (21) receptors, increasing glutamate receptor currents and/or trafficking/insertion, thereby strengthening bulbospinal synaptic inputs to phrenic motor neurons.

With severe AIH, A2A receptors activate adenylyl cyclase, increasing cyclic adenosine monophosphate (cAMP) formation. High cAMP levels activate a pathway including cAMP exchange protein (EPAC), PI3 kinase/Akt, and mammalian target of rapamycin (mTOR). Subsequent synthesis of an immature TrkB isoform strengthens synaptic inputs to phrenic motor neurons. Activation of 5-HT7 receptors initiates the same mechanism (22).

The Q- and S-pathways interact via powerful cross-talk inhibition (23). Thus, net plasticity reflects the inhibitory balance of opposing pathways. Where moderate AIH favors the Q-pathway (24), subthreshold S-pathway activation constrains its expression (23). On the other hand, severe AIH shifts the balance to S-pathway dominance (18,22); Q-pathway activation constrains its expression (22). With intermediate AIH, the two pathways cancel one another, and phrenic LTF is not observed (22). Similarly, with sustained moderate hypoxia, time-dependent adenosine accumulation and S-pathway activation undermines the Q-pathway, canceling phrenic LTF (25) (please see Devinney et al. (26) and Perim et al. (22) for further information).

Due to these complex interactions, protocol details are a key factor determining phrenic LTF expression. Harnessing AIH as a therapeutic modality requires understanding of this balance and how to minimize pathway competition. Translational efforts to date have focused on moderate AIH because of fears of tissue hypoxia; hence, successful treatment strategies using AIH must optimize BDNF synthesis. Because TST and moderate AIH both increase spinal BDNF, AIH-induced BDNF synthesis may prime the nervous system, accentuating therapeutic benefits of subsequent TST — a concept elaborated hereafter.


Physical exercise promotes plasticity and improves neuromotor function with neurological injury or disease. One established mechanism of exercise-induced neuroplasticity is activity-dependent production of growth/trophic factors. However, exercise alone is not always sufficient to support complete functional recovery of specific tasks. The success of exercise-based neurorehabilitation depends on congruency between the trained motor function and the targeted motor networks (i.e., TST).

TST is a repetitive functional practice of a motor task specific to the intended outcome. For example, when cats with SCI are trained to stand, standing ability improves but not stepping ability (27). This form of task-specific exercise can result in the acquisition of new motor skills, refinement of existing motor skills, or partial restoration of lost motor skills. Furthermore, TST may promote recovery by triggering neuroplasticity in specific neural circuits, including changes in pre- and postsynaptic function (28), neuromodulation via monoamines (e.g., serotonin, norepinephrine) (29), and new synapse formation (30).

Supraspinal Plasticity

TST-induced plasticity may occur at supraspinal sites. Indeed, sensory and motor cortices are reorganized by TST (31). Synaptic plasticity changes cortical maps (32) and may strengthen specific motor circuits practiced during TST. Other supraspinal structures, such as the cerebellum, also may exhibit TST-induced plasticity.

The neurotrophin, BDNF, is implicated in TST-induced cortical/cerebellar plasticity and motor learning. During skilled motor training, cortical neurons upregulate BDNF and undergo synaptic potentiation and synaptogenesis, thereby reinforcing practiced skills. Although exogenous BDNF applied to the motor cortex of rats with unilateral SCI fails to elicit functional recovery, exogenous BDNF coupled with TST improves task performance (33). Thus, in certain situations, combined BDNF and TST elicit recovery of motor function.

Spinal Plasticity

TST also strengthens spinal functions, including altered spinal reflex strength. For example, strength training augments Hoffmann (H) reflex amplitude during a maximum voluntary contraction (34). Increased spinal reflex strength after 4 weeks of training at least partially results from a net increase in excitatory synaptic inputs onto motor neurons (35). H-reflex amplitude also exhibits plasticity after operant conditioning (36). Down-conditioning affects motor neuron axon conduction velocity and firing threshold (37), affects the size and number of γ-aminobutyric acid (GABA) terminals on motor neurons, and alters ventral spinal GABA interneurons (38). The mechanisms of H-reflex up-conditioning are less well understood but may involve increased strength in glutamatergic terminals on motor neurons. Thus, acquiring or improving motor skill performance may necessitate refinement of a specific motor plan (supraspinal), increased spinal motor output, or both.

Brain-Derived Neurotrophic Factor

Treadmill and running wheel exercise increase BDNF in the hippocampus (39) and spinal cord (40–42). Increased BDNF synthesis may relate to increased neural network activity or medullary serotonergic neuron activation during exercise (43,44). Lumbar BDNF mRNA is upregulated by reflex training in spinally transected rats (42). Therefore, TST may increase spinal BDNF expression in specific neurons via activity and serotonin-dependent mechanisms, potentially strengthening excitatory synapses onto motor neurons. Although both AIH and TST can increase spinal BDNF synthesis, their specific effects, in magnitude and distribution, are not known. In our model, we suggest that AIH has a more widespread impact on BDNF expression, priming the system for additional, localized BDNF synthesis in specific neural circuits activated by TST. Combined interventions may elevate BDNF within neurons/synapses above a functional “threshold” for plasticity, giving rise to unique therapeutic opportunities. Such task-specific, BDNF-dependent plasticity synergy may occur in cortical and spinal elements of the motor circuit. For the purposes of this review, we focus on spinal elements; however, because AIH increases BDNF in the motor cortex (45), we do not exclude the possibility of additional cortical mechanisms.


Although there has been steady growth in our understanding of AIH-induced plasticity and its implications for neurorehabilitation, recent evidence suggests that therapeutic benefits are greatest with paired AIH-TST, at least with nonrespiratory motor behaviors (8,11,46). Unfortunately, few studies have combined AIH and TST (see the Table), and even fewer directly demonstrate that combined training is more effective than either treatment alone (8,11). More research in this area is warranted. Nevertheless, we base our hypothesis on these carefully designed experiments.

Previous investigations of AIH-TST

In a rodent model of forelimb function after cervical SCI, AIH and TST (horizontal ladder walking) were ineffective when applied separately; functional benefits were observed only when applied together (8). In this same study, non–task-specific treadmill training failed to improve ladder walking performance, even when combined with AIH (8). Thus, AIH was effective when presented in combination with TST, but not nonspecific exercise training. In humans with chronic, incomplete SCI, five consecutive days of AIH marginally improved walking speed, whereas daily walking practice had minimal impact (11). However, when AIH was presented 30 min before over-ground walking practice, functional benefits in walking endurance were greatly enhanced 3 d posttreatment (11). These two studies (one rodent, one human) were the inspiration for the working model and hypothesis presented here.

Neither rodent nor human experiments of combinatorial therapy have been thoroughly investigated. Many other motor functions remain to be explored, including the respiratory motor pools that formed much of our current understanding of AIH-induced motor plasticity. Because AIH exerts widespread effects on motor neuron BDNF expression (7,45), our contention is that AIH acts as a global priming mechanism for motor plasticity, enhancing localized plasticity mechanisms triggered by TST. By priming the system, AIH augments subsequent activity-dependent plasticity, improving motor recovery in specific tasks in people with neurological injury/disease. We hypothesize that combined AIH-TST is more effective than the sum of benefits elicited by each treatment presented alone (i.e., they are synergistic) because 1) AIH broadly elevates BDNF synthesis in diverse neural circuits, and 2) TST further elevates BDNF above a functional threshold for motor plasticity only in neural circuits activated during the specific motor task practiced. In Figure 1, a neural network model of AIH- and TST-activated pathways and their convergence onto spinal motor neurons is illustrated. In Figure 2, we illustrate a working intracellular model of signaling cascades within task-targeted motor neurons. Functional outcomes from the neurochemical convergence of sequenced AIH-TST are illustrated in Figure 3.

Figure 1
Figure 1:
Neural network model. Postulated neural circuits activated by AIH and TST. Hypoxia stimulates carotid body chemoreceptors and their chemoafferent neurons, activating second-order medullary neurons in the NTS. These second-order neurons project to the central pattern generator of the ventral respiratory column (signified by an oscillator and integral sign). Either direct (NTS) or indirect projections from ventral respiratory group neurons activate serotonergic neurons of the caudal raphe nuclei. Descending projections to spinal motor nuclei and subsequent serotonin release initiate and orchestrate plasticity in phrenic and limb motor neurons. During TST, primary motor cortex neurons with direct projections to phrenic motor neurons, limb motor neurons, and their premotor neurons are activated; although they could activate raphe and spinal central pattern generator neurons, we do not postulate that those relay projections play a major role in TST-induced motor plasticity. Dashed lines indicate unknown pathways. CB indicates carotid body; CPG, central pattern generator; LMn, limb motor neurons; M1, primary motor cortex; PMn, phrenic motor neurons.
Figure 2
Figure 2:
Cellular model. Postulated intracellular signaling cascades giving rise to AIH- and TST-induced motor plasticity. A. AIH-activated cellular cascades (elevated BDNF) and increased motor neuron output (bottom trace). Serotonin release and receptor activation are the major drivers of the AIH-induced cellular cascade. B. TST-activated cellular cascades (elevated BDNF) and increased motor neuron output (bottom trace). We postulate that serotonin plays a less prominent role, and that the more relevant cascade is activity-dependent increases in intracellular calcium and CaMK activation—an alternate mechanism of increased BDNF synthesis; this activity-driven effect would occur exclusively in neurons relevant to TST. C. Combined AIH-TST converges both mechanisms, but only in neurons activated during TST. Thus, BDNF and motor activity in these task-relevant neurons are greater than that elicited by either treatment alone. TrkB indicates tyrosine kinase B.
Figure 3
Figure 3:
Working hypothetical model of motor plasticity. With a postulated sigmoid BDNF dose-response curve, AIH and TST each elicit modest plasticity when delivered alone. However, combined AIH-TST elicits greater elevations in BDNF (but only within task-relevant neurons), increasing plasticity and improving motor function more than the combined value predicted from each alone. TrkB indicates tyrosine kinase B.

Neural Network Model

The network model consists of peripheral and central nervous system elements activated by AIH and TST, converging on targeted alpha-motor neurons (Fig. 1). We focus on serotonergic and glutamatergic pathways.

During hypoxia, spinal serotonin release in motor nuclei is triggered by carotid chemoreceptor activation and synaptic activation of second-order neurons in the NTS, ventral respiratory group, and caudal raphe nuclei. Serotonergic raphe neurons, which send descending projections to spinal gray matter, are activated by rhythmic motor behaviors such as walking, chewing, and breathing (43). While serotonin could play a common role with both AIH and TST, accounting for their synergy, we do not favor this idea because treadmill training (i.e., non–task-specific exercise) increases spinal serotonin release but does not enhance horizontal ladder walking (8). To explain available data, we propose that TST-induced neural activity (at glutamatergic synapses) is a necessary component of the AIH-TST synergy.

During TST, descending cortical projections activate alpha-motor neurons directly or indirectly via glutamatergic synaptic transmission. The resulting neuronal activity may upregulate BDNF within alpha-motor neurons. This increase in BDNF would occur exclusively within neurons activated by the specific task. Combining AIH- and TST-induced BDNF effects within specific neuron populations may evoke selective, task-appropriate plasticity.

Cellular Model

The cellular model (Fig. 2) describes potential mechanisms of AIH and TST convergence on targeted alpha-motor neurons from an intracellular signaling perspective. We posit that this convergence encapsulates the synergistic plasticity-inducing relationship of AIH and TST. Our model is not intended to portray comprehensive cellular pathways but emphasizes principle aspects of our working hypothesis — that convergence onto common cellular elements in the same (task-specific) cells elevates BDNF levels above an effective threshold necessary for plasticity. We propose that these cascades are initiated via AIH-activated serotonergic and TST-activated glutamatergic receptors. Despite initial activation of different neurochemical receptors (serotonin vs glutamate), both AIH and TST ultimately converge on postsynaptic glutamate receptor density and current conductance for the maintenance of plasticity.

Knowledge concerning cellular mechanisms of AIH-induced motor plasticity is based on decades of studies on the phrenic motor system; relatively little is known concerning AIH effects on nonrespiratory motor systems. In humans, AIH enhances motor-evoked potentials in the first dorsal interosseous muscle at a site between the medulla and motor neuron synapse (47), consistent with our understanding of AIH-induced phrenic LTF. The effects of AIH appear to be synergistic with spike-timing–dependent plasticity because their combined effects are greater than their individual effects in most subjects. In a rodent model of cervical SCI, 7 d of AIH elevates BDNF and TrkB in both phrenic and nonrespiratory motor nuclei (7), similar to intact rats. Despite parallels in neurochemical plasticity, direct comparisons of AIH-induced plasticity in respiratory versus nonrespiratory motor systems are not available.

We propose that AIH-induced serotonin-dependent BDNF synthesis is general to alpha-motor neurons, and that the independent effects of AIH on BDNF synthesis are sufficient to induce motor plasticity in some (e.g., phrenic, interosseous) but not all motor systems. We hypothesize that in both situations, plasticity synergy and specificity can be achieved when TST triggers additional BDNF synthesis exclusively within the task-specific motor circuit.

TST elevates BDNF in the brain (48) and spinal cord (40). Increased BDNF/TrkB signaling triggers new synthesis of synapsin I, indicating possible synaptogenesis. Exercise also increases BDNF in skeletal muscles, including the neuromuscular junction (41), which can be retrogradely transported from muscle to the spinal cord through motor neuron axons, potentially supporting BDNF-induced plasticity.

TST activates postsynaptic glutamatergic NMDA and AMPA receptors. When glutamate receptors are phosphorylated, NMDA receptor currents and AMPA receptor trafficking to the postsynaptic membrane are augmented. Calcium influx and activation of Ca2+-calmodulin–dependent kinase (CaMK) can lead to glutamate receptor phosphorylation, thereby enhancing NMDA and AMPA receptor conductances (49). This mechanism contributes to a classic example of activity-dependent synaptic plasticity, known as hippocampal long-term potentiation (LTP). Activity-dependent LTP is associated with spatial learning and memory (50). BDNF is required for at least some forms of hippocampal LTP (51). Thus, both activity-dependent hippocampal synaptic plasticity and AIH-induced phrenic motor plasticity require BDNF/TrkB signaling. Similar activity- and BDNF-dependent plasticity could occur in task-specific motor circuits because calcium is increased and CaMK activated exclusively in TST-relevant neuronal populations.

In our model, TST alone increases BDNF expression, but only to a limited extent (Fig. 3). Activity-dependent increases in BDNF expression would be localized within the relevant (activated) motor circuit, and may or may not be sufficient to elicit some minor degree of motor plasticity (depending upon the shape/slope of the stimulus dose-response curve). AIH preconditioning may establish a higher baseline level of spinal BDNF, eliciting greater nonspecific plasticity in multiple (i.e., non–task-specific) motor systems. On the other hand, a global increase in baseline BDNF would set the stage for enhanced TST effects, but only in neural circuits activated during the practiced task. In this way, plasticity would be unique to the specific task, accounting for synergy between AIH and TST. Because AIH elevates BDNF via translational regulation, it requires many minutes to hours to function as a plasticity-priming stimulus. As a result, paired AIH-TST would ideally be separated by 30 to 60 min, giving adequate time for BDNF protein synthesis (8,11).

Hypothesis Testing

Our model and hypothesis were developed to explain recent rodent and human experiments demonstrating AIH-TST synergy (8,11). We propose that AIH and TST mechanisms converge at the level of alpha-motor neurons, increasing BDNF synthesis more than in other motor pools (i.e., those not activated by TST). There is currently no empirical evidence supporting the notion that TST elevates BDNF within specific activated versus nonactivated motor neurons. This premise is the crux of our hypothesis and thus a prime target for future mechanistic experimentation.

From another perspective, respiratory infections and ventilatory insufficiency are leading causes of morbidity and mortality after cervical SCI. Improvements in respiratory function following daily AIH have been observed (13), yet the impact of combined treatments have not been explored. The use of respiratory strength training, voluntary (isocapnic) or CO2-driven chemoreflex hyperpnea, offers simple task-specific strategies to target neural pathways involved in breathing. Combined AIH preconditioning and respiratory TST may improve functional benefits.

Finally, although our discussion has focused on spinal alpha-motor neuron plasticity, the possibility remains that synergy may result instead in the motor cortex or other regions affected by both AIH and TST. These possibilities await further study and experimental verification.


After spinal trauma, motor functions are impaired. Recovery depends on multiple factors, including the level and extent of injury. Because spontaneous compensatory mechanisms and conventional rehabilitation are insufficient to fully restore function, new strategies are needed.

AIH is a novel intervention with promise to improve respiratory and nonrespiratory motor function after SCI by promoting plasticity through targeted increases in BDNF/TrkB signaling. In nonrespiratory motor systems, two studies (one rodent, one human) demonstrate that AIH effects are amplified considerably when combined with TST (8,11). The AIH-TST synergy is specific to the engaged motor function and does not appear to be generalizable to off-target motor functions. Although AIH-TST synergy has been demonstrated in limb functions (ladder walking in rats, over-ground walking in humans), we do not know if similar synergies occur in breathing function.

Because mechanisms of AIH-TST synergy are only poorly understood, we offer hypothetical working models of both neural network and cellular mechanisms. Our model is highly speculative, although grounded in the neuroscience literature; our hope is that these models will provoke future experiments. In this review, we proposed that AIH primes the nervous system for plasticity by broadly elevating BDNF/TrkB signaling in diverse neuron pools, whereas TST selectively activates neural circuits, further increasing BDNF only in task-relevant neuronal circuits via activity-dependent mechanisms. Due to the limited number of studies on this topic, we advise caution — the model requires extensive testing.


This study was supported by the UF McKnight Brain Institute, the Craig H. Neilsen Foundation, DoD (SCIRP), NIH R01 HL147554, Brooks Rehabilitation, and the Brooks-PHHP Research Collaboration.


1. Gonzalez-Rothi EJ, Lee K, Dale EA, Reier PJ, Mitchell GS, Fuller DD. Intermittent hypoxia and neurorehabilitation. J. Appl. Physiol. 2015; 119:1455–65.
2. Hayashi F, Coles SK, Mitchell GS, McCrimmon DR. Time-dependent phrenic nerve responses to carotid afferent activation: intact vs. decerebrate rats. Am. J. Physiol. Regul. Integr. Comp. Physiol. 1993; 34:R811–9.
3. Kinkead R, Bach KB, Johnson SM, Hodgeman BA, Mitchell GS. Plasticity in respiratory motor control: intermittent hypoxia and hypercapnia activate opposing serotonergic and noradrenergic modulatory systems. Comp. Biochem. Physiol. 2001; 130:207–18.
4. Dale EA, Ben Mabrouk F, Mitchell GS. Unexpected benefits of intermittent hypoxia: enhanced respiratory and nonrespiratory motor function. Phys. Ther. 2014; 29:1–10.
5. Golder FJ, Mitchell GS. Spinal synaptic enhancement with acute intermittent hypoxia improves respiratory function after chronic cervical spinal cord injury. J. Neurosci. 2005; 25:2925–32.
6. Bach KB, Mitchell GS. Hypoxia-induced long-term facilitation of respiratory activity is serotonin dependent. Respir. Physiol. 1996; 104:251–60.
7. Lovett-Barr MR, Satriotomo I, Muir GD, et al. Repetitive intermittent hypoxia induces respiratory and somatic motor recovery after chronic cervical spinal injury. J. Neurosci. 2012; 32:3591–600.
8. Prosser-Loose EJ, Hassan A, Mitchell GS, Muir GD. Delayed intervention with intermittent hypoxia and task training improves forelimb function in a rat model of cervical spinal injury. J. Neurotrauma. 2015; 32:1403–12.
9. Trumbower RD, Jayaraman A, Mitchell GS, Rymer WZ. Exposure to acute intermittent hypoxia augments somatic motor function in humans with incomplete spinal cord injury. Neurorehabil. Neural Repair. 2012; 26(2):163–72.
10. Trumbower RD, Hayes HB, Mitchell GS, Wolf SL, Stahl VA. Effects of acute intermittent hypoxia on hand use after spinal cord trauma: a preliminary study. Neurology. 2017; 89(18):1904–7.
11. Hayes HB, Jayaraman A, Herrmann M, Mitchell GS, Rymer WZ, Trumbower RD. Daily intermittent hypoxia enhances walking after chronic spinal cord injury: a randomized trial. Neurology. 2014; 82:104–13.
12. Navarrete-Opazo A, Alcayaga J, Sepulveda O, Varas G. Intermittent hypoxia and locomotor training enhances dynamic but not standing balance in patients with incomplete spinal cord injury. Arch. Phys. Med. Rehabil. 2017; 98(3):415–24.
13. Tester NJ, Fuller DD, Fromm JS, Spiess MR, Behrman AL, Mateika JH. Long-term facilitation of ventilation in humans with chronic spinal cord injury. Am. J. Respir. Crit. Care Med. 2014; 189:57–65.
14. Dale-Nagle EA, Hoffman MS, MacFarlane PM, Mitchell GS. Multiple pathways to long-lasting phrenic motor facilitation. Adv. Exp. Med. Biol. 2010; 669:225–30.
15. Baker-Herman TL, Mitchell GS. Phrenic long-term facilitation requires spinal serotonin receptor activation and protein synthesis. J. Neurosci. 2002; 22:6239–49.
16. MacFarlane PM, Mitchell GS. Episodic spinal serotonin receptor activation elicits long-lasting phrenic motor facilitation by an NADPH oxidase-dependent mechanism. J. Physiol. 2009; 587(22):5469–81.
17. Hoffman MS, Mitchell GS. Spinal 5-HT7 receptor activation induces long-lasting phrenic motor facilitation. J. Physiol. 2011; 589:1397–407.
18. Nichols NL, Dale EA, Mitchell GS. Severe acute intermittent hypoxia elicits phrenic long-term facilitation by a novel adenosine-dependent mechanism. J. Appl. Physiol. 2012; 112:1678–88.
19. Hoffman MS, Nichols NL, MacFarlane PM, Mitchell GS. Phrenic long-term facilitation after acute intermittent hypoxia requires spinal ERK activation but not TrkB synthesis. J. Appl. Physiol. 2012; 113:1184–93.
20. Baker-Herman TL, Fuller DD, Bavis RW, et al. BDNF is necessary and sufficient for spinal respiratory plasticity following intermittent hypoxia. Nat. Neurosci. 2004; 7(1):48–55.
21. Bocchiaro CM, Feldman JL. Synaptic activity-independent persistent plasticity in endogenously active mammalian motoneurons. Proc. Natl. Acad. Sci. U. S. A. 2004; 101(12):4292–5.
22. Perim RR, Fields DP, Mitchell GS. Protein kinase Cdelta constrains the S-pathway to phrenic motor facilitation elicited by spinal 5-HT7 receptors or severe acute intermittent hypoxia. J. Physiol. 2019; 597(2):481–98.
23. Devinney MJ, Huxtable AG, Nichols NL, Mitchell GS. Hypoxia-induced phrenic long-term facilitation: emergent properties. Ann. N. Y. Acad. Sci. 2013; 1279:143–53.
24. Tadjalli A, Mitchell GS. Cervical spinal 5-HT2A and 5-HT2B receptors are both necessary for moderate acute intermittent hypoxia-induced phrenic long-term facilitation. J. Appl. Physiol. 2019; 127(2):432–43.
25. Devinney MJ, Nichols NL, Mitchell GS. Sustained hypoxia elicits competing spinal mechanisms of phrenic motor facilitation. J. Neurosci. 2016; 36:7877–85.
26. Devinney MJ, Fields DP, Huxtable AG, Peterson TJ, Dale EA, Mitchell GS. Phrenic long-term facilitation requires PKCtheta activity within phrenic motor neurons. J. Neurosci. 2015; 35:8107–17.
27. De Leon RD, Hodgson JA, Roy RR, Edgerton VR. Full weight-bearing hindlimb standing following stand training in the adult spinal cat. J. Neurophysiol. 1998; 80:83–91.
28. Bliss TV, Collingridge GL. Expression of NMDA receptor-dependent LTP in the hippocampus: bridging the divide. Mol. Brain. 2013; 6:5.
29. Mauelshagen J, Sherff CM, Carew TJ. Differential induction of long-term synaptic facilitation by spaced and massed applications of serotonin at sensory neuron synapses at Aplysia californica. Learn. Mem. 1998; 5:246–56.
30. Kleim JA, Barbay S, Cooper NR, et al. Motor learning-dependent synaptogenesis is localized to functionally reorganized motor cortex. Neurobiol. Learn. Mem. 2002; 77:63–77.
31. Foffani G, Shumsky J, Knudsen EB, Ganzer PD, Moxon KA. Interactive effects between exercise and serotonergic pharmacotherapy on cortical reorganization after spinal cord injury. Neurorehabil. Neural Repair. 2016; 30(5):479–89.
32. Hess G, Donoghue JP. Long-term potentiation of horizontal connections provides a mechanism to reorganize cortical motor maps. J. Neurophysiol. 1994; 71(6):2543–7.
33. Weishaupt N, Li S, Di Pardo A, Sipione S, Fouad K. Synergistic effects of BDNF and rehabilitative training on recovery after cervical spinal cord injury. Behav. Brain Res. 2013; 239:31–42.
34. Aagaard P, Simonsen EB, Andersen JL, Magnusson P, Dyhre-Poulsen P. Neural adaptation to resistance training: changes in evoked V-wave and H-reflex responses. J. Appl. Physiol. 2002; 92(6):2309–18.
35. Del Vecchio A, Casolo A, Negro F, et al. The increase in muscle force after 4 weeks of strength training is mediated by adaptations in motor unit recruitment and rate coding. J. Physiol. 2019; 597(7):1873–87.
36. Wolpaw JR. What can the spinal cord teach us about learning and memory. Neuroscientist. 2010; 16(5):532–49.
37. Carp JS, Wolpaw JR. Motoneuron plasticity underlying operantly conditioned decrease in primate H-reflex. J. Neurophysiol. 1994; 72:431–42.
38. Wang Y, Pillai S, Wolpaw JR, Chen XY. H-reflex down-conditioning greatly increases the number of identifiable GABAergic interneurons in rat ventral horn. Neurosci. Lett. 2009; 452(2):124–9.
39. Mitchell GS, Johnson SM. Neuroplasticity in respiratory motor control. J. Appl. Physiol. 2003; 94:358–74.
40. Gomez-Pinilla F, Ying Z, Opazo P, Roy RR, Edgerton VR. Differential regulation by exercise of BDNF and NT-3 in rat spinal cord and skeletal muscle. Eur. J. Neurosci. 2001; 13:1078–84.
41. Gomez-Pinilla F, Ying Z, Roy RR, Molteni R, Edgerton VR. Voluntary exercise induces a BDNF-mediated mechanism that promotes neuroplasticity. J. Neurophysiol. 2002; 88(5):2187–95.
42. Gomez-Pinilla F, Huie JR, Ying Z, et al. BDNF and learning: evidence that instrumental training promotes learning within the spinal cord by up-regulating BDNF expression. Neuroscience. 2007; 148(4):893–906.
43. Jacobs BL, Martin-Cora FJ, Fornal CA. Activity of medullary serotonergic neurons in freely moving animals. Brain Res. Brain Res. Rev. 2002; 40(1–3):45–52.
44. Gerin C, Legrand A, Privat A. Study of 5-HT release with a chronically implanted microdialysis probe in the ventral horn of the spinal cord of unrestrained rats during exercise on a treadmill. J. Neurosci. Methods. 1994; 52(2):129–41.
45. Satriotomo I, Nichols NL, Dale EA, Emery AT, Dahlberg JM, Mitchell GS. Repetitive acute intermittent hypoxia increases growth/neurotrophic factor expression in non-respiratory motor neurons. Neuroscience. 2016; 322:479–88.
46. Navarrete-Opazo A, Alcayaga J, Sepulveda O, Rojas E, Astudillo C. Repetitive intermittent hypoxia and locomotor training enhances walking function in incomplete spinal cord injury subjects: a randomized, triple-blind, placebo-controlled trial. J. Neurotrauma. 2017; 34(9):1803–12.
47. Christiansen L, Urbin MA, Mitchell GS, Perez MA. Acute intermittent hypoxia enhances corticospinal synaptic plasticity in humans. Elife. 2018; 7:e34303.
48. Neeper SA, Gomez-Pinilla F, Choi J, Cotman C. Exercise and brain neurotrophins. Nature. 1995; 373:109.
49. Poncer JC, Esteban JA, Malinow R. Multiple mechanisms for the potentiation of AMPA receptor-mediated transmission by alpha-Ca2+/calmodulin-dependent protein kinase II. J. Neurosci. 2002; 22(11):4406–11.
50. Bliss TV, Collingridge GL. A synaptic model of memory: long-term potentiation in the hippocampus. Nature. 1993; 361(6407):31–9.
51. Korte M, Carroll P, Wolf E, Brem G, Thoenen H, Bonhoeffer T. Hippocampal long-term potentiation is impaired in mice lacking brain-derived neurotrophic factor. Proc. Natl. Acad. Sci. U. S. A. 1995; 92(19):8856–60.

acute intermittent hypoxia; exercise training; neuroplasticity; spinal plasticity, phrenic; motor neuron; rehabilitation; spinal cord injury

Copyright © 2020 by the American College of Sports Medicine