V2a Neurons Pattern Respiratory Muscle Activity in Health and Disease

by Victoria N. Jensen

A dissertation submitted to the University of Cincinnati in partial fulfillment of the requirements for the degree of Doctor of Philosophy

University of Cincinnati College of Medicine Neuroscience Graduate Program December 2019

Committee Chairs Mark Baccei, Ph.D. Steven Crone, Ph.D.

Abstract

Respiratory failure is the leading cause of death in in amyotrophic lateral sclerosis (ALS) patients and spinal cord injury patients. Therefore, it is important to identify neural substrates that may be targeted to improve breathing following disease and injury. We show that a class of ipsilaterally projecting, excitatory interneurons located in the brainstem and spinal cord – V2a neurons – play key roles in controlling respiratory muscle activity in health and disease. We used chemogenetic approaches to increase or decrease V2a excitability in healthy mice and following disease or injury. First, we showed that silencing

V2a neurons in neonatal mice caused slow and irregular breathing. However, silencing V2a neurons in adult mice did not alter the regularity of respiration and actually increased breathing frequency, suggesting that

V2a neurons play different roles in controlling breathing at different stages of development. V2a neurons also pattern respiratory muscle activity. Our lab has previously shown that increasing V2a excitability activates accessory respiratory muscles at rest in healthy mice. Surprisingly, we show that silencing V2a neurons also activated accessory respiratory muscles. These data suggest that two types of V2a neurons exist: Type I V2a neurons activate accessory respiratory muscles at rest whereas Type II V2a neurons prevent their activation when they are not needed. Moreover, altering the excitability of just cervical spinal neurons using viral strategies suggests that these two V2a subtypes are both located in the cervical spinal cord. The effect of increasing and decreasing V2a excitability on accessory respiratory muscle activity was also tested in ALS model mice. Increasing or decreasing V2a excitability activates accessory respiratory muscles throughout disease progression in SOD1(G93A) ALS model mice. Finally, we demonstrate that increasing V2a excitability promotes recovery of diaphragm function following a high level C2 hemisection spinal cord injury. Moreover, silencing V2a neurons two weeks following injury impairs recovery of diaphragm function. Thus, we conclude that V2a neurons control accessory respiratory muscle activity and promote recovery of respiratory muscle function following disease and spinal cord injury.

iii

Acknowledgments

Many thanks to the following funding sources and personnel that have supported this research:

Funding Sources

University of Cincinnati University of Cincinnati Neuroscience Graduate Program T32NS5007453 University of Cincinnati Dean’s Dissertation Completion Fellowship University of Cincinnati/Kentucky Center for Clinical and Translational Science and Training

Cincinnati Children’s Hospital Medical Center Dissertation Committee CCHMC Division of Neurosurgery CCHMC Trustee Award Mark Baccei, Ph.D. (Chair) CCHMC RIP Grant Steve Davidson, Ph.D. Steven Crone, Ph.D. External Funding Warren Alilain, Ph.D. Albert J. Ryan Fellowship Timothy Weaver, Ph.D. L.I.F.E Foundation Research Grant Amyotrophic Lateral Sclerosis Starter Grant Craig H. Neilson Foundation NINDS1RO1NS112255

Yutaka Yoshida Lab (CCHMC) Crone Lab (current and former members) Masaki Ueno Steven A. Crone (Mentor) Fumiyasu Imai Shannon Romer

Kari Seedle Steve Danzer Lab (CCHMC) Sarah Turner Raymond Pun Emma Dwenger Candi LeSarge *Brooke Chastain

Ashley Lengel Mark Baccei Lab (University of Cincinnati) *Azl Saeed Ji Lie *Omar Barazi

Bailey Clark John Lorenz Lab (University of Cincinnati Kanika Segal

Warren Alilain Lab (University of Kentucky) *These undergraduates worked directly on *Emily Huffman research projects presented in this thesis. Daimen Stoltz

Aaron Silverstein Lydia Hager Rachel Jolly

TABLE OF CONTENTS

Page

CHAPTER I: Introduction: Nervous System Control of Breathing A. Significance Statement 2 B. Role of Primary and Accessory Respiratory Muscles in Ventilation 2-4 C. Neural Control of Respiration 4-16 C1. Three brainstem oscillators interact to generate respiratory rhythm 5-6 C2. Descending input from the brainstem to the spinal cord generates rhythmic 6-7 contraction of the diaphragm during resting ventilation. C3. Chemosensory respiratory neurons alter ventilation to maintain physiological pH 8-11 C3.1 Peripheral chemoreceptors: The carotid bodies 8-9 C3.2 Central chemoreceptors: The Retrotrapezoid Nucleus 9-10 C3.3 Central chemoreception: C1 Neurons 10-11 C4. Spinal interneurons are important for breathing 11-12 C5. Sensory spinal neurons mediate sensory afferent feedback 12-14 C6. Coordinating ventilation to motor activity during exercise 14-19 C6.1 Central feed forward mechanism 14-19 C6.2 Neural interactions between locomotor and respiratory circuits 15-16 C6.3 Activation of accessory respiratory muscles 16-18 C6.3.1 Descending neural drive modulates activation of accessory 17 respiratory muscles C6.3.2 Sensory afferent feedback modulates accessory respiratory 18 muscle activity C6.3.3. Spinal interneurons control accessory respiratory muscle 18-19 activity

D. Breathing is Impaired in Neuromuscular Diseases 19-27 D1. Respiratory deficits in ALS 19-23 D2. Compensatory plasticity in ALS 23-25 D3. Treatment for ALS: New Therapies are Needed 26-27

E. Breathing is Impaired Following High-Level Spinal Cord Injury 27-39 E1. Respiratory deficits following spinal cord injury 27-28 E2. Compensatory plasticity following spinal cord injury 29-37 E2.1 Compensatory plasticity following spinal cord injury: The Crossed 29-33 Phrenic Phenomenon E2.2 Interneurons play key roles in promoting respiratory plasticity following 33-36 spinal cord injury E2.3 Compensatory plasticity following spinal cord injury: Accessory 36-37 respiratory muscle activation E3. Treatment for high level spinal cord injury 38-39

F. V2a Neurons Control Locomotion and Breathing 39-47

F1. Distinct classes of interneurons exist with distinct properties 39-40 F2. Properties of V2a interneurons 40-41 F3. V2a neurons coordinate locomotion 41-43 F4. V2a neurons modulate breathing 44-45 F5. V2a neurons may promote recovery of breathing following disease or injury 45-47

CHAPTER II: Brainstem and Spinal Cord V2a Neurons Regulate Respiratory Muscle Activity 47-88 and Ventilation in Healthy Mice 2.1 Introduction 48-50 2.2 Methods 50-58 2.3 Results 58-80 2.4 Discussion 80-88

CHAPTER III: Cervical Spinal V2a Neurons Control Respiratory Muscle Activity and Increase 89-120 Ventilation in Healthy Mice 3.1 Introduction 90-91 3.2 Methods 91-95 3.3 Results 95-113 3.4 Discussion 113-120

CHAPTER IV: Altering V2a Excitability Increases ARM Activity and Enhances Ventilation in 121-143 ALS Model Mice 4.1 Introduction 122-123 4.2 Methods 123-126 4.3 Results 126-137 4.4 Discussion 137-143

CHAPTER V: Increasing the Excitability of V2a Neurons Promotes Recovery of Respiratory 144-181 Muscle Activity Following Spinal Cord Injury 5.1 Introduction 145-147 5.2 Methods 147-152 5.3 Results 152-172 5.4 Discussion 172-181

CHAPTER VI: Conclusions and Future Directions 182-190

List of Abbreviations 191-192

References 193-208

Chapter Figure Figure Title Page Chapter Figure 1 Analyzing the respiratory cycle 3 I Figure 2 Complex interaction among brainstem and spinal cord respiratory structures 5 controls respiratory rhythm and pattern generation Figure 3 The Crossed Phrenic Phenomenon 31 Chapter Figure 1 (Gi)DREADD is expressed in V2a neurons in the spinal cord and brainstem of V2a- 61 II (Gi)DREADD mice Figure 2 Decreasing the excitability of V2a neurons causes irregular breathing in neonatal 64 V2a-(Gi)DREADD mice Figure 3 Decreasing the excitability of V2a neurons increases breathing frequency without 66 causing irregular breathing in adult V2a-(Gi)DREADD mice Figure 4 Decreasing the excitability of V2a neurons increases extradiaphragmatic respiratory 68 muscle activity Figure 5 Diaphragm EMG peak amplitude does not change during ARM bouts 70 Figure 6 Ventilation is increased during bouts of ARM activity 73 Figure 7 Decreasing V2a neuron excitability does not alter arterial oxygen saturation 75 Figure 8 Brainstem and spinal cord V2a neurons are sufficient but not necessary to increase 77 heart rate and blood pressur Figure 9 Increasing and decreasing the excitability of V2a neurons differentially affects 79 motor functions Figure 10 Hypothetical Model: V2a neuron diversity may explain how increasing or 80 decreasing the excitability can activate ARMs Chapter Figure 1 Cervical spinal V2a neurons express the excitatory (Gq)DREADD receptor in AAV- 96 III (Gq)DREADD mice Figure 2 Cervical spinal V2a neurons express the inhibitory (Gi)DREADD receptor in AAV- 98 (Gi)DREADD mice Figure 3 Cervical spinal V2a neurons express eGFP in non-DREADD expressing AAV-eGFP 99 mice Figure 4 Increasing or decreasing the excitability of cervical spinal V2a neurons increases 102 scalene ARM activity Figure 5 Cervical spinal V2a neurons are sufficient but not required for diaphragm activity 104 Figure 6 Increasing the excitability of cervical spinal V2a neurons increases ventilation 106 during bouts of ARM activity Figure 7 Decreasing the excitability of cervical spinal V2a neurons increases ventilation 109 during bouts of ARM activity Figure 8 Cervical spinal V2a neurons project to respiratory centers in the brainste 110 Figure 9 Altering cervical spinal V2a excitability does not change heart rate 111 Figure 10 Increasing the excitability of cervical spinal V2a neurons on side of the spinal cord 112 bilaterally activates the right and left scalene muscle Figure 11 Hypothetical model of cervical spinal V2a neuron subtypes and their organization in 113 respiratory circuits Chapter Figure 1 DREADD-expressing SOD1(G93A) ALS model mice show characteristic weight loss 126 IV due to disease progression

vii

Figure 2 Increasing the excitability of V2a neurons activates ARMs throughout disease 128 progression Figure 3 Decreasing the excitability of V2a neurons activates ARMs throughout disease 130 progression Figure 4 Activating V2a neurons enhances ventilation during ARM bouts during early stages 132 of ALS whereas silencing V2a neurons enhances ventilation during ARM bouts during late stages of ALS

Figure 5 Increasing or decreasing V2a excitability may acutely reverse ALS induced changes 135 in ventilation Figure 6 Hypothetical Model of ARM Activation 137 Table 1 Summary of motor deficits in SOD1(G93A) ALS model mice 124 Table 2 Report of Statistics 136 Chapter Figure 1 Increasing the excitability of V2a neurons restores bursting activity to a previously 156 V paralyzed diaphragm Figure 2 Increasing the excitability of V2a neurons restores diaphragm function two weeks 159 following a C2Hx Figure 3 Cycle triggered averaging reveals that the predominant pattern of diaphragm 162 activity after activating V2a neurons is rhythmic bursting during inspiratio Figure 4 Silencing V2a neurons impairs the ability to restore diaphragm function via nasal 164 occlusion two weeks following injur Figure 5 V2a neurons may be one of many mediators of spontaneous recovery 166 Figure 6 Increasing the excitability of V2a neurons does not adversely affect respiratory 167 rhythm generation Figure 7 Repeatedly increasing V2a excitability following injury may restore diaphragm 168 function back to pre-injury levels faster than spontaneous recovery alone

Figure 8 The frequency of ipsilateral scalene activity changes following spinal cord injury and 170 V2a neuron activation Figure 9 Ventilation is enhanced during scalene bouts 171 Figure 10 Multiple subtypes of V2a neurons may contribute to recovery of diaphragm 172 function following spinal cord injury

viii

CHAPTER I: Introduction

Nervous System Control of Breathing

A. Significance Statement

Breathing allows for the exchange of O2 and CO2, maintenance of blood pH levels, swallowing, and vocalization/communication. As a result, the neuronal circuitry involved in regulating and patterning ventilation is complex. There are multiple interacting respiratory centers in the brainstem, distinct respiratory motor neuron pools in the spinal cord, multiple respiratory muscles, mechanisms to coordinate ventilation to increased motor activity, and compensatory mechanisms of neural plasticity to maintain breathing following disease or injury. As such, it is critical to gain a better understanding of the neural circuitry controlling respiratory muscle activity and ventilation both in healthy mice and following neuromuscular diseases or spinal cord injury. This will help identify specific neural substrates that can be activated to enhance ventilation and restore a better quality of life to patients suffering from respiratory insufficiency. The following research investigates how excitatory V2a neurons - a class of ipsilaterally projecting, excitatory neurons located in the brainstem and spinal cord - control respiratory muscle activity and ventilation in healthy mice and how they may be recruited following neuromuscular disease and spinal cord injury to restore impaired respiratory muscle activity.

B. Role of Primary and Accessory Respiratory Muscles in Ventilation

Primary and accessory respiratory muscles have different functions in coordinating ventilation. As their name suggests, primary respiratory muscles play a primary role in ventilation whereas accessory respiratory muscles (ARMs) assist primary muscles to enhance ventilation during increased ventilatory effort (Sieck & Gransee, 2012). Ventilation is dependent on changes in transthoracic pressure due to coordinated activity of primary chest wall “pump muscles” (Sieck & Gransee, 2012; Sieck et al. 2013). Two inspiratory pump muscles, the diaphragm and external intercostals, contract during inspiration and expand the chest wall (Sieck et al. 2013). The diaphragm alone accounts for 65% of vital capacity in healthy humans (DiMarco 2009). This muscle has an upside-down U-shape whose activity contributes to the three

phases of the respiratory cycle: inspiration

(Phase I), post-inspiration (Phase II), and

active expiration (Phase III) (Anderson and

Ramirez 2017). The diaphragm contracts

downward during inspiration, which

increases the volume in intraabdominal space

and decreases the intrathoracic pressure. This

causes the ambient atmospheric pressure to

become greater than the thoracic pressure,

which drives air into the airways and lungs to

restore equilibrium and allows the chest wall

Figure 1. Analyzing the respiratory cycle. (A) (Left) During inspiration, the to expand during inspiration (Phase I) diaphragm contracts downward to increase intrathoracic pressure and allow inspiration of air. (Middle ) During post-inspiration (or passive expiration), the (Figure 1A, left). The diaphragm acts in diaphragm relaxes, pushing air back out. (Right) During active expiration, expiratory muscles, like the abdominals, become active to increase expulsive force, such as during a cough or sneeze. (B) Example of a breath from a whole concert with the external intercostals, which breath plethysmography (WBP) trace. Inspiration is outlined in orange and expiration is outlined in blue. (C) Peak inspiratory flow, tidal volume, breathing are responsible for forced and quiet frequency, and minute volume are all measures of ventilation that can be measured from WBP. inhalation by raising the ribs and expanding the chest cavity. When inspiratory pump muscles relax, the diaphragm passively moves back upward and pushes air back out through the airway during expiration (Figure 1A, middle).

Post-inspiration and expiration are merged in the mammalian respiratory system. Post inspiration occurs during the early stages of expiration and slows expiratory airflow to maximize the time for pulmonary gas exchange and prevent collapse of the lungs (Dutschmann et al. 2014). This post-inspiratory activity causes activation of the laryngeal adductors and the crural diaphragm to allow voluntary breath- holding and non-ventilatory behaviors such as vocalization, swallowing, defecation, and vomiting.

Exhalation then resumes when the inspiratory pump muscles have completely relaxed and passively push air back out through the airways.

It is only during forced or active expiration, such as coughing, sneezing, or whistling, that expiratory ARMs (such as the abdominals, obliques, and internal intercostals) become recruited to increase intraabdominal pressure and expel air with a greater force (Figure 1A, right) (de Paleville et al.

2011a; Sieck et al. 2013; De Troyer and Boriek 2011). Similarly, inspiratory ARMs include the trapezius, scalenes, and sternocleidomastoid and are recruited under conditions of increased inspiratory effort to further expand the chest wall, such as during augmented breaths (sighs) and exercise (Sieck & Gransee,

2012; de Paleville et al. 2011a). They also stabilize the chest wall and prevent paradoxical breathing, a phenomenon caused by instability of the ribcage and abdomen resulting in the collapse of the chest wall

(instead of expansion) during inspiration (Urmey et al. 1985). Finally, in cases where neuromuscular diseases or spinal cord injury impairs primary muscle function, ARMs may be recruited to compensate for impaired ventilation (Bennett et al. 2004; Dougherty et al. 2012; Nichols et al. 2013). Thus, both primary and accessory respiratory muscles are critical for maintaining respiratory health.

As expected, impairment to any of these respiratory muscles can severely impact health. Because breathing is required for life, the neuronal circuitry controlling respiration is necessarily complex.

Therefore, a better understanding of the neural control of primary and accessory respiratory muscles under normal conditions is critical for understanding why neuromuscular diseases and injury impairs respiratory function.

C. Neural Control of Respiration

Both respiratory and locomotor circuits are governed by central pattern generators (CPGs) and/or neural input into CPGs, which control the frequency and pattern of rhythmic movements such as walking, swimming, chewing, swallowing and breathing (Grillner, 2006; Grillner & Jessell, 2009; Grillner & Manira,

2015; Ramirez et al. 2016). These CPGs control two distinct aspects of respiration and locomotion: rhythm generation and pattern generation. Rhythm generation refers to the frequency and regularity of respiration whereas pattern generation refers to which muscles are activated and when they are activated

(McCrea and Rybak 2008; Kam et al. 2013; Dick et al. 2018). Section C discusses both aspects of the respiratory CPG as well as additional physiological and neural inputs that modify respiration.

C1. Three brainstem oscillators interact to generate respiratory rhythm

Over the last century, research aiming to understand central control of breathing has identified multiple brainstem regions necessary for regulating respiratory rhythm. As described above, the respiratory cycle is divided into three phases: inspiration, post-inspiration, and active expiration

(Figure 2). The triple oscillator model of respiratory rhythm generation states that each of these phases is controlled by distinct brainstem oscillators in the central nervous system to ensure rhythmic breathing

(Anderson & Ramirez, 2017; Del Negro et al. 2018; Figure 2. Complex interaction among brainstem and spinal cord respiratory structures controls respiratory rhythm and pattern generation. The main respiratory structures in the brainstem are shown Baertsch et al. 2019). Inspiration is controlled by the and labeled in different colors. Brainstem respiratory centers in the ventral respiratory group (VRG) send descending projections to respiratory motor neurons in the spinal cord to control respiratory Ö PreB tzinger Complex (Figure 2, green structure), muscle activity. The phrenic motor neurons (PMNs) that control the diaphragm are shown. post-inspiration is controlled by the post-inspiratory complex (PiCo) (Figure 2, black structure), and active expiration is controlled by cells within the parafacial respiratory group (pFRG) (Figure 2, purple structure). Interaction among these central pattern generators

(CPGs) pattern breathing.

All three respiratory oscillators respond to excitatory neural inputs. The PreBÖtzinger Complex contains autonomous pace-maker neurons that are sufficient to drive inspiratory rhythm generation in the absence of additional excitatory input (Johnson et al. 2001; Gray et al. 2010) while the PiCo provides excitatory drive to generate post-inspiratory activity (Ramirez and Anderson 2017; Anderson et al. 2016).

It is likely that these two respiratory rhythms are coordinated through GABAergic connections. In fact, stimulation of neurons in the PreBÖtzinger Complex causes PiCo neurons to hyperpolarize, presumably to prevent co-activated inspiratory and post-inspiratory phases (Gray et al. 2010; Vann et al. 2018). In contrast, active expiration is mainly modulated by the pFRG and is only stimulated under conditions of high oxygen demand (e.g. exercise) or expulsive behaviors such as coughing (Pagliardini et al. 2011;

Huckstepp et al. 2015; Guyenet and Bayliss 2015).

It is also necessary to alternate activation of inspiratory and expiratory brainstem centers in a phase-dependent manner to generate the respiratory cycle. The respiratory system does this through inhibitory interactions. The BÖtzinger Complex is an expiratory brainstem center located in the ventral respiratory group (VRG). The BÖtzinger Complex is believed to drive expiration by inhibiting inspiratory activity from the PreBÖtzinger Complex and rostral VRG. The PreBÖtzinger Complex also contains early- inspiratory neurons that inhibit all expiratory neurons during inspiration (Smith et al. 2007). Thus, inspiratory and expiratory respiratory brainstem centers produce a repetitive and rhythmic respiratory cycle comprised of inspiration, post-inspiration and expiration.

C2. Descending input from the brainstem to the spinal cord generates rhythmic contraction of the diaphragm during resting ventilation

A hallmark of eupnea (normal, unlabored breathing with a resting respiratory rate) is a regular breathing pattern with synchronous contraction of the right and left hemi-diaphragms. Therefore, respiratory centers are bilaterally distributed. As shown in Figure 2, each hemi-diaphragm receives

6 ipsilateral descending input from the ventral respiratory column (VRC) in the brainstem. The VRC extends from just below the facial nucleus within the pontine tegmentum to C1 of the cervical spinal cord and is composed of the PreBÖtzinger Complex, the Bӧtzinger Complex, the rostral ventral respiratory group

(rVRG) and the caudal ventral respiratory group (cVRG) (Figure 2) (Smith et al. 2007). The PreBÖtzinger

Complex is known to contain pacemaker neurons that project to premotor neurons within the rVRG

(Smith et al. 1991; Feldman and Del Negro 2006; Cui et al. 2016; Vann et al. 2018). These premotor neurons within the rVRG project to phrenic motor neurons in the ventral horn of the spinal cord via monosynaptic and polysynaptic pathways (Ellenberger et al. 1990; Dobbins and Feldman 1994; Lane et al.

2008) (black line projections Figure 2). Phrenic motor neurons in the phrenic nucleus (located between

C3 and C5) then relay neural signals to the diaphragm muscle via the phrenic nerve to stimulate diaphragm contraction and start inspiration (Fogarty, Mantilla, and Sieck 2018) (Figure 2).

Since the respiratory circuit is a bilateral system, it is necessary to coordinate left and right respiratory rhythm generation so both sides of the diaphragm contract and relax simultaneously to maximize ventilatory efficiency. V0 neurons within the PreBotzinger Complex arise from Dbx1+ neural progenitor cells. Glutamatergic V0 commissural interneurons within the PreBÖtzinger Complex cross the midline of the spinal cord to contact V0 neurons on the contralateral side of the spinal cord to coordinate synchronous firing between the two sides. Additionally, there are excitatory V0 neurons that project from the PreBÖtzinger Complex to the opposite rVRG that coordinates the synchronous activation of pre- phrenic motor neurons (Wu et al. 2017) (green projections Figure 2). Thus, the VRC of the brainstem bilaterally coordinates synchronous activation of respiratory muscles while ensuring alternating activity of inspiratory and expiratory respiratory brainstem centers to produce a repetitive, rhythmic, and efficient respiratory cycle.

C3. Chemosensory respiratory neurons alter ventilation to maintain physiological pH

A regular respiratory rhythm is characteristic of normal breathing (eupnea) during resting states.

However, the respiratory system has to adapt to physiological stimuli that alter homeostasis, including changes in atmospheric oxygen levels (hypoxia or hypercapnia), emotional states (e.g. stress associated hyperventilation), and changes in the cardiovascular system such as blood pressure, pH and gasses (O2 and CO2). Therefore, while the PreBÖtzinger Complex, PiCo, and the pFRG are critical regulators of respiratory rhythm generation, there are also multiple additional respiratory nuclei that alter ventilation to adapt to homeostatic changes. Ventilation is the movement of air in and out of the system and is largely dependent on the activity of airway patency and primary and accessory respiratory muscles. However, respiration is a dynamic cellular process of gas exchange that responds to homeostatic perturbations

(Merico et al., 2018). In fact, central respiratory drive is highly sensitive to changes in blood gasses, namely oxygen (O2) and carbon dioxide (CO2). These two blood gas concentrations dictate blood pH, which has a narrowly defined physiological range that can support life. Both hypoxia (oxygen deficiency) and hypercapnia (excessive carbon dioxide in the bloodstream) alter blood gas concentrations to shift the equilibrium in favor of CO2 and correspondingly enhance ventilation to remove that excess CO2 (Guyenet and Bayliss 2015). Chemoreceptors that sense these changes and appropriately enhance ventilation to restore blood pH levels are located in peripheral and central nervous system.

C3.1 Peripheral chemoreceptors: The carotid bodies

The carotid bodies are peripheral chemoreceptors that primarily detect changes in arterial oxygen levels as well as blood pH (Prabhakar and Peng, 2017). Carotid bodies are located at the bifurcation of the common carotid artery in the neck and are innervated by the sinus branch of the glossopharyngeal nerve

(De Castro 1926). These nerves carry cardiorespiratory sensory information to the commissural and medial subnuclei of the nucleus tractus solitarius (NTS) in the brainstem, which comprises part of the

8 dorsal respiratory group neurons (Kumar et al. 2012) and may help adapt the respiratory system to chronic hypoxia (Chen et al. 2002; Chen et al. 2007). When arterial PO2 decreases, the carotid bodies cause a chemo-afferent discharge in the carotid sinus nerve, matching the pattern and duration of discharge changes to the severity and duration of the hypoxic stimulus (Kumar et al. 2012). Furthermore, removing carotid body chemoreception via a carotid body denervation abolishes acute ventilatory responses to hypoxia (Timmers et al. 2003). Interestingly, although carotid bodies are peripheral chemoreceptors, they appear to regulate ventilation in conjunction with central chemoreceptors. Final integration between central and peripheral chemoreflexes occurs via convergence of second-order carotid body neuron projections and projections from the main central chemoreceptor located in the brainstem: the retrotrapezoid nucleus (Song et al. 2012; Guyenet 2014).

C3.2 Central chemoreception: The Retrotrapezoid Nucleus

Central chemoreception originates from CO2 sensing neurons in the retrotrapezoid nucleus (RTN)

+ located in the brainstem. The RTN is a group of CO2/H sensitive neurons that express Phox2b and are located within the parafacial respiratory group (pFRG) in the medullary medial reticular formation

(Lazarenko et al. 2009) (Figure 2). Together, the RTN/pFRG contribute to active expiration (Feldman et al.

2013; Huckstepp et al. 2015). They express VGlut2 (and are therefore excitatory neurons) and respond to increases in both PaCO2 (arterial) and increases in local PCO2, suggesting they may sense hypercapnia

+ (Mulkey et al. 2004). Indeed, knock out mice that are lacking Phox2b RTN neurons do not respond to CO2 and suffer fatal central apneas at birth, likely because ventilation is not appropriately stimulated (Dubreuil et al. 2009). Optogenetic stimulation of RTN neurons produces the same ventilatory response as hypercapnia: increased respiratory rate and tidal volume to enhance alveolar ventilation and expel excess

CO2 from the lungs (Abbott et al. 2011). Conversely, Basting et al. 2015 and Marina et al. 2010 showed that inhibiting RTN neurons decreases respiratory rate and tidal volume and abolishes the CO2-dependent onset of active expiration. Furthermore, this reduction in respiratory rate and tidal volume is proportional

9 to arterial blood pH (Basting et al. 2015), consistent with the RTN functioning as a chemoreceptor and receiving both second-order and direct projections from peripheral CBs (Takakura et al. 2006). In order to produce these changes in ventilation, these Phox2b+ neurons in the RTN project to four brainstem regions critical for controlling breathing: the VRC, the NTS, the Kolliker-Fuse nucleus (coordinates expiratory rhythm generator), and the lateral parabrachial nucleus (lPBN) (Bochorishvili et al. 2012; Smith et al. 2013).

There are several ways that the RTN may cause the observed changes in breathing frequency and amplitude by projecting to these other respiratory brainstem centers. RTN neurons may increase excitatory input to the PreBÖtzinger Complex (respiratory rhythm generator) via monosynaptic projections (Bochorishvili et al. 2012) or disinhibit the PreBÖtzinger Complex neurons via complex polysynaptic pathways and increase respiratory rate (Guyenet and Bayliss 2015). Thus, the peripheral carotid body chemoreceptors and RTN central chemoreceptors increase respiratory rate, inspiratory amplitude, and active expiration to help expel excess CO2 from the lungs to restore physiological CO2, O2, and pH levels back to normal following hypoxic and hypercapnic stimuli.

C3.3 Central chemoreception: C1 Neurons

The third major source of central nervous system chemosensation are the C1 neurons located in the rostral ventrolateral medulla (RVLM) of the brainstem. This group of neurons produces short term autonomic, metabolic and neuroendocrine responses to acute physical stressors including (but not limited to) inflammation and infection, blood loss, and central nervous system hypoxia (Guyenet et al. 2013). C1 cells are catecholaminergic (Dahlstrom and Fuxe 1964) and mediate multiple autonomic responses by innervating sympathetic preganglionic neurons, the locus coeruleus, the dorsal motor nucleus of the

Vagus nerve, periacqueductal gray matter, hypothalamic paraventricular nucleus (PVH), perifornical

10 region, and the dorsomedial nucleus of the hypothalamus (Hökfelt et al. 1974). For an extensive review on C1 neuronal control of all autonomic functions, please see (Guyenet et al. 2013).

In summary, three major classes of chemoreceptors exist that help maintain blood pH levels by altering ventilation to expel excess CO2 during hypercapnia or hypoxia: carotid bodies, the RTN, and C1 cells. It is important to understand the role these different chemoreceptors play in altering respiratory rhythm generation and descending brainstem drive as a mechanism to appropriately match ventilation to changes in homeostasis.

C4. Spinal interneurons are important for breathing

The canonical view of the control of phrenic motor neurons primarily focuses on bulbospinal monosynaptic input from the rVRG. Early anterograde and retrograde tracing studies concluded that bulbospinal pathways were exclusively monosynaptic in healthy animals (Ellenberger et al. 1990; Dobbins and Feldman 1994), since transsynaptic labeling failed to identify intermediary interneurons between the rVRG and phrenic motor neurons. However, later studies used a pseudorabies virus (PRV) to identify interneurons in laminae VII and X of the cervical spinal cord that connected the brainstem rVRG and spinal cord phrenic motor neurons in adult rats (Lane et al. 2008). Pre-phrenic interneurons were also discovered in cats (Lois et al. 2008) and ferrets (Yates et al. 1999), suggesting polysynaptic pathways exist to modulate phrenic motor neuron activity (and by extension, other respiratory motor neuron pools).

Electrophysiology has been pivotal in confirming the existence of respiratory interneurons in the spinal cord. For example, some spinal interneurons subsets have been shown to exhibit inspiratory or expiratory related phasic bursting. Merrill and Lipski 1987 showed that less than 4% of synaptic connections from medullary respiratory neurons to intercostal motor neurons in the thoracic cord of the cat were monosynaptic. Instead, thoracic excitatory and inhibitory interneurons showed spontaneous respiratory discharge, and could even be excited following stimulation of segmental afferents. Bellingham

11 and Lipski 1990 later recorded from interneurons at the C5 segment of the spinal cord in cats and discovered interneurons that preferentially fired during inspiration or expiration, even in the absence of afferent stimulation. Interestingly, most of these actually did not fire in response to laryngeal or phrenic afferent stimulation, suggesting they may have been relaying central respiratory drive. Respiratory interneurons have been shown to have heterogenous firing properties. Duffin and Iscoe 1996 showed that out of 26 interneurons they recorded from at C5, 10 interneurons rhythmically discharged during inspiration whereas 16 interneurons discharged tonically with increased firing during inspiration.

Therefore, multiple types of respiratory interneurons exist in the spinal cord.

Multiple studies have shown that respiratory motor neuron output may be modulated by spinal interneuron activity. Spinal interneurons may relay, amplify, or modulate drive to respiratory motor neurons (Lee and Fuller 2011). For example, research in cats showed that inhibitory Renshaw cells receive excitatory input from phrenic motor neurons. In response, these Renshaw cells provide recurrent inhibition to both phrenic and intercostal motor neuron pools in cats as a potential mechanism to coordinate phrenic motor neuron activation only during inspiration (Hilaire et al. 1983; Kirkwood et al.

1981; Lipski et al. 1985; Makito Iizuka et al. 2018). Marchenko et al. 2015 later demonstrated that blocking

GABergic spinal neuron inhibition into the phrenic motor neuron pool increased activity during expiration and altered the shape of the phrenic neurogram. Spinal interneurons also alter their activity in response to acute intermittent hypoxia (a stimulus that increases descending brainstem drive) (K. A. Streeter et al.

2017). Specifically, a subset of pre-phrenic excitatory interneurons received increased excitatory input during hypoxia, suggesting that some interneurons respond to and relay input from descending drive.

Thus, these studies provide convincing evidence that spinal interneurons are involved in controlling respiratory motor neuron output and function.

C5. Spinal neurons mediate sensory afferent feedback

Sensory afferent feedback is another important mediator of respiration which acts through polysynaptic pathways. Like most nerves, the phrenic nerve relays both motor information to the diaphragm muscle (efferents) as well as sensory information back to the phrenic motor neurons

(afferents). These afferents were identified by a transganglionic tracer that localized in lamina IX in the

C3-C5 region of the spinal cord where the phrenic nucleus is located (Nair et al. 2017) (Figure 2).

Interestingly, very few muscle spindles (stretch receptors) providing direct feedback to phrenic motor neurons were observed, and may not contribute substantially to phrenic motor neuron excitability or output (Corda et al. 1965; Road 1990). Instead, spinal interneurons within the dorsal horn and intermediate laminae (I-III, IV, VII and X) receive input from phrenic afferents (Nair et al. 2017) and can respond to phrenic afferent stimulation (Road 1990), suggesting that spinal interneurons mediate responses to phrenic afferent feedback via polysynaptic pathways. In fact, these spinal neurons that respond to phrenic afferent stimulation are largely distinct from the neurons that receive rhythmic drive

(Cleland and Getting 1993; Iscoe and Duffin 1996). These phrenic afferents can even inhibit the activation of the inspiratory external intercostals, showing that phrenic afferents can also modulate other inspiratory pump muscles (De Troyer 1998; De Troyer et al. 1999). Similarly, the intercostal afferents suppress phrenic nerve activity in the intercostal-to-phrenic reflex during inspiratory loading (Decima et al. 1969;

Remmers 1973).

More indirect sensory afferent pathways also exist to modulate diaphragm activity.

Mechanoreceptors within the airway respond to lung inflation and prevent over-inflation by peaking at the end of inspiration. These afferents travel through the Vagus nerve to the NTS in the brainstem that in turn projects to brainstem respiratory centers to inhibit further contraction of the diaphragm (Canning et al. 2014; Narula et al. 2014). Laryngeal mechanoreceptors also increase inspiratory drive during upper airway collapse (Sant’Ambrogio et al. 1983). Therefore, while it is unclear exactly which interneurons

13 mediate laryngeal sensory feedback, these mechanisms act through polysynaptic pathways to relay signals to the spinal cord to alter respiratory muscle activity.

Therefore, while monosynaptic bulbospinal inputs are important for controlling primary pump muscle activity and corresponding ventilation, interneurons are critical mediators for controlling respiratory muscle activity. However, multiple classes of interneurons exist and additional studies are needed to determine which interneuron classes modulate which respiratory muscles and how they help pattern activity. For example, do they relay descending brainstem drive? Or, do they provide tonic activity to motor neurons so they can respond to synaptic inputs from other respiratory circuits? Future studies are required to investigate these questions further.

C6. Coordinating ventilation to motor activity during exercise

The respiratory system has to adapt to many different activities such as climbing, swimming, running, jumping – in other words, exercise. Exercise increases oxygen consumption, the concentration

+ of CO2 and H in muscles, and enhances ventilation in an attempt to restore depleted oxygen stores and expel excess CO2 (Forster et al. 2012). Here, we discuss several mechanisms that appropriately coordinate ventilation to locomotor activity and how interneurons may act as critical mediators between respiratory and locomotor circuits. These mechanisms include a central feed forward mechanism of exercise hyperpnea, neural interactions between locomotor and respiratory circuits, and the activation of accessory respiratory muscles.

C6.1 Central feed forward mechanism

One typical response to exercise is enhanced ventilation prior to or at the onset of exercise. This exercise induced enhancement of ventilation is called exercise hyperpnea (Forster et al. 2012).

Interestingly, this is accomplished without directly sensing the rate of gas exchange in muscles or lungs.

The central feed-forward mechanism of exercise hyperpnea refers to “a signal generated in the brain that

14 initiates the hyperpnea simultaneous with or in advance of locomotion” (Waldrop et al. 1996; Eldridge and Gill-Kumar 1980; Forster et al. 2012). In 1913, Krogh and Lindhard noticed that rapid breathing coincided with the onset or the very beginning of bicycle exercise (Krogh and Lindhard 1913). Because the onset of rapid ventilation was too quick to be explained by changes in metabolic signaling, they concluded it must be due to a central nervous system mechanism. This concept was further tested by ruling out a role for circulatory feedback (Asmussen and Nielsen 1964). Participants were asked to bicycle exercise with cuffs around their upper thighs to block circulation to the legs, yet the increase in minute ventilation was still observed at the onset of exercise. Moreover, even though the cuffs were left on after the cessation of exercise, minute ventilation still decreased. Therefore, it was hypothesized that a neural signal from a suprapontine brain structure coordinated respiration and spinal locomotion and respiration to match ventilation to physical demand. The primary motor cortex, premotor and supplementary motor cortex, subthalamic, mesencephalic, and hypothalamic locomotor areas were all tested, but the only one that elicited locomotor and ventilatory responses to stimulation was the hypothalamic motor region

(Waldrop et al. 1996). It was even shown that stimulating this hypothalamic region increased blood pressure, heart rate, and ventilation before locomotion began, consistent with previous studies (Waldrop et al. 1988) Moreover, this region may mediate hyperpnea by stimulating ventilation via projections to the RTN (Eldridge et al. 1981). Thus, a central feed forward neural mechanism is likely important for matching ventilation to locomotion.

C6.2 Neural interactions between locomotor and respiratory circuits

In addition to the feedforward mechanisms, there is mounting evidence that feedback from locomotor circuits influence respiratory circuits. Limb afferents play important roles in 1) matching cardiorespiratory activity levels to locomotor levels and 2) coupling the respiratory and locomotor cycles to maximize the efficiency of ventilation and movement (Shevtsova et al. 2019). In some cases, entrainment of these two circuits appears to occur via ascending limb afferent projections directly to

15 brainstem respiratory centers (Morin and Viala 2002; Giraudin et al. 2008; Giraudin et al. 2012). For example, in isolated brainstem/spinal cord preparations, activation of fictive locomotion can cause an increase in phrenic burst frequency as well as entrainment of locomotor and respiratory periods (Le Gal et al. 2016; Yazawa 2014). Sensory afferents from the limbs also appear to play an important role, since

Yu and Younes 1999 showed that high intensity stimulation of phrenic group III-IV afferents can stimulate ventilation up to 4-5 times baseline values. Moreover, the stimulation of limb afferents and phrenic afferents can increase ventilation in an additive manner, suggesting they act through different mechanisms (Ward et al. 1992).

There are also intersegmental interactions within the spinal cord that link the locomotor and respiratory circuits that appears to be independent of brainstem drive. For example, spinalized rabbits exhibit entrainment (a 1:1 coupling between successive periods) between locomotor and phrenic motor output, with inspiratory activity highly correlated to extensor activity (Viala et al. 1987). Bimodal spinal neurons were also recently described in a neonatal brainstem spinal cord preparation. These neurons were expiratory and also received flexor-related drive via propriospinal pathways (Le Gal et al. 2016). It was also shown that phrenic motor neurons can be driven by lumbar spinal circuits when local inhibition is blocked (Cregg et al. 2017), suggesting that spinal interneurons are key mediators between respiratory and locomotor circuits. Thus, the nervous system uses propriospinal neural pathways to appropriately match ventilation to the intensity of locomotor tasks. However, additional studies are needed to understand how propriospinal pathways mediate these signals and which classes of interneurons are involved.

C6.3 Activation of accessory respiratory muscles

An additional mechanism that helps maintain ventilation during exercise is the recruitment of accessory respiratory muscles (ARMs). Even though the diaphragm accounts for 65% of tidal volume at

16 rest (Di Marco, 2009), inspiratory ARMs (scalenes, trapezius, pectoralis, sternocleidomastoid, and parasternal muscles) are used to stabilize the chest wall during eupnea and increase thoracic volume to ensure sufficient ventilation during increased oxygen demand (Gransee and Sieck, 2012). There is evidence for at least three sources of neural control for ARM activity: 1) descending neural drive from the brainstem 2) sensory afferent feedback and 3) spinal interneurons.

C6.3.1 Descending neural drive modulates activation of accessory respiratory muscles

Descending commands from the brainstem help coordinate inspiratory ARM activity. Like the diaphragm and external intercostals, inspiratory ARMs receive rhythmic drive from the rVRG of the medulla (De Troyer et al. 2005; Butler 2007; Johnson and Mitchell, 2013; Butler et al. 2014). In contrast, expiratory internal intercostal and abdominal motor neurons receive bulbospinal input from the cVRG.

Each set of muscles has its own distinct motor neuron pool within the spinal cord. This allows each muscle to have distinct neural pathways to control the timing, discharge frequency, and patterns of activity

(Butler et al. 2014). One proposed explanation for this observation is that “neuromechanical matching” of drive to the inspiratory motor neurons ensures the most efficient contraction of inspiratory muscles based on their mechanical advantage. In humans, for example, rostral intercostal muscles are preferentially recruited for inspiration because they provide higher mechanical advantage compared to caudal intercostal muscles. This pattern cannot be attributed solely to intrinsic motor neuron properties and does not require afferent input. Moreover, neuromechanical matching of drive to inspiratory neurons is likely mediated by propriospinal circuits (Oka et al. 2019). This neuromechanical matching is likely important for activating ARMs during exercise, especially since central drive to accessory respiratory muscles is increased during exercise to ensure sufficient ventilation (Aliverti 2016). Thus, descending brainstem drive helps regulate ARM activity by activating them under conditions of high oxygen demand while preventing their activation when they do not need to be active to preserve energy.

C6.3.2 Sensory afferent feedback modulates accessory respiratory muscle activity

Sensory afferent feedback has also been shown to modulate ARM activity. De Troyer 1988 and De

Troyer et al. 1999 showed that stimulation of phrenic nerve sensory afferents inhibited inspiratory intercostal activity. Additionally, short-latency intersegmental reflexes have also been described between intercostal and scalene muscles in humans (McBain et al. 2016). The significance of spinal intersegmental reflexes for coordinating respiratory muscle activity may best be illustrated by the observation that proprioceptive afferents can entrain phrenic nerve activity to chest movements driven by a ventilator even in spinalized animals (Persegol et al. 1987). Thus, sensory afferents are likely important modulators of ARM activity.

C6.3.3 Spinal interneurons control accessory respiratory muscle activity

Emerging evidence that interneurons might play a key role in regulating the activity of accessory respiratory muscles (ARMs) comes from tracing and functional studies. First, spinal interneurons in the dorsal or intermediate laminae have been shown to receive respiratory muscle afferents (Nair et al. 2017).

The Crone lab has even revealed an important role for a specific class of interneuron. The scalene ARM was injected with PRV, which was retrogradely transported back to scalene motor neurons as well as premotor V2a interneurons (unpublished observations). Functional studies then used chemogenetics to activate brainstem and spinal V2a interneurons and demonstrate that they are sufficient to activate the scalene and trapezius ARMs and increase ventilation in healthy mice (Romer et al. 2017). However, additional studies are necessary to 1) further elucidate how many different types of interneurons control

ARM activity 2) determine which neurons are upstream of interneurons that control ARM activity 3) determine how these neurons are organized to regulate ARM activity so they are only active when needed and 4) determine how interneurons control recruitment of ARMs following disease or injury.

In summary, the neural control of breathing is multifaceted, involving complex interactions among brainstem rhythm generating centers, chemoreceptors, sensory feedback, and interneurons to constantly adjust ventilation to different behaviors and tasks.

D. Breathing is Impaired in Neuromuscular Diseases

Respiratory failure is the leading cause of death in amyotrophic lateral sclerosis (ALS) (Hardiman 2011;

Lyall 2001). In extreme cases, patients need to be placed on a mechanical ventilator, which is costly, not always effective, and can severely compromise independence, sleep, speech, sense of smell, mental state and overall of quality of life (Hachmann et al. 2017). Moreover, mechanical ventilation can lead to respiratory muscle decompensation (including diaphragm dysfunction) and a weakening of the synaptic pathways controlling respiratory muscles (Petrof and Hussain 2016). Breathing impairment is caused by the dysfunction and death of respiratory motor neurons and interneurons in neuromuscular disorders.

However, patients can exhibit respiratory circuit plasticity to compensate for respiratory dysfunction. Well established mechanisms of plasticity include, but are not limited to: 1) increased central respiratory drive

2) motor neuron plasticity 3) plasticity at the neuromuscular junction and 4) activation of ARMs (Nichols et al. 2013). Section D will review why breathing is disrupted in amyotrophic lateral sclerosis (ALS), how the respiratory system compensates for impaired breathing, and treatments for respiratory insufficiency.

D1. Respiratory deficits in ALS

ALS is a fatal neuromuscular disease that is characterized by progressive muscle weakness due to the death of motor neurons (Nichols et al. 2013). To date, no cure has been found for ALS. Two types of

ALS exist: sporadic ALS and familial ALS. Approximately 90% of all ALS cases are sporadic, meaning the cause of the disease is unknown (Cleveland and Rothstein 2001). 5-10% of ALS cases are familial and are caused by known genetic mutations that have been passed onto at least two family members. Gene mutations encoding TARDBP, SQSTM1, VCP, FUS, PFN1, OPTN, and superoxide dismutase 1 (SOD1) are

19 only some of the estimated 25 ALS genes that have been discovered (Nichols et al. 2013; Renton et al.

2014). Regardless of etiology, ALS is a fast-acting disease with a high fatality rate within 2-5 years of adult onset (Ludolph and Knirsch 1999). Patients who opt to be placed on a mechanical ventilator via a tracheostomy may extend their lifetime up to an additional two years; however, few choose to do so because mechanical ventilation is costly and severely impairs the quality of life for the patient. Outside of respiratory management near the end of the disease, ALS is particularly difficult to treat due to a wide spectrum of neuropathology, with distinctive neuropathological and molecular signatures (Johnson and

Mitchell, 2013; Nichols et al. 2013).

ALS is primarily characterized by a progressive dysfunction and loss of upper (cortical) and lower

(brainstem and spinal cord) motor neurons, leading to muscle spasticity, weakness, and eventually paralysis (Charcot and Joffroy, 1869; Mayeux R 2003). Other symptoms of motor neuron loss include hyperreflexia, swallowing and speaking (bulbar) deficits (Cleveland and Rothstein, 2001), central sleep apnea (Bourke et al. 2001), as well as respiratory insufficiency. Specifically, patients exhibit reduced inspiratory capacity and maximum voluntary ventilation that affects the ability to not only maintain sufficient ventilation at rest, but also to perform higher ventilatory effort behaviors such as sighing

(augmented breath) and forceful coughing (Fallat et al. 1979; Lyall 2001; Stewart et al. 2001; Talakad et al. 2009). This is particularly detrimental to patients because sighing can re-open collapsed alveoli and prevent atelectasis (Li and Yackle 2017) while coughing clears the airways from infection. It is actually this deficit in producing a forceful cough that often leads to respiratory failure because patients cannot expel debris from the airway, increasing risk for infections (Gregory 2007). Coughing requires inspiratory and expiratory motor neurons to coordinate activity. The diaphragm has a greater contraction to increase intrathoracic pressure against a closed epiglottis while expiratory muscles become active to expel the air at a greater force once the epiglottis opens (Polverino et al. 2012). Therefore, weakness and atrophy of

20 both the primary and accessory respiratory muscles contribute to respiratory failure in ALS. As a result,

ALS research has investigated mechanisms that lead to the degeneration of respiratory motor neurons.

The diaphragm is controlled by both upper and lower motor neurons, both of which degenerate in ALS. In healthy individuals, a rapidly conducting corticospinal pathway connects upper motor neurons in the motor cortex with phrenic motor neurons in the spinal cord (Gandevia and Rothwell 1987). Using

EEG, this corticospinal pathway has further been shown to control voluntary control of breathing, including self-paced inhalations and forced exhalation (but it does not contribute to resting ventilation)

(Macefield and Gandeviat 1991). In fact, it is these neurons help prevent apnea under hypercapnic conditions in waking states but not in sleeping states (Datta et al. 1991). Thus, it is not surprising that the loss of upper motor neurons is associated with a defect in voluntary breathing, sleep disorders, and bulbar deficits in ALS patients (Shimizu et al. 2010) and that degeneration of these motor neurons may predict respiratory insufficiency (Miscio et al. 2006).

In contrast, it is the brainstem VRC and lower phrenic motor neurons that largely control inspiration during resting states. Loss of these neurons contributes to disease progression, especially since the neuromuscular junction and phrenic nerve can degenerate, leading to neurotransmission failure and diaphragm atrophy (Rizzuto et al. 2015). Animal research has revealed that there is drastic degeneration of phrenic motor neurons (Llado et al. 2006), intercostal motor neurons (Nichols et al., 2013), V2a interneurons (Romer et al. 2017), reduction of diaphragm action potentials, impaired transmission of via the neuromuscular junction (Rizzuto et al. 2015), phrenic nerve fiber loss, and diaphragm atrophy (Llado et al. 2006). Thus, it is clear that one component of respiratory failure in ALS is diaphragm dysfunction.

In addition to the phrenic motor neurons controlling the diaphragm, ARM motor neuron pools also degenerate during ALS progression (Gregory 2007; Pinto et al. 2007; De Carvalho et al. 2010). This can cause instability of the ribcage and abdomen, leading to paradoxical breathing (Smith et al. 1987; Park et al. 2010). However, different muscles degenerate at different rates, depending on the muscle’s motor

21 neuron size, their axons, and motor unit size (Hegedus et al. 2007). For example, the accessory respiratory muscles degenerate faster than airway muscles in patients with spinal onset ALS (Marti-Fabregas et al.

1995; Lechtzin 2006).

Several types of respiratory motor units/muscle fibers exist. A motor unit is composed of a single motor neuron and the numerous muscle fibers it innervates. There are four types of muscle fibers: slow twitch (Type 1) and three subtypes of fast twitch (Type II). Type I slow twitch fibers efficiently use oxygen to generate ATP for continuous, extended muscle contractions over a long time. As a result, these are also known as fatigue resistant (FR). Conversely, fast twitch fibers use anaerobic metabolism to generate short bursts of contractions in rapid succession. Specifically, Type IIa fast twitch fibers are called intermediate fast-twitch fibers and use both aerobic and anaerobic metabolism to create energy. Type IIb fibers use anaerobic metabolism to produce quick, powerful bursts of speed with a high firing rate that are quick to fatigue (Mendell 2005). Henneman’s Sizing Principle states that motor units are recruited from smallest to largest, meaning that slow-twitch, low-force, fatigue-resistant Type I muscle fibers become activated before fast-twitch, high-force, fatigue Type II fibers (hypothesized to be a mechanism meant to conserve energy) (Mendell 2005). Normal breathing recruits slow fatigue-resistant fibers, increased ventilatory demand recruits additional fast fatigue resistant motor units, and expulsive behaviors (coughing) requires fast fatigable motor units (Polla et al. 2004; Sieck et al. 2013). In ALS, these different muscle fiber types exhibit differential susceptibility to atrophy during disease progression.

Research in ALS model rats has shown that fast twitch, fast-fatigue Type IIb fibers are most susceptible to degeneration, followed by type IIa and finally to Type I fatigue resistant motor units (Nichols et al., 2013). Most respiratory muscles are comprised of a mixture of each fiber type. For example, 46.8% of the human costal diaphragm is composed of Type I slow twitch muscle fibers while 45.1% are fast twitch muscle fibers (Meznaric and Cvetko 2016). Similarly, the trapezius inspiratory ARM is composed of 66%

Type I fibers and 34% Type II fibers (Larsson et al. 2001). Interestingly, Valdez et al., 2012 showed that the

22 diaphragm and limb muscles show equal susceptibility to the effects of denervation in ALS. However, ventilation is preserved until late stages of disease progression in ALS patients. It is only near the end of the disease that ALS patients show drastic decreases in forced vital capacity, maximum voluntary ventilation, and residual volume (Fallat et al. 1979; Stewart et al. 2001; Talakad et al. 2009; Nichols et al.,

2013).

Similar results were observed in the SOD1G93A ALS mouse model. This mouse model encodes a mutation accounting for 20% of familial ALS cases and faithfully recapitulates human disease progression

(Rosen et al. 1993; Gurney et al. 1994; Wong et al. 1995; Liu et al. 1999; Nagai et al. 2001; Howland et al.

2002; Andersen et al. 2003; Bruijn et al. 2004). Like humans, despite the consistent and progressive decline of skeletal muscle function and motor neuron loss, ventilation does not change until late stages of the disease. Tankersely et al. 2007 showed that SOD1G93A mice maintain breathing frequency, tidal volume and minute ventilation until late stages of disease (18 weeks old). It was only when body weight decreased by 20-30% that minute volume and tidal volume significantly increased that ALS severely impacted ventilation. Specifically, there was a sharp decrease in ventilation immediately preceding endstage when animals are no longer mobile enough to feed themselves (Tankersely et al. 2007; Romer et al. 2017). Taken together, these data suggest that although skeletal (including respiratory) motor neurons progressively degenerate and respiratory muscles atrophy, the respiratory system is somehow able to compensate and maintain ventilation until late stages of ALS.

D2. Compensatory plasticity in ALS

The respiratory system is necessarily complex to preserve life, and therefore has great potential for compensatory plasticity following impairment. In ALS, multiple forms of plasticity are implicated in compensating for the loss of respiratory motor neurons. Mechanisms of respiratory plasticity include 1) altered morphology of spared respiratory motor neurons and their neuromuscular junctions 2) increased

23 central respiratory drive from the brainstem rVRG to respiratory motor neurons and 3) increased activation of ARMs.

Because the documented loss of phrenic motor neurons exceeds what would be predicted based on the ventilatory capacity that is maintained in rodent models, it is likely that a combination of these compensatory mechanisms maintains normal ventilation during early to mid-stages of disease progression (Nichols et al. 2013). For example, spared phrenic motor neurons show upregulated expression of neurotrophic factors, indicative of spinal motor plasticity (Satriotomo et al. 2012).

Additionally, Gordon et al. 2004 discovered that each spared locomotor motor neurons shows an increased number of associated neuromuscular junctions, a phenomenon which may also occur in respiratory motor neurons. Increased central respiratory drive may also compensate for loss of respiratory motor neurons. Acute intermittent hypoxia (AIH) increases respiratory drive to induce long term phrenic facilitation, a prolonged increase in phrenic nerve burst amplitude lasting several hours after the final hypoxic episode (Devinney et al. 2013). Interestingly, AIH can induce long term phrenic facilitation at all stages in SOD1(G93A) rats to combat respiratory failure (Nichols et al. 2015) and together with stem cell transplantation can even preserve breathing (Nichols et al. 2013). However, increasing brainstem drive likely does not account for all compensatory mechanisms to preserve breathing, especially since the apneic threshold remains unchanged (Nashold et al. 2006). In fact, chronic intermittent hypoxia may actually induce respiratory insufficiency by causing respiratory muscle fatigue due to over activation

(Strickland et al. 2019). Therefore, maintaining ventilatory capacity during late stages of disease may require another mechanism, such as the activation of ARMs.

Increased ARM activation has been observed in both ALS patients and rodent models. Specifically, rhythmic activation of the parasternal intercostals can fully drive ventilation under complete diaphragm paralysis (Bennett et al. 2004; Johnson and Mitchell 2013; Nichols et al. 2013) and patients who recruited

ARMs in their sleep survived longer than patients that did not recruit ARMs (Arnulf et al. 2000). Animal

24 research in SOD1(G93A) ALS model mice also exhibit ARM activation. Specifically, young asymptomatic mice and healthy wild-type mice with strong diaphragm activity do not show scalene and trapezius inspiratory ARM activation during normal, resting breaths. Instead, ARMs are only recruited during higher inspiratory behaviors (e.g. sighs and sniffing). In contrast, ARM activation is observed at rest in rodent models of ALS. ARM activity substantially increases during early- to mid-stage disease progression and is associated with enhanced ventilation (with no change in minute ventilation) despite progressive phrenic motor neuron death. However, ARM activation returns to baseline levels (almost no activation) during late stages of the disease when animals have to be humanely sacrificed to prevent unnecessary suffering

(Romer et al. 2017). This is consistent with other studies that report degeneration of intercostal motor neurons (Nichols et al., 2013) as well as drastic denervation and atrophy of the sternocleidomastoid muscle at endstage (Smittkamp et al. 2010). Therefore, preserving ARM activation as a compensatory mechanism to maintain ventilation may be a promising strategy to prevent respiratory insufficiency.

Future studies are needed to determine how ARM activation is controlled by the nervous system, especially since it is unknown why some ALS patients recruit ARMs to compensate for impaired diaphragm function while others do not. The loss of respiratory motor neurons, including ARM motor neurons, has been well documented in ALS patients and animals. Less is known about how interneuron loss alters ARM activity in ALS. Dysfunction of cortical and spinal inhibitory interneurons has also been reported and is thought to contribute to degeneration in ALS (Turner and Kiernan, 2012). Romer et al. 2017 later discovered that excitatory V2a interneurons in the brainstem and spinal cord also degenerate at the same rate as limb and phrenic motor neurons in a mouse model of ALS, the same type of interneuron that is sufficient to activate ARMs in healthy mice at rest. Therefore, it is important to identify interneuron substrates (such as V2a neurons) that control ARM activation and determine if altering their excitability throughout disease progression can increase ARM activity and enhance ventilation.

D3. Treatment for ALS: New Therapies are Needed

To date, there is no cure for ALS. Instead, treatment for inevitable respiratory failure relies on management practices. It is recommended that ALS patients undergo respiratory muscle function evaluation every three months, using common tests of respiratory muscle strength such as forced vital capacity, sniff nasal pressure, maximal inspiratory pressure, and cough peak flow (Gregory et al., 2007).

Current standard of care practices includes noninvasive ventilation via a face mask or nasal pillow, tracheostomy and invasive ventilation, medications for patients opting out of invasive ventilation, and prevention of pneumonia. Noninvasive ventilation is initially prescribed for patients suffering from nocturnal hypoventilation or who have an FVC <50%. As ALS progresses, patients suffering from daytime dyspnea may be asked to utilize it continuously. Moreover, noninvasive ventilation may eventually become ineffective at maintaining ventilation or bulbar dysfunction. This prevents further use and necessitates the use of a tracheostomy and/or mechanical ventilation to maintain breathing. However, less than 10% of ALS patients in the United States choose to undergo mechanical ventilation because it is expensive, severely impairs independence, and there is a lack of health care coverage and trained attending personnel (Lechtzin 2006; Gregory et al., 2007). Furthermore, while this management practice ventilates patients, it does not halt disease progression or degeneration. As such, the majority of patients who choose to withdraw from a noninvasive ventilator instead of being placed on a mechanical ventilator are prescribed opioids and benzodiazepine medication primarily for symptom and discomfort relief during end of life care (Gregory et al., 2007; Niedermeyer et al. 2019).

Diaphragm pacing is another option that was tested to improve diaphragm function; however, early clinical trials revealed increased mortality and a reduction in ventilator free time (Neidermeyer et al. 2018) and is no longer an approved treatment option for ALS. There are four FDA approved drugs for

ALS on the market (Radicava, Rilutek, Tiglutik, and Nuedexta). However, these only slow disease

26 progression without reversing its effects, and are only predicted to extend life by 2 months – 1 year

(Dharmadasa and Kiernan 2018; Cruz 2018).

There is a clear need for new therapies to prevent respiratory insufficiency in ALS. As mentioned earlier, ALS is primarily characterized as a motor neuron disease. However, emerging evidence suggests that interneuron dysfunction and degeneration may be just as important in contributing to ALS pathology and respiratory failure. Therefore, it is important to develop ways to 1) improve respiratory motor neuron and interneuron survival and 2) improve respiratory muscle activity.

E. Breathing is Impaired Following High-Level Spinal Cord Injury

E1. Respiratory deficits following spinal cord injury

Spinal cord injury is an acute, traumatic injury that disrupts descending neural input into spinal circuits. The severity of the injury is dependent on the level of the injury because higher level (more rostral) injury sites affect more locomotor and respiratory motor neurons than more caudal injury sites.

For example, patients with an injury above C5 are likely to be placed on a mechanical ventilator whereas a complete transection at T12 will not affect respiration, balance, or trunk/upper body strength (Berlowitz et al., 2016). The majority of spinal cord injury research is focused on regaining locomotor ability, yet over half of all spinal cord injuries occur at cervical levels which impairs the ability to breathe. Often times, this leads to patient dependence on a mechanical ventilator which is costly, severely impairs independence and quality of life, and can even lead to respiratory muscle decompensation (Vazquez et al. 2013).

Therefore, it is important to understand how spinal cord injury disrupts neural pathways that control respiration in order to develop therapies to improve breathing.

A multitude of respiratory complications ensue following spinal cord injury. Atelectasis (collapse of alveoli in the lungs), pneumonia, pulmonary edema, pulmonary embolism, paradoxical breathing, sleep disordered breathing, sleep apnea syndrome, and respiratory insufficiency all contribute to respiratory

27 failure, which is the leading cause of death for spinal cord injury patients (Fishburn et al. 1990; Berlly and

Shem 2007; Fuller et al. 2013; Berlowitz et al. 2016). Flaccid paralysis of muscles below the site of injury occurs immediately following injury. Specifically, lost descending bulbospinal input into phrenic motor neurons paralyzes the diaphragm following high level spinal cord injury. Paralysis of intercostal and abdominal muscles causes chest wall instability and leads to paradoxical breathing, resulting in increased respiratory effort and a significant reduction in vital capacity, both of which contribute to respiratory muscle fatigue. Moreover, the ability to cough is severely impaired, leading to decreased clearance of airway secretions that increases patient vulnerability for atelectasis, pulmonary infections and mucus retention (Schilero et al. 2009; Fishburn et al. 1990; DeVivo, Black, and Stover 1993; Cotton et al. 2005).

Normal, unlabored breathing is also significantly affected following spinal cord injury. Common ventilation changes observed in quadriplegics include a decrease in tidal volume, vital capacity (reduced to 20-60% of predicted levels in tetraplegics), and expiratory flow rates and an increase in breathing frequency (Loveridge and Dubo 1990; Schilero et al. 2009). Some studies also report a decrease in the frequency of sighs (Schilero et al. 2009), likely due to impairment of the diaphragm, inspiratory pump muscles, and ARMs that support the generation of these augmented breaths. Decreased sigh frequency may lead to atelectasis as sighs normally function to re-open collapsed alveoli. A reversal of these ventilatory changes may be observed within the first 12 months following injury, likely resulting from decreased inflammation and edema (Haas et al., 1985), compensatory activation of ARMs (Loveridge et al. 1992), and increased rib cage stability (McMichan 1980; Haas et al. 1985). However, lingering intercostal and scalene muscle activity is associated with higher pulmonary function and decreased rib cage instability (Guttmann and Silver 1965;De Troyer and Heilporn 1980; Estenne and De Troyer 1985).

Thus, spinal cord injury affects both inspiratory and expiratory primary and accessory respiratory muscles to impair ventilation and lead to respiratory failure.

E2. Compensatory plasticity following spinal cord injury

High level cervical spinal cord injury impairs the ability to breathe by disrupting brainstem descending input into respiratory motor neurons. Because breathing is essential to life, the respiratory system necessarily needs to adapt to impaired breathing following injury to preserve life. It is important to note that respiratory plasticity does occur, as evidenced by the recovery of breathing over time in some patients (Hoh et al. 2013). Compensatory plasticity following spinal cord injury takes on many forms, including activation of latent neural pathways, morphological and signaling changes in the phrenic motor neurons, altered synaptic input into respiratory motor neurons, altered interneuron connectivity and activity, and the activation of ARMs.

E2.1 Compensatory Plasticity Following Spinal Cord Injury: The Crossed Phrenic Phenomenon

High level cervical spinal cord injury disrupts descending bulbospinal input from the rVRG premotor neurons to respiratory motor neurons located in the spinal cord, impairing respiratory muscles below the site of injury (Berlly and Shem, 2007; de Paleville et al. 2011; Berlowitz et al. 2016). Since breathing is required for life, multiple compensatory mechanisms exist in the respiratory system to adapt to impaired breathing following disease or injury. A promising strategy to restore respiratory and locomotor function to spinal cord injury patients involves activating spared or latent pathways

(Goshgarian et al. 2003; Alilain et al. 2008; Courtine et al. 2008; Smith and Knikou 2016; Cregg et al. 2017;

Warren et al. 2018). Multiple models of spinal cord injury exist to study respiratory circuit plasticity, including spinal contusions, spinal transections, and spinal hemisections. Here, we focus on a well- established experimental model of spinal cord injury in which a latent pathway – the Crossed Phrenic

Pathway (CPP) – can be activated to restore breathing following a high-level spinal cord injury.

The CPP was first discovered in 1895 and termed the CPP in 1936 (Rosenblueth and Ortiz 1936).

Porter performed a spinal hemisection at C2 (C2Hx) to sever descending bulbospinal input into phrenic

29 motor neurons on one half of the spinal cord, paralyzing the ipsilateral hemidiaphragm while leaving contralateral hemidiaphragm function intact (Figure 3B). A subsequent phrenicotomy on the contralateral side cut off phrenic nerve input into the contralateral diaphragm muscle, causing complete paralysis of the diaphragm (Figure 3C). However, this diaphragm paralysis was not fatal. Instead, the previously paralyzed diaphragm resumed rhythmic bursting activity after a brief period of asphyxia. (Figure 3D).

The restoration of ipsilateral diaphragm activity was hypothesized to result from the activation of latent crossed phrenic pathways (Goshgarian 2003). The CPP is a neural pathway that originates from the contralateral rVRG and decussates below the site of the C2Hx injury to contact the ipsilateral phrenic motor neuron pool (Figure 3B-D) (Goshgarian et al. 1991; Moreno et al. 1992). It is considered to be a latent pathway, typically silent or functionally inactive under healthy, uninjured conditions. The CPP has been discovered in mice, dogs, cats, rabbits, rats, guinea pigs, and woodchucks, and can be induced within minutes to hours following a C2Hx depending on the species (Goshgarian et al. 2003). For example, it takes 3.5 hours to induce the CPP in guinea pigs (Goshgarian and Guth 1977) whereas 86% of mice can induce the CPP within 4-8 hours following a C2Hx (Minor et al. 2006). Because the CPP has been discovered in multiple species and is dependent on a spinal cord injury that can be consistently reproduced, it is well established as an experimental model of respiratory circuit plasticity.

Porter hypothesized that anatomical substrates of the CPP were phrenic motor neuron dendrites that crossed the spinal cord to receive descending input from the contralateral brainstem (Porter 1895).

However, the number of commissural phrenic motor neuron dendrites is substantially decreased from neonates to adults during development (Prakash et al. 2000). Instead, the activation of the CPP and any subsequent neuroplasticity is reliant on changes at multiple levels of respiratory control, including altered central respiratory drive from the brainstem, altered morphology and signaling properties of phrenic motor neurons, and altered interneuron connectivity and excitability. Here, we briefly review the mechanisms underlying activation of the CPP following a C2Hx spinal cord injury.

Figure 3. The Crossed Phrenic Phenomenon. (A) Descending projections (solid blue lines) from the ventral respiratory group (VRG) contact phrenic motor neurons (PMNs) to cause rhythmic bursting from the diaphragm. (B) A C2Hx spinal cord injury disrupts descending brainstem input on one half of the spinal cord, paralyzing the ipsilateral side of the diaphragm. There is a latent pathway that crosses below the site of injury to contact the ipsilateral phrenic motor neuron pool, called the Crossed Phrenic Phenomenon (CPP) (gray lines). (C) A contralateral phrenicotomy (severing the phrenic nerve) paralyzes the contralateral diaphragm. (D) The contralateral phrenicotomy asphyxiates the animal, increases contralateral descending brainstem drive (lightning bolt), and activates the CPP (black lines) to restore rhythmic bursting activity to the previously paralyzed diaphragm.

Lewis and Brookhart 1951 showed that any condition that enhanced central respiratory drive enhanced the output of the CPP. In contrast, decreasing central respiratory discharge attenuated the CPP.

Additional evidence comes from Erickson and Millhorn 1994, who showed that stimulation of the carotid sinus nerve can induce the CPP, but additional chemoreceptor activation via hypoxia/hypercapnia further increased ipsilateral diaphragm activity. Any form of temporary asphyxia (including nasal occlusion, hypoxia and hypercapnia) increases brainstem respiratory drive to temporarily activate the CPP and restores bursting activity to the previously paralyzed diaphragm (Fuller et al. 2003; Golder and Mitchell

2005; Golder et al. 2008; Hernandez-Torres et al. 2017). Since paralysis of the ipsilateral diaphragm returns shortly after removing asphyxia, temporary induction of the CPP is considered neuromodulation.

In order to be considered neuroplasticity, the CPP must be strengthened to permanently restore ipsilateral diaphragm activity. One way to strengthen the CPP is with chronic intermittent hypoxia. Chronic intermittent hypoxia (CIH) over the course of 7 days drastically improved phrenic output during normoxia, hypoxia and hypercapnia in rats two weeks following a C2Hx compared to rats who were not treated with

CIH (Fuller et al. 2003). Even acute intermittent hypoxia can improve functional recovery long after spinal

31 cord injury had occurred (Golder and Mitchell 2005). Thus, increased central respiratory discharge from the brainstem is sufficient to induce the CPP and restore diaphragm function following a C2Hx spinal cord injury.

Altered release of neurotransmitters have also been observed following a C2Hx. Several studies have identified serotonin (5-HT) and adenosine as two important neural modulators of neuroplasticity

(Bach and Mitchell 1996; Baker-Herman and Mitchell 2002; Petrov et al. 2007; Navarrete-Opazo et al.

2017; Wen et al. 2019). Even in healthy, uninjured rodents, serotonin modulates a phenomenon called phrenic long-term facilitation (pLTF). pLTF is is a form of respiratory plasticity characterized by a long- lasting increase in respiratory motor output or drive lasting at least one hour (Fuller et al. 2000). 5-HT2A/C receptor activation and subsequent protein synthesis is necessary for pLTF (Millhorn et al. 1980; Kinkead et al. 1998; Baker-Herman and Mitchell 2002). In fact, both serotonin agonists (specifically for 5HT2A receptors) and adenosine antagonists have even been shown to experimentally enhance respiratory plasticity following spinal cord injury (Goshgarian 2003). In contrast, blocking the tryptophan hydroxylase enzyme necessary for making serotonin reduced the ability to induce the CPP (Hadley et al. 1999).

Phrenic motor neurons themselves also display altered morphology and firing patterns following a high level C2Hx. Castro-Moure and Goshgarian (1996 and 1997) first discovered three main changes, including increased length and number of dendro-dendritic appositions, increased number of multiple synapses onto the phrenic nucleus, and increased length of synaptic contacts onto phrenic motor neurons.

Interestingly, it was hypothesized that the increased number of dendro-dendritic appositions resulted from the retraction of glial processes, which occurred within hours of a C2Hx (Goshgarian et al. 1989).

The loss of this separation between cell bodies and primary dendrites may have even allowed coupling between phrenic motor neurons to amplify bulbophrenic synaptic inputs. Phrenic motor neurons even upregulate the serotonin receptor 5HT2A as a result of the C2Hx (Fuller et al. 2005) and blocking serotonin synthesis with a tryptophan hydroxylase antagonist (Hadley et al. 1999) attenuated these morphological

32 changes. Moreover, 5HT2A activation triggers protein synthesis for BDNF and TrkB (Meller et al. 2002), components of a signaling pathway that promotes neurological recovery (Xiangzhe et al. 2019). Phrenic motor neurons also show increased expression of receptors that are activated by glutamate, including

GluR3/4, NMDAR, and AMPAR that are thought to enhance LTP like mechanisms (another form of neuroplasticity). For a detailed review, see Goshgarian 2003. Thus, the induction of the CPP may be partially mediated by changes in the phrenic motor neurons themselves following a C2Hx.

The activation of the CPP appears to mediated by many mechanisms. This section has focused on research aiming to activate the CPP by increasing brainstem respiratory drive or by altering the excitability of phrenic motor neurons themselves. However, the neuronal circuitry underlying respiratory plasticity is comprised of more than just bulbospinal projections to respiratory motor neurons. Interneurons play key roles in controlling respiration in healthy mice and following disease and injury, suggesting they may be important modulators for the recovery of diaphragm function via the CPP. Therefore, we next review the roles of excitatory and inhibitory interneurons in promoting recovery following spinal cord injury.

E2.2 Interneurons play key roles in promoting respiratory plasticity following spinal cord injury

More and more evidence supporting the role of spinal neurons in respiratory plasticity following spinal cord injury suggests they may be key players in promoting recovery of breathing. Since respiratory motor neurons also receive excitatory and inhibitory inputs from spinal neurons, spinal circuits may serve as substrates to improve breathing following injury (Lane et al. 2011; Marchenko et al., 2015; Zhouludeva et al. 2018).

Early studies suggested that descending pathways from the rVRG to the phrenic motor neurons were exclusively monosynaptic (Ellenberger et al. 1990; Moreno et al. 1992; Dobbins and Feldman 1994).

However, later studies suggested that the CPP could also include polysynaptic pathways comprised of propriospinal neurons. Specifically, Lane et al. 2008 used a transynaptic tracing agent (pseudorabies virus)

33 to identify spinal neurons presynaptic to phrenic motor neurons that also receive input from anterogradely labeled rVRG neurons in both uninjured and spinal cord injured rats. These results suggested that propriospinal neurons could serve as an anatomical substrate to restore respiratory function following injury.

Functional studies have also provided indirect evidence that spinal neurons play a role in respiratory plasticity. Blocking inhibitory neurotransmission with a GABA receptor antagonist applied directly to the cervical spinal cord 1 week after a C2 hemisection activates the CPP (Ballanyi et al. 1999;

Zimmer et al. 2007). These results indicated that GABAergic neurons (most likely spinal interneurons or bulbospinal neurons spared by injury) normally provide tonic inhibition to phrenic motor neurons and/or crossed phrenic pathways. Multi-unit recordings of spinal neurons have also identified pre-phrenic interneurons in the rat cervical cord (Sandhu et al. 2015; Streeter et al., 2017). Functional connections between interneurons were examined more closely in the rat spinal cord following a C2Hx spinal cord injury. Interestingly, the functional connectivity of excitatory contralateral-to-ipsilateral interneurons was much greater than observed from the opposite direction, suggesting that interneurons may “bridge” the disconnect between the brainstem drive and respiratory motor neurons following spinal cord injury.

Moreover, increasing brainstem respiratory drive with hypoxia abolished this directionality, suggesting that interneurons act as substrates for hypoxia induced neuroplasticity (Streeter et al. 2019). Additional studies are necessary to investigate how functional connectivity within phrenic circuitry might be altered following spinal cord injury.

Over time, spinal neurons exhibit anatomical and functional changes that suggest that they are important for recovery of function following injury. For example, commissural interneurons within the cervical spinal cord show substantial axonal regeneration across the midline of the spinal cord within 56-

72 days following a mid-saggital axotomy without any therapeutic intervention. Moreover, these commissural interneurons form functional synapses with phrenic motor neurons, as confirmed by

34 electrophysiology (Fenrich and Rose 2009). Ipsilateral neurons also form new connections following injury, as evidenced by the increased connectivity between spinal V2a neurons and phrenic motor neurons in rats two weeks following a C2 hemisection injury (Zhouludeva et al. 2017). Spinal plasticity is not unique to respiratory interneurons, as the emergence of novel intraspinal circuits to enhance motor behavior has been reported across multiple levels of the spinal cord (Bareyre et al. 2004; Courtine et al. 2008). Thus, a better understanding of the mechanisms leading to recovery of respiratory function following injury may lead to improvements in locomotor, fine motor control, or autonomic functions as well.

Optogenetic and chemogenetic techniques have allowed investigators to target propriospinal neurons and directly demonstrate their ability to drive respiratory muscle activity following spinal cord injury. Optogenetic activation of all cervical neurons (phrenic motor neurons and interneurons) at the level of the phrenic nucleus was able to restore activity to the ipsilateral diaphragm 4 days following a C2 hemisection (Alilain et al. 2008). Interestingly, a rhythmic waxing and waning of rhythmic bursting activity and tonic activity was observed even after cessation of photostimulation, hypothesized to result from spinal interneuron activity. Later, Cregg et al. 2017 showed that glutamatergic (VGlut2+) interneurons within the spinal cord are sufficient to drive phrenic motor neuron activity from a neonatal ex vivo spinal cord preparation following a complete bilateral C1 transection. However, persistent rhythmic and synchronous bilateral phrenic motor neuron bursting was only elicited after silencing inhibitory interneurons, suggesting that both excitatory and inhibitory spinal interneurons are important modulators of respiratory motor neuron activity (Cregg et al., 2017). This was further confirmed when disinhibiting the spinal cord alone following a C2 hemisection restored rhythmic bursting activity to the ipsilateral diaphragm. Since a recurrent excitatory spinal circuit was observed following a complete C1 transection (and therefore completely independent of descending medullary input), it is likely that propriospinal neurons may promote respiratory plasticity via mechanisms other than just activation of crossed phrenic pathways.

Spinal neurons are critical mediators of respiratory plasticity that enable some animals to maintain ventilation after spinal cord injury. For example, Satkunendrarajah et al. 2018 used a chemogenetic strategy to inhibit cervical excitatory neurons and showed that cervical excitatory (vGluT2+) neurons sustain breathing despite phrenic motor neuron loss and are required for adequate ventilation in a non-traumatic spinal cord injury mouse model of cervical myopathy. Importantly, modulating the activity of excitatory spinal interneurons does not appear to produce adverse effects on breathing in healthy mice at rest, suggesting that this strategy could be safer than approaches that alter brainstem function (Romer et al. 2017; Satkunendrarajah et al. 2018; Jensen et al. 2019). In addition, increasing the excitability of these interneurons restored bursting activity to the ipsilateral diaphragm within hours of a

C2 hemisection. Thus, modulating propriospinal neuron activity is a promising strategy to restore breathing at both chronic and acute stages of injury, but additional experiments are needed to determine which specific subsets of excitatory interneurons are sufficient or required for the CPP. Section F4 and

Section F5 highlight why V2a neurons in particular may serve as important mediators of the CPP to restore diaphragm function following spinal cord injury.

E2.3 Compensatory Plasticity Following Spinal Cord Injury: Accessory respiratory muscle activation

While diaphragm function is absolutely critical for maintaining normal ventilation following spinal cord injury, ARMs assist the diaphragm during sighs, coughs, and under conditions of high oxygen demand such as exercise (Gransee and Sieck 2012). Other inspiratory pump muscle activity (i.e. external intercostals) and expiratory accessory respiratory muscles (i.e. abdominals) are critical to preserve quality of life. Fortunately, increased inspiratory and expiratory ARM activity is observed both in spinal cord injury patients and animal models to enhance ventilation (Tamplin et al. 2011; Dougherty et al. 2012; Zimmer et al. 2015). It has been shown that the external intercostals are impaired following a C2 hemisection in rats.

However, EMG recordings showed that recovery of the most rostral external intercostals occurred

36 spontaneously as soon as one week following injury and all 10 intercostals had regained function by 16 weeks post injury. More importantly, this recovery of intercostal function was associated with improved tidal volume (Dougherty et al. 2012). Other inspiratory ARMs, such as the trapezius and sternocleidomastoid, also become activated in spinal cord injury patients with an impaired forced vital capacity for soft and loud volume vocalizations (Tamplin et al. 2011). In fact, lingering intercostal and scalene muscle activity is associated with higher pulmonary function and decreased rib cage instability

(Guttmann and Silver 1965; De Troyer and Heilporn 1980). Thus, the respiratory system is able to compensate for impaired inspiration following spinal cord injury via activation of inspiratory ARMs.

It is critical for expiratory ARMs to also become active following spinal cord injury, since expiratory muscles are required to produce a forceful cough to clear the airways of infection. Unfortunately, few studies report compensatory activation of expiratory ARMs such as the abdominals and internal intercostals. However, non-traditional accessory respiratory muscles can become active. A study from

1999 showed that some tetraplegics can cough even though they cannot contract their abdominal muscles (Fujiwara et al. 1999). Interestingly, the latissimus dorsi and pectoralis major muscles showed increased compensatory activation during coughing that increased peak expiratory flow rate and maximal expiratory mouth pressure. Therefore, the impairment and neuroplastic potential for the activation of

ARMs following spinal cord injury necessitates the understanding of circuits and specific neural substrates patterning all types of respiratory muscle activity. Understanding how these muscles are activated could lead to therapeutic targets to improve ventilation.

E3. Treatment for high level spinal cord injury

Treatment of respiratory deficits following high level spinal cord injury can be broadly classified as invasive and noninvasive. Invasive treatment options include diaphragm or phrenic nerve pacing and mechanical ventilation via tracheostomy. Both of these are often required when respiratory deficits are

37 most severe. If the phrenic nerve is intact, patients can opt to undergo implantation of a phrenic nerve pacer (or direct diaphragm pacer). This treatment allows greater mobility, improved speech, and more independence, but it does include the inherent risk of undergoing a major surgery with the risk of surgical damage to the phrenic nerve (Posluszny et al. 2014; Dalal and DiMarco 2014). Alternatively, patients may be mechanically ventilated with a tracheostomy, which can lead to ventilator induced diaphragm dysfunction (Petrof and Hussain 2016). Some patients are able to be gradually weaned off of invasive mechanical ventilation, a process that requires transitioning to noninvasive mechanical ventilation.

Noninvasive respiratory management practices following high level spinal cord injury include noninvasive ventilation (using intermittent positive pressure breathing (IPPB), breath stacking, bilevel positive airway pressure (BiPAP), and continuous positive airway pressure (CPAP)), glossopharyngeal breathing, assisted cough, and respiratory training. IPPB uses a mouthpiece or facemask to support inspiration by augmenting lung volume whereas breath stacking aids in secretion clearance by delivering two or more breaths prior to exhalation to increase lung volume. Assisted cough is equally necessary for airway clearance and requires a practiced care provider to generate force below the diaphragm to increase expiratory pressure. Respiratory training on the other hand significantly improves respiratory muscle strength and function in tetraplegics by improving maximal expiratory pressure, maximum voluntary ventilation, and inspiratory capacity. Although respiratory training shows promise, the benefits are dependent on continual practice and will wean with disuse. For an extensive review of respiratory management in spinal cord injury patients, see Berlowitz et al. 2016.

Research has focused on multiple strategies for promoting recovery of breathing following spinal cord injury, which include (but are not limited to) spinal stimulation (Hachmann et al. 2017), nerve/axonal regeneration (Lee and Zheng 2008; Urban et al. 2019), cell transplantation into the spinal cord

(Zhouludeva et al. 2018), increasing brainstem respiratory drive to increase drive to respiratory muscles

(Vinit et al. 2009), and identifying novel neural substrates to increase respiratory motor neuron activity

(Zhouludeva et al. 2018; Satkunendrarjah et al. 2018). While advances have been made in both axonal regeneration and cell transplantation, work still needs to be done to ensure anatomical connectivity translates into functionally improving respiratory circuits. Hypoxia has also been considered clinically

(Hayes et al. 2014), but cognitive side effects of chronic intermittent hypoxia need to be carefully considered for the patient’s safety (Navarrete-Opazo et al. 2016). Finally, recent treatment options have used spinal stimulation and training to help patients regain locomotor function and may be applied to help improve breathing (Hachmann et al. 2017). Therefore, although treatment options are advancing, there is still a clear need to restore breathing following spinal cord injury. Section F focuses on a neural candidate that may promote recovery of respiratory muscle activity following neuromuscular disease and spinal cord injury: V2a neurons.

F. V2a Neurons Control Locomotion and Breathing

F1. Distinct classes of interneuron exist with distinct properties

Developmental studies investigating morphogenic signaling during embryogenesis identified 10 progenitor domains that give rise to 6 dorsal (dI1-6) and 4 ventral (V0-V3) classes of spinal interneurons

(Goulding 2009; Alaynick et al. 2011; Kiehn 2016; Gosgnach et al. 2017). These developmental classes have different properties, including specific molecular markers (i.e. transcription factors), neurotransmitter identity, and projection pattern and perform different roles in locomotion and respiration (Lu et al. 2015; Ziskind-Conhaim and Hochman 2017). Although recent research has made advances in determining the role of interneurons in locomotion and other behaviors (i.e. breathing, goal directed movements, postural control, autonomic regulation), elucidating specific roles for each neuronal subtype is an ongoing area of research. Here, we focus on the role of ipsilaterally projecting, glutamatergic

(excitatory) V2a neurons in regulating locomotion and breathing.

F2. Properties of V2a Interneurons

The p2 progenitor domain in the ventral spinal cord gives rise to two ipsilaterally projecting interneuron subtypes: V2a and V2b. V2b neurons are inhibitory interneurons that express Gata2 and

Gata3 (Karunaratne et al. 2002; Li et al. 2005). In contrast, V2a neurons are glutamatergic (excitatory) and marked by their expression of the transcription factor Chx10 (as well as Lhx3, Sox14, and Shox2) (Ericson et al. 1997; Dougherty et al. 2013; Clovis et al. 2016; Butts et al. 2017). Their distribution in the nervous system is diverse. They are located in the eye, the brainstem, and the spinal cord and have been studied extensively in multiple species, including zebrafish (Kimura et al. 2006; Barabino et al. 1997), mice (Crone et al. 2008, 2009, 2012; Romer et al. 2017; Jensen et al. 2019), and rats (Zhouludeva et al. 2017, 2018).

Because V2a neurons are located at multiple levels of the central nervous system, their projection patterns and firing properties are diverse. Briefly, V2a neurons can have bulbospinal projections (from the brainstem to the spinal cord (Bouvier et al. 2015), spinobulbar projections (from the spinal cord to the brainstem) (Azim et al. 2014; Hayashi et al. 2018), direct local connections to motor neurons (Dougherty et al. 2013), and local connections to neighboring inhibitory or commissural interneurons (interneurons that cross the midline of the spinal cord) (Crone et al. 2008; Dougherty et al. 2013). Some V2a neurons with similar firing patterns (e.g. rhythmic vs tonic) are even electrically coupled to coordinate V2a neuron output (Zhong et al. 2010; Prendergast and Wyart 2016; Menelaou and McLean 2019). Not surprisingly, several subsets of V2a neurons have been discovered in zebrafish (McLean et al. 2008; Ampatzis et al.

2014; Song et al. 2018) and in mice (Dougherty et al. 2013; Hayashi et al. 2018). In fact, Hayashi et al. 2018 used single cell RNA sequencing to identify at least 11 different subsets of V2a neurons in the lumbar and cervical spinal cord of neonatal mice. Thus, the class of V2a interneurons is diverse and likely have diverse functions in controlling locomotor and respiratory motor functions.

Numerous studies have investigated the role of V2a interneurons in generating the frequency and pattern of locomotion, but less is known about their role in respiratory circuits. Understanding how V2a neurons control rhythm and pattern generation in locomotor spinal circuits may inform researchers how they are organized in respiratory spinal circuits. Therefore, the next two sections provide an overview of how V2a neurons control locomotion and respiration.

F3. V2a neurons coordinate locomotion

Precise control of locomotion is dependent on the complex interplay between descending commands from the brain and interneuron/motor neuron activity in the spinal cord. Current models of the mammalian locomotor CPG are often based on a two-level functional organization containing a rhythm generator level and a pattern forming level (McCrea and Rybak 2008; Rybak et al. 2015). The mammalian locomotor rhythm generator is distributed across multiple spinal segments as multiple rhythmogenic flexor and extensor modules that can produce alternation of flexors and extensors and alternation of left and right limbs even in the absence of supraspinal or sensory afferent input (Zhong et al. 2012; Hägglund et al. 2013; Rybak et al., 2015; Kiehn 2016). Here, we briefly review how V2a neurons in both the brainstem and spinal cord play important roles in controlling locomotion.

Studies in zebrafish first demonstrated that reticulospinal V2a neurons (that project from the brainstem to the spinal cord) pattern locomotion. In zebrafish, hindbrain V2a neurons project to motor neurons and provide cycle-by-cycle excitation during swimming (Eklöf-Ljunggren et al. 2012; Kimura et al.

2006; McLean et al. 2008). Optogenetics were later used to discover that activating hindbrain V2a neurons in larval zebrafish induced swimming whereas silencing hindbrain V2a neurons halted swimming

(Kimura et al. 2013). Electrophysiology of these V2a neurons further showed that they provide both tonic and rhythmic input to locomotor circuits, suggesting excitation from V2a neurons to spinal motor neurons

41 is critical for controlling swimming in zebrafish. Importantly, these glutamatergic reticulospinal pathways also exist in mice.

Reticulospinal V2a neurons are also critical for patterning locomotion in mice. Lhx3+/Chx10+ V2a neurons are found in the medial reticular formation of the medulla in the mouse, receive rhythmic input from the mesencephalic locomotor region, and project to the spinal cord (Bretzner and Brownstone 2013).

A neonatal brainstem-spinal cord preparation showed that brainstem or spinal V2a neurons are required for locomotion, since ablating them prevented the initiation of fictive locomotion following stimulation of the caudal medulla (Crone et al. 2008). Moreover, these neurons showed increased expression of c-fos during walking, suggesting they are active during locomotion (Bretzner and Brownstone 2013). Capelli et al. 2017 later showed that optogenetic activation of VGlut2+ neurons in the lateral paragigantocellular nucleus (which include V2a neurons) in vivo initiated full-body locomotion in mice. Thus, V2a neurons appear to be key players in initiating locomotion. However, different subsets of reticulospinal V2a neurons may perform different locomotor functions, since activating reticulospinal V2a neurons in the rostral gigantocellular nucleus in the medulla of mice actually halted locomotion via descending projections to lumbar spinal cord inhibitory networks (Bouvier et al. 2015). Together, these studies demonstrate that reticulospinal V2a neurons modulate locomotor activity, but different subsets of V2a neurons may produce distinct effects on motor activity.

V2a neurons are also located in the spinal cord, pattern locomotor CPGs, and may be functionally divided into subtypes. Zebrafish V2a neurons comprise an important source of motor neuron excitation in the zebrafish spinal cord, and these motor neurons become gradually recruited as the swimming speed increases from slow to fast (Ampatzis et al. 2013). Interestingly, distinct V2a subtypes have been observed in zebrafish to control the speed of swimming. V2a neurons can be divided into three functionally distinct classes that selectively synapse onto slow, intermediate, or fast motor neurons (McLean et al. 2008;

Ampatzis et al. 2014). Moreover, these distinct subclasses of V2a neurons have distinct firing properties.

V2a neurons that target slow motor neurons fire in bursts and provide strong non-linear excitation whereas V2a neurons that target fast motor neurons provide weaker excitation (Song et al. 2018). This type of organization is postulated to ensure that slow motor neurons are recruited before fast motor neurons during swimming.

The role of spinal V2a neurons in controlling locomotion has also been studied in mice. Crone et al. 2008 demonstrated that ipsilaterally projecting spinal V2a neurons coordinate the left-right alternation of locomotion via projections to commissural interneurons that cross the midline of the spinal cord.

Moreover, ablating these V2a neurons actually resulted in the loss of left-right alternation during high speed running and caused mice to display a galloping gait instead (Crone et al. 2009). Interestingly, at least two subtypes of V2a neurons exist in the lumbar spinal cord of mice based on their transcription factor expression: Chx10+/Shox2+ (Type I) and Chx10+/Shox2- (Type II) (Dougherty 2013; Rybak et al. 2015).

Both types of V2a neurons receive input from the rhythm generating kernel in the lumbar spinal cord.

However, Type II V2a neurons project to commissural interneurons to help control left-right alternation during locomotion whereas Type I V2a neurons project directly to hindlimb motor neurons to control locomotor motor neuron output. Some (but not all) V2a neurons coordinate their firing with rhythmic firing of locomotor motor neurons (Dougherty and Kiehn 2010). Intrinsic membrane properties of individual lumbar V2a neurons are heterogeneous (including phasic, tonic, and delayed onset intrinsic firing properties), but their firing properties appear to be dependent on the pattern of synaptic input they receive (Dougherty et al. 2013). Thus, these studies investigating the role of V2a neurons in locomotion have provided invaluable information on intrinsic firing properties, synaptic connections, and their roles in modulating limb motor neuron output.

F4. V2a neurons modulate breathing

V2a neurons have been implicated in controlling both respiratory rhythm and pattern generation.

Respiratory rhythm refers to the regularity and rate of breathing (e.g. breaths/minute) whereas pattern

43 generation refers to the coordination and timing of respiratory muscle activation during each cycle. V2a neurons in the brainstem and spinal cord have roles in controlling both aspects of respiration.

Brainstem V2a neurons regulate respiratory rhythm in neonatal mice, and ablation of V2a neurons causes severe respiratory deficits. Crone et al. 2012 used the Chx10:DTA mouse model to selectively ablate >98% of Chx10+ V2a neurons in the brainstem and spinal cord on a Bl6 background as well as a mixed Bl6/ICR background. All mice showed a drastic reduction in respiratory rate and significant increase in irregularity (including apneas) at P0. Ablating V2a neurons is fatal for 100% of Chx10:DTA mice on a Bl6 background at P0. On the other hand, it is fatal for only 60% of Chx10:DTA mice on a Bl6/ICR mixed background at P0. Brainstem V2a neurons located in the medial reticular formation of the medulla project to and provide tonic excitatory input to the PreBÖtzinger Complex (the respiratory rhythm generator).

Ablation of these neurons removes excitatory input into the central respiratory rhythm generator, resulting in slow and irregular breathing (rhythm generation) without affecting phrenic nerve bursting amplitude (pattern generation). Interestingly, the Chx10:TA mice on the mixed Bl6/ICR background that survived ablation of V2a neurons actually recovered regularity of respiration 4 days after birth (without recovering normal frequency). However, it is unknown whether this recovery occurs because 1) V2a neurons are not required for maintaining regularity of breathing in adult mice or 2) compensatory mechanisms make up for the lack of V2a neuron development during embryogenesis. This will be addressed in Chapter II, Section 2.3.

In addition to regulating respiratory rhythm and frequency, V2a neurons also pattern inspiratory

ARM activation in adult mice. Our lab previously injected a PRV tracer into the scalene muscle, which was retrogradely transported back to the scalene motor neuron pool in the spinal cord and to any premotor neurons. Using this technique, we demonstrated that V2a neurons are mono- or polysynaptically connected to ARM motor neuron pools in the cervical cord (unpublished observations). The Excitatory

Designer Receptor Exclusively Activated by Designer Drugs (DREADD) receptor was then expressed in

+ Chx10 V2a neurons in the brainstem and spinal cord of V2a-(Gq)DREADD mice. These (Gq)-DREADD receptors are G-protein coupled receptors that increase neuronal excitability by activating the intracellular Gq signaling pathway to release intracellular calcium from the endoplasmic reticulum and depolarize the neuron following treatment with the synthetic ligand Clozapine-N-Oxide (CNO) (Roth

2016). These experiments showed that V2a neurons are sufficient to pattern respiratory muscle activity in adult mice since increasing the excitability of V2a neurons increased scalene and trapezius ARM activity

(Romer et al. 2017). However, several important questions remain unanswered. First, it is unknown whether V2a neurons are required for ARM activity and enhanced ventilation. Second, V2a neurons are synaptically connected to phrenic motor neurons (Zhouludeva et al. 2017), but the impact of activating or silencing V2a neurons on diaphragm function has not been tested. Experiments to address each of these questions are described in Chapter II, Section 2.3. Finally, DREADDs are expressed in all brainstem and spinal cord V2a neurons in our transgenic mouse lines, so it is not possible to determine which population of V2a neurons – brainstem or spinal – is sufficient to alter respiratory muscle activity and ventilation.

Experiments targeting DREADD receptors to cervical spinal V2a neurons only are described in Chapter III,

Section 3.3.

F5. V2a neurons may promote recovery of breathing following disease or injury

Here, we focus on the potential for V2a neurons to improve respiratory muscle activity and ventilation in two conditions where respiratory failure is the leading cause of death: ALS and high-level spinal cord injury. As described in Section 4.1, ALS is a neurodegenerative disease that affects upper and lower motor neurons. Dysfunction of cortical and spinal interneurons in ALS mouse models has also been postulated to contribute to motor neuron degeneration (Turner and Kiernan 2012; Zhang et al. 2016; Romer et al.

2017). Specifically, Romer et al. 2017 showed that excitatory V2a interneurons in the brainstem and spinal cord are sufficient to activate ARMs and enhance ventilation in healthy mice. Moreover, these neurons degenerate at the same rate as limb and phrenic motor neurons in the SOD1(G93A) ALS model mouse.

Therefore, we investigated whether altering brainstem and spinal cord V2a neuron excitability can activate respiratory circuits to compensate for neurodegeneration at late stages of ALS in the SOD1(G93A)

ALS mouse model. These results are described in Chapter IV, Section 4.3.

High level spinal cord injury also severely impairs the ability to breathe. Although many mechanisms underlying recovery of diaphragm function have been uncovered, the neural substrates themselves (outside of phrenic motor neurons) still need to be identified. Zhouludeva et al. 2017 showed that V2a neurons are synaptically connected to phrenic motor neurons and display increased connectivity to phrenic motor neurons following a C2Hx spinal cord injury. An important study by Butts et al. 2017 showed that transplantation of human iPSC-derived V2a neurons into the spinal cord of a healthy mouse could survive. Follow-up experiments by Zholoudeva et al. 2018 then transplanted iPSC-derived V2a neurons and neural progenitor cells (NPCs) into the injured cervical spinal cord of rats. Not only did these transplanted V2a neurons survive in the injured cervical cord, but diaphragm EMG recordings one month following injury showed that rats who had received NPCs and iPSC derived V2a neurons showed a greater recovery than rats who received NPCs alone (Zhouludeva et al. 2018). Together, these results suggest that

V2a neurons may contribute to respiratory plasticity following a high-level spinal cord injury. Therefore, we tested the functional impact of increasing V2a excitability on diaphragm activity following a C2Hx.

These results are described in Chapter V, Section 5.3.

CHAPTER II

Brainstem and Spinal Cord V2a Neurons

Regulate Respiratory Muscle Activity and

Ventilation in Healthy Mice

2.1 Introduction A major challenge in the study of respiratory control is to understand how different neurons modulate respiratory motor output to maintain ventilation during different behaviors, stages of development, or following disease and injury. For example, healthy mammals at rest rely predominantly on the diaphragm for inspiration. When ventilatory demands increase (i.e. during exercise), there is an increase in the activity of extradiaphragmatic respiratory muscles that include inspiratory pump muscles

(e.g. external intercostal and scalenes) as well as accessory respiratory muscles (e.g. trapezius, pectoralis, and sternocleidomastoid (Sieck & Gransee, 2012) that we will refer to collectively as auxiliary respiratory muscles (ARMs). ARMs are also recruited in patients with neuromuscular disease, spinal cord injury, or diaphragm paralysis to compensate for impaired diaphragm function (Bennett et al., 2004; Sieck &

Gransee, 2012; Johson and Mitchell, 2013).However, it is not known why some patients with diaphragm impairment use ARMs more than others, as the neural mechanisms leading to recruitment of ARMs during exercise or following disease and injury are poorly understood. A better understanding of circuits patterning respiratory muscle activity would facilitate the development of therapies to improve ventilation, reduce respiratory infections, and prevent ventilation failure in patients with neuromuscular disease and injury.

The present study investigates the roles of V2a neurons, a class of glutamatergic neurons in the brainstem and spinal cord marked by Chx10 expression, in the control of respiratory rhythm and pattern generation in adult mice. A prior study showed that ablating V2a neurons during development results in slow and irregular breathing in neonates, likely due to a loss of excitatory drive from V2a neurons in the medial reticular formation to neurons in the PreBötzinger complex (Crone et al. 2012). However, these experiments could not conclusively distinguish whether the changes in breathing were the direct result of loss of V2a neuron function versus developmental defects resulting from the loss of V2a neurons in the embryo. In addition, a subset of these mice were able to survive past the neonatal period and their

48 breathing pattern appears regular after the first week, indicating that V2a neurons might be dispensable for normal breathing in mature mice. Alternatively, ablation of V2a neurons in the embryo may lead to developmental compensation by other cell types to maintain ventilation in mature animals. Thus, it is not yet clear whether V2a neuron function is critical for a normal respiratory rhythm in mature animals.

In addition to modulating respiratory rhythm and frequency, previous studies suggest that V2a neurons may be important for patterning respiratory muscle activity. Cervical spinal V2a neurons are synaptically connected to phrenic motor neurons and thus could modulate diaphragm function

(Zholudeva et al. 2017). The number of V2a neurons connected to phrenic motor neurons increases following spinal cord injury, indicating that V2a neurons may be particularly important for respiratory compensation following injury. In addition, it has been suggested that degeneration of V2a neurons in a mouse model of ALS could account for the failure of accessory respiratory muscles to be activated for breathing at late stages of disease (Romer et al. 2017). This hypothesis was based on the observation that increasing the excitability of V2a neurons using an excitatory DREADD (Designer Receptors Exclusively

Activated by Designer Drugs) in healthy adult mice activates scalene and trapezius respiratory muscles at rest, when these muscles are mostly inactive (Romer et al. 2017). Thus, V2a neurons have the potential to influence the number and type of respiratory motor neurons recruited for breathing. However, it is not yet known whether V2a neural activity is required for patterning respiratory motor activity in adult mice.

Here, we target expression of an inhibitory DREADD (hM4Di or “(Gi)DREADD”) (D. J. Urban and

Roth 2015) to Chx10 expressing neurons to test whether normal activity of V2a neurons is required for a regular breathing rhythm and frequency in healthy neonatal and adult mice and to evaluate their role in patterning activity of respiratory muscles in adult mice. We find that the activity of V2a neurons is important for modulating the frequency and regularity of respiratory rhythm in neonatal mice, but only modulates the frequency of respiration in adult mice. In addition, we show that silencing V2a neurons

49 results in activation of ARMs in adult mice at rest, suggesting that these neurons play an important role in patterning extradiaphragmatic respiratory muscle activity.

2.2 Materials and Methods

Animals

All animal procedures were performed according the National Institutes of Health guidelines and approved by the institution’s animal care committee’s regulations. The Chx10Cre/+ mouse is a knock-in mouse model that allows targeted expression of Cre recombinase in the locus where Chx10 is located in the mouse (Azim et al. 2014; Romer et al. 2017). Therefore, breeding Chx10Cre/+ mice (Azim et al. 2014;

Romer et al. 2017) to ROSAPNP-(Gi)DREADD/+ (B6N.129-Gt(ROSA)26Sortm1(CAG-CHRM4*, -mCitrine)Ute/JJax, stock #026219 Jackson Laboratory) mice generated Chx10Cre/+; ROSAPNP-(Gi)DREADD/+ mice (hereafter referred

+ to as V2a-(Gi)DREADD mice) in which only Chx10 V2a neurons express the inhibitory (Gi)DREADD receptor. Adult male Chx10Cre/+ mice (lacking the ROSAPNP-(Gi)DREADD/+ allele) were used as non-DREADD expressing controls. On rare occasions, we observe recombination outside of the nervous system as a result of “leaky” germline or early embryo expression of Cre recombinase. Therefore, we perform PCR on

DNA isolated from the tail of all animals to identify any with the recombined ROSAP-(Gi)DREADD/+ allele and these animals are excluded from all experiments and breeding (primers: 5'-CTCTG CTAAC CATGT TCATG

C-3' and 5'-GAAGG CGCCT ATGAT GAGAT C-3'). Genotyping was performed by PCR using primers to detect the Chx10Cre (5’-GCATT AGACA CCGGA GGG-3’ and 5’-GGACA GAAGC ATTTT CCAG-3’), Chx10 wild type

(5’-GCATT AGACA CCGGA GGG-3’ and 5’-CTCCC GACTG TGACT TTCC-3’), ROSAPNP-(Gi)DREADD (5'-ATGTC

TGGAT CCCCA TCAAG-3' and 5'-GAAGG CGCCT ATGAT GAGAT C-3' ) and ROSA wild type (5'-AAGGG AGCTG

CAGTG GAGTA-3' and 5'-GAAAA TCTGT GGGAA GTC-3') alleles.

Histology, Imaging and Quantification

Immunohistochemical and imaging analysis was performed on spinal cords and medulla of neonatal and adult V2a-(Gi)DREADD mice. Specificity of the HA antibody was confirmed by lack of staining in

Cre/+ PNP-CHRM4/+ Chx10 and ROSA mice that lack expression of the HA-tagged (Gi)DREADD. Neonatal mice were placed on ice for four minutes and transcardially perfused with 4% paraformaldehyde (PFA) in phosphate

buffer (PB: 770mM Na2HPO4٠7H20, 231mM NaH2PO4٠H20, pH 7.4). Brains and spinal cords were dissected and submerged in 4% PFA until 2 hours after the start time of perfusion. Tissue was then rinsed in phosphate buffered saline solution (PBS: 137mM NaCl, 2.7mM KCl, 10mM Na2HPO4, 2mM KH2PO4, pH 7.2) overnight to remove PFA, cryoprotected in 30% sucrose overnight (all at 4°C), mounted in OCT medium and stored at -80º C. Transverse sections of cervical spinal cord and medulla were cut (14 m) on a cryostat microtome and collected directly onto slides. Immunohistochemistry was performed as described (Thaler et al. 1999; Crone et al. 2009)) using the following primary antibodies (and DAPI nuclear stain): Chx10 (Guinea pig @ 1:10,000), and HA-Tag C29F4 (Rabbit, Cell Signaling #3724, @1:1000). Images were obtained on a Nikon A1 Plus inverted (Melville, NY) confocal microscope. The number of HA-Tag+,

Chx10+, and HA-Tag+/Chx10+ cells with DAPI+ nuclei were counted from single plane confocal 20x images using NIS Elements Software. Every 4th section at C4 (16-18 hemisections/animal) was counted for cervical sections and every 4th section at the level of the nucleus ambiguous for medulla sections (18-20 hemisections/animal). The average number of cells per hemisection was calculated for each animal. C4 was used as a standard reference point based on the distinct morphology of the phrenic nucleus in this segment to ensure counts were made at a consistent axial level. Similarly, the distinct morphology of the nucleus ambiguous was used as a reference in the medulla.

Adult V2a-(Gi)DREADD mice were anesthetized with pentobarbital (0.1 mg/g, IP) and were transcardially perfused with PB followed by 4% PFA in PB. Tissue was rinsed in PBS overnight, cryoprotected in 30% sucrose overnight (all at 4°C), mounted in OCT medium and stored at -80º C.

Transverse sections of adult cervical spinal cords were cut (14 m) on a cryostat microtome and collected directly onto slides. Transverse sections of adult medulla were cut (50 m) on a cryostat microtome and collected as floating sections in PBS + 30% v/v ethylene glycol + 25% w/v glycerol. Immunohistochemistry was performed on adult tissue as described previously (Romer et al. 2017) using the following primary antibodies (and DAPI nuclear stain): Chx10 (Sheep, abcam #ab16141, @1:2000), and HA-Tag C29F4

(Rabbit, Cell Signaling #3724, @1:1000). The number of HA-Tag+ cells, Chx10+ cells, and HA-Tag+/Chx10+ cells with DAPI+ nuclei were counted as described above in every 8th section at C4 (9-14 hemisections/animal) and in every 3rd section at the level of the nucleus ambiguous (12-13 hemisections/animal). The average number of cells per hemisection was calculated for each animal.

Surgical Implantation of Telemetry Devices. Telemetry devices (either single (TAE-F10) or double channel

(F20-EET) transmitters from Data Sciences International) were implanted as early as 8 weeks of age to chronically record EMG from respiratory muscles of V2a-(Gi)DREADD and non-DREADD control mice.

Surgeries to implant devices subcutaneously for recording EMG from the trapezius and scalene muscles were performed as described (Jensen et al. 2017; Romer et al. 2017) Briefly, mice were anesthetized using isoflurane anesthesia (induction 4-5%; maintenance 1-2%, both in 100% O2). A 3cm long incision was made between the right shoulder and ear. Forceps (#2) were used to separate the tissue covering the brachial plexus and phrenic nerve, landmarks used to identify the trapezius and middle scalene. The transmitter was the placed subcutaneously on the back of the mouse. Two sets of biopotential leads were inserted approximately 1mm apart from one another into the trapezius and scalene muscles using a 25G needle.

A cyanoacrylate adhesive was used to secure the implanted leads. Excess lead length was coiled and tucked near the transmitter before closing the incision with cyanoacrylate adhesive. Mice were allowed to recover for one week prior to recording.

Surgeries to implant devices intraperitoneally (IP) to record EMG from diaphragm and/or trapezius muscles were performed as follows. Mice were placed supine under isoflurane anesthesia. A “T” incision was made in the skin by making a 4cm long lateral incision directly below the xyphoid process and an adjoining 2cm long vertical incision was made from the middle of the lateral incision to the middle of the rib cage. Next, a 3cm long lateral incision was made through the abdomen muscle just below the rib cage. The xyphoid process was clamped with Hemostats to expose the diaphragm. The transmitter was placed on the left side of the abdominal cavity next to the intestines. One set of leads was set aside for later use (trapezius insertion) while the other set was positioned near the diaphragm. Forceps were used to gently lift up on the right side of the costal diaphragm. Forceps were used to insert the exposed wire of one biopotential lead through the first muscle layer of the diaphragm muscle without puncturing the second muscle layer. Extreme care was taken to ensure that the diaphragm was not punctured to prevent a fatal injury. The exposed wire was secured in place against the diaphragm muscle with a small drop of cyanoacrylate adhesive. A second biopotential lead was inserted in the same manner ~1mm caudal to the position of the first lead. Excess lead length was coiled and placed just below the abdominal muscle and above the implanted transmitter. The abdominal muscle was closed over the transmitter and excess coiled leads with suture, with care taken to externalize the second pair of biopotential leads for trapezius insertion. The skin incision was closed using inside out sutures, leaving approximately 1mm open closest to the mouse’ head where the second set of leads remained exposed. A second incision was made between the mouse’ right shoulder and ear to expose the trapezius muscle. A trocar was used to tunnel the second set of exposed leads subcutaneously from the IP implanted transmitter to the open incision near the mouse’ shoulder. The leads were inserted into the trapezius muscle as described above. Mice were allowed to recover for 1-2 weeks prior to EMG data acquisition.

Neonatal Whole Body Plethysmography. Whole-body plethysmography was performed on age P2 non-

DREADD expressing mice and V2a-(Gi)DREADD mice. The barometric chambers were standardized by

53 injecting 10 consecutive 10uL pumps of air into an open port on the experimental chamber using a 100uL

Hamilton syringe. Animals were placed in a barometric chamber and equilibrated with open air flow for 5 min. The pressure difference between the sealed experimental and reference chambers was measured with a differential pressure transducer for two 60s recording sessions: one baseline session and one session 20 minutes after an IP injection of 10uL of 10.0 mg/kg*bw CNO. Data are expressed in relative units (R.U.) normalized to the average maximum value obtained during standardization with 10uL pumps of air from a syringe. CNO solutions were made by dissolving CNO directly into saline. At least 10 minutes of open air flow was allowed between trials. Signals were amplified, digitized, and low pass-filtered (50Hz).

Data were collected and analyzed using LabChart 7.0 software (AD Instruments).

Neonatal Plethsymography Analysis. LabChart 7.0 software (AD Instruments) was used to mark individual resting breaths to be analyzed for breathing frequency. At least 70 resting breaths were analyzed before and after CNO administration. An irregularity score (IS= ABS[(Tn-T(n-1))/T(n-1)]*100) was calculated for each breath (Viemari et al. 2011; Sheikhbahaei et al. 2017). The average IS of at least 50 consecutive breaths for each animal under each condition (without and with CNO) is reported. Breaths during movement were excluded from analysis.

Adult Whole Body Plethysmography and EMG Acquisition in Conscious Animals. Plethysmography, EMG acquisition, and digital video recording was performed as described (Jensen et al. 2017; Romer et al. 2017) using DSI Ponemah Physiology Platform Acquisition software v.5.20. Briefly, mice were placed in a plethysmography chamber on top of a transmitter receiving pad and allowed to acclimate for 30 minutes.

The transmitter was turned on by placing a strong magnet near the animal inside the chamber prior to data collection. A vehicle control (0.1mL of saline) was injected IP. Mice were placed in their home cage for 2 minutes immediately following all injections before being placed into the plethysmography chamber to prevent association of a painful injection with the chamber. After two minutes of re-acclimating to the chamber, the transmitter was turned on and at least 20 minutes of resting data was collected

(characterized by mouse inactivity and regular breathing). Next, mice were injected IP with either 1.0,

5.0, 10.0 or 15.0 mg/kg*bw CNO, placed in their home cage for two minutes, and then re-acclimated to the chamber for two minutes before data acquisition began for one hour. The same animals were tested with multiple CNO doses, with at least two days between each dose to avoid desensitization of DREADD receptors. A vehicle control recording was performed prior to each dose of CNO. Only periods in which the animal is at rest (not moving in video recordings) are used for analysis.

Adult Plethysmography Analysis. The Ponemah Physiology Platform Analysis software v.5.20 was used to mark individual resting breaths to be analyzed for breathing frequency (f), tidal volume (VT), minute volume (MV), peak inspiratory flow (PIF), inspiratory time (Ti), expiratory time (Te), and total breathing cycle time (Tt) as described (Jensen et al. 2017; Romer et al. 2017). VT/Ti, a measure of respiratory drive, was calculated from these derived values. Periods of ARM activity (trapezius bouts) were identified as described above. Interbout intervals used for analysis excluded all bouts as well as the 4 breaths immediately preceeding or following a bout. At least 100 resting breaths were analyzed for each vehicle control and CNO treatment period. The average IS of 100 consecutive breaths (the same breaths analyzed for ventilation parameters) for each animal under each condition (vehicle or CNO) is reported.

EMG Analysis in Conscious Animals. The raw EMG signal was high pass filtered at 1.5Hz and the root mean square (RMS) was calculated using a 30ms window. The frequency of bouts of trapezius or scalene activity was counted as described (Jensen et al. 2017). Briefly, ARM EMG activity is scored as a “bout” if it satisfies the following three criteria: 1) the RMS EMG trace must exhibit at least 3 consecutive points that are at least 50% higher than the surrounding baseline (each point represents a 30 ms window) 2) EMG activity occurs during resting breaths (i.e. sighs and movement artifacts are excluded) and 3) EMG activity occurs when the mouse is still (i.e. no movement or changes in posture) based on the synchronized video recordings. ARM bout frequency was calculated by dividing the total number of bouts observed by the total time analyzed. Interbout intervals used for analysis excluded all bouts as well as the 4 breaths

55 immediately preceeding or following a bout. For analysis of diaphragm EMG, the RMS was used to measure the peak amplitude, duration, and instantaneous frequency of each breath (burst of activity).

Diaphragm EMG peak amplitude was normalized to the maximum ventilatory effort reached during sighs to reduce intra-animal variability (Mantilla et al. 2011). At least 50 resting breaths were analyzed for each vehicle control and CNO treatment period.

Diaphragm and ARM EMG Acquisition in Anesthetized Animals

Terminal bilateral diaphragm EMG recordings were performed on adult V2a-(Gq)DREADD and V2a-

(Gi)DREADD mice before and after CNO administration. Animals were anesthetized under isoflurane anesthesia and placed supine on a heating pad. A 4cm lateral incision through the skin and abdominal muscle just below the xyphoid process exposed the intraperitoneal cavity and diaphragm. Bi-polar platinum electrodes (Grass Technology, Middleton, WI, USA) connected to an amplifier (BMA-400 AC/DC

Bioamplifier, CWE Inc.) were inserted into the diaphragm and secured with cyanoacrylate adhesive to record diaphragmatic activity via Spike2 Data Analysis software (Cambridge Electronic Design Limited,

Cambridge, England). Electrodes were grounded with an additional lead inserted into the abdominal muscle. A subset of animals also had electrodes inserted into the scalene ARM. A 10-minute baseline of bilateral diaphragm activity was recorded. CNO was then topically applied to the exposed intraperitoneal cavity and diaphragm EMG was recorded for one hour.

EMG Analysis in Anesthetized Animals

The diaphragm EMG signal was digitally amplified (5000x gain) and band pass filtered (30 – 3000Hz) with a sampling frequency of 6.25kHz. EMG signals were further processed to remove DC noise and the root mean square was calculated over a 30ms window. Electrocardiogram (ECG) artifact was digitally filtered out using the ECGDelete02 Spike2 script (Cambridge Electronic Design). Diaphragm EMG peak amplitude was normalized to the maximum ventilatory effort reached during sighs to reduce intra-animal variability

(Mantilla et al. 2011). At least 50 bursts were analyzed for diaphragm EMG peak amplitude and regularity of bursting frequency for the baseline recording and following CNO treatment.

Blood Pressure Acquisition and Analysis

Blood pressure was noninvasively measured with a tail cuff using the CODA system by Kent Scientific before and after CNO treatment. Non-implanted mice were placed in a restrainer onto a warming pad and a blood pressure tail cuff was positioned around their exposed tail. Mice were positioned so their tail rested on a warming pad to prevent artificial drops in blood pressure due to temperature changes. Since movement artifact can also alter blood pressure, mice received three days of training to acclimate them to the restrainer and blood pressure cuff tail system. Each training day consisted of 3 acclimation cycles

(one set/cycle and 10 blood pressure measurements/set) in which diastolic, systolic and mean blood pressure were measured. On the fourth day, 3 acclimation cycles (with 10 measurements/1 acclimation cycle) were taken before CNO treatment and 20 minutes following CNO treatment. The average systolic, diastolic, and mean blood pressure was averaged for each of the 10 cycles. The average of these 10 cycles was then calculated to report an average blood pressure measurement/animal. Data are reported as the mean +/- SEM for all animals of each genotype. To avoid variations in baseline values due to circadian rhythm induced changes in blood pressure, all blood pressure recordings were taken during the middle of the light cycle at noon.

Heart Rate Analysis

Electrocardiogram (ECG) signal was present in the EMG signal. Heart rate was measured by counting the number of ECG waves during the same time period that plethysmography was analyzed before and after

CNO treatment in transgenic mice and Chx10Cre/+ mice injected. Data are reported as heart beats/minute.

Pulse Oximetry Data Acquisition and Analysis

Blood oxygen saturation was noninvasively measured in anesthetized mice using Mouse Ox S collar sensors (Mouse Ox Plus, Starr Life Sciences Corp.). Mice were anesthetized with 3.5% isoflurane/1% oxygen and maintained at 1% isoflurane/1% oxygen. As described above, electrodes were inserted into the diaphragm and scalene ARM muscle to record EMG. The S collar was placed around the mouse neck so that probes on the inside of the collar covered the carotid arteries. These probes calculated the ratio of oxygenated to deoxygenated blood (hereby referred to as arterial oxygen saturation) by measuring the amount of red or blue light emitted from the carotid arteries, respectively. As soon as the baseline recording began, the arterial oxygen saturation acquisition began to monitor arterial oxygen saturation every 0.5s throughout the entire recording. The average arterial oxygen saturation (over two minutes) was reported during the baseline recording, 50 minutes after CNO treatment when ARM bouts were present, and during nasal occlusion.

Statistical Analysis. Statistical tests from SigmaStat 13 (Systat Software, SPSS Inc., Chicago, IL) were used to analyze the data. All data passed the Shapiro-Wilk test for normality, except as noted. A one-way repeated measures ANOVA (Holm-Sidak all pairwise comparisons post-hoc analysis) was used to analyze differences within animals when comparing values from three or more conditions tested within the same animal (e.g. multiple doses of CNO on ARM recruitment). Paired t-tests were used to analyze differences in treatments within the same animal where only two conditions existed (e.g. vehicle versus CNO).

Cre/+ Student t-tests were used to analyze differences between groups (e.g. V2a-(Gi)DREADD vs. Chx10 mice). Non-parameteric Wilcoxin signed rank tests were performed on data sets that failed the test for normality (coefficient of variation of breathing frequency (CVf) and irregularity score (IS) between vehicle control and CNO treated conditions). Statistical significance was set at alpha level 0.05 and values in figures are reported as means ± SE.

2.3 Results

(Gi)DREADD is expressed in V2a neurons in the spinal cord and brainstem of V2a-(Gi)DREADD mice.

In order to test the role of V2a neurons in modulating respiratory rhythm and motor pattern, we

Cre/+ PNP-(Gi)DREADD/+ generated V2a-(Gi)DREADD mice (Chx10 ; ROSA ) to decrease the excitability of V2a neurons following acute injections of clozapine-N-oxide (CNO) (Urban and Roth 2015; Urban and Roth

2015). Chx10Cre/+ mice, in which Cre recombinase has been inserted into the endogenous Chx10 locus by homologous recombination, have previously been used to reliably drive reporter gene expression in V2a neurons that express Chx10 during development or in the adult (Azim et al., 2014; Bouvier et al. 2015;

Romer et al., 2017; Hayashi et al. 2018). Viruses expressing the (Gi)DREADD receptor have previously been used to inhibit V2a neuron activity in spinal neurons to assess their role in fine motor control (Ueno et al.

2018), but breathing was not assessed in this study. We used antibodies to the hemagglutinin tag (HA-

Tag) on the (Gi)DREADD receptor and antibodies to Chx10 to mark the nuclei of V2a neurons to assess the distribution of (Gi)DREADD in the cervical spinal cord (C4) and medulla of neonatal and adult V2a-

(Gi)DREADD mice (Figure 1). The pattern of (Gi)DREADD expression is similar in neonatal (postnatal day 2) and adult animals. HA-Tag+ cells have neuronal morphology and are located where V2a neurons are found: the intermediate laminae (but not dorsal horn) of the spinal cord and medial reticular formation of the medulla, consistent with prior studies using Chx10Cre/+ mice to selectively drive reporter expression in V2a neurons (Azim et al., 2014; Bouvier et al., 2015; Romer et al., 2017; Hayashi et al., 2018). We detected HA-

Tag+ immunoreactivity in an average of 24.1 ± 0.6 (n=5), 11.1 ± 0.4 (n=6), 100.6 ± 10.9 (n=3), and 112.9 ±

7.5 (n=3) cells/hemisection in neonatal spinal cord, adult spinal cord, neonatal medulla, and adult

+ medulla, respectively, of V2a-(Gi)DREADD mice. This represents 77 ± 0.8% of Chx10 neurons in neonatal and 80 ± 1.3% of Chx10+ neurons in adult cervical cord. In the medulla (at the level of the nucleus ambiguous) we detect HA-Tag+ immunoreactivity in 65 ± 4% of Chx10+ neurons in neonatal and 58 ± 3%

+ + of Chx10 neurons in adult V2a-(Gi)DREADD mice. HA-Tag cells are not observed in mice lacking the

ROSAPNP-(Gi)DREADD/+ allele in P2 neonates (n=5) or adults (n= 4), confirming the specificity of our antibody.

As expected from prior studies showing that ~50% of V2a neurons in the cervical cord that express Chx10 during embryonic development have low (detectable by RNA sequencing but undetected by antibodies)

Chx10 expression after birth (Hayashi et al. 2018), we found that 47 ± 2% of HA-Tag+ neurons of P2 neonatal mice did not express detectable levels of Chx10. Similarly, 35 ± 6% of HA-Tag+ cells in neonatal medulla, 23.3 ± 3% in adult cervical spinal cord, and 23 ± 3% in adult medulla do not express detectable levels of Chx10. These results demonstrate that V2a-(Gi)DREADD mice express the inhibitory DREADD receptor in the majority of V2a neurons and the extent and pattern of expression is similar in neonatal and adult mice.

Figure 1: (Gi)DREADD is expressed in V2a neurons in the spinal cord and brainstem of V2a-(Gi)DREADD mice. Immunohistochemistry using antibodies to Chx10 and the HA tag on (Gi)DREADD were used to label the nuclei of V2a neurons and (Gi)DREADD expressing cells, respectively in V2a-(Gi)DREADD mice. (A) C4 cervical spinal cord section (spinal cord boundary outlined) of a neonatal (P2) mouse showing (Gi)DREADD expression in the intermediate lamina where V2a neurons are found. Scale bar = 100um. (B) Coronal section of the medulla of a neonatal (P2) mouse at the level of the nucleus ambiguous showing (Gi)DREADD expression in the medial reticular formation outlined (dotted line) where V2a neurons are found. Scale bar = 200um. (C) C4 cervical spinal cord section (gray matter outlined) of an adult mouse showing (Gi)DREADD expression in the intermediate lamina where V2a neurons are found. Scale bar = 100um. (D) Coronal section of the medulla of an adult mouse at the level of the nucleus ambiguous showing (Gi)DREADD expression in the medial reticular formation outlined (dotted line) where V2a neurons are found. + + Scale bar = 200um. Insets: boxed regions in A-D. (Gi)DREADD cells (green) predominantly co-label with Chx10 nuclei (red) (examples indicated + by white arrows). Chx10 neurons without (Gi)DREADD expression (examples indicated by red arrowheads) and (Gi)DREADD cells without detectable Chx10+ (examples indicated by green arrowheads) are infrequently observed. Scale bars = 20um.

Decreasing the excitability of V2a neurons results in slow and irregular breathing in V2a-(Gi)DREADD neonatal mice.

Previous experiments in which V2a neurons were ablated during embryonic development could not conclusively distinguish if the changes in breathing were the direct result of loss of V2a neuron function versus developmental defects resulting from the loss of V2a neurons in the embryo. Therefore, we used whole body plethysmography (WBP) to measure respiratory frequency and regularity in postnatal day 2 V2a-(Gi)DREADD mice before and after CNO treatment to determine the effects of acutely silencing V2a neurons on breathing (Figure 2). The number of respiratory cycles/minute was calculated before and after CNO treatment in both V2a-(Gi)DREADD mice and littermate non-DREADD expressing controls. Decreasing the excitability of V2a neurons slowed the respiratory rate from 91.3 ± 4.3 cycles/min. to 53.5 ± 3.1 cycles/min (n=6, p=0.023) (Figure 2D) whereas no change was observed in non-

DREADD controls (before CNO: 99.2 ± 5.1 cycles/min. vs. after CNO 94.0 ± 5.3 cycles/min., n=5, p=0.135)

(Figure 2F). There is no significant difference in the respiratory rate between V2a-(Gi)DREADD mice and controls before administration of CNO (p=0.308).

To examine breath-to-breath consistency of the breathing frequency, we plotted the instantaneous frequency for at least 50 consecutive breaths before and after 10.0 mg/kg*bw CNO treatment. V2a-(Gi)DREADD mice injected with CNO showed a highly variable breathing frequency (Figure

2C) that was not seen in CNO treated non-DREADD controls (Figure 2E). In addition, we generated

Poincare maps by plotting the breathing period (Tn) versus the subsequent period (Tn+1). When breathing is consistent, the points in a Poincare plot form a tight cluster (Del Negro et al. 2002; Crone et al. 2012), as seen in non-DREADD controls before and after administration of CNO (Figure 2J). In contrast, decreasing the excitability of V2a neurons in neonatal V2a-(Gi)DREADD mice resulted in disperse Poincare maps

(Figure 2G), indicating that breathing period fluctuates from breath to breath, similar to mice in which

V2a neurons are ablated (Crone et al. 2012). To further quantify the regularity of breathing in V2a-

(Gi)DREADD mice, we compared the coefficient of variation of breathing frequency (CVf) in each animal before and after CNO treatment. We also calculated the irregularity score (IS), which has previously been used to measure changes in respiratory rhythm regularity in medullary slices treated with pharmacological agents and whole animals following carotid body ablation (Viemari et al. 2011; Sheikhbahaei et al. 2017).

The CVf and IS were both significantly increased following CNO treatment in V2a-(Gi)DREADD mice (CVf:

0.22±0.03 vehicle vs. 0.73±0.15 CNO, p=0.030, n=6; IS: 19.22±2.79% vehicle vs. 68.81±19.40% CNO, p=0.031, n=5) (Figure 2H and 2I). This is compared to no change observed in non-DREADD expressing control mice (CVf: 0.23±0.03 vehicle vs. 0.26±0.02 CNO, p=0.619, n=6 IS: 16.67±3.31% vehicle vs.

21.65±3.31% CNO, p=0.363, n=5) (Figure 2K and 2L). No fatalities occurred in either genotype

following CNO treatment, despite the occurrence of apneas in V2a-(Gi)DREADD pups. Thus, V2a neuron activity is required to maintain both the frequency and regularity of breathing in neonatal mice.

Decreasing V2a neuron excitability increases breathing frequency without altering regularity in adult mice.

We next assessed whether acutely silencing V2a neurons in adult mice alters the frequency or regularity of breathing. Following IP injection of either vehicle or CNO, V2a-(Gi)DREADD mice (n=9) were placed in a plethysmography chamber to measure ventilation. A subset of V2a-(Gi)DREADD mice (4 of the

9) were instrumented with telemetry devices to record EMG from the diaphragm. A representative trace showing simultaneous recordings of WBP and diaphragm EMG following vehicle and CNO injections is shown in Figure 3A-B. Instantaneous breathing frequency (f) was measured from plethysmography recordings taken before and after CNO treatment (n=9). Surprisingly, silencing V2a neurons in adult mice increased respiratory frequency 15.3% (Figure 3C) whereas no change was observed in the control mice lacking the (Gi)DREADD receptor (0.2 ± 4.5%; p=1.00, n=8) (Figure 3D). Consistent with WBP, the

63 instantaneous bursting frequency of the diaphragm in V2a-(Gi)DREADD mice was significantly increased after CNO treatment (188.0 ± 11.4 vehicle vs. 237.553 ± 20.2 bursts/min. CNO, p=0.04, n=4) (Figure 3E).

No change was observed after CNO treatment in non-DREADD controls (193.7 ± 8.6 vehicle vs. 187.7 ±

14.6 bursts/min. CNO, p=0.733, n=5) (Figure 3F). There is no difference in bursting frequency between genotypes following administration of vehicle only (p=0.735). Thus, adult mice show a small increase in breathing frequency when V2a neurons are silenced rather than a dramatic decrease in frequency observed in neonatal mice.

Figure 2: Decreasing the excitability of V2a neurons causes irregular breathing in neonatal V2a-(Gi)DREADD mice. Representative traces of whole body plethysmography before (top trace) and after (bottom trace) 10.0 mg/kg*bw CNO treatment in a (A) V2a-Gi(DREADD) mouse and a (B) non-DREADD control (Chx10Cre/+) mouse at P2. (C and E) Instantaneous respiratory frequency is plotted for at least 50 consecutive breaths in

V2a-(Gi)DREADD mice and non-DREADD control mice before (left) and after (right) CNO treatment to illustrate the breath to breath consistency of respiratory frequency. (D and F) Respiratory rate is decreased after silencing V2a neurons in V2a-(Gi)DREADD mice but remains unchanged following CNO treatment in non-DREADD control mice. Poincare maps generated by plotting the respiratory cycle period (Tn) versus the subsequent respiratory cycle period (Tn+1) from the same trials shown in C and E show an inconsistent period following CNO treatment in V2a-

(Gi)DREADD mouse (G) but not in the non-DREADD control (J). (H) CNO treatment (10.0 mg/kg*bw) increases the coefficient of variation of breathing frequency (CVf) compared to baseline in V2a-(Gi)DREADD mice (p=0.030, n=6 but not in non-DREADD control mice (p=0.619, n=5) (K). (I) The average irregularity score of the breathing frequency is increased in V2a-(Gi)DREADD mice following CNO (10.0 mg/kg*bw) (p=0.031, n=6) but remains unchanged in non-DREADD control mice (p=0.363, n=5) (L). Gray dashed lines show individual animals. Solid black line shows the mean ± SE. * p<0.05, paired t-test or Wilcoxin signed-rank test (Figure 1I).

We next assessed whether silencing V2a neurons impacts the regularity of breathing in adult V2a-

(Gi)DREADD mice. To exclude potential effects caused by changes in ARM activity (see below), we plotted the instantaneous frequency for 100 consecutive breaths during intervals in which ARMs are not active

(interbout intervals). We observed a stable, consistent breathing frequency for each animal during vehicle and CNO treatment (Figure 3G-H). Breathing regularity is evident from Poincare plots that form a tight cluster in vehicle and CNO treated V2a-(Gi)DREADD mice (Figure 2I-J) as well as no effect of CNO treatment on the CVf (0.146 ± 0.014 vehicle vs. 0.130 ± 0.010 CNO, p=0.492, n=9) (Figure 3K) or IS (12.8 ± 0.6% vehicle vs. 11.6 ± 0.8% CNO, p=0.378, n=9) (Figure 3L). These results demonstrate that, unlike in neonatal mice,

V2a neuron activity is not critical for maintaining a regular respiratory rhythm in adult mice.

Decreasing V2a neuron excitability increases breathing frequency without altering regularity in adult mice.

Decreasing V2a neuron excitability increases ARM EMG activity.

We next assessed the effect of decreasing V2a excitability on extradiaphragmatic respiratory muscle activity. Mice were instrumented with devices to non-invasively measure EMG from trapezius

(n=9) and either scalene (n=5) or diaphragm (n=4) muscles. V2a-(Gi)DREADD mice were placed in a plethysmography chamber and muscle activity was measured after a vehicle injection and then after injecting CNO. The root mean square (RMS) of the EMG signal was used to identify bouts of increased muscle activity in mice at rest (C. B. Mantilla et al. 2011). There was no change in the percentage of time spent in a resting state before and after silencing V2a neurons (47.3 ± 5.2% vehicle vs. 50.7±5.8% CNO, n=9, p=0.670). Unexpectedly, we observed an increased incidence of bouts of trapezius and scalene activity after V2a neurons are inhibited (Figure 4A). The frequency of trapezius and scalene bouts is dependent on the dose of CNO, with the highest frequency in both muscles observed at a dose of 10.0

65 mg/kg*bw CNO (Figure 4B and C). A 9.3-fold increase in trapezius and 7.5-fold increase in scalene activity was observed following CNO treatment at this dose. Trapezius and scalene muscles were usually

Figure 3. Decreasing the excitability of V2a neurons increases breathing frequency without causing irregular breathing in adult V2a- (Gi)DREADD mice. (A-B) Representative traces of whole body plethysmography (WBP) and diaphragm electromyography (EMG) following injections of vehicle (A) and subsequently 10.0 mg/kg*bw CNO (B) into the same V2a-(Gi)DREADD mouse on the same day. (C-D) Instantaneous Cre/+ respiratory frequency was assessed in V2a-(Gi)DREADD mice (C) and non-DREADD (Chx10 ) control mice (D) by WBP after injection of vehicle peak (-CNO) and CNO. (E-F) The instantaneous bursting frequency was calculated from the RMS of the diaphragm EMG signal in V2a-(Gi)DREADD mice (E) and non-DREADD control mice (F) after injection of vehicle (-CNO) and CNO. (G-H) Instantaneous respiratory frequency is plotted for 100 consecutive breaths in a V2a-(Gi)DREADD mouse injected with vehicle (G) or 10.0 mg/kg*bw CNO (H) to illustrate the breath to breath consistency of respiratory frequency. (I-J) Poincare maps generated by plotting the respiratory cycle period (Tn) versus the subsequent respiratory cycle period (Tn+1) from the same trials shown in G and H show a consistent period following both vehicle (I) and CNO (J) treatments. (K) CNO treatment (10.0 mg/kg*bw) does not alter the coefficient of variation of breathing frequency (CVf) compared to vehicle (-CNO) treatment in V2a-(Gi)DREADD mice (p=0.492, n=9). (L) The average breathing frequency irregularity score of V2a-(Gi)DREADD mice following vehicle and CNO (10.0 mg/kg*bw) is not significantly different (p=0.378, n=9). Gray dashed lines show individual animals. Solid black line shows the mean ± SE. * p<0.05, paired t-test or Wilcoxin signed-rank test (Figure 2K).

66 co-activated (87±4% of scalene bouts coincided with trapezius bouts, n=5) following treatment with 10.0 mg/kg*bw CNO, signifying that the bouts are a form of patterned activity. CNO does not increase trapezius or scalene activity in non-DREADD expressing control mice at any dose tested (Figure 4D), demonstrating that the effects of CNO are the result of (Gi)DREADD inhibition of V2a neurons. We observed a small but statistically significant difference in baseline trapezius ARM activity in V2a-

(Gi)DREADD mice not treated with CNO compared to non-DREADD controls at rest (0.203 ± 0.04 bouts/min. (Gi)DREADD (n= 9) vs. 0.080 ± 0.021 bouts/min. control (n= 5), p = 0.043). However, a difference of 0.1 bouts/min. is not biologically significant compared to the increase we see after CNO administration (1.46 bouts/min.). Further, trapezius ARM activity in vehicle treated V2a(Gi)DREADD mice is comparable to what has previously been reported in wild type mice (0.176 ± 0.038 bouts/min.) (Romer et al. 2017), indicating that this group of non-DREADD control mice had a particularly low level of baseline

ARM activity. Finally, we plotted the occurrence of each trapezius bout over the hour-long recording period for each animal to characterize the timing of ARM activation, with time spent in the resting state indicated with gray bars (Figure 4E). Results show that ARM bouts can occur as soon as 5 minutes following

CNO treatment (accounting for the two-minute re-acclimation to the chamber following CNO injection) and substantially declines within the last 15 minutes of the hour long recording. In summary, inhibition of

V2a neuron activity causes an increase in the incidence of trapezius and scalene muscle activity in mice at rest.

Figure 4. Decreasing the excitability of V2a neurons increases extradiaphragmatic respiratory muscle activity. (A) Representative traces showing three different examples of bouts of trapezius and scalene EMG activity (recorded simultaneously) following injection of 10.0 mg/kg*bw CNO into a V2a-(Gi)DREADD mouse. The root mean square of the EMG signal (RMS) is shown above the EMG trace for each muscle. The frequency of trapezius (B) (n=9) and scalene (C) (n=5) bouts was measured prior to CNO treatment (white bars) as well as following 1.0 – 15.0 mg/kg*bw CNO (black bars) in V2a-(Gi)DREADD mice. Experiments using different doses of CNO were performed on separate days within the same animals. (D) CNO treatment (1.0- 15.0 mg/kg*bw) does not increase trapezius bout frequency in non-DREADD control (Chx10Cre/+) mice lacking the V2a- (Gi)DREADD receptor (n=5). (E) The time point when each ARM bout occurred for the trapezius (n=9) is plotted for each mouse (black dot) across the hour long EMG recording following CNO treatment. Gray highlighted regions respresent time periods each mouse spent in the resting state when ARM bout frequency could be analyzed. (*) Treatment group is significantly different than vehicle control (p<0.05, one-way repeated measures ANOVA p<0.05); (┼) Group is different than 10.0 mg/kg*bw CNO (p<0.05, one-way repeated measures ANOVA). Gray dashed lines show individual animals. Solid black line shows the mean ± SE.

Decreasing V2a neuron excitability does not alter diaphragm EMG peak amplitude

In order to address the possibility that increased ARM activity resulting from silencing V2a neurons is a compensatory response to impaired diaphragm function, we analyzed the diaphragm EMG signal in V2a-

(Gi)DREADD mice and controls implanted with leads in the diaphragm and trapezius muscles (Figure 5A).

The peak amplitude of the root mean square of the EMG signal (RMSpeak) was used to measure the amplitude of each burst of diaphragm activity (each breath) as well as trapezius activity during bouts and interbout intervals. The RMSpeak amplitude was normalized to the amplitude during near maximal ventilatory behavior (sighs) for each muscle in order to reduce intra-animal variability (C. B. Mantilla et al.

2011). The trapezius RMSpeak is not significantly different between animals implanted in the trapezius and diaphragm (n= 4) compared to trapezius and scalene (n= 5) (t-test p= 0.938), so these two groups were pooled for analysis of trapezius RMSpeak. We observe a 4.2-fold increase in trapezius RMSpeak during a bout compared to interbout intervals (Figure 5C). In contrast, we observe no significant change in diaphragm

RMSpeak during bouts of trapezius activity compared to interbout intervals (Figure 5D) and no change in diaphragm RMSpeak during interbout intervals following CNO injection compared to vehicle only (0.51 ±

0.05 vehicle vs. 0.52 ± 0.05 CNO, n=4, paired t-test p=0.889). However, we verified that our devices are capable of measuring changes in diaphragm RMSpeak during different behaviors by performing nasal occlusion for four minutes on the same V2a-(Gi)DREADD mice (Figure 5B). Nasal occlusion is known to increase ventilatory drive and increase diaphragm RMSpeak (C. B. Mantilla et al. 2011). We measured a 2.1- fold increase in diaphragm RMSpeak following nasal occlusion compared to eupnea (Figure 5E), consistent with previously published literature. Thus, despite our ability to detect increases in the peak amplitude of diaphragm muscle activity during sighs and nasal occlusion, we did not observe a change in diaphragm peak amplitude in V2a-(Gi)DREADD mice following CNO treatment.

Figure 5. Diaphragm EMG peak amplitude does not change during ARM bouts. (A) Representative traces showing whole body plethysmography (WBP) as well as diaphragm and trapezius EMG signals recorded simultaneously from a V2a-(Gi)DREADD mouse. The root mean square of the EMG signal (RMS) is shown above the EMG trace for each muscle. (Left) The gray box indicates a bout of trapezius activity in a CNO treated mouse. (Right) WBP, EMG, and RMS signals from the same mouse during a sigh (near maximal ventilation, arrow). (B) Representative trace showing WBP and diaphragm EMG during eupnea and during nasal occlusion. (C and D) Diaphragm activity is not increased during bouts of increased trapezius activity. (C) Trapezius RMSpeak values were normalized to trapezius RMSpeak values during sighs to calculate the average normalized peak amplitude of trapezius activity for each animal (n= 9 animals) during bouts of trapezius activity (ARM Bout) and during interbout intervals (No Bout). (D) Diaphragm RMSpeak during bouts of trapezius activity and during interbout intervals, normalized to diaphragm RMSpeak during sighs for each animal (n=4). (E) Diaphragm RMSpeak is increased during nasal occlusion (n= 4, *p<0.05, paired t-test). Gray dashed lines represent individual animals. Solid black lines show the mean ± SE for all animals.

Decreasing V2a neuron excitability increases ventilation.

To assess the impact on ventilation of decreasing V2a neuron excitability, we analyzed WBP to compare ventilation following CNO injection to vehicle control injections. We first focused our analysis on interbout intervals in order to exclude potential effects on ventilation caused by increased ARM recruitment. We observed a 24.6% increase in peak inspiratory flow (PIF), 28.6% increase in minute volume (MV), and 15.3% increase in breathing frequency (f) in V2a-(Gi)DREADD mice after silencing V2a neurons, but no change was observed in tidal volume (VT) (Figure 6A-F). In addition, there was no significant difference in sigh frequency following CNO injection compared to vehicle (0.54 ± 0.14 vehicle vs. 0.54 ± 0.05 CNO; p=0.812). We assessed ventilatory drive by measuring VT/Ti (Morgan et al. 2014) and found that it is increased following CNO treatment, primarily driven by a decrease in Ti (Figure 6G-H).

Importantly, the effects of CNO on ventilation are caused by altering Gi protein signaling in V2a neurons

because we observed no CNO-dependent changes in PIF, VT, MV, or f in non-DREADDcontrol mice (Figure

6C-F). We also investigated any potential effects of (Gi)DREADD expression alone on breathing by

comparing ventilation parameters in V2a-(Gi)DREADD and non-DREADD controls following vehicle injection. We did not find a statistically significant difference in PIF (V2a-(Gi)DREADD: 2.562 ± 0.120 vs. non-DREADD 2.306 ± 0.139; p=0.182), VT (0.192 ± 0.014 vs. 0.176 ± 0.010 ; p=0.716), MV (1.24 ± 0.06 vs.

1.29 ± 0.10; p=0.728), f (199.2 ± 9.4 vs. 195.6 ± 6.3; p=0.066) or Ti (0.134 ± 0.006 vs. 0.151 ± 0.004; p=0.136). Our results are the first to demonstrate that inhibiting V2a neurons in adult animals impacts ventilation, even during the intervals between bouts of ARM activity.

We further investigated the impact of increased ARM activity on breathing by comparing ventilation during bouts of trapezius activity to ventilation during interbout intervals in V2a-(Gi)DREADD mice implanted with leads in the diaphragm and trapezius (n=4). Significant increases are observed in PIF

(30.6%), MV (28.0%), and f (22.2%) during ARM bouts compared to interbout intervals in V2a-(Gi)DREADD mice treated with CNO (Figure 6C, E-F). There is a modest (15.6 ± 2.9%) increase in VT during ARM bouts

71 that is statistically significant compared to VT during interbout intervals in the absence of CNO (Figure 6D).

VT/Ti is increased during ARM bouts, primarily due to a significant decrease in Ti (Figure 6G-H). The decrease in Ti and increase in f during bouts is consistent with a decrease in diaphragm EMG burst duration

(64 ± 3ms bout, 97 ± 3ms no bout; p=0.003) and increase in f (298 ± 10 bursts/min. bout, 238 ± 18 no bout; p=0.012) following CNO injections observed in V2a-(Gi)DREADD mice. We did not attempt to quantify ventilatory parameters during trapezius activity in non-DREADD expressing mice or V2a-

(Gi)DREADD mice during vehicle injections due to the rarity of ARM bouts in control mice. These data show that the ARM activity resulting from inhibition of V2a neurons is associated with a further increase in ventilation over that observed during interbout intervals.

In order to determine if ARM activity might be triggered by a change in ventilation, we compared minute volume during the four breaths immediately preceding each bout of ARM activity to the average minute volume during interbout intervals and found no significant difference (1.52 ± 0.06 vs. 1.61 ± 0.09 mL/min./g*bw; p=0.392). We also detected no significant differences in PIF (3.22 ± 0.11 vs. 3.18 ± 0.14 max mL/s; p=0.823), VT (0.19 ± 0.01 vs. 0.21 ± 0.01 mL/breath; p=0.282), or f (227.1 ± 9.8 vs. 227.7 ± 9.7 breaths/min.; p=0.922) during breaths immediately preceding a bout compared to interbout intervals.

However, we did observe a small but significant decrease in inspiratory time immediately preceding a bout (0.11 ± 0.01 vs. 0.19 ± 0.01 s; p=0.030). Thus, the majority of ventilation parameters are not significantly altered immediately prior to the onset of ARM bouts.

Figure 6. Ventilation is increased during bouts of ARM activity. (A) Representative trace showing WBP and trapezius EMG following injection of vehicle. (B) Representative trace showing WBP and trapezius EMG after injection of 10.0 mg/kg*bw CNO. Gray boxes outline trapezius bouts and corresponding changes in WBP. (C) Peak inspiratory flow (PIF), (D) tidal volume (VT), (E) breathing frequency (BPM), (F) minute volume (MV), (G) inspiratory time (Ti), and (H) VT/Ti were measured within the same V2a-(Gi)DREADD mice (n=9) and non-DREADD control mice (n=8) during interbout intervals in the absence of CNO, during interbout intervals with 10.0 mg/kg*bw CNO, and during trapezius bouts with 10.0 mg/kg*bw CNO. The low frequency of ARM activity in non-DREADD controls precluded measurement of WBP during ARM bouts. (*) Group is different than -CNO/-bout (vehicle control); (┼) +CNO/+bout is different than +CNO/-bout (p<0.05, one-way repeated measures ANOVA with all comparisons Holm-Sidak post-hoc test for V2a-(Gi)DREADD mice or paired t-test for non-DREADD controls). Gray dashed lines show individual animals. Solid black line shows the mean ± SE.

Silencing V2a neurons does not alter arterial oxygen saturation

Ventilation is tightly coupled to blood gas levels (O2 and CO2) to maintain blood pH within a narrow physiological range. It is possible that silencing V2a neurons may decrease blood oxygen levels while increasing blood carbon dioxide levels. This decrease may be detected indirectly by chemosensors (such as the peripheral carotid bodies or central retrotrapezoid nucleus Phox2b+ neurons) and activate ARMs to enhance ventilation in an attempt to expel the excess CO2 from the body. In order to assess whether silencing V2a neurons decreases blood oxygen saturation, we recorded pulse oximetry from anesthetized

V2a-(Gi)DREADD mice to measure any potential changes in arterial oxygen saturation.

We first confirmed that silencing V2a neurons with CNO in V2a-(Gi)DREADD mice increases ARM activity in 2/4 isoflurane anesthetized mice (Figure 7A). Trapezius ARM EMG activity, diaphragm EMG activity, and arterial blood oxygen saturation were simultaneously measured before and after CNO administration to determine whether arterial oxygen saturation decreases at the onset of ARM activity. A representative plot of the continuous oxygen saturation levels from a V2a-(Gi)DREADD animal throughout the entire recording is shown in Figure 7B. The average blood oxygen saturation over the course of two minutes was averaged during the baseline recording and 50 minutes following CNO treatment (following the onset of ARM activity). No change was observed before and after CNO treatment in V2a-(Gi)DREADD mice (99.5 ± 0.1 vehicle vs. 99.5 ±, 0.2 CNO, p=0.944, n=4). However, we verified that we can detect a decrease in oxygen saturation during nasal occlusion (p=0.016, n=4) (Figure 7C). Therefore, silencing V2a neurons does not appear to indirectly activate ARMS by decreasing arterial blood oxygen saturation.

Figure 7. Decreasing V2a neuron excitability does not alter arterial oxygen saturation. (A) Representative trace showing scalene and diaphragm EMG terminal recordings in anesthetized V2a-(Gi)DREADD mice before and after CNO treatment (10.0 mg/kg*bw). (B) Representative plot of arterial oxygen saturation in a V2a-(Gi)DREADD mouse. Black arrowheads indicate application of CNO, the onset of scalene bouts, and nasal occlusion. (C) Arterial oxygen saturation in the absence of CNO, with CNO following the onset of scalene bouts, and during nasal occlusion is shown for V2a-(Gi)DREADD mice (n=4). (*) Group is different from -CNO and +CNO. Gray dashed lines show individual animals. Solid black line shows the mean ± SE.

Decreasing V2a neuron excitability does not alter blood pressure or heart rate

Similar to arterial oxygen saturation, heart rate and blood pressure is tightly coupled to ventilation to adequately match cardiac output with ventilatory demand during different behaviors (e.g. resting vs. during strenuous exercise). Therefore, it is possible that silencing all V2a neurons may cause peripheral changes in the cardiovascular system that feedback to alter ventilation. We investigated whether the increase in ventilation after silencing V2a neurons may be caused by changes in blood pressure or heart rate. There was no change before and after CNO treatment in V2a-(Gi)DREADD mice in systolic blood pressure (104.8 ± 3.8 mmHg vehicle vs. 107.7 ± 2.3 mmHg CNO, n=6, paired t-test, p=0.459) (Figure 8A), diastolic blood pressure (73.4 ± 2.4 mmHg vehicle vs. 77.9 ± 2.3 mmHg CNO, n=6, paired t-test, p=0.252)

(Figure 8D), or mean blood pressure (83.5 ± 2.7 mmHg vehicle vs. 87.5 ± 2.2 mmHg CNO, n=6, paired t-

75 test, p=0.293) (Figure 8G). Similarly, no change was observed before and after CNO treatment in non-

DREADD expressing controls in systolic blood pressure (108.7 ± 3.7 mmHg vehicle vs. 108.7 ± 2.8 mmHg

CNO, n=7, paired t-test, p=0.999) (Figure 8B), diastolic blood pressure (78.3 ± 3.8 mmHg vehicle vs. 78.6

± 2.3 mmHg CNO, n=7, paired t-test, p=0.995) (Figure 8E), or mean blood pressure (88.2 ± 3.7 mmHg vehicle vs. 88.3 ± 2.4 mmHg CNO, n=6, paired t-test, p=0.995) (Figure 8H). There was also no difference observed between V2a-(Gi)DREADD mice and non-DREADD expressing controls in baseline systolic blood pressure (t-test, p=0.518) , diastolic blood pressure (t-test, p=0.352), or mean blood pressure (t-test, p=0.383). Similarly, ECG analysis from implanted mice revealed no change in heart rate in V2a-(Gi)DREADD mice (573 ± 22 beats/min. vehicle vs. 576 ± 48 beats/min. CNO, n=8, paired t-test, p=0.742) (Figure 7J) and non-DREADD expressing controls (618 ± 27 beats/min. vehicle vs. 615 ± 22 beats/min. CNO, n=6, paired t-test, p=0.293) (Figure 8K). Therefore, we conclude that changes in ARM activity and ventilation does not result from indirect effects of silencing V2a neurons on cardiovascular function.

Although we do not detect a change in blood pressure or heart rate following CNO treatment in

V2a-(Gi)DREADD mice and non-DREADD controls, we verified that we can detect changes in these autonomic parameters. We increased the excitability of V2a neurons in V2a-(Gq)DREADD mice where brainstem and spinal cord V2a neurons express the excitatory DREADD receptor and assessed the effect on heart rate and blood pressure. Interestingly, we observe an increase in systolic blood pressure ((114.1

± 2.4 mmHg vehicle vs. 132.9 ± 2.5 mmHg CNO, n=7, paired t-test, p=0.003) (Figure 8C), diastolic blood pressure (84.9 ± 3.8 mmHg vehicle vs. 98.0 ± 2.4 mmHg CNO, n=7, paired t-test, p=0.01) (Figure 8F), mean blood pressure (94.3 ± 3.3 mmHg vehicle vs. 109.6 ± 2.4 mmHg CNO, n=7, paired t-test, p=0.0.007) (Figure

8I) and heart rate (470 ± 8 beats/min. vehicle vs. 584 ± 11 beats/min. CNO, n=3, paired t-test, p=0.0.01)

(Figure 8L). Thus, we show that 1) our blood pressure and heart rate measurement methods are sensitive to detect changes even though we do not detect any change after silencing V2a neurons or in non-

DREADD expressing controls following CNO treatment and 2) brainstem and spinal cord V2a neurons are sufficient but not necessary to alter blood pressure and heart rate.

Figure 8. Brainstem and spinal cord V2a neurons are sufficient but not necessary to increase heart rate and blood pressure. Systolic blood pressure, diastolic blood pressure, mean blood pressure, and heart rate were measured in conscious V2a-(Gi)DREADD (7A, D, G, and L, respectively), non-DREADD expressing control mice (7B, E, H, and K, respectively), and V2a-(Gq)DREADD mice (7C, F, I, and L, respectively). All data were analyzed with a paired t test. (*) p<0.05. Gray dashed lines show individual animals. Solid black line shows the mean ± SE.

Increasing and decreasing the excitability of V2a neurons differentially affects motor functions

It has previously been shown that increasing the excitability of V2a neurons with CNO in V2a-

(Gq)DREADD mice activates ARMs (Morgan et al. 2014). Since increasing and decreasing the excitability of

V2a neurons both increase the incidence of ARM bouts, we examined the effects of altering the excitability of V2a neurons on other motor activities. Wild type mice can fully splay their hindlimbs when lifted by the tail (Figure 9A) whereas mice lacking V2a neurons do not extend their hindlimbs (n=3) (Figure 9B). Prior to CNO treatment, V2a-(Gi)DREADD mice fully extend their hindlimbs when suspended by the tail, like wild type mice. However, following CNO treatment (10 mg/kg*bw), V2a-(Gi)DREADD mice (n=3) lose this hindlimb extension reflex and their limbs hang in a relaxed manner (Figure 9C). V2a-(Gq)DREADD mice

(n=3) show normal hindlimb extension prior to and after treatment with CNO (1.0 mg/kg*bw) (Figure 9D).

The selected doses of CNO were those that elicited the greatest increase in ARM activity for each mouse line. Thus, increasing and decreasing the excitability of V2a neurons has distinct effects on a hindlimb extension reflex.

Next, we examined the effects of increasing and decreasing the excitability of V2a neurons on diaphragm activity by performing terminal diaphragm EMG recordings under isoflurane anesthesia.

Similar to what is observed in conscious V2a-(Gi)DREADD mice instrumented with telemetry transmitters, silencing V2a neurons does not alter the diaphragm EMG peak amplitude (0.61 ± 0.07 vehicle vs. 0.52 ±

0.03 CNO, n=3, paired t-test, p=0.376) or regularity of diaphragm bursting in anesthetized mice (CVf: 0.04

± 0.004 vehicle vs. 0.04 ± 0.003 CNO, n=3, paired t-test, p=0.903; IS: 3.94 ± 0.27 vehicle vs. 3.97 ± 0.27

CNO, n=3, paired t-test, p=0.935) (Figure 9E-F). We then used the same protocol to assess diaphragm EMG activity after increasing the excitability of V2a neurons in V2a-(Gq)DREADD mice. CNO does not alter the normalized diaphragm EMG peak amplitude in anesthetized mice (0.46 ± 0.04 pre CNO vs. 0.44 ± 0.03 post CNO, n=5, paired t-test, p=0.984) (Figure 9G-I). However, we observed additional activity during the expiratory phase of the respiratory cycle in 5 out of 5 V2a-(Gq)DREADD mice (Figure 9I). Expiratory

78 diaphragm activity is observed intermittently during 44 ± 8% of the total recording time following CNO treatment with an average duration of 308 ± 112 s. In contrast, activity during the expiratory phase is not observed in V2a-(Gi)DREADD mice (n=3) (Figure 9F). These results demonstrate that although increasing and decreasing the excitability of V2a neurons has similar effects on ARM activity, the two manipulations have distinct effects on diaphragm activity during the expiratory phase of respiration.

Figure 9. Increasing and decreasing the excitability of V2a neurons differentially affects motor functions. (A-D) Mice were suspended by the tail in order to assess hindlimb extension. (A) Wild type mice extend their hindlimbs. (B) Chx10::DTA mice (in which V2a neuron were ablated during development) do not splay their hindlimbs. (C) V2a-(Gi)DREADD mice treated with 10 mg/kg*bw CNO do not splay their hindlimbs. (D) V2a-(Gq)DREADD mice treated with 1.0 mg/kg*bw CNO splay their hindlimbs. (E-I) Diaphragm EMG was recorded under isoflurane anesthesia in V2a-(Gi)DREADD and V2a-(Gq)DREADD mice. Normal diaphragm bursting activity is observed in V2a-(Gi)DREADD mice before (E) and after (F) silencing V2a neurons with 10 mg/kg*bw CNO, as well as in V2a-(Gq)DREADD mice prior to CNO treatment (G). V2a-(Gq)DREADD mice treated with 1.0 mg/kg*bw CNO alternate between a normal diaphragm bursting pattern (H) and a pattern that also includes diaphragm activity during the expiratory phase (I).

Figure 10. Hypothetical Model: V2a neuron diversity may explain how increasing or decreasing the excitability can activate ARMs.

We propose that at least two subtypes of V2a neurons exist in respiratory circuits. Type I V2a neurons are part of an excitatory pathway (green circle) to increase ARM activity. Type II V2a neurons are part of an inhibitory pathway (red circle) that prevents ARM activation when not needed. When V2a excitability is increased, the Type I neurons (normally inactive at rest) drive ARM activity. When V2a excitability is decreased, Type II V2a (normally active at rest) are inactivated, disinhibiting ARM motor neurons (purple diamond). Descending brainstem drive from the ventral respiratory group to ARM motor neurons is shown by the blue arrow.

2.4 Discussion

In this study, we tested the role of brainstem and spinal cord V2a neurons in respiratory rhythm generation and pattern formation by acutely silencing V2a neurons in conscious neonatal or adult mice using an inhibitory DREADD while measuring respiratory muscle activity and ventilation. We show that

V2a neural activity is required to maintain the frequency and regularity of breathing in neonatal mice.

However, V2a neuron activity is not required to maintain respiratory rhythm regularity in adult mice. In addition, we show that decreasing the excitability of V2a neurons results in activation of extradiaphragmatic respiratory muscles and increased ventilation, with little effect on diaphragm activity.

Thus, during the neonatal period V2a neuron function is critical for respiratory rhythm generation, whereas in adults V2a neurons play an important role in patterning the activity of extradiaphragmatic respiratory muscles.

Our study also suggests that inhibiting V2a neuron activity could be a safe, effective method to increase ARM activity and ventilation without impairing diaphragm function in adult patients with neuromuscular disease or spinal cord injury. This study is the first to demonstrate that acutely silencing

V2a neurons in neonatal mice causes slow and irregular breathing. A limitation of a prior study in which

V2a neurons were ablated during development (Crone et al. 2012) is that it was not possible to distinguish between the acute functions of V2a neurons in neonatal mice versus potential developmental defects that might have resulted from the loss of V2a neurons in the embryo. Importantly, CNO does not alter the frequency or regularity of breathing in non-DREADD expressing mice, demonstrating that our results are not due to off-target effects of CNO on other cell types. Further, we observe no difference in breathing frequency or pattern in V2a-(Gi)DREADD mice and non-DREADD control mice before CNO treatment, indicating that DREADD expression alone does not appear to impair the development or function of V2a neurons. Within the medulla, V2a neurons are located in the medial reticular formation and are distinct from several other classes of neurons known to influence respiration, including the Dbx1+ V0 neurons of the ventral respiratory group, Phox2b+ and Atoh1+ neurons of retrotrapezoid nucleus/parafacial respiratory group, Lmx1b+ and Pet1+ catecholaminergic and serotonergic neurons, and Pax2+, Lbx1+, and

FoxP2+ neurons of the brainstem (Gray 2013; Dick et al. 2018). Our study is consistent with prior studies indicating that V2a neurons in the medial reticular formation of the brainstem provide excitatory drive to neurons in the PreBötzinger complex (likely V0 neurons) required for normal respiratory rhythm generation in neonatal mice (Crone et al. 2012).

By performing acute silencing experiments in adult mice, we show that V2a neuron activity is not required in mature animals for regular breathing rhythm. In fact, instead of slowing respiration, we observe a small increase in frequency when V2a neurons are silenced in the adult. Prior studies in which

V2a neurons were ablated could not distinguish whether the recovery of a normal breathing rhythm after the first week of birth was because V2a neurons are no longer needed at older ages or the result of compensatory changes caused by ablation. Our results demonstrate that the role of V2a neurons in modulating respiratory rhythm changes during normal postnatal development. It is unlikely that the lack of slow and irregular breathing after silencing V2a neurons in adult mice is due to downregulation of the

(Gi)DREADD receptor in V2a neurons during maturation since there is no difference in the percentage of

V2a neurons that express the (Gi)DREADD receptor between neonatal and adult mice in both the brainstem and spinal cord. It is also unlikely that this difference is the result of dramatic changes in connectivity between V2a neurons and brainstem respiratory centers since increasing V2a neuron excitability can increase respiratory frequency in adult mice (Romer et al. 2017). Moreover, V2a neurons appear to project throughout the ventral respiratory column in adult mice (Romer et al. 2017; our unpublished observations). Many changes occur in the neural control of respiration during the perinatal period, including changes in chemosensation (Guyenet and Bayliss 2015), neuromodulator signaling (Gray et al. 2015; Hilaire and Duron 1999; Hilaire et al. 2004; Doi and Ramirez 2008; Wong-Riley and Liu 2008), inhibitory neurotransmission (Hubner et al. 2001; Okabe et al. 2015; Dubois et al. 2018), afferent regulation (Hilaire and Duron 1999; Dutschmann et al. 2004), pontine-medullary interactions (Oku,

Masumiya, and Okada 2007; Dutschmann and Dick 2012; Del Negro, Funk, and Feldman 2018), as well as morphology and properties of respiratory network neurons (Denavit-Saubie et al. 1994; Di Pasquale et al.

1996; Richter and Smith 2014). Thus, the loss of excitatory drive provided by V2a neurons may be largely mitigated by other sources of drive in the adult that are not present or sufficient to restore breathing rhythm in newborn mice. Alternatively, drive from V2a neurons may only be required under specific conditions or behaviors in adult mice that were not tested in the present study (e.g., running).

Although we cannot rule out the possibility that the small population of V2a neurons that apparently lack DREADD expression are sufficient to maintain the frequency and regularity of breathing in adult mice, this seems unlikely given our results that this population is insufficient to maintain normal breathing in neonatal mice. Despite the fact that we are unable to detect DREADD expression in a significant fraction of medullary V2a neurons (35% of Chx10+ cells), we observe a strikingly similar severity of phenotype in neonatal V2a-(Gi)DREADD mice treated with CNO as observed in Chx10::DTA mice in which >98% of V2a neurons were ablated during development (Crone et al. 2012). We do not know

82 whether this is because silencing a majority of V2a neurons is sufficient to impair breathing or whether we are underestimating the fraction of V2a neurons expressing (Gi)DREADD due to limitations in the sensitivity of our HA antibody. Further, we observe specific motor deficits (including failure of hindlimbs to splay when mice are suspended by the tail and increased ankle extension when walking) in adult V2a-

(Gi)DREADD mice treated with CNO that are consistent with loss of V2a neuron function because they are also observed in mice in which V2a neurons are ablated (our unpublished observations).

Our study has also revealed an unexpected role for V2a neurons in constraining extradiaphragmatic respiratory muscle activity at rest. We observed coordinated activation of scalene and trapezius muscles at rest when V2a neurons were silenced. Interestingly, these results mimic the effects of activating V2a neurons at rest in healthy V2a-(Gq)DREADD mice with the excitatory (Gq)DREADD receptor (Romer et al. 2017). It is unlikely that one DREADD strategy (activating or silencing) is failing because we observe different effects of CNO on the diaphragm in V2a-(Gi)DREADD mice versus V2a-

(Gq)DREADD mice. Specifically, activating V2a neurons causes activity during the expiratory phase of the respiratory cycle, whereas silencing V2a neurons does not. We also observe that V2a-(Gi)DREADD mice, but not V2a-(Gq)DREADD mice, fail to splay their legs when suspended by the tail after CNO treatment, a phenotype observed after V2a neurons have been ablated. Together, these results demonstrate that although silencing and activating V2a neurons both increase ARM activity in healthy mice, each manipulation has distinct effects on diaphragm activity and hindlimb reflexes.

Alternatively, it is possible that decreasing V2a neuron excitability impairs diaphragm function and thereby causes a compensatory increase in ARM activity. However, this mechanism is not consistent with our results as diaphragm RMSpeak is not altered during ARM bout activity or interbout intervals in

V2a-(Gi)DREADD mice treated with CNO. Thus, diaphragm function at rest does not appear to be impaired by silencing V2a neurons, although we cannot rule out a role for V2a neurons in modulating diaphragm activity during exercise, cough, or other behaviors. We do not think that this result is due to our inability

83 to detect changes in diaphragm activity as we clearly see changes in diaphragm RMSpeak during sighs and following nasal occlusion, behaviors that require higher force generation by the diaphragm (C. B. Mantilla et al. 2011). We observed no changes in the frequency of sighs or arousals from sleep, as would be expected if animals experienced hypoxia or hypercapnia, and no changes in ventilation immediately preceding ARM activity. Thus, although we cannot rule out subtle changes in diaphragm function that are not detectable by our methods, it is unlikely that increased ARM activity is secondary to diaphragm impairment caused by silencing V2a neurons.

Another potential mechanism leading to ARM recruitment could be an increase in respiratory drive caused by silencing V2a neurons. We do, in fact, observe changes in ventilation even during periods in which ARMs are not active that are consistent with changes in the activity of brainstem respiratory networks, namely a higher frequency of breathing, shorter Ti and increased VT/Ti (an indicator of neural ventilatory drive). A higher PIF appears to compensate for the shorter Ti to maintain VT. We see an even greater increase in f, PIF, and VT/Ti during ARM activity compared to interbout intervals, as well as an increase in VT (as expected when ARMs are recruited). However, it is not clear from these experiments if the ventilatory changes are due to direct actions of V2a neurons on brainstem respiratory centers or the result of chemosensory or cardiovascular feedback. Despite the changes in respiratory timing, the phrenic motor neurons appear to receive similar levels of overall drive because we observe no detectable change in peak diaphragm activity when V2a neurons are silenced. Moreover, the peak diaphragm activity does not change even when ARMs become active, suggesting that the increased ARM activity is driven by circuits independent of (or with minimal impact on) diaphragm activity. This concept is consistent with studies in human patients showing that neural drive is not uniform to all inspiratory muscles, and may include inspiratory, expiratory and tonic components differentially distributed across motor neuron pools

(Jane E. Butler, 2007; J.E. Butler, Hudson, & Gandevia, 2014).

Changes in cardiovascular function (such as heart rate or blood pressure) can alter respiratory rate. Therefore, it is possible that silencing V2a neurons enhances ventilation by indirectly affecting these autonomic processes. This may happen via a couple mechanisms. First, the autonomic nervous system involuntarily regulates bodily functions such as heart rate, blood pressure, respiratory rate, digestion, and urination. V2a neurons may be synaptically connected to sympathetic preganglionic neurons and relay/provide excitatory input to them to control cardiovascular output. Alternatively, we observe projections from cervical spinal V2a neurons to the lateral paragigantocellular (LPGi) nucleus (unpublished observations), which contains inhibitory neurons that project to inhibitory parasympathetic cardiac vagal neurons that control heart rate. Although we cannot distinguish between these hypotheses in this study, both are consistent with the increase in heart rate and blood pressure observed after activating V2a neurons. Vice versa, we would expect that decreasing V2a excitability would decrease heart rate and blood pressure. However, it is unlikely V2a neurons are required to maintain cardiovascular function since silencing V2a neurons does not alter heart rate or blood pressure. Therefore, these data suggest that silencing V2a neurons does not indirectly enhance ventilation (when ARMs are not active) in response to altered cardiovascular output.

Altered blood flow can also change blood gas composition (oxygen and carbon dioxide saturation), which will activate chemoreceptors. Blood pH and gasses must be kept within a very narrow physiological range to preserve life. Therefore, peripheral chemosensory carotid bodies and central chemosensory neurons in the retrotrapezoid nucleus detect changes in arterial oxygen saturation and carbon dioxide, respectively, to provide neural feedback to alter ventilation and restore equilibrium. For example, a decrease in arterial oxygen saturation and an increase in carbon dioxide will stimulate both the carotid bodies and the retrotrapezoid nucleus in the brainstem to enhance ventilation to expel the excess carbon dioxide from the body (Guyenet and Bayliss 2015). However, it is unlikely that silencing

V2a neurons indirectly activates ARMs and enhances ventilation via altered chemosensory feedback since

85 silencing V2a neurons did not change cardiovascular output or arterial oxygen saturation. Future experiments should directly measure blood gasses before and after silencing V2a neurons in conscious mice to corroborate these results.

A previous study demonstrated that increasing the excitability of V2a neurons is able to activate

ARMs (Romer et al. 2017). To explain our findings that either increasing or decreasing V2a excitability can increase ARM activity, we hypothesize that a distinct subset of V2a neurons promotes ARM activity (i.e., during exercise) via an excitatory spinal pathway whereas another distinct subset inhibits ARM activity at rest (i.e., to prevent activation when they are not needed) via an inhibitory spinal pathway (Figure 10). An inhibitory pathway for ARMs may serve to reduce unnecessary (and energetically expensive) respiratory muscle activity during behaviors that require low force generation when diaphragm activity is sufficient.

The reticulospinal system provides another example where different subsets of V2a neurons appear to produce distinct effects on motor activity. Activating V2a neurons in the rostral medulla can halt ongoing locomotion, indicating that these neurons inhibit locomotor circuits (Romer et al. 2017). On the other hand, ablating V2a neurons prevents initiation of fictive locomotion following stimulation of the caudal medulla in a neonatal brainstem-spinal cord preparation, indicating that caudal medullary or spinal V2a are important for activating locomotor circuits (Crone et al. 2008). Further, even within spinal locomotor circuits, different molecularly defined subsets of V2a neurons appear to have different roles in controlling locomotion. For example, Shox2+/Chx10+ V2a neurons provide excitatory drive to ipsilateral limb motor neurons important for maintaining consistent motor output whereas Shox2-/Chx10+ V2a neurons have connections to inhibitory commissural neurons that maintain alternation of left and right sides at high locomotor speeds (Dougherty et al. 2013; Rybak et al. 2015). An alternative hypothesis is that silencing

V2a neurons induces a compensatory change in downstream neurons (i.e., respiratory motor neurons) which renders them more excitable to other inputs, leading to ARM activation. Distinguishing between these two possibilities would require identifying specific V2a subsets that only activate ARMs when

86 silenced but not activated (or vice versa) or demonstrating that either silencing or inhibiting the same premotor V2a neurons could activate ARMs.

These experiments used the inhibitory DREADD receptor and treatment with CNO to decrease

V2a excitability, a chemogenetic technique that has several limitations that may influence the interpretation of our results. First, Manvich et al. 2018 recently showed that the ligand clozapine (a metabolite of clozapine-N-oxide (CNO)) binds to DREADD receptors to activate intracellular signaling pathways instead of CNO. However, we observe effects of CNO on ARM recruitment within 5 minutes of

CNO administration, suggesting that treatment with CNO instead of clozapine still alters V2a excitability.

Second, high concentrations of clozapine may cause sedation. This is likely not a concern for this study because we do not see a change in the amount of time spent in resting states before and after CNO treatment. Finally, it is unclear how the Gi signaling pathway affects downstream signaling cascades to alter V2a excitability. The canonical Gi signaling pathway silences V2a neurons by activating inward rectifying potassium channels to hyperpolarize the neuron. However, Harris et al. 2018 showed that Gi signaling pathways can activate hyperpolarization activated cyclic nucleotide cation (HCN) channels to actually increase the overall excitability of neurons in the dorsal bed nucleus of the stria terminalis in the brain, and activation of the inhibitory DREADD receptor mimics this excitatory effect. Although additional studies should determine whether HCN channels are present in cervical spinal V2a neurons, it is unlikely that this would account for activation of ARMs in V2a-(Gi)DREADD mice. First, CNO treatment in neonatal

V2a-(Gi)DREADD mice reproduces respiratory deficits observed after embryonically ablating V2a neurons in Chx10::DTA mice. Moreover, CNO has different effects on motor function in V2a-(Gi)DREADD mice and

V2a-(Gq)DREADD mice, suggesting V2a neurons are not activated by the Gi signaling pathway. In order to bypass these limitation and validate our results, V2a neurons can be ablated in adult mice to determine if the same increase in respiratory frequency and activation of ARM bouts occurs that is observed after V2a-

(Gi)DREADD mice are treated with CNO.

Another important limitation of this study is that our mouse model targets V2a neurons in both the brainstem and spinal cord, two populations that likely play different roles in the control of breathing.

Crone et al. 2012 suggested that V2a neurons in the medulla provide excitatory drive necessary for respiratory rhythm generation, since pontine and spinal cord neurons are not present in the transverse medullary slice preparation used in some experiments. However, they did not rule out additional roles for pontine or spinal V2a neurons, and neither do the experiments reported here. Additional studies using transverse medullary slices or pontomedullary preparations from neonatal V2a-(Gi)DREADD mice could further test the importance of brainstem V2a neurons for respiratory rhythm generation as well as investigate the mechanism(s) by which excitatory drive from these neurons promotes frequent, regular inspiratory burst activity. Preparations from older animals could be used to further investigate whether

V2a neurons have a less significant impact on inspiratory rhythm generating networks at older ages or whether other neurons/structures compensate for the loss of excitatory drive from V2a neurons. We hypothesize that spinal V2a neurons may be responsible for patterning respiratory motor output, including control of ARM activity. However, we cannot rule out a role for brainstem neurons, such as reticulospinal V2a neurons, in either promoting or inhibiting ARM activity. Two approaches can be used to drive DREADD expression to only subsets of spinal or brainstem V2a neurons to test their roles in controlling respiratory rhythm and pattern: 1) intersectional genetic approaches that require Cre and Flp recombinases to drive DREADD expression (Ray et al. 2011) or 2) Cre dependent viruses to drive DREADD expression. In order to determine whether cervical spinal V2a neurons are sufficient to control respiratory muscle activity and enhance ventilation, we next used a Cre-dependent virus to target excitatory or inhibitory DREADD expression to only cervical spinal V2a neurons.

Chapter III.

Cervical Spinal V2a Neurons Control Respiratory

Muscle Activity and Increase Ventilation in

Healthy Mice

3.1 Introduction

We have shown that increasing (Romer et al. 2017) or decreasing (Jensen et al. 2019) the excitability of all brainstem and spinal cord V2a neurons increases ARM activity and enhances ventilation, but it is unknown which population of V2a neurons – brainstem or spinal – drives this respiratory activity.

It is well established that respiratory centers in the brainstem are critical for generating respiratory rhythm and providing descending brainstem drive to respiratory motor neurons (Ramirez and Anderson 2017; Del

Negro et al. 2018), but less is known about how spinal circuits may generate respiratory rhythm and pattern respiratory muscle activation.

Increasing or decreasing brainstem and spinal cord V2a excitability activates ARMs. We previously proposed a model to explain these results in which two subtypes of V2a neurons exist, one that participates in an excitatory spinal pathway to activate ARMs when they are needed and one that participates in an inhibitory pathway to constrain ARM activity when they are not needed. This model is based off of a lumbar spinal circuit in which one subtype of V2a neuron (Shox2+/Chx10+) provides excitatory drive to motor neurons on the same side of the spinal cord. On the other hand, a second type of V2a neuron (Shox2-/Chx10+) regulates locomotor motor neurons on the opposite side of the spinal cord by projecting to commissural interneurons (that cross the midline of the spinal cord) to coordinate left- right alternation of locomotion (Dougherty et al. 2013). Although ARMs on opposite sides of the spinal cord do not need to alternate activation, they do require precise, coordinated activation to enhance ventilation under conditions of high oxygen demand (Sieck and Gransee, 2012). Therefore, it is possible that two V2a neuron subtypes may exist in the cervical spinal cord to pattern their activity and enhance ventilation.

Although the functional roles of spinal neurons in the control of breathing are not well defined, there is increasing evidence that they are critical for patterning the activity of respiratory motor neurons.

For example, at least a subset of spinal neurons are innervated by inspiratory rVRG neurons or display

90 inspiratory-related phasic bursting, suggesting that they may relay, amplify, or modulate central respiratory drive to motor neurons (K. Z. Lee and Fuller 2011). Spinal neurons also play an important role in patterning output within a motor pool. For example, studies of the rat phrenic nucleus have identified inspiratory, expiratory, and tonic firing inhibitory (GABAergic) interneurons (Marchenko et al. 2015; Ghali

2018) and blocking this GABAergic inhibition by spinal neurons alters the shape of the phrenic neurogram, increasing activity during expiration (Marchenko et al. 2015). Excitatory and inhibitory respiratory neurons have also been identified in the thoracic cord of rats and cats and likely modulate extra-diaphragmatic muscle activity, such as the intercostals (Saywell et al. 2011; Iizuka, Onimaru, and Izumizaki 2016; Iizuka et al. 2018).

Together, these studies suggest that spinal neurons play important roles in shaping both phrenic and extra-diaphragmatic respiratory motor neuron output. Therefore, we hypothesize that increasing or decreasing the excitability of cervical spinal V2a neurons alone is sufficient to pattern respiratory muscle activity and enhance ventilation during ARM activity.

3.2 Materials and Methods

Animals

All animal procedures were performed according the National Institutes of Health guidelines and approved by the institution’s animal care committee’s regulations. Chx10Cre/+ mice were bred to express

Cre recombinase in Chx10+ V2a neurons (Azim et al., 2014; Romer et al., 2017). Genotyping was performed by PCR using primers to detect the Chx10Cre (5’-GCATT AGACA CCGGA GGG-3’ and 5’-GGACA GAAGC

ATTTT CCAG-3’) and Chx10 wild type (5’-GCATT AGACA CCGGA GGG-3’ and 5’-CTCCC GACTG TGACT TTCC-

3’).

Histology, Imaging and Quantification

Adult Chx10Cre/+ mice were transcardially perfused and their spinal cords and brains harvested as previously described (see Chapter II, Section 2.2). Immunohistochemistry was performed on adult tissue

91 in mice injected with AAV-hSYN-DIO-(Gq)DREADD-mCherry and AAV-hSYN-DIO-(Gi)DREADD-mCherry using the following antibodies (and DAPI nuclear stain): Chx10 (Sheep at 1:2000, abcam #161641, Lot

GR3197-1606-13), and Red Flourescent Protein (Rabbit at 1:25,000, Rockland, Lot 600-401-379). The following antibodies were used for mice injected with rAAV-hSYN-DIO-HA-hM4D(Gi)-IRES-mCitrene:

Chx10 (Sheep at 1:2000, abcam #16141) and HA tag (Rabbit at 1:1000, Cell Signaling #3724). The following antibodies were used for mice injected with AAV-hSYN-DIO-eGFP: Chx10 (Sheep at 1:2000, abcam#16141) and Green Flourescent Protein (GFP) (Rabbit at 1:5000, Life Technologies, A-6455). The number of AAV+ cells, Chx10+ cells, and AAV+/Chx10+ cells with DAPI+ nuclei were counted in every 8th section from C1-C8

(77-116 hemisections/animal) and in every 3rd section from the midbrain (hippocampus) to the hindbrain

(at the level of the nucleus ambiguous) (8-16 hemisections/animal). The total number of cells for each hemisection and spinal segment was calculated for each animal. Results are reported as mean ± SE.

Cervical spinal cord injections of Cre-dependent adeno-associated viruses

A Cre-dependent adeno-associated virus (AAV) was injected into the cervical spinal cord of postnatal day

Cre/+ age 14 (P14) Chx10 mice to target either the excitatory (Gq)DREADD receptor, inhibitory (Gi)DREADD receptor, or (non-DREADD) eGFP to cervical spinal V2a neurons only. Chx10Cre/+ mice, in which all Chx10+

V2a neurons express Cre-recombinase, were unilaterally or bilaterally injected with 1uL of the Cre- dependent AAV-hSYN-DIO-(Gq)DREADD-mCherry [Addgene, serotype AAV8, Lot v27924, titre 4x10e12 vg/ml], AAV-hSYN-DIO-(Gi)DREADD-mCherry [Addgene, serotype AAV8, Lot v27924, titre 4x10e12 vg/ml], rAAV-hSYN-DIO-HA-hM4D(Gi)-IRES-mCitrene [UNC Viral Core, serotype AAV8, Lot AV4930, titre

4x10e12vg/ml], or rAAV8-hSYN-DIO-eGFP [UC Vector Core, serotype AAV8, Lot AV4937, titre 8x10e12 vg/ml].

A glass capillary needle was connected to a 1uL syringe with a plastic tube that was pressurized with mineral oil. This glass capillary needle was mounted on a Stereotaxic frame. Mice were then induced with 3.5% isoflurane/2% oxygen anesthesia gas mixture and subsequently maintained at 1.5%

92 isoflurane/1% oxygen. Mice were shaved from the top of their head to their shoulders. The incision site was sterilized with chlorohexidine surgical scrub and 70% ethanol and placed on a heating pad in the

Stereotax. A 5cm incision was made from the base of the skull to the shoulders. The trapezius muscle was separated down the midline with a small incision and sterile cotton tipped swabs were used to carefully tease apart the remaining splenius capitis muscles to expose the spinal column. #5 forceps were used to carefully remove the vertebrae between C1 and C4 and expose the spinal cord. Any bleeding was staunched with a sterile pad of gauze before the injection. 2uL of virus was drawn into the glass capillary needle and positioned at the midline of C3/4. The needle was then moved 0.5mm mediolateral (ML) to the right. The needle was slowly lowered down until it punctured the dura and subsequently raised back to the surface of the spinal cord. The glass capillary needle was then re-lowered into the spinal cord 0.9mm dorsal-ventral (DV). Once the needle was in position and a visual check confirmed there was no blood being drawn into the needle, the virus was injected at a rate of 0.1uL/20s until 1uL was injected on the right side of the cord. The needle was left in the spinal cord for 2 minutes to prevent the virus being vacuumed back up into the capillary needle. The needle was then slowly raised at a rate of 0.2mm DV/20s.

The needle was returned to the midline and the procedure repeated on the left side for animals receiving bilateral injections. The two layers of muscle were separately sutured back together using absorbable suture (size 6.0) and a cyanoacrylate adhesive closed the skin incision. 0.05mL of the analgesic Carprofen was administered subcutaneously. Mice were placed in an incubator set at 32◦C for 30 minutes following surgery before being returned to their home cage. Mice were monitored for the next 48 hours.

Surgical Implantation of Telemetry Devices.

Telemetry implants were performed in P65 (and older) adult mice as described above (see Chapter II,

Section 2.2). Briefly, double channel telemetry (F20-EET) transmitters from Data Sciences International were implanted to chronically record EMG from respiratory muscles of Chx10Cre/+ mice injected with Cre- dependent AAV viruses. Three different sets of muscles were implanted. 1) Mice who received bilateral

AAV injections with a weight greater than 27g had biopotential leads implanted into the diaphragm and scalene muscles. 2) Mice who received bilateral AAV injections with a weight less than 27g had biopotential leads implanted into the trapezius and scalene muscles. 3) Mice who received unilateral AAV injections had the left and right scalene muscles implanted.

Adult Whole Body Plethysmography and EMG Acquisition in Conscious Animals. Plethysmography, EMG acquisition, and digital video recording was performed as described previously using DSI Ponemah

Physiology Platform Acquisition software v.5.20 (see Chapter II, section 2.2). Animals only received the optimal dose of CNO for the specific DREADD receptor expressed (1.0 mg/kg*bw for the (Gq)DREADD

(Romer et al., 2017) and 10.0 mg/kg*bw for the (Gi)DREADD) (Jensen et al., 2019). Non-DREADD expressing AAV-eGFP animals served as controls for both groups by receiving both 1.0 and 10.0 mg/kg*bw

CNO on different days. At least 48 hours passed in between each CNO trial.

Adult Plethysmography Analysis.

The Ponemah Physiology Platform Analysis software v.5.20 was used to analyze plethysmography as described previously (see Chapter II, section 2.2).

EMG Analysis in Conscious Animals.

The raw diaphragm and ARM EMG signals were filtered and analyzed as described previously (see Chapter

II, section 2.2).

Nasal Occlusion.

Nasal occlusion was performed to increase respiratory drive and increase diaphragm EMG peak amplitude and ventilation. A small piece of tape was placed over the nose of the mouse for four minutes to force the mouse to breathe through its mouth and increase central respiratory drive, which is reflected in the increased diaphragm EMG peak amplitude (Mantilla et al., 2011).

Heart Rate Analysis. ECG signal was analyzed before and after CNO treatment (during the same periods that plethysmography was analyzed) as described previously (see Chapter II, section 2.2).

3.3 Results

Cervical spinal V2a neurons express the excitatory (Gq)DREADD receptor in AAV-(Gq)DREADD mice.

Cre/+ Chx10 mice were bilaterally injected with a Cre-dependent AAV-hSYN-DIO-(Gq)DREADD- mCherry virus to target the excitatory (Gq)DREADD to cervical spinal V2a neurons. Injections targeted the mid-cervical spinal cord, but post-hoc analysis showed that injection sites varied from C4-C6 (n=4 mice at

C4, n=3 mice at C5, n=1 mouse at C6). A representative image of peak transduction at C4 is shown in

Figure 1A. The distribution of the percent of Chx10+ V2a neurons/section that expresses the AAV-

(Gq)DREADD virus is graphed in Figure 1D, where zero denotes the site of peak AAV expression (83-99 hemisections analyzed/animal). On average, 90 ± 5% of V2a neurons/section are AAV+ and express the excitatory (Gq)DREADD receptor. As the distance from the injection site decreases (both rostrally and caudally), so does the percent of V2a neurons that express the (Gq)DREADD virus.

Next, we analyzed the distribution of (Gq)DREADD expression at each cervical spinal segment. On

+ average, a total of 4,430 ± 45 Chx10 V2a neurons were transduced with the AAV-(Gq)DREADD virus (n=8).

Figure 1C shows the total number of Chx10+/AAV+ V2a neurons per spinal segment throughout the cervical cord, with C4 showing the highest transduction rate of 896 ± 134 Chx10+/AAV+ neurons. In contrast, C1 and C8 show low AAV expression, with transduction rates of 78 ± 39 AAV+ V2a neurons and 169 ± 43 AAV+

V2a neurons, respectively. On the other hand, not every AAV+ neuron is Chx10+, as 20.1 ± 4.2% of AAV+ neurons do not express Chx10 in the cervical spinal cord in AAV-(Gq)DREADD mice. However, all

AAV+/Chx10- neurons are found in the intermediate lamina where V2a neurons are typically found and are not seen in the ventral horn near respiratory motor neurons. These results are similar to what has been previously published in (Gi)DREADD expressing transgenic mice, and likely represent the population of V2a neurons that downregulate Chx10 expression during development (Jensen et al., 2019; Hayashi et al. 2018). Therefore, this subset of V2a neurons still expresses Chx10. The levels are simply too low to prevent detection by antibodies but still sufficient to drive Cre expression and cause recombination.

To verify that brainstem V2a neurons do not express the excitatory (Gq)DREADD receptor, we quantified the percent of V2a neurons that express AAV-(Gq)DREADD at the level of the nucleus ambiguous in the brainstem (Crone et al. 2012) (Figure 1B). Although projections from cervical spinal V2a

+ neurons can be seen in this area, less than 2% of all Chx10 V2a neuron cell bodies express the (Gq)DREADD receptor in the brainstem (1.3 ± 0.7 AAV+/Chx10+ V2a neurons/hemisection) (n=8). Furthermore, more rostral areas of the brain were analyzed for AAV+ cell bodies, but none were observed.

Figure 1. Cervical spinal V2a neurons express the excitatory (Gq)DREADD receptor in AAV-(Gq)DREADD mice. A Cre-dependent adeno- associated virus (AAV) was injected into the spinal cord between C4 and C6 of Chx10Cre/+ mice to target the excitatory (Gq)DREADD receptor to cervical spinal V2a neurons. (A) Representative image (20X magnification) of the spinal cord at C4 showing that the majority of Chx10+ V2a neurons (green) express the excitatory (Gq)DREADD receptor (red). The gray matter is outlined with a white dotted line. Scale bar = 100um. The inset (60X magnification) shows two AAV+ Chx10+ cells (white arrows). AAV+ neuronal processes (non-circular, thin lines) are also observed. Scale bar = 20um. (B) Representative image (20X) of the brainstem at the level of the nucleus ambiguous (NA) (dotted white circle) showing that very few Chx10+ V2a neurons (green) express the excitatory (Gq)DREADD receptor (red). Solid white arrow indicates one of the few AAV+/Chx10+ V2a neurons. Open white arrow indicates one of many AAV-/Chx10+ V2a neurons in the brainstem. The brainstem boundary is outlined with a white dotted line. Scale bar = 200um. The inset (60X) magnification shows one AAV+ Chx10+ cell (green and red) (white arrow) and multiple AAV- Chx10+ cells (green). Scale bar = 20um. (C) The total number of AAV+ Chx10+ cells were quantified at each spinal cervical spinal segment 1-8 (n=8). Gray dashed lines show individual animals. Red solid line shows the average ± SEM. (D) The percent of Chx10+ V2a neurons that express AAV was quantified per section and graphed as a function of the distance from the injection site (site of peak transduction), which is denoted as zero.

Cervical spinal V2a neurons express the inhibitory (Gi)DREADD receptor in AAV-(Gi)DREADD mice

Cre/+ Chx10 mice were bilaterally injected with a Cre-dependent rAAV-hSYN-DIO-HA-hM4D(Gi)-

IRES-mCitrene (n=2) or rAAV-hSYN-DIO-(Gi)DREADD-mCherry (n=5) to target the inhibitory (Gi)DREADD receptor to cervical spinal V2a neurons. The mid-cervical cord was targeted but post-hoc analysis showed

96 that injection sites varied from C4-C6 (n=3 mice at C4, n=1 mouse at C5, n=3 mice at C6). A representative image of peak transduction at C4 is shown in Figure 2A. The distribution of the percent of Chx10+ V2a neurons that express the AAV-(Gi)DREADD virus per section is graphed in Figure 2D, where zero denotes the site of peak AAV expression (77-116 hemisections analyzed/animal). At the site of peak AAV

+ expression, 100 ± 0% of Chx10 V2a neurons/section express the inhibitory (Gi)DREADD receptor (n=7).

The number of V2a neurons that express the inhibitory (Gi)DREADD receptor is drastically decreased as the distance from the injection increases. When the distribution of AAV expression was analyzed throughout the entire cervical cord, a total of 3,505 ± 305 Chx10+ cells were shown to express the

+ inhibitory (Gi)DREADD receptor. Figure 2C shows the total average number of AAV V2a neurons at each cervical spinal segment. The peak transduction is observed at C4, with 1,219 ± 50 AAV+/Chx10+ V2a neurons. In contrast, C1 and C8 show low AAV expression, with transduction rates of 7 ± 2 and 176 ± 37

+ AAV V2a neurons, respectively. Similar to AAV-(Gq)DREADD mice, 12.3 ± 0.9% of AAV+ neurons do not express Chx10 in the cervical spinal cord. However, all AAV+/Chx10- neurons are found in the intermediate lamina where V2a neurons are typically found and likely represent the class of V2a neurons that downregulate Chx10 expression during development (Hayashi et al. 2018).

Brainstem V2a neurons do not express the inhibitory (Gi)DREADD receptor (Figure 2B). Cervical spinal V2a neuron projections are observed at the level of the nucleus ambiguous, but less than 1% of V2a neuron cell bodies are transduced with the (Gi)DREADD receptor (n=5). More rostral brainstem and midbrain structures also do not show AAV expression in cell bodies.

Figure 2. Cervical spinal V2a neurons express the inhibitory (Gi)DREADD receptor in AAV-(Gi)DREADD mice. A Cre-dependent adeno-associated virus (AAV) was injected into the spinal cord between C4 and C6 of Chx10Cre/+ mice to target the inhibitory (Gi)DREAD receptor to cervical spinal V2a neurons. (A) Representative image (20X magnification) of the spinal cord at C4 showing that the majority of Chx10+ V2a neurons (green) express the excitatory (Gq)DREADD receptor (red). The gray matter is outlined with a white dotted line. Scale bar = 100um. The inset (60X magnification) shows two AAV+ Chx10+ cells (white arrows). AAV+ neuronal processes (non-circular, thin lines) are also observed. Scale bar = 20um. (B) Representative image (20X) of the brainstem at the level of the nucleus ambiguous (NA) (dotted white circle) showing that very few Chx10+ V2a neurons (green) express the inhibitory (Gi)DREADD receptor (red). Solid white arrow indicates one of the few AAV+/Chx10+ V2a neurons. Open white arrow indicates one of many AAV-/Chx10+ V2a neurons in the brainstem. The brainstem boundary is outlined with a white dotted line. Scale bar = 200um. The inset (60X) magnification shows one AAV+ Chx10+ cell (green and red) (white arrow) and multiple AAV- Chx10+ cells (green). Scale bar = 20um. (C) The total number of AAV+ Chx10+ cells were quantified at each spinal cervical spinal segment 1-8 (n=8). Gray dashed lines show individual animals. Blue solid line shows the average ± SEM. (D) The percent of Chx10+ V2a neurons that express AAV was quantified per section and graphed as a function of the distance from the injection site (site of peak transduction), which is denoted as zero.

Cervical spinal V2a neurons express eGFP in non-DREADD expressing AAV-eGFP mice

Chx10Cre/+ mice were bilaterally injected with a Cre-dependent AAV-hSYN-DIO-eGFP virus to target eGFP to cervical spinal V2a neurons and serve as non-DREADD expressing controls. Injection sites varied from

C4-C6 (n=3 mice at C4, n=1 mouse at C5, n=2 mice at C6). A representative image of peak transduction at

C4 is shown in Figure 3A. The distribution of the percent of Chx10+ V2a neurons that express the AAV- eGFP virus per section is graphed in Figure 3C, where zero denotes the site of peak AAV expression (81-

101 hemisections analyzed/animal). At the injection site, 97.8 ± 2% of V2a neurons/section express eGFP.

On average, a total of 3,416 ± 57 Chx10+ V2a neurons express the AAV-hSYN-DIO-eGFP throughout the cervical cord of Chx10Cre/+ mice (n=6). Figure 3B shows the total number of Chx10+/AAV+ V2a neurons per spinal segment throughout the cervical cord, with C5 showing the highest transduction rate of 791 ± 116

+ + AAV V2a neurons. Similar to AAV-(Gq)DREADD mice and AAV-(Gi)DREADD mice, 14.2 ± 2.2% of AAV

98 neurons do not express Chx10 in the cervical spinal cord and are found in the intermediate lamina of the spinal cord.

Figure 3. Cervical spinal V2a neurons express eGFP in non-DREADD expressing AAV-eGFP mice. A Cre-dependent adeno-associated virus (AAV) was injected into the spinal cord between C4 and C6 of Chx10Cre/+ mice to target enhanced green fluorescent protein (eGFP) to cervical spinal V2a neurons to serve as non-DREADD expressing controls. (A) Representative image (20X magnification) of the spinal cord at C4 showing that the majority of Chx10+ V2a neurons (red) express eGFP (green). The gray matter is outlined with a white dotted line. Scale bar = 100um. The inset (60X magnification) shows AAV+ Chx10+ cells (white arrows). AAV+ neuronal processes (non-circular, thin lines) are also observed. Scale bar = 20um. (B) The total number of AAV+ Chx10+ cells were quantified at each spinal cervical spinal segment 1-8 (n=8). Gray dashed lines show individual animals. The green solid line shows the average ± SEM. (C) The percent of Chx10+ V2a neurons that express AAV was quantified per section and graphed as a function of the distance from the injection site (site of peak transduction), which is denoted as zero.

Increasing or decreasing the excitability of cervical spinal V2a neurons increases scalene ARM activity.

We have previously shown that either increasing or decreasing the excitability of all brainstem and spinal cord V2a neurons increases ARM activity. We proposed a model in which two subtypes of V2a neurons work together to tightly constrain ARM activity, based off the organization of V2a neurons in the lumbar spinal cord (Dougherty et al. 2013). Therefore, we tested whether a similar cervical spinal V2a neurons alone regulate ARM activity at rest by measuring ARM EMG before and after increasing or decreasing cervical spinal V2a excitability.

Similar to increasing brainstem and spinal cord V2a excitability (Figure 4E), increasing cervical spinal V2a excitability alone is sufficient to increase scalene muscle activity from 0.3 scalene bouts/min. pre CNO to 10.1 bouts/min. following 1.0 mg/kg*bw CNO (n=9, p<0.001) (representative traces in Figure

4A-B and quantified in Figure 4C). In non-DREADD AAV-eGFP controls, there is not a change in scalene activity before and after treatment with 1.0 mg/kg*bw CNO (n=6, p=0.220) (Figure 4D). Since the number of AAV+ V2a neurons was variable among animals, we plotted the total number of AAV+/Chx10+ V2a neurons in the cervical cord against scalene bout frequency post CNO for each animal to determine if the magnitude of ARM activation is correlated with transduction efficiency. We did not detect a significant

2 correlation between the number of cells transduced with the (Gq)DREADD and bout frequency (r = 0.0329, p=0.951) (n=9) or the number of cells transduced with eGFP and bout frequency (r2 = 0.0858 , p=0.872)

(Figure 4F). However, these data show that transduction efficiency varies from animal to animal. Since a high viral transduction efficiency can be achieved, we excluded any animals (n=1) from further analyses that do not have at least 1500 AAV+ V2a neurons within the cervical cord.

Interestingly, increasing all brainstem and spinal cord V2a neurons vs. cervical spinal V2a neurons results in different patterns of ARM activation. Activating brainstem and spinal V2a neurons causes bouts of scalene EMG activity that spans multiple breaths (arrhythmic) (Romer et al., 2017), whereas activating only cervical spinal V2a neurons causes both arrhythmic (Figure 4A) and rhythmic activation of the scalene muscle (Figure 4B). 3/8 AAV-(Gq)DREADD animals show a mix of arrhythmic and rhythmic ARM activation and 5/8 AAV-(Gq)DREADD animals only show arrhythmic ARM activation. However, there does not appear to be a relationship between the injection site, total number of AAV+ V2a neurons in the cervical spinal cord, or the total number of AAV+ V2a neurons at C4 near the phrenic nucleus and the presence of rhythmic ARM activity. Additionally, it is unlikely that the amount of muscle surrounding the electrode impacts the presence of rhythmic ARM activation, since both the trapezius and scalene muscle can show synchronous rhythmic ARM activation when recorded together in the same animal. These data suggest

100 that although increasing brainstem and spinal V2a neurons and cervical spinal only V2a neurons both activate ARMs, they may play different roles in patterning ARM activity.

Similar to decreasing brainstem and spinal cord V2a excitability (Figure 4J), silencing cervical spinal V2a neurons alone increases ARM activity from 0.3 scalene bouts/min. pre CNO to 2.6 bouts/min. following treatment with 10.0 mg/kg*bw CNO (n=8, p=0.008) (representative trace in Figure 4G and quantified in Figure 4H). The correlation plot of the #AAV+/Chx10+ cells vs. scalene bout frequency did not show a significant correlation (r2 = 0.0775, p=0.869) (n=7). There were not any animals excluded from analysis due to low transduction. However, one animal was excluded because damage to the spinal cord while harvesting tissue prevented accurate analysis of viral expression. Unlike activating V2a neurons, silencing brainstem/spinal vs. just cervical spinal V2a neurons both result in arrhythmic ARM activation that typically spans multiple breaths. Therefore, although the pattern of ARM activation is different, both increasing and decreasing cervical spinal V2a excitability is sufficient to increase ARM activity.

Cervical spinal V2a neurons are sufficient but not required for diaphragm activity.

It is known that some cervical spinal V2a neurons project to phrenic motor neurons that control the diaphragm (Zholudeva et al. 2017). Therefore, a subset of mice had recording electrodes inserted into the scalene and diaphragm muscle so we could analyze whether altering cervical spinal V2a excitability altered diaphragm activity. Interestingly, increasing cervical spinal V2a excitability increases diaphragm

RMSpeak amplitude by 18 ± 3% (n=5, p=0.018) even in the absence of ARM activity and is further increased by another 18 ± 3% (p=0.013) during bouts of scalene activity (representative trace in Figure 5A and quantified in Figure 5B). This increase in diaphragm RMSpeak amplitude differs from the effect of increasing brainstem and spinal cord V2a excitability, where CNO does not have an effect on diaphragm RMSpeak amplitude (n=3, p=0.101) (Figure 5E). It is unlikely expression of the AAV vector itself or treatment with

101

Figure 4. Increasing or decreasing the excitability of cervical spinal V2a neurons increases scalene ARM activity. Whole breath plethysmography (WBP) and electromyography (EMG) were recorded from animals injected with AAV into the cervical spinal cord to measure ventilation and ARM activity, respectively, before and after altering cervical spinal V2a excitability. (A) Representative trace showing a lack of scalene ARM activity following vehicle treatment (-CNO). Increasing the excitability of cervical spinal V2a neurons increases ARM activity. Two distinct patterns of scalene ARM activation is observed. Bouts of ARM activity that span multiple breaths are shown in (A) (+CNO, gray box outlines scalene EMG activity). A representative trace of rhythmic ARM activation alternating with a bout of scalene activity that spans multiple breaths is shown in (B). (C) Scalene bout frequency before and after treatment with 1.0 mg/kg*bw CNO is quantified (n=9 animals). (D) Scalene bout frequency in non- DREADD expressing AAV-eGFP mice was quantified before and after treatment with 1.0 mg/kg*bw CNO (n=6). (E) The effect of increasing brainstem and spinal cord V2a neurons in V2a-(Gq)DREADD mice (black line) vs. increasing cervical spinal excitability in AAV-(Gq)DREADD mice (gray line) on scalene activity is compared. (F) The amount of total AAV+ V2a neurons in the cervical cord vs. scalene bout frequency following treatment with 1.0 mg/kg*bw CNO is graphed for AAV-(Gq)DREADD animals (red dots) and AAV-eGFP animals (green dots). (G) Representative trace showing a lack of scalene ARM activity following vehicle treatment (-CNO). Decreasing the excitability of cervical spinal V2a neurons increases ARM activity. Only one pattern of ARM activation is observed (arrhythmic) (+CNO) (gray box outlines scalene EMG activity). (H) Scalene bout frequency before and after treatment with 10.0 mg/kg*bw CNO is quantified (n=7 animals). (I) Scalene bout frequency in non-DREADD expressing AAV-eGFP mice was quantified before and after treatment with 10.0 mg/kg*bw CNO (n=5). (J) The effect of decreasing brainstem and spinal cord V2a neurons in V2a-(Gi)DREADD mice (black line) vs. decreasing cervical spinal excitability in AAV-(Gi)DREADD mice (gray line) is compared. (F) The amount of total AAV+ V2a neurons in the cervical cord vs. scalene bout frequency following treatment with 10.0 mg/kg*bw CNO is graphed for AAV-(Gq)DREADD animals (blue dots) and AAV-eGFP animals (green dots). Gray dashed lines indicated individual animals. Black solid lines show the average ± SEM. Paired t-tests were used to analyze C, D, H, and I. A Pearson’s correlation was used to analyze F and K. (*)p<0.05.

102

CNO itself causes this change because there is no change in diaphragm RMSpeak amplitude in AAV-eGFP mice following treatment with 1.0 mg/kg*bw CNO (Wilcoxin signed-rank, n=4, p=0.100) (Figure 5C), suggesting that increased cervical spinal V2a excitability enhances diaphragm function. However, even though we do not detect an increase in diaphragm EMG peak amplitude in AAV-eGFP mice following CNO treatment, the F20-EET transmitters can detect changes in diaphragm activity as shown by the increased diaphragm RMSpeak amplitude following nasal occlusion (n=3, p<0.001) (Figure 5D).

Previous studies in V2a-(Gi)DREADD mice suggest that silencing brainstem and spinal cord V2a neurons does not activate ARMs as a compensatory mechanism for impaired diaphragm function because diaphragm RMSpeak amplitude is not altered following CNO treatment during interbout intervals or during

ARM bouts. Since silencing cervical spinal V2a neurons also activates ARMs, we assessed diaphragm activity following CNO treatment. Similar to V2a-(Gi)DREADD mice (Figure 5J), silencing cervical spinal V2a neurons does not alter diaphragm RMSpeak amplitude, even during ARM bouts (one-way repeated measures ANOVA, n=5, p=0.152) (Figure 5G). However, we can detect a significant increase in diaphragm

RMSpeak amplitude during nasal occlusion (paired t-test, n=5, p<0.001) (Figure 5I). Non-DREADD expressing

AAV-eGFP control mice do not show a change in diaphragm RMSpeak amplitude following treatment with

10.0 mg/kg*bw CNO (paired t-test, n=3, p=0.816) (Figure 5H).

Therefore, even though increasing or decreasing the excitability of cervical spinal V2a neurons increases scalene ARM activity, they have different effects on the diaphragm. Activating spinal V2a neurons increases diaphragm EMG peak amplitude during interbout intervals and ARM activity, whereas silencing cervical spinal V2a neurons does not have an effect on diaphragm EMG peak amplitude, suggesting that cervical spinal V2a neurons not required for (but can enhance) diaphragm function.

103

Figure 5. Cervical spinal V2a neurons are sufficient but not required for diaphragm activity. (A) WBP and EMG were recorded from animals injected with AAV into the cervical spinal cord to measure ventilation, diaphragm and scalene EMG activity before and after increasing cervical spinal V2a excitability. The gray box outlines a bout of scalene activity. (B) Diaphragm RMSpeak amplitude normalized to the maximal amplitude reached during sighs was quantified before 1.0 mg/kg*bw CNO treatment (-CNO/-Bout), after CNO treatment during interbout intervals (+CNO/- Bout), and after CNO treatment during bouts of scalene activity (+CNO/+Bout) in AAV-(Gq)DREADD animals (n=5). Diaphragm RMSpeak amplitude was not measured during bouts in non-DREADD expressing AAV-eGFP animals (C) due the rarity of scalene bouts (n=4). (D) Diaphragm RMSpeak amplitude was measured in AAV-eGFP animals (n=3) before (eupnea) and after nasal occlusion. (E) The effect of increasing brainstem and spinal cord V2a neurons in V2a-(Gq)DREADD mice (black line) vs. increasing cervical spinal excitability in AAV-(Gq)DREADD mice (gray line) on diaphragm RMSpeak amplitude is compared. (F) WBP and EMG were recorded from animals injected with AAV into the cervical spinal cord to measure ventilation, diaphragm and scalene EMG activity before and after decreasing cervical spinal V2a excitability. The gray box outlines a bout of scalene activity. (G) Diaphragm EMG peak amplitude normalized to the maximal amplitude reached during sighs was quantified before 10.0 mg/kg*bw CNO treatment (-CNO/-Bout), after CNO treatment during interbout intervals (+CNO/-Bout), and after CNO treatment during bouts of scalene activity (+CNO/+Bout) in AAV-(Gq)DREADD animals (n=5). Diaphragm RMSpeak amplitude was not measured during bouts in non-DREADD expressing AAV-eGFP animals (H) due the rarity of scalene bouts (n=4). (I) Diaphragm RMSpeak amplitude was measured in AAV-(Gi)DREADD animals (n=5) before (eupnea) and after nasal occlusion. (J) The effect of increasing brainstem and spinal cord V2a neurons in V2a-(Gi)DREADD mice (black line) vs. increasing cervical spinal excitability in AAV-(Gi)DREADD mice (gray line) on diaphragm RMSpeak amplitude is compared. Gray dashed lines indicated individual animals. Black solid lines show the average ± SEM. A one-way repeated measures ANOVA was used to analyze B and G. Paired t-tests were used to analyze C, D, H, and I. (*) Group is different than -CNO/-bout (vehicle control); (┼) +CNO/+bout is different than +CNO/-bout, p<0.05.

104

Increasing the excitability of cervical spinal V2a neurons increases ventilation during bouts of ARM activity.

Our previous studies have shown that increasing brainstem and spinal cord V2a excitability enhances ventilation during ARM activity as well during interbout intervals (absence of ARM activity)

(Romer et al. 2017). These results were not surprising because 1) it is known that brainstem V2a neurons project to respiratory centers (i.e. the PreBÖtzinger Complex and rVRG) that control respiratory frequency and brainstem drive and 2) the function of ARMs is to enhance ventilation. Less is known about how spinal circuits and spinal V2a neurons control ventilation, so we analyzed whole breath plethysmography (WBP) to assess the impact of increasing the excitability of cervical spinal V2a neurons on ventilation.

We first focused our analysis on interbout intervals to exclude the potential effects on ventilation caused by ARM recruitment (non-shaded areas in Figure 6). There is no change in PIF (p=0.124) (Figure

6C), VT (p=0.060) (Figure 6F), f (p=0.055) (Figure 6I), MV (p=0.882) (Figure 6L), Ti (p=0.980) (Figure 6O), or

VT /Ti (p=0.590) (Figure 6R) before and after CNO treatment during interbout intervals. The expression of the (Gq)DREADD in cervical spinal V2a neurons alone does not affect ventilation since there is no difference between AAV-(Gq)DREADD mice and AAV-eGFP mice in the following baseline ventilation values: PIF (AAV-(Gq)DREADD: 2.4 ± 0.1 vs AAV-eGFP: 2.4 ± 0.1, p=0.900), VT (AAV-(Gq)DREADD: 0.16 ±

0.01 vs AAV-eGFP: 0.16 ± 0.01, p=0.788), MV (AAV-(Gq)DREADD: 31.1 ± 2.1 vs AAV-eGFP: 31.6 ± 2.4, p=0.874), f (AAV-(Gq)DREADD: 195 ± 7 vs AAV-eGFP: 202 ± 9, p=0.601), or Ti (AAV-(Gq)DREADD: 114 ± 4 vs

AAV-eGFP: 109 ± 5, p=1.000). Moreover, neither AAV transduction nor CNO treatment alters ventilation because there is no change in PIF (p=0.361) (Figure 6D), VT (p=0247) (Figure 6G), MV (p=0.278) (Figure

6M), f (p=0.697) (Figure 6J), Ti (p=0.632) (Figure 6P), or VT /Ti (p=0.318) (Figure 6S) in AAV-eGFP mice following treatment with 1.0 mg/kg*bw CNO. Therefore, in contrast to what is observed after increasing brainstem and spinal V2a excitability, increasing cervical spinal V2a excitability alone does not alter baseline ventilation.

105

We next assessed the impact of ARM activation on ventilation. As expected, there was a 23.7% increase in PIF (p=0.005) (Figure 6C, gray box), 13.3% increase in VT (p=0.111) (Figure 6F, gray box),

12.0% in f (p=0.034) (Figure 6I, gray box), 24.3% increase in MV (p=0.055) (Figure 6L, gray box), and

28.7% increase in VT /Ti (p=0.002) (Figure 6R, gray box) that is partially driven by an 11.1% decrease in Ti

(p=0.029) (Figure 6O, gray box). Thus, increasing cervical spinal V2a excitability enhances ventilation during ARM bouts.

Figure 6. Increasing the excitability of cervical spinal V2a neurons increases ventilation during bouts of ARM activity. (A) Representative trace showing WBP and scalene EMG following injection of vehicle saline. (B) Representative trace showing WBP and scalene EMG after injection of 1.0 mg/kg*bw CNO. (C) Peak inspiratory flow (PIF), (F) tidal volume (VT), (I) breathing frequency (BPM), (L) minute volume (MV), (O) inspiratory time (Ti), and (R) VT/Ti were measured within the same AAV-(Gq)DREADD mice (n=8) during interbout intervals in the absence of CNO, during interbout intervals with 1.0 mg/kg*bw CNO, and during scalene bouts with 1.0 mg/kg*bw CNO (n=8). Similarly, PIF (D), Vt (G), BPM or f (J), MV (M), Ti (P), and Vt/Ti (S) were measured in non-DREADD expressing AAV-eGFP control mice (n=6). The low frequency of ARM activity in non- DREADD controls precluded measurement of WBP during ARM bouts. The effect of increasing brainstem and spinal V2a excitability in V2a- (Gq)DREADD mice (gray lines) vs. cervical spinal V2a neurons only in AAV-(Gq)DREADD mice (black lines) is compared for PIF (E), Vt (H), BPM or f (K), MV (N), Ti (Q) and Vt/Ti (T). (*) Group is different than -CNO/-bout (vehicle control); (┼) +CNO/+bout is different than +CNO/-bout (p<0.05), one-way repeated measures ANOVA with all comparisons Holm-Sidak post-hoc test for V2a-(Gi)DREADD mice or paired t-test for non-DREADD

106 controls). A two-way repeated measures ANOVA was used to analyze E, H, K, N, Q, and T. Gray dashed lines show individual animals. Solid black line shows the mean ± SE.

Decreasing the excitability of cervical spinal V2a neurons increases ventilation.

We analyzed WBP to 1) assess the impact of decreasing the excitability of cervical spinal V2a neurons on ventilation and 2) determine if we observe the same effects as seen after silencing all brainstem and spinal cord V2a neurons. We first focused our analysis on interbout intervals to exclude the potential effects on ventilation caused by ARM recruitment (non-shaded areas in Figure 7). There is no change in PIF (11.4%; p=0.511) (Figure7C), VT (10.8%; p=0.329) (Figure 7F), MV (8.0%; p=0.414) (Figure

7L), or ventilatory drive VT/Ti (0.5%; p=0.977) (Figure 7R). Interestingly, we observe a significant 9.7% increase in breathing frequency during eupnea (p=0.003) (Figure 7I) and a corresponding significant 6.3% decrease in inspiratory time (p=0.046) (Figure 7O) after silencing cervical spinal V2a neurons. The expression of the (Gi)DREADD in cervical spinal V2a neurons alone does not appear to affect ventilation parameters since there is no difference between AAV-(Gi)DREADD mice and AAV-eGFP mice in the following baseline ventilation values: PIF (AAV-(Gi)DREADD: 2.6 ± 0.2 vs AAV-eGFP: 2.6 ± 0.1, p=0.316), VT

(AAV-(Gi)DREADD: 0.17 ± 0.01 vs AAV-eGFP: 0.16 ± 0.01, p=0.316, p=0.513), MV (AAV-(Gi)DREADD: 31.8

± 2.9 vs AAV-eGFP: 33.6 ± 2.0, p=0.959), f (AAV-(Gi)DREADD: 199 ± 3 vs AAV-eGFP: 207 ± 7, p=0.823), or

Ti (AAV-(Gi)DREADD: 111 ± 2 vs AAV-eGFP: 109 ± 5, p=0.171). Moreover, the expression of AAV in Chx10+ neurons itself does not alter ventilation in response to CNO since there is no change observed in any ventilatory parameters in AAV-eGFP (non-DREADD expressing controls) following administration of 10.0 mg/kg*bw CNO (Figure 7D, G, M, J, P and S).

We next assessed the impact of increased ARM activity on breathing by comparing ventilation during bouts of scalene activity to ventilation during interbout intervals in AAV-(Gi)DREADD mice implanted with leads in the diaphragm and scalene. There is a 66.1% increase in PIF (p<0.001) (Figure 7C, gray box), 28.9% increase in VT (p=0.002) (Figure 7F, gray box), 57.9% increase in MV (p<0.001) (Figure 7L,

107 gray box), 19.5% increase in f (p<0.001) (Figure 7I, gray box), and 71.7% increase in VT /Ti (p=0.001) (Figure

7R, gray box). A modest but significant 17.1% decrease was observed in inspiratory time (p<0.001) (Figure

7O, gray box) that is consistent with the increase in breathing frequency. We did not attempt to quantify ventilation during ARM bouts in AAV-eGFP control mice due to the infrequent occurrence of ARM activity.

It is important to note that decreasing the excitability of cervical spinal V2a neurons recapitulates the changes in ventilation during ARM bouts that we observe when silencing all brainstem and spinal cord

V2a neurons; however, there are notable differences produced when comparing the effect of CNO treatment on ventilation during eupnea pre to post CNO. Namely, silencing all V2a neurons increases PIF, f, MV, and VT/Ti whereas no change is observed after silencing cervical spinal V2a neurons only in AAV-

(Gi)DREADD animals. Surprisingly, AAV-(Gi)DREADD mice show an increase in breathing frequency during interbout intervals, despite the distal location of their cell bodies from respiratory centers in the brainstem.

Cervical spinal V2a neurons project to respiratory centers in the brainstem

It has been shown that upper cervical neurons may influence brainstem respiratory centers via ascending projections (Jones et al. 2012). V2a interneurons have been shown to project into the brainstem

(Azim et al., 2014; Menelaou and McLean 2014; Pivetta et al. 2014). Therefore, it is possible cervical spinal

V2a neurons may modulate ventilation via have ascending projections to respiratory brainstem centers.

Although less than 2% of brainstem V2a neuronal bodies express the excitatory DREADD receptor in AAV-

(Gq)DREADD and less than 1% of brainstem V2a neuronal bodies express the inhibitory (Gi)DREADD receptor in AAV-(Gi)DREADD mice, there are projections from cervical spinal V2a neurons observed near respiratory centers in the brainstem that control breathing frequency. These structures include the parafacial respiratory group (pFRG) (Figure 8A), the PreBÖtzinger Complex (Figure 8B), and the parabrachial nucleus/Kolliker Fuse nucleus (Figure 8C). Thus, it is possible that cervical spinal V2a neurons regulate baseline breathing via ascending projections to respiratory centers in the brainstem.

108

Figure 7. Decreasing the excitability of cervical spinal V2a neurons increases ventilation during bouts of ARM activity. (A) Representative trace showing WBP and scalene EMG following injection of vehicle saline in an AAV-(Gi)DREADD mouse. (B) Representative trace showing WBP and scalene EMG after injection of 10.0 mg/kg*bw CNO in the same AAV-(Gi)DREADD mouse. (C) Peak inspiratory flow (PIF), (F) tidal volume (VT), (I) breathing frequency (BPM), (L) minute volume (MV), (O) inspiratory time (Ti), and (R) VT/Ti were measured within the same AAV-(Gi)DREADD mice (n=8) during interbout intervals in the absence of CNO, during interbout intervals with 1.0 mg/kg*bw CNO, and during scalene bouts with 10.0 mg/kg*bw CNO (n=7). Similarly, PIF (D), Vt (G), BPM or f (J), MV (M), Ti (P), and Vt/Ti (S) were measured in non-DREADD expressing AAV-eGFP control mice before and after treatment with 10.0 mg/kg*bw CNO (n=5). The low frequency of ARM activity in non-DREADD controls precluded measurement of WBP during ARM bouts. The effect of decreasing brainstem and spinal V2a excitability in V2a-(Gi)DREADD mice (gray lines) vs. cervical spinal V2a neurons only in AAV-(Gi)DREADD mice (black lines) is compared for PIF (E), Vt (H), BPM or f (K), MV (N), Ti (Q) and Vt/Ti (T). (*) Group is different than -CNO/-bout (vehicle control); (┼) +CNO/+bout is different than +CNO/-bout (p<0.05, one-way repeated measures ANOVA with all comparisons Holm-Sidak post-hoc test for V2a-(Gi)DREADD mice or paired t-test for non-DREADD controls). A two-way repeated measures ANOVA was used to analyze E, H, K, N, Q, and T. Gray dashed lines show individual animals. Solid black line shows the mean ± SE.

109

(

Figure 8. Cervical spinal V2a neurons project to respiratory centers in the brainstem. Brains of Chx10Cre/+ mice injected with an AAV-(Gq)DREADD virus into the cervical spinal cord were analyzed to determine if cervical spinal V2a neurons project to respiratory centers in the brainstem. Cervical spinal V2a neuron projections are observed in (A) the parafacial respiratory group (pFRG), (B) PreBotzinger Complex, and (C) the Kolliker Fuse (KF)/Parabrachial nucleus (PB).

Altering cervical spinal V2a excitability does not change heart rate

Cardiovascular function can alter breathing. Therefore, it is possible that altering V2a excitability indirectly increases breathing frequency during interbout intervals by increasing cardiac output. We estimated the heart rate by counting the frequency of electrocardiogram (ECG) waves from the EMG signal but observed no change in heart rate during interbout intervals after activating cervical spinal V2a neurons (n=5, p=0.609) (Figure 9A) or silencing cervical spinal V2a neurons (n=5, p=0.0527) (Figure 9C).

Similarly, no change was observed in AAV-eGFP mice treated with 1.0 mg/kg*bw CNO (n=4, p=0.693)

(Figure 9B) or 10.0 mg/kg*bw (n=4, p=0.311) (Figure 9D). Therefore, increasing or decreasing cervical spinal V2a excitability does not change heart rate. Animals that did not show a clear ECG signal were excluded from analyses.

110

Figure 9. Altering cervical spinal V2a excitability does not change heart rate. Heart rate was measured using electrocardiogram (ECG) signal from the EMG traces recorded from (A) AAV-(Gq)DREADD mice (n=4) and (B) AAV-eGFP (n=4) before and after treatment with 1.0 mg/kg*bw CNO and (C) AAV-(Gi)DREADD mice and (D) AAV-eGFP before and after treatment with 10.0 mg/kg*bw CNO. Animals were excluded if the ECG signal was not strong enough to accurately analyze. Gray dashed lines show individual animals. Solid black line shows the mean ± SE.

Increasing the excitability of V2a neurons on one side of the spinal cord bilaterally activates the right and left scalene muscles.

Ipsilaterally projecting interneurons (that do not cross the midline of the spinal cord) can influence motor neurons on the opposite side of the spinal cord by projecting to commissural interneurons (e.g. V0 and some V3 interneurons) that do cross the midline of the spinal cord. It has been specifically shown that one subtype of ipsilaterally projecting V2a neurons (Chx10+/Shox2-) synapses onto an excitatory commissural interneuron in the lumbar spinal cord to control motor neuron activity on the opposite side of the spinal cord (Dougherty et al., 2013). Therefore, we hypothesized that cervical spinal V2a neurons may also project to commissural interneurons. The excitatory AAV-(Gq)DREADD virus was unilaterally injected into one side of the cervical spinal cord. These animals were then implanted with F20-EET transmitters to record EMG from the contralateral and ipsilateral scalene muscles to determine if activating or silencing V2a neurons on one side of the spinal cord can activate the scalene muscle on the contralateral side.

Figure 10A shows that AAV-(Gq)DREADD expression was confined to the ipsilateral side of the spinal cord. 1/6 animals had to be excluded because AAV expression was observed on both sides of the spinal cord. A representative trace in Figure 10B shows that increasing V2a excitability on the right side

111 of the spinal cord increases left (p=0.0506) (n=5) and right scalene bout frequency (p=0.0703) (n=5)

(quantified in Figure 10C), with 95.6 ± 1.8% of ipsilateral scalene bouts co-activated with contralateral scalene activity and 100% of contralateral scalene bouts co-activated with ipsilateral scalene activity.

Thus, these data demonstrate that increasing the excitability of ipsilaterally projecting cervical spinal V2a neurons on one side of the spinal cord is sufficient to bilaterally activate the right and left scalene muscle.

Figure 10. Increasing the excitability of cervical spinal V2a neurons on side of the spinal cord bilaterally activates the right and left scalene muscles. (A) Representative image (20X) showing that only cervical spinal V2a neurons on one side of the spinal cord (ipsilateral to the injection site) express the excitatory (Gq)DREADD receptor following a unilateral injection of 1.0 uL AAV-hSYN-DIO-(Gq)DREADD-mCherry. Scale bar = 200um. The insets show representative images (60X) of Chx10+ V2a neurons on the ipsilateral (left) and contralateral (right) sides of the spinal cord relative to the injection site. Scale bar = 20um. (B) Representative trace showing a lack of scalene ARM activity following vehicle treatment (-CNO) in the ipsilateral (top EMG trace) and contralateral (bottom EMG trace) scalene muscles in mice unilaterally injected with the AAV-hSYN- DIO-(Gq)DREADD-mCherry virus. Increasing the excitability of cervical spinal V2a neurons on one side of the spinal cord with 1.0 mg/kg*bw (+CNO) synchronously activates the scalene muscle on the ipsilateral and contralateral side (relative to the site of the AAV injection) (scalene bouts are outlined with a gray box). (C) Scalene bout frequency is quantified for the ipsilateral (white bars) and contralateral (black bars) scalene muscles before (-CNO) and after (+CNO) treatment. Data were analyzed with paired t-tests, (*)p<0.05.

112

Figure 11. Hypothetical model of cervical spinal V2a neuron subtypes and their organization in respiratory circuits. We propose that Type I V2a neurons are typically inactive at rest and participate in excitatory pathways to control diaphragm and ARM activity by providing excitatory drive to phrenic motor neurons and ARM motor neurons, respectively. Type II V2a neurons may be active at rest and only participate in inhibitory pathways to constrain the activation of ARMs at rest. Furthermore, ipsilaterally projecting cervical spinal V2a neurons on one side of the spinal cord may control ARM activity on the opposite side of the spinal cord by projecting to commissural interneurons that cross the midline of the spinal cord. This allows for synchronized and efficient control of respiratory muscles.

3.4 Discussion

These data show that cervical spinal V2a neurons are sufficient to modulate respiratory muscle activity and enhance ventilation. Specifically, activating cervical spinal V2a excitability increases diaphragm activity, scalene ARM activity, and enhances ventilation during ARM bouts. On the other hand, silencing cervical spinal V2a neurons activates ARMs without altering diaphragm activity and enhances ventilation during ARM bouts. Unexpectedly, we show that silencing cervical spinal V2a neurons even increases breathing frequency in the absence of ARM bouts. There was no noticeable change in heart rate after silencing V2a neurons, but cervical spinal projections to respiratory brainstem centers were observed, suggesting that cervical spinal V2a neurons may modulate respiratory frequency via ascending supraspinal projections. Finally, preliminary data shows that increasing V2a excitability on one side of the cervical spinal cord can activate both the left and right scalene muscle, suggesting V2a neurons may also project to commissural interneurons to coordinate synchronous activation of ARMs.

These data show that both increasing or decreasing cervical spinal V2a neurons alone increases

ARM activity. We previously proposed a model in which two subtypes of V2a neurons work together to constrain ARM activity. This model is based off of Dougherty et al. 2013, in which two subtypes of V2a

113 neurons participate in a lumbar spinal circuit to regulate locomotor motor neuron activity. In previous studies (in which either the excitatory or inhibitory DREADD receptor was expressed in all brainstem and spinal cord V2a neurons) (Jensen et al., 2019), it was not possible to rule out a role for brainstem neurons in either promoting or inhibiting ARM activity. For example, reticulospinal V2a neurons have been shown to halt locomotion (Bouvier et al. 2015). Therefore, this same set of neurons may also actively prevent the activation of ARMs in the cervical cord when they are not needed. However, the present study shows that increasing or decreasing cervical spinal V2a excitability still activates ARMs, even in the absence of altered brainstem V2a excitability, suggesting that their ability to tightly regulate ARM activity can be confined to

V2a neurons whose cell bodies are located in the cervical spinal cord. Future studies should target

DREADDs to brainstem V2a neurons to specifically test the role of brainstem V2a neurons in controlling

ARM activity. Interestingly, there was not a correlation between the number of cervical spinal V2a neurons infected with the excitatory or inhibitory DREADD receptor and bout frequency following CNO treatment.

This may be due to inter-animal variability regarding the injection site. Alternatively, different numbers of the proposed Type I vs. Type II V2a neurons may have been transduced with DREADDs. There may also be a ceiling effect, where the maximum number of V2a neurons controlling respiratory muscle activity expresses DREADDs, and the additional cells expressing DREADDs are not incorporated into respiratory circuitry. Instead, these “extra” cervical spinal V2a neurons may control skilled movements or locomotion

(Azim et al. 2014). Therefore, increasing the excitability of more non-respiratory V2a neurons would not increase ARM activity further.

Although increasing and decreasing cervical spinal V2a excitability both activates ARMs, they have different effects on the diaphragm. First, we show that increasing cervical spinal V2a excitability slightly increases diaphragm RMSpeak amplitude during interbout intervals (in the absence of ARM activity). This is not surprising because we know that V2a neurons are synaptically connected to phrenic motor neurons that control the diaphragm (Zholudeva et al. 2017), and may relay, amplify, or modulate descending

114 brainstem input into phrenic motor neurons like other interneurons (Lee and Fuller 2011). At least three different types of connections can explain these results: 1) cervical spinal V2a neurons may act as phrenic motor neuron pre-amplifiers for rhythmic supraspinal input 2) spinal V2a neurons provide tonic excitatory drive to phrenic motor neurons and descending brainstem drive regulates respiratory rhythm generation or 3) V2a neurons provide intrinsic rhythmic drive to phrenic motor neurons. V2a neuron firing properties are heterogenous, as they can fire rhythmically or tonically in the lumbar spinal cord (Zhong et al. 2010;

Dougherty and Kiehn 2010), but electrophysiology is required to investigate intrinsic firing properties of cervical spinal V2a neurons. However, since tidal volume and respiratory drive (VT/Ti) remained unchanged during interbout intervals, this increased diaphragm RMSpeak amplitude may not be biologically relevant enough to enhance ventilation. We also do not observe a change in breathing frequency, suggesting that exciting cervical spinal V2a neurons does not affect the respiratory rhythm generator in the absence of ARM activity.

In contrast to interbout intervals, diaphragm RMSpeak amplitude is substantially increased during

ARM bouts after increasing cervical spinal V2a excitability. Associated with this increase is enhanced ventilation, including increased PIF, Vt, f, MV, and VT/Ti. However, it is not possible to determine which respiratory muscle – diaphragm or scalene – contributes more to this enhanced ventilation. Two different patterns of connectivity are possible: 1) separate Type I V2a neurons independently project to either phrenic motor neurons or ARM motor neurons, but not both. Alternatively, the same Type I V2a neuron may bifurcate and project to both phrenic and ARM motor neurons. For example, Lane et al. 2008 used dual tracing studies to describe individual cervical neurons connected to both phrenic and intercostal motor neurons, possibly as a mechanism to coordinate the activity of different respiratory motor pools.

Future tracing experiments are needed to determine whether the same V2a neuron bifurcates to contact both types of respiratory motor neurons. This can be done by injecting two pseudorabies viruses (PRV) into a Chx10Cre/+;Td-Tomato mouse, a mouse line where all Chx10+ V2a neurons express the red

115 fluorescent marker. First, PRV-290 (labeled with the Turquoise2 fluorophore) should be injected into the scalene muscle and 2) PRV-152 (labeled with the green fluorescent protein) (Hogue et al. 2018) should be injected into the diaphragm. Both viruses will retrogradely label their motor neuron pools and any premotor neurons (Zholudeva et al. 2017; Ueno et al. 2018). If there are any Tomato+ V2a neurons that are co-labeled with both red and blue, this suggests that at least some of the same V2a neurons project to both the scalene and diaphragm muscle.

Despite the sufficiency of cervical spinal V2a neurons to increase diaphragm EMG peak amplitude, we also show that they are not necessary for diaphragm activity in healthy mice. Namely, silencing cervical spinal V2a neurons does not alter diaphragm RMSpeak amplitude during interbout intervals or during ARM activity. It is possible that there were too few cervical spinal V2a neurons that expressed the inhibitory

(Gi)DREADD receptor. However, these data support previous studies (see Chapter II) in which decreasing brainstem and spinal V2a neurons activates ARMs but does not alter diaphragm EMG peak amplitude, suggesting ARMs are not activated as a compensatory mechanism for impaired diaphragm function

(Jensen et al., 2019). Alternatively, it is possible that silencing cervical spinal V2a neurons may alter cardiovascular output to indirectly activate ARMs and enhance ventilation.

These data also suggest that not all V2a neurons serve as necessary relays or amplifiers for descending brainstem drive to control phrenic motor neuron output. In fact, some cervical spinal V2a neurons have been shown to control fine motor control instead of breathing (Azim et al. 2014). Taken together, these data suggest that direct monosynaptic connections from the rVRG to phrenic motor neurons (or other polysynaptic pathways) is sufficient to maintain resting phrenic motor neuron output.

Since cervical spinal V2a neurons are sufficient but not required for diaphragm activity, we propose an important amendment to our model of V2a control of respiratory muscles. We propose that Type I V2a neurons in the cervical cord are inactive at rest and project to ARM and phrenic motor neurons via excitatory pathways. Therefore, increasing V2a excitability enhances diaphragm and ARM activity. On the

116 other hand, we propose that Type II V2a neurons in the cervical spinal cord are active at rest and participate in an inhibitory spinal pathway to prevent ARM activation at rest (Figure 11). Therefore, silencing cervical spinal V2a neurons does not alter phrenic motor neuron activity but removes inhibition onto ARM motor neurons to activate the scalene muscle.

In order to test this, it is critical to identify distinct molecular markers of the proposed Type I and

Type II cervical spinal V2a neurons and selectively target DREADD expression to each subtype in different cohorts of animals. If our model is correct, increasing the excitability of Type I V2a neurons will activate

ARMs and increase diaphragm EMG peak amplitude, but silencing Type I V2a neurons will not alter respiratory muscle activity. In contrast, increasing Type II V2a excitability will not alter respiratory muscle activity but silencing Type II V2a neurons will result in increased ARM activity.

Interestingly, silencing cervical spinal V2a neurons differentially affects ventilation during interbout intervals compared to silencing brainstem and spinal cord V2a neurons. Decreasing brainstem and spinal cord V2a excitability increased PIF, breathing frequency, MV, and VT/Ti, even in the absence of

ARM activity. In contrast, silencing cervical spinal V2a neurons did not alter PIF, Vt, MV, or VT/Ti. Two potential explanations exist to explain these discrepancies. First, it is possible that decreasing cervical spinal V2a excitability may alter these ventilatory parameters in healthy mice at rest, but there were not enough V2a neurons in the cervical spinal cord that expressed the inhibitory DREADD receptor in these experiments. The peak AAV expression is found a at a distinct injection site, and tapers off as the distance from the injection site increases to form a histogram-like distribution of AAV expression. To test whether increasing cervical spinal V2a expression of the inhibitory (Gi)DREADD receptor would alter ventilation in the absence of ARM bouts, future studies should perform two injections into each side of the spinal cord

(e.g. at C3 and C6). Alternatively, silencing cervical spinal V2a neurons may not enhance ventilation at rest because brainstem V2a neurons may be required. In order to test this, a Cre-dependent AAV virus can be injected into the brainstem to target inhibitory (Gi)DREADD expression to brainstem V2a neurons.

117

Despite the differences in altered ventilation during interbout intervals between V2a-(Gi)DREADD mice and AAV-(Gi)DREADD mice, silencing V2a neurons in both mouse models resulted in a drastic increase in PIF, Vt, f, MV, and VT/Ti during bouts of ARM activity (Jensen et al., 2019). Because silencing cervical spinal V2a neurons does not alter diaphragm EMG peak amplitude, we conclude this enhanced ventilation is due to the activation of ARMs.

The one change observed in baseline ventilation after silencing cervical spinal V2a neurons is increased breathing frequency, even though cervical spinal V2a neuron bodies are located distal from the respiratory rhythm generating centers in the brainstem. One explanation for this increased breathing rate may be the influence of spinobulbar projections from cervical spinal V2a neurons to brainstem rhythm generators, including the pFRG, PreBotzinger Complex, and Kolliker Fuse/Parabrachial (KF/PBN) nucleus, which regulate respiratory rhythm (e.g. breathing regularity and frequency). The least amount of projections is observed in the KF/PBN and the PreBotzinger Complex. The KF/PBN is critical for controlling respiratory rhythm variability. Therefore, if spinal projections to this brainstem area significantly affect respiratory frequency, it is likely that the regularity of respiration would be altered after silencing V2a neurons; however, this is not observed in AAV-(Gi)DREADD mice. On the other hand, the pFRG also showed a high density of cervical spinal AAV projections, which is important for chemosensation, generating expiratory rhythm to expiratory muscles, and providing excitatory input into the PreBÖtzinger

Complex (main inspiratory rhythm generator) (Guyenet and Mulkey 2010; Guyenet and Bayliss 2015).

Finally, we show that increasing cervical spinal V2a neuron excitability on one side of the spinal cord bilaterally increases scalene ARM activity on the left and right side. These data suggest that the proposed Type I ipsilaterally projecting V2a neurons participate in polysynaptic pathways to influence

ARM activity on the contralateral side of the spinal cord. At least three different possibilities exist to explain these results. First, it is possible V2a neurons on one side of the spinal cord project to sympathetic preganglionic neurons that control cardiac output and may systemically alter heart rate or blood pressure

118 that feeds back to bilaterally activate ARMs. However, we do not detect a change in heart rate following

CNO treatment in AAV-(Gq)DREADD mice. Alternatively, cervical spinal V2a neurons may participate in a spino-bulbar-spino loop. Specifically, cervical spinal V2a neurons on one side of the spinal cord may project to supraspinal structures in the brain that send bilateral descending input to the spinal cord to alter both left and right scalene activity. Lastly, cervical spinal V2a neurons may project to commissural interneurons that cross the midline of the spinal cord to influence (directly or indirectly) ARM motor neurons on the opposite side of the spinal cord. This hypothesis is consistent with spinal circuitry controlling locomotion. Specifically, V2a neurons in the lumbar spinal cord have been shown to project to commissural interneurons to coordinate left-right alternation of locomotion (Crone et al., 2008;

Dougherty et al., 2013). Furthermore, injection of PRV into the left hemidiaphragm results in PRV+ labeled interneurons on the right side of the spinal cord (Zholudeva et al. 2017). Therefore, it is likely that V2a neurons project to commissural interneuron to coordinate bilateral activation of respiratory muscles. In order to test this, PRV should be injected into the scalene muscle on one side of the spinal to analyze whether PRV+ cells are observed on both the ipsilateral and contralateral side.

If cervical spinal V2a neurons project to commissural neurons to pattern ARM activity on the opposite side of the spinal, it is important to note the differences between the function of the respiratory system and the locomotor system. Crone et al. 2008 and Dougherty et al. 2013 showed that lumbar V2a neurons project to commissural interneurons and participate in inhibitory pathways to prevent simultaneous activation of limb muscles on the opposite side of the spinal cord to allow for coordinated left-right alternation of locomotion. However, our data shows that increasing cervical spinal V2a excitability on one side coordinates synchronous activation of the left and right scalene muscle. This is not surprising because the respiratory system is organized bilaterally. Because the activation of ARMs helps stabilize and expand the thoracic cavity, coordinated activation of respiratory muscles is important for ensuring efficient drive to respiratory muscles and efficient breathing. While these data show that

119 increasing the excitability of ipsilateral cervical spinal V2a neurons bilaterally activates accessory respiratory muscles, future studies should test whether unilaterally silencing cervical spinal V2a neurons also bilaterally activates ARMs.

Together, these data show that cervical spinal V2a neurons are sufficient to control respiratory muscle activation and ventilation in healthy mice. Moreover, it is likely that at least two different subsets of V2a neurons exist in the cervical spinal cord to tightly regulate this activity. Future experiments should identify molecular markers specific to different subsets of V2a neurons and determine which subtype is important for enhancing diaphragm activity and controlling ARM activation.

120

Chapter IV.

Altering V2a Excitability Increases ARM Activity

and Enhances Ventilation in ALS Model Mice

121

4.1 Introduction

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease characterized by the loss of motor neurons, including respiratory motor neurons (Cleveland and Rothstein 2001; Boillée, Vande Velde, and Cleveland 2006). As a result, respiratory failure is the leading cause of death in ALS patients (Lyall

2001; Hardiman 2011). Despite progressive denervation and dysfunction of the diaphragm, patients and

ALS models have the ability to maintain normal ventilation until late stages of disease due to mechanisms of respiratory compensation (Kennel et al., 1996; Barnéoud et al. 1997; Hegedus et al. 2007; Cappello et al., 2012; Valdez et al. 2012). Therefore, understanding how these compensatory mechanisms are activated is critical for identifying novel therapeutic approaches to keep patients off of a mechanical ventilator to improve their quality of life.

Recruitment of accessory respiratory muscles (ARMs), such as the trapezius, scalenes, and sternocleidomastoid, is important for enhancing ventilation during conditions of high oxygen demand

(e.g. exercise) or when the diaphragm is impaired due to disease or injury (Pinto & De Carvalho, 2008;

Smittkamp et al., 2010; Johnson & Mitchell, 2013). Notably, ARMs are sufficient to drive respiration in patients suffering from diaphragm paralysis (Bennett et al. 2004) and ALS patients that recruit ARMs sleep better and survive longer than patients who do not recruit ARMs (Arnulf et al. 2000). However, it is unknown why some patients recruit ARMs and others do not. Therefore, a better understanding of how

ARMs are activated in ALS is important for identifying potential therapeutic targets to improve ventilation.

Romer et al. 2017 previously showed that the SOD1(G93A) mouse model of ALS showed increased

ARM activity to enhance ventilation during early to mid-stages of disease progression, but failed to activate these ARMs at later stages. This same study showed that V2a neurons degenerate throughout disease progression, are synaptically connected to ARM motor neurons, and are sufficient to activate

ARMs at rest in healthy mice. Because SOD1(G93A) ALS model mice were still able to activate accessory respiratory muscles during other behaviors (like grooming) during late stages of ALS, it was postulated

122 that the loss of ARM activity at late stages of ALS may be due to the loss of V2a neurons instead of the loss of motor neurons (Romer et al. 2017).

Increasing (Romer et al. 2017) or decreasing (Jensen et al. 2019) the excitability of V2a neurons is sufficient to increase ARM activity and enhance ventilation in healthy mice. These data suggest that two different subtypes of V2a neurons may exist and may be important modulators of ARM activity following neuromuscular disease when primary pump inspiratory muscles become impaired. However, it is unknown which V2a subtype – Type I, Type II, or both – is able to activate ARMs and enhance ventilation at late stages of ALS disease progression. Therefore, we investigated whether increasing or decreasing the excitability of V2a neurons throughout ALS disease progression activates ARMs and enhances ventilation, despite progressive respiratory motor neuron and V2a interneuron loss. Here, preliminary data shows that activating or silencing V2a neurons can increase ARM activity throughout disease progression in ALS model mice. Interestingly, increasing vs. decreasing the excitability of V2a neurons differentially affects ventilation at different disease stages, suggesting that different V2a subtypes may play different roles in controlling breathing throughout ALS disease progression.

4.2 Methods

Animals

All procedures were performed according to National Institutes of Health guidelines and approved by

Cincinnati Children’s Hospital Medical Center Animal Care and Use Committee.

G93A/+ Cre/+; PNP-tTA/+ G93A V2a-(Gq)DREADD mice (SOD1 ; Chx10 ROSA ; TetO-CHRM3/+), in which the excitatory

(Gq)DREADD receptor is expressed in all brainstem and spinal cord V2a neurons, were generated by first breeding SOD1G93A/+ (B6.Cg-Tg (SOD1*G93A)1Gur/J; Stock #004435 Jackson Laboratory, Bar Harbor, ME) mice with ROSAPNP-tTA/+; TetO-CHRM3/+ mice [mouse line generation described in Romer et al., 2017].

These SOD1G93A/+; ROSAPNP-tTA/+; TetO-CHRM3/+ mice were then bred to Chx10Cre/+ mice to confine

123

+ G93A/+ Cre/+; PNP-tTA/+ excitatory (Gq)DREADD expression to Chx10 V2a neurons in SOD1 ; Chx10 ROSA ; TetO-

CHRM3/+ mice. The optimal 1.0 mg/kg*bw dose of CNO was used to depolarize V2a neurons (Romer et al. 2017).

G93A/+ Cre/+; CHRM4/+), Similarly, G93A V2a-(Gi)DREADD mice (SOD1 ; Chx10 ROSA in which the inhibitory

(Gi)DREADD receptor is expressed in all brainstem and spinal cord V2a neurons, were generated by first breeding SOD1G93A (B6.Cg-Tg (SOD1*G93A)1Gur/J; Stock #004435 Jackson Laboratory, Bar Harbor, ME) mice to ROSACHRM4/+ (B6N. 129-Gt(ROSA)26Sortm1(CAG-CHRM4*,-mCitrene)Ute/J; Stock # 026219

Jackson Laboratory, Bar Harbor, ME) to generate SOD1G93A/+; ROSACHRM4/+ mice. These mice were then bred

Cre/+ + G93A/+ to Chx10 mice to confine inhibitory (Gi)DREADD expression to Chx10 V2a neurons in SOD1 ;

Chx10Cre/+;ROSACHRM4/+ mice. The optimal 10.0 mg/kg*bw dose of CNO was used to hyperpolarize V2a neurons (Jensen et al. 2019).

ALS-like Disease Progression

SOD1G93A; tTA/+;TetOCHRM3/+ and

SOD1G93A; ROSACHRM4/+ mice were assessed for hindlimb and neurological deficits beginning at 11-

12 weeks of age, as previously described in (Hatzipetros et al. 2015;

Romer et al. 2017). Mice were classified into six different stages based on the appearance of motor deficits (Table 1; reproduced from

Romer et al. 2017). Mice were sacrificed immediately after reaching endstage (Stage 5) when they were unable to right themselves within 30s after being placed on their back/unable to reach food and water du e to motor deficits.

124

Surgical Implantation of Telemetry Devices

G93A V2a-(Gq)DREADD and G93A V2a-(Gi)DREADD mice were implanted with F20-EET transmitters (Data

Sciences International) beginning at age postnatal day 56 (ALS stage 0) to record EMG from the scalene and trapezius ARMs, as previously described in (Romer et al. 2017; Jensen et al. 2019). Mice recovered for at least one week following surgery before being used in experiments.

Electromyography and Plethysmography Acquisition and Analysis

Electromyography (EMG), whole breath plethysmography (WBP), and video were simultaneously recorded in non-anesthetized, freely moving mice before (vehicle saline treatment) and after CNO treatment, as previously described in (Romer et al. 2019 and Jensen et al. 2019). Briefly, WBP, ARM EMG, and video were recorded for at least one hour prior to and following CNO treatment at least one time during each ALS disease stage. Bouts of ARM activity were scored based on three criteria from (Jensen et al. 2017): 1) the RMS EMG trace must exhibit at least three consecutive points that are at least 50% higher than the surrounding baseline 2) EMG activity occurring during resting breaths (i.e. sighs and movement artifacts are excluded) and 3) EMG activity occurs when the mouse is still based on the synchronized video recordings (Jensen et al., 2017). WBP was measured before CNO treatment (vehicle saline injection) during resting breaths, after CNO treatment during resting breaths, and after CNO during bouts of ARM activity. The average peak inspiratory flow (PIF), tidal volume (Vt), minute volume (MV), and breathing frequency (f) was recorded for each animal under all three conditions for at least 100 breaths from each animal.

Statistical Analysis

Data are reported as mean ± SEM. A two-way repeated measures ANOVA with post-hoc Holm-Sidak all pair-wise comparisons was used to analyze bout frequency and WBP based on two conditions: 1) ALS disease stage 0-4 and 2) CNO treatment (e.g. vehicle control vs. CNO). A one-way repeated measures

125

ANOVA was used to analyze changes in baseline ventilation due to disease progression. The p-value was set at 0.05. Since the data presented below constitute preliminary data and are not sufficiently powered, all statistical tests, p-values, and power values are reported in Table 2 at the end of the results section.

4.3 Results DREADD expressing SOD1(G93A) ALS Model Mice Exhibit Normal Disease Progression All mice were monitored at least twice a week to check for disease progression based on hindlimb motor deficits. Similar to non-DREADD expressing SOD1(G93A) mice, DREADD expressing SOD1(G93A) ALS model mice showed age-dependent induced hindlimb motor deficits. Moreover, we observed a drastic decrease in body weight at stage 4 in G93A V2a-(Gq)DREADD (one-way repeated measures ANOVA, n=3, p=0.015) (Figure 1A) and G93A V2a-(Gi)DREADD (one-way repeated measures ANOVA, n=3, p=0.015)

(Figure 1B) mice that is consistent with previously published literature (Tankersley et al. 2007).

Figure 1. DREADD-expressing SOD1(G93A) ALS model mice show characteristic weight loss due to disease progression. Body weight was measured at each stage of disease progression in ALS model mice to confirm that DREADD expression did not alter ALS induced hindlimb motor deficits (not shown) or weight loss. Weight is plotted at each ALS stage for G93A V2a-(Gq)DREADD mice (A) and G93A V2a-(Gi)DREADD mice (B). Gray dashed lines indicate individual animals. Black solid lines indicate average ± SEM. Data are analyzed with a one-way repeated measures ANOVA. (*)p<0.05.

Increasing the excitability of V2a neurons activates ARMs throughout disease progression. We next used EMG leads inserted into the trapezius and scalene muscles of G93A V2a-

(Gq)DREADD (in which the excitatory (Gq)DREADD receptor is expressed in brainstem and spinal cord V2a neurons) to measure bouts of ARM activity throughout ALS disease progression. Similar to the pattern observed in non-DREADD expressing SOD1G93A ALS model mice (Romer et al. 2017), there is a significant main effect of ALS disease stage on baseline (vehicle treated) scalene and trapezius ARM activity in G93A

V2a-(Gq)DREADD mice throughout disease progression [Trapezius: F(4, 1) = 8.302, p=0.006; Scalene: F(4,

126

1) =6.861, p=0.011]. Representative traces are shown in Figure 2A-E (left side) and quantified in Figure 2F

(gray bars). Specifically, ARM activity progressively increases from Stage 0 (pre-symptomatic) until Stage

2, when peak scalene and trapezius activity is observed (trapezius HL0: 0.3 bouts/min vs trapezius HL2:

1.9 bouts/min., n=3, p=0.041; scalene HL0: 0.2 bouts/min. vs HL2: 1.5 bouts/min., n=3, p=0.042). This

ARM activity then declines beginning at Stage 3 (at hindlimb paralysis onset) until ARM activity is comparable to pre-symptomatic levels (n=3, p=1.00). Thus, excitatory (Gq)DREADD expression in ALS model mice does not appear to alter baseline ARM recruitment throughout disease progression in ALS model mice.

Since increasing the excitability of V2a neurons with 1.0 mg/kg*bw CNO activates ARMs in healthy

V2a-(Gq)DREADD mice (Romer et al. 2017), we next assessed whether activating brainstem and spinal cord

V2a neurons with CNO further increases ARM activity at each stage of disease progression, despite the progressive loss of V2a neurons. Interestingly, CNO treatment caused increased ARM activity at all stages of ALS disease progression (representative traces in Figure 2A-E (right side)). The greatest increase over baseline ARM activity is observed pre-symptomatically with a 6.7-fold increase at Stage 0, 2.0-fold increase at Stage 1, 1.6-fold increase at Stage, 2, 1.7-fold increase at Stage 3, and 1.8-fold increase at Stage

4 (Figure 2F, black boxes). Similarly, CNO increases scalene bout frequency across all disease stages, with a 5.2-fold increase at Stage 0, 3.2-fold increase at Stage 1, 1.7-fold increase at Stage 2, 4.8-fold increase at Stage 3, and 2.2-fold increase at Stage 4 (Figure 2G, black boxes). There is not a significant main effect of CNO treatment on ARM activity [two-way repeated measures ANOVA, F(4,1)= 9.89, p=0.088] because this study is under-powered (power=0.384) and requires a greater n to reach significance (Table 2). There is a high percentage of scalene bout co-activation with trapezius bouts from stages 0-3 (96% at stage 0,

97% at stage 1, 96% at stage 2, and 95% at stage 3) but this co-activation drastically decreases to 56% at

Stage 4 (p=0.064, power=0.443), likely due to significant ARM motor neuron death and V2a neuron degeneration that prevents synchronous and patterned activity.

127

Interestingly, despite evidence of a strong trend that shows increasing V2a neuron excitability activates ARMs at all stages of disease progression, we observed differences in the trapezius and scalene

EMG bouts at different stages in G93A V2a-(Gq)DREADD mice. We quantified ARM bout EMG peak amplitude at each stage and normalized it to the maximal amplitude reached during sighs. We show that there is an initial increase in trapezius and scalene EMG peak amplitude between Stage 0 and Stage 1, followed by a gradual decrease that persists until Stage 4 (Figure 2H). Analysis of burst duration of trapezius and scalene EMG bouts shows there is no difference throughout disease progression (p=0.080 and p=0.076, respectively) (Figure 2I). Thus, increasing V2a excitability can further activate ARMs at all stages of disease progression, but the EMG peak amplitude continually decreases as the disease progresses and respiratory motor neurons continue to degenerate.

Figure 2. Increasing the excitability of V2a neurons activates ARMs throughout disease progression. Scalene and trapezius EMG was recorded throughout ALS disease progression (stages 0 – 4) before and after activating V2a neurons with 1.0 mg/kg*bw CNO treatment in G93A V2a- (Gq)DREADD mice (n=3). Representative traces at each stage during baseline recordings and following CNO treatment are shown for the scalene muscle (A-E). Trapezius (F) and scalene (G) bouts/minute are quantified before (gray bars) and after (black bars) CNO treatment. (H) Trapezius (black bars) and scalene (white bars) EMG peak amplitude normalized to the maximal amplitude reached during sighs was measured through disease progression. (I) Trapezius (black bars) and scalene (white bars) burst duration was measured across disease progression. Data are reported as mean ± SEM. Statistics: A two-way repeated measures ANOVA was used to analyze data in F-G and a one-way ANOVA was used to analyze data in H-I. (*) p<0.05.

128

Decreasing the excitability of V2a neurons activates ARMs throughout disease progression. We have previously shown that decreasing the excitability of V2a neurons activates ARMs in healthy mice (Jensen et al. 2019). Therefore, we assessed 1) whether silencing V2a neurons in G93A V2a-

(Gi)DREADD mice activates ARMs throughout ALS disease progression and 2) whether the effect of activating and silencing V2a neurons differentially affects ARM activity at different stages of ALS. Similar to G93A V2a-(Gq)DREADD mice, we observe a 32% increase in baseline scalene bouts between Stage 0 and Stage 1 that is once again decreased by 31% from Stage 3 to Stage 4, reaching pre-symptomatic levels at Stage 4 as seen in (Romer et al. 2017) (representative traces (left) in Figure 3A-E and quantified in

Figure 3F (gray bars)). More animals must be assessed and analyzed to detect a main effect of ALS disease stage on baseline trapezius activity across disease progression [F(4,1)=0.853, p=0.53] (power=0.050)

(Table 2).

Since decreasing the excitability of V2a neurons with 10.0 mg/kg*bw CNO activates ARMs in healthy V2a-(Gi)DREADD mice (Jensen et al. 2019), we next assessed whether activating brainstem and spinal cord V2a neurons with CNO further increases ARM activity at each stage of disease progression, despite the progressive loss of motor neurons and V2a interneurons. There is a significant main effect of

CNO on ARM activation throughout disease progression in G93A V2a-(Gi)DREADD mice [F(4,1)=42.132, p=0.023]. We observe a 6.5-fold increase in scalene activity at Stage 0, 2.2-fold increase at Stage 1, 2.9- fold increase at Stage 2, 1.3-fold increase at Stage 3, and 1.4-fold increase at Stage 4 (Figure 3F (gray bars)). In contrast to G93A V2a-(Gq)DREADD animals, both scalene EMG peak amplitude (Figure 3G) and scalene bout burst duration (Figure 3H) incrementally increases as ALS progresses and reaches its peak at

Stage 4, despite progressive motor neuron and V2a neuron loss. Thus, increasing or decreasing the excitability of V2a neurons throughout disease progression can increase ARM activity in SOD1(G93A) ALS model mice.

129

Figure 3. Decreasing the excitability of V2a neurons activates ARMs throughout disease progression. Scalene EMG was recorded throughout ALS disease progression (stages 0 – 4) before and after silencing V2a neurons with 10.0 mg/kg*bw CNO treatment in G93A V2a-(Gi)DREADD mice (n=3). Representative traces at each stage during baseline recordings and following CNO treatment are shown for the scalene muscle (3A-E). Scalene (3F) bouts/minute are quantified before (gray bars) and after (black bars) CNO treatment. (G) ARM EMG peak amplitude normalized to the maximal amplitude reached during sighs was measured through disease progression. (H) Scalene burst duration was measured across disease progression. Gray dashed lines show individual animals. Solid black lines show the mean ± SEM. A two-way repeated measures ANOVA was used to analyze data in 3F and a one-way ANOVA was used to analyze data in 3G. (*) p<0.05.

Activating V2a neurons enhances ventilation during ARM bouts during early stages of ALS whereas silencing V2a neurons enhances ventilation during ARM bouts during late stages of ALS

We next assessed whether the activation of ARMs following CNO treatment is productive muscle activity (i.e. enhances ventilation) or unproductive muscle activity (does not enhance ventilation). It is known that ARMs become active to enhance ventilation under conditions of high oxygen demand, such as during exercise or following disease/injury when the diaphragm is impaired (Sieck and Gransee, 2012;

Bennett et al. 2004; Arnulf et al. 2000). Romer et al. 2017 postulated that ARMs become active during ALS to enhance ventilation and compensate for progressive respiratory muscle impairment. Therefore, we

130 assessed whether DREADD induced ARM activation was sufficient to enhance ventilation in ALS model mice, even at later stages when the bout amplitude is decreased.

We measured the percent change in four ventilation parameters following CNO treatment in

G93A V2a-(Gq)DREADD mice from interbout intervals (ARM bouts absent) to ARM bouts: peak inspiratory flow (PIF), tidal volume (Vt), breathing frequency (f), and minute volume (MV). We show that ARM activation increases PIF by 67% at Stage 0, 75% at Stage 1, 68% at Stage 2, 40% at Stage 3, and only 5% at

Stage 4 (Figure 4A). A similar pattern is seen in MV with a 56% increase at Stage 0, 64% increase at Stage

1, 47% increase at Stage 2, 33% increase at Stage 3, and a 1% decrease at Stage 4 (Figure 4D). There is a negligible effect on breathing frequency throughout disease progression between interbout intervals compared to ARM activation (Figure 4C). Thus, enhanced ventilation during ARM bouts after activating

V2a neurons is greatest early on in disease progression and gradually decreases beginning at stage 2, which is consistent with the gradual decrease in trapezius and scalene EMG peak amplitude and burst duration beginning at Stage 2 in G93A V2a-(Gq)DREADD mice (Figure 3H).

We next assessed the impact of ARM activation on ventilation after silencing V2a neurons in G93A

V2a-(Gi)DREADD mice. Interestingly, there is only a modest increase in PIF by 33% at Stage 0, 20% increase at Stage 1, 8% increase at Stage 2, 11% increase at Stage 3, and 38% increase at Stage 4 (Figure 4E). ARM activity similarly enhances Vt and MV the most at stage 4 with a 36% (Figure 4F) and 40% (Figure 4H) increase, respectively. There is a negligible effect of ARM activity on breathing frequency throughout disease progression.

These data reveal two important differences between G93A V2a-(Gq)DREADD mice and G93A V2a-

(Gi)DREADD mice in how ARM activity modulates ventilation. First, both increasing and decreasing the

V2a excitability enhances ventilation at early stages of disease progression, but activating V2a neurons enhances ventilation to a greater extent than silencing V2a neurons does. For example, there is a

131 significant difference in the percent increase in PIF at stage 1 in G93A V2a-(Gq)DREADD mice compared to G93A V2a-(Gi)DREADD mice (75 ± 13% vs. 20 ± 6% student t-test, p=0.038, n=3). Second, the timing of the peak effect of ARM activity on ventilation differs between the two mouse lines. Activating V2a neurons causes ARM activity to enhance ventilation the most during early stages of disease progression (and does not enhance ventilation at stage 4 even though CNO increases ARM bout frequency). There is even an increase observed in PIF and MV in G93A V2a-(Gq)DREADD mice at Stage 3 that is not seen in G93a V2a-

(Gi)DREADD mice (PIF: 45 ± 2% vs. 10 ± 6%, student t-test, p=0.029, n=3; MV: 40 ± 0.6% vs. -3 ± 7%, student t-test, p=0.0351, n=3). In contrast, silencing V2a neurons causes ARM activity to enhance ventilation the most at stage 4 in ALS model mice, as shown by a significant difference in the percent change at stage 4 in PIF (5 ± 8% vs. 38 ± 14%, student t-test, p=0.047, n=3) and MV (-4 ± 7% vs. 40 ± 7%, student t-test, p=0.0458, n=3). Together, these data suggest that increasing and decreasing the excitability of V2a neurons may modulate ventilation throughout disease progression via different mechanisms.

Figure 4. Activating V2a neurons enhances ventilation during ARM bouts during early stages of ALS whereas silencing V2a neurons enhances ventilation during ARM bouts during late stages of ALS.

The percent change in ventilation parameters from interbout intervals to ARM bouts was calculated for G93A V2a-(Gq)DREADD mice and G93A V2a-(Gi)DREADD at all stages of disease progression. PIF (4A and E), Vt (4B and F), f (4C and G), and MV (4D and H) were analyzed. Data are reported as mean ± SEM. Each data set was analyzed with a one-way repeated measures ANOVA. (*) p<0.05.

132

Increasing or decreasing V2a excitability reverses ALS induced changes in ventilation

Despite the loss of phrenic motor neurons and subsequent diaphragm impairment during ALS disease progression, patients and rodent models alike maintain minute ventilation (MV) up until late stages of the disease (Fallat et al. 1979; Tankersley et al. 2007; Romer et al., 2017). We have previously shown that altering V2a neuron excitability enhances ventilation even in the absence of ARM activity (Romer et al.

2017; Jensen et al. 2019). Therefore, we assessed whether increasing or decreasing the excitability of V2a neurons in ALS model mice alters baseline ventilation.

First, we measured baseline ventilation throughout disease progression to verify that our

DREADD-expressing G93A mice faithfully recapitulate the maintenance of ventilation up until later stages of ALS. Similar to Tankersely et al. 2007 and Romer et al. 2017, both G93A V2a-(Gq)DREADD (Figure 5A-D) and G93A V2a-(Gi)DREADD mice (Figure 5I – L) maintain MV until ALS disease stage 3 when MV starts to increase. We measured the percent increase between stage 3 and stage 4 in baseline (-CNO) ventilation parameters in G93A V2a-(Gq)DREADD mice and observed a 35% increase in PIF (Figure 5A), 9% increase in Vt (Figure 5B), 41% increase in MV (Figure 5D), and a 31% increase in f (Figure 5C). Similarly, we observed a 29% increase in PIF (Figure 5I), 25% increase in Vt (Figure 5J), and 14% increase in MV (Figure

5L) with a negligible change in f (7% decrease) (Figure 5K) between stages 3 and 4 in G93A V2a-

(Gi)DREADD mice. Thus, the expression of excitatory or inhibitory DREADDs does not appear to alter the progression of motor or ventilatory deficits in our G93A mice.

We have previously shown that increasing or decreasing V2a excitability with CNO can enhance ventilation even in the absence of ARM activity (interbout intervals) in healthy mice. Therefore, we assessed the impact of increasing or decreasing V2a excitability on ventilation during interbout intervals at all stages of ALS disease progression. Interestingly, increasing V2a excitability with 1.0 mg/kg*bw CNO in G93A V2a-(Gq)DREADD mice (n=3) enhances PIF, Vt, and MV during early ALS stages (0-2), has a

133 negligible effect at stage 3, and actually decreases PIF, Vt, and MV at stage 4. Specifically, PIF is increased by 23% at stage 0, 32% at stage 1, 10% at stage 2, 2% at stage 3, and decreases PIF by 32% at stage 4

(Figure 5E). Vt is increased by 15% at stage 0, 53% at stage 1, and decreased by 1% at Stage 2, 0.6% at stage 3, and 19% at stage 4 (Figure 5F). MV is increased by 19% at stage 0, 19% at stage 1, 9% at stage 2, and decreased by 3% at stage 3 and 30% at stage 4 (Figure 5H). Little change following CNO treatment is observed throughout disease progression in breathing frequency [F(2) = 1.308, p=0.345)(Figure 5G). Thus, increasing V2a excitability enhances PIF, Vt, and MV the most at Stage 1 and depresses ventilation the most at Stage 4, likely to counteract ALS induced increases in ventilation and to restore MV.

Decreasing V2a excitability with 10.0 mg/kg*bw CNO in G93A V2a-(Gi)DREADD mice (n=3) has a negligible effect on ventilation during early ALS stages (0-1), enhances ventilation the most at Stage 2, and depresses ventilation at stage 4. Specifically, PIF is increased by 4% at stage 0, 7% at stage 1, 20% at stage

2, and decreases PIF by 11% at stage 3, and 45% at stage 4 (Figure 5M). Vt is increased by 13% at stage 0,

11% at stage 1, increased by 27% at Stage 2, decreased by 13% stage 3, and decreased by 50% at stage 4

(Figure 5N). MV is increased by 4% at stage 0, 13% at stage 1, 25% at stage 2, and decreased by 10% at stage 3 and 50% at stage 4 (p=0.007) (Figure 5P). Little change following CNO treatment is observed throughout disease progression in breathing frequency [F(2)=0.41, p=0.797)(Figure 5O). Thus, silencing

V2a neurons enhances PIF, Vt, and MV the most at Stage 2 and depresses ventilation the most at Stage 4.

134

Figure 5. Increasing or decreasing V2a excitability may acutely reverse ALS induced changes in ventilation during interbout intervals.

Ventilation was measured with plethysmography before and after exciting V2a neurons with 1.0 mg/kg*bw CNO in G93A V2a-(Gq)DREADD mice (n=3) and G93A V2a-(Gi)DREADD mice (n=3). The effect of ALS disease progression on baseline peak inspiratory flow (PIF) (5A and I), tidal volume (Vt) (5B and J), breathing frequency (f) (5C and K), and minute volume (MV) (5D and L) from stages 0-4 was measured. The percent increase in ventilation from pre CNO to post CNO (in the absence of ARM activity) was then calculated at each stage of disease progression for PIF (5E and M), Vt (5F and N), f (5G and O), and MV (5H and P). Data are reported as mean ± SEM. Each data set was analyzed with a one-way repeated measures ANOVA. (*) p<0.05.

135

Statistics Main Effect of Figure Statistical Test Main Effect of Stage Power Power CNO 1A One-way RM ANOVA p=0.015 0.778 NA NA 1B One-way RM ANOVA p=0.004 0.952 NA NA 2F Two-way RM ANOVA p=0.006 0.913 p=0.088 0.384 2G Two-way RM ANOVA p=0.011 0.838 p=0.087 0.389 2H (Trpz) One-way RM ANOVA p=0.117 0.299 NA NA 2H (Scalene) One-way RM ANOVA p=0.736 0.05 NA NA 2I (Trpz) One-way RM ANOVA p=0.080 0.389 NA NA 2I (Scalene) One-way RM ANOVA p=0.076 0.406 NA NA 3F Two-way RM ANOVA p<0.001 1.000 p=0.23 0.872 3G One-way RM ANOVA p=0.002 0.977 NA NA 3H One-way RM ANOVA p=0.002 0.977 NA NA 4A One-way RM ANOVA p=0.063 0.451 NA NA 4B One-way RM ANOVA p=0.543 0.05 NA NA 4C One-way RM ANOVA p=0.807 0.05 NA NA 4D One-way RM ANOVA p=0.248 0.144 NA NA 4E One-way RM ANOVA p=0.258 0.137 NA NA 4F One-way RM ANOVA p=0.006 0.91 NA NA 4G One-way RM ANOVA p=0.843 0.05 NA NA 4H One-way RM ANOVA p=0.039 0.57 NA NA 5A One-way RM ANOVA p=0.038 0.577 NA NA 5B One-way RM ANOVA p=0.26 0.136 NA NA 5C One-way RM ANOVA p=0.354 0.086 NA NA 5D One-way RM ANOVA p=0.364 0.082 NA NA 5E One-way RM ANOVA p=0.148 0.248 NA NA 5F One-way RM ANOVA p=0.189 0.197 NA NA 5G One-way RM ANOVA p=0.274 0.127 NA NA 5H One-way RM ANOVA p=0.345 0.09 NA NA 5I One-way RM ANOVA p=0.004 0.955 NA NA 5J One-way RM ANOVA p=0.025 0.672 NA NA 5K One-way RM ANOVA p=0.363 0.082 NA NA 5L One-way RM ANOVA p=0.035 0.596 NA NA 5M One-way RM ANOVA p=0.045 0.535 NA NA 5N One-way RM ANOVA p=0.049 0.514 NA NA 5O One-way RM ANOVA p=0.797 0.05 NA NA 5P One-way RM ANOVA p=0.048 0.517 NA NA

Table 2. Report of Statistics. A summary of all statistical tests, p-values, and power calculated for each data set in this study is reported. Analyses may fail to reach statistical significance due to a lack of power and need for increased sample size.

136

Figure 6. Hypothetical Model of ARM Activation. We propose that at least two different types of V2a neurons exist in the cervical spinal cord: Type I V2a neurons participate in an excitatory pathway (green) to activate ARMs under conditions of high oxygen demand and are inactive at rest whereas Type II V2a neurons participate in an inhibitory pathway (red) to prevent ARM activation when they are not needed and are active at rest. Our data suggests that Type I V2a neurons may preferentially project to fast ARM motor neurons (dark purple diamonds) and intermediate ARM motor neurons (plum diamonds) whereas Type II V2a neurons may preferentially project to slow ARM motor neurons (light purple diamonds) and intermediate ARM motor neurons. (Left) All motor neurons and V2a neurons are present before symptom onset in ALS. (Middle) All fast motor neurons, some intermediate ARM motor neurons, and some V2a neurons have degenerated by late stage ALS. Increasing V2a excitability activates intermediate ARM motor neurons to increase ARM activity without affecting slow ARM motor neuron activity. (Right) Decreasing V2a excitability removes inhibition onto slow ARM motor neurons to increase ARM activity without affecting intermediate ARM motor neuron activity. Dotted lines = inactive. Solid lines = active.

4.4 Discussion

In this study, we tested the role of brainstem and spinal cord V2a neurons in activating ARMs and enhancing ventilation throughout disease progression in ALS model mice by increasing or decreasing the excitability of V2a neurons. We first showed that similar to healthy mice, increasing or decreasing the excitability of V2a neurons activates ARMs throughout ALS disease progression. We have previously postulated that at least two different subtypes of V2a neurons exist in the cervical cord to tightly regulate

ARM activity in healthy mice so they become active when needed but remain inactive when they are not to preserve energy (Jensen et al. 2019). Briefly, we propose that Type I V2a neurons are inactive at rest and participate in an excitatory pathway to increase ARM motor neuron activity under conditions of high oxygen demand. Type II V2a neurons may participate in an inhibitory pathway to prevent ARM activation

137 and are typically active at rest (Figure 6, left panel). Assuming this model is correct, our data showing that activating and silencing V2a neurons throughout disease progression activates ARMS corroborates our previous findings in healthy mice. Furthermore, it is not surprising the greatest increase in ARM activity occurs when ALS model mice are still pre-symptomatic and there are approximately 90% of V2a neurons

(and respiratory motor neurons) remaining (Romer et al. 2017). We observe the peak ARM bout frequency but also the least percent increase in ARM activity at Stage 2, likely because there is a ceiling effect of

ARM recruitment. Additionally, our data suggests that at least some Type I and Type II V2a neurons are still present at Stage 4 of ALS disease progression (despite 50% of V2a neuron degeneration observed in

G93A mice)(Romer et al. 2017) because both exciting and silencing V2a neurons activates ARMs at late stages. However, even though we do see an increase in ARM activity at Stage 4, the maximum ARM bout frequency reached is only 1 bout/min., results that likely reflect the substantial ARM motor neuron loss as well as V2a neuron loss (Charcot and Joffrey 1869; Romer et al. 2017) These data suggest that while

V2a neurons can be targeted to increase ARM activity, the increase in ARM activity may not be biologically significant, especially if activation is not productive at late stages of ALS (i.e. does not enhance ventilation).

This information is especially important for determining which strategy – activating or silencing V2a neurons – would be most beneficial in a therapeutic setting.

We showed that ARM bouts enhance PIF, Vt, and MV the most during early stages of ALS after increasing the excitability of V2a neurons. This finding is consistent with the gradual decline in ARM EMG bout amplitude observed throughout disease progression. Furthermore, it must be considered that both motor neuron dysfunction at late stages of ALS is high and V2a neurons are also degenerating, suggesting that although the same dose of CNO is used to increase the excitability of V2a neurons, there are less V2a neurons to excite and less motor neurons to receive that excitatory input to generate enough respiratory muscle force to drastically enhance ventilation. In contrast to activating V2a neurons, silencing V2a neurons in G93A V2a-(Gi)DREADD mice enhanced ventilation the most during ARM bouts at Stage 4 of ALS

138 when motor and V2a neuron numbers are substantially reduced. The peak scalene EMG amplitude and burst duration observed at Stage 4 is consistent with this enhanced ventilation. Therefore, we propose that activating V2a neurons is the best strategy for enhancing respiratory muscle activity at early stages of ALS, but therapeutic intervention at later stages may favor silencing V2a neurons, especially since activating V2a neurons at later stages may contribute to excitotoxicity and exacerbate degeneration (Van

Den Bosch et al. 2006). Taken together, these data suggest that if two subsets of V2a neurons exist, they likely play different roles in controlling respiratory muscle activity and ventilation at different stages of

ALS disease progression.

We propose several hypothetical models to explain these results. First, Type I and Type II V2a neurons may preferentially project to different types of motor units (model outlined in Figure 6). Three types of motor units exist: Fast Fatigue (FF) or fast motor neurons (recruited last and during high force behaviors), Fast Resistant (FR) (recruited second and recruited for medium force behaviors) or intermediate motor neurons, and Slow (S) (recruited first and during low force behaviors) (Fournier &

Sieck, 1988; Mantilla and Sieck, 2003; Peter et al. 1972). In ALS, FF motor units are the most vulnerable to disease and have all degenerated before symptom onset whereas S motor units are least vulnerable to degeneration (Hegedus et al., 2007; Hegedus et al. 2008). We propose that Type I V2a neurons

(hypothesized to provide direct excitatory input into respiratory motor neurons) may preferentially project to FF and FR motor units and provide excitatory input into motor neurons to produce higher force behaviors. On the other hand, we hypothesize that Type II V2a neurons (hypothesized to participate in an inhibitory pathway to constrain ARMs at rest) preferentially project to slow motor units (and some FR motor units).

Before symptom onset, all types of motor units and V2a neurons are present. Therefore, both activating and silencing V2a neurons will increase ARM activity. However, because Type I V2a neurons preferentially project to FF and FR motor units, activating V2a neurons produces a high ARM EMG peak

139 amplitude and drastically enhanced ventilation during ARM bouts. On the other hand, silencing V2a neurons will also increase ARM activity, but the preferential projection to slow motor neurons produces a lower ARM EMG peak amplitude and milder increase in ventilation during ARM bouts. At late stages of

ALS, the FF motor units have all degenerated as well as some FR motor units, meaning that activating Type

I V2a neurons has a smaller effect on ARM EMG peak amplitude and ventilation during ARM bouts. In contrast, most slow motor units are still present and receiving input from Type II V2a neurons, which when silenced, activates ARMs. It is possible that the severe respiratory dysfunction at late stage ALS actually causes a greater reliance on the remaining slow motor neurons for ventilation or that these slow motor units may show compensatory increased excitability, either of which would account for the increased

ARM burst duration, amplitude, and enhanced ventilation after silencing V2a neurons at stage 4 in ALS model mice. Similar compensatory mechanisms have been observed in the tibialis anterior and medial gastrocnemius (limb muscles). Rapid and progressive loss of fast motor units was compensated for by an activity dependent re-innervation of faster fiber types with slow motor neurons. (Hegedus et al. 2008).

A similar type of V2a organization has been observed in zebrafish. V2a neurons comprise an important source of motor neuron excitation in the zebrafish spinal cord, and these motor neurons become gradually recruited as the intensity of swimming increases (slow vs fast) (Ampatzis et al. 2013).

Interestingly, V2a neurons can be divided into three functionally distinct classes that selectively synapse onto slow, intermediate, or fast motor neurons (Kimura et al. 2006; McLean et al. 2008; Ampatzis et al.

2014). Moreover, these distinct subclasses of V2a neurons have distinct projection patterns and firing properties. V2a neurons that target slow motor neurons fire in bursts and provide strong non-linear excitation whereas V2a neurons that target fast motor neurons provide weaker excitation (Song et al.

2018). This type of organization is postulated to ensure recruitment order of slow to fast motor neurons during swimming, according to Henneman’s Size Principle. Therefore, a similar circuit may exist in the cervical cord to coordinate ordered recruitment of ARM motor neurons.

140

Alternatively, Type I and Type II V2a neurons may not preferentially project to different types of motor units. Instead, exciting V2a neurons at later stages of ALS may (1) simply have less of an effect on

ARM activity and ventilation because there are simply less V2a neurons and motor neurons left to excite

(Charcot and Joffrey, 1869; Romer et al. 2017) or (2) Type I V2a neurons may be more vulnerable to degeneration and the respiratory system compensates to be reliant on Type II V2a neurons at late stages of ALS to regulate ARM activity. It is possible to test this hypothesis by perfusing the transgenic mouse lines described above at each stage of ALS and performing counts from Stage 0 to Stage 4 to determine if

V2a Type I degenerates faster than V2a Type II if distinct molecular markers for each proposed subtype is discovered.

We also observe substantial effects on ventilation following CNO treatment in G93A V2a-

(Gq)DREADD and G93A V2a-(Gi)DREADD mice in the absence of ARM activity, suggesting that V2a neurons are not dependent on ARM recruitment to alter ventilation. At least four potential explanations exist: 1)

V2a neurons in the brainstem directly or indirectly alter respiratory frequency/brainstem descending drive

2) V2a neuron activity directly or indirectly alters cardiovascular output that can feedback to indirectly alter ventilation 3) V2a neurons directly or indirectly alters primary respiratory muscle activity, such as the diaphragm or 4) V2a neurons directly or indirectly alter airway patency. It is unlikely that the respiratory rhythm generator is altered because there is no change in breathing frequency pre to post

CNO throughout disease progression after activating or silencing V2a neurons. Similarly, Tankersley et al.

2007 showed that blood gasses are not significantly altered throughout disease progression in

SOD1(G93A) ALS model mice and likely do not contribute to increased ventilation at later stages. Since we do not observe any changes in blood oxygen saturation, blood pressure, or heart rate after silencing V2a neurons in healthy mice, it is therefore unlikely that altering V2a excitability in SOD1(G93A) ALS model mice reverses ALS induced changes in ventilation through this mechanism. However, these parameters

141 may be altered due to disease progression and should be measured before and after CNO treatment in

DREADD expressing ALS model mice.

On the other hand, it is possible that V2a neurons may alter the activity of non-ARM respiratory muscles that help maintain ventilation, including the diaphragm and airway muscles. First, we consider the hypothesis that altering V2a neuron excitability differentially affects diaphragm function at early vs. late stages of ALS. It is known that V2a neurons project to phrenic motor neurons in addition to ARM motor neurons (Zhouludeva et al. 2017). Our studies in healthy mice have shown that brainstem and spinal cord V2a neurons are sufficient to cause tonic activity during expiration and cervical spinal V2a neurons are sufficient to increase diaphragm EMG peak amplitude, showing that V2a neurons can alter diaphragm EMG activity. Further, Gordon et al. 2004 showed that phrenic motor neurons show altered morphology as ALS progresses (e.g. increased neuromuscular junction number per spared motor neuron).

V2a connectivity to phrenic motor neurons is also increased two weeks following spinal cord injury, suggesting that both interneurons and phrenic motor can show disease/injury induced plasticity

(Zhouludeva et al. 2017). Similarly, it is possible that V2a neurons also modulate airway muscle motor neurons (i.e. pharyngeal muscles) that control airway patency and may also be subject to plasticity during

ALS disease progression, especially since airway muscles are resistant to weakness until late stages of ALS

(Oliver 1996). Therefore, future studies should simultaneously record diaphragm EMG, airway muscle

EMG, and ARM EMG to determine how altered V2a excitability and ALS stage together affect diaphragm function/airway patency. This is particularly important because loss of airway tone contributes to additional respiratory impairments (Kuhnlein et al. 2008).

Several concerns need to be considered when evaluating the therapeutic potential of V2a neurons to restore or preserve ventilation during ALS progression. Although these data suggest that increasing V2a excitability enhances ventilation the most at early ALS stages, patients and rodent models actually maintain minute ventilation until late stages of the disease – so is activating V2a neurons at early stages

142 really necessary? Some studies suggest that excitotoxicity actually contributes to motor neuron death, so activating V2a neurons might actually exacerbate motor neuron excitotoxicity and diminish survival (for a detailed review, please see Van Den Bosch et al. 2006). For example, Riluzole is one FDA approved drug for ALS, and has been hypothesized (although not proven) to prolong survival due to its anti-glutaminergic modulation properties (Dharmadasa and Kiernan 2018). On the other hand, chronically activating V2a neurons at early stages may prevent the loss of motor neurons by continually providing excitatory input into respiratory motor and strengthening those efferent pathways to respiratory muscles (Schütz 2005) .

For example, exercise training has been shown to delay muscle atrophy in ALS patients (Merico et al.

2018). In order to test whether increasing the excitability of glutamatergic V2a neurons would have a negative or positive effect on motor neuron survival, V2a neurons should be chronically activated during early stages of ALS. The number of both V2a and motor neurons should then be counted at mid and late stages of ALS and Kaplan-Meier survival curves generated to determine if chronically activating V2a neurons shortens or prolongs survival of ALS model mice. Alternatively, silencing V2a neurons benefits ventilation the most at late stages, and should be considered as a therapeutic target for patients with a late/delayed diagnosis. Survival studies should also be performed after chronically silencing V2a neurons.

It is possible that spinal V2a neurons act as an amplifier for descending brainstem drive and silencing V2a neurons may block crucial excitation of phrenic motor neurons in a disease state (despite their ability to still activate ARMs).

Although future work is needed to elucidate the role of V2a neurons in preserving, restoring, or enhancing ventilation in ALS, their established role in controlling respiratory muscle activity and enhancing ventilation both in healthy and SOD1(G93A) ALS model mice suggests they may be viable therapeutic targets to improve breathing following ALS disease onset.

143

Chapter V:

Increasing the Excitability of V2a Neurons

Promotes Recovery of Respiratory Muscle

Activity Following Spinal Cord Injury

144

5.1 Introduction

Respiratory failure is the leading cause of death in spinal cord injury (Berlly & Shem, 2007; de

Paleville et al. 2011; Berlowitz et al. 2016). Connections between brainstem respiratory centers and respiratory motor neurons in the spinal cord are disrupted following injury, leading to loss of respiratory drive. Specifically, high level cervical spinal cord injuries interrupts descending brainstem drive into respiratory motor neurons that control primary pump muscles and accessory respiratory muscles (ARMs)

(e.g. trapezius, scalenes, sternocleidomastoid), contributing to respiratory insufficiency (Zimmer et al.

2007). While mechanical ventilation and diaphragm pacing can be used as treatment options to maintain ventilation acutely following injury, it is costly, severely impairs independence/quality of life, and complications such as pneumonia, atelectasis, sleep apnea, sleep-disordered breathing, respiratory muscle decompensation, and paradoxical breathing can occur (Fishburn et al. 1990; Berlly & Shem, 2007;

Fuller et al. 2013; Berlowitz et al., 2016).

A promising strategy to restore respiratory and locomotor function to spinal cord injury patients involves activating spared or latent pathways (Courtine et al. 2008; Smith and Knikou 2016; Cregg et al.

2017; Warren et al. 2018). One such respiratory pathway is called the crossed phrenic phenomenon (CPP)

(see Chapter I: Introduction, Figure 3). When one half of the spinal cord is lesioned at C2 following a C2 hemisection injury (C2Hx), only the side of the diaphragm ipsilateral to the injury is paralyzed while the contralateral diaphragm remains functionally intact. In 1895, Porter performed a contralateral phrenicotomy (severing the phrenic nerve) following a C2Hx and paralyzed the other half of the diaphragm. Interestingly, this phrenicotomy was not fatal. Instead, it induced asphyxia and restored rhythmic bursting activity to the previously paralyzed ipsilateral diaphragm by activating the latent CPP

(Porter 1895). Other forms of asphyxiation, such as hypoxia/hypercapnia or nasal occlusion, also increase central respiratory drive to temporarily restore function to the previously paralyzed diaphragm (Lee et al.

2015; Ghali 2017; Hernandez-Torres et al. 2017).

145

This pathway has been discovered in multiple species, is easily reproduced, and its activation can restore diaphragm function following a C2Hx. Therefore, the CPP is a well-established model of experimental respiratory circuit plasticity following a high-level spinal cord injury. However, despite being discovered over 100 years ago, it is unknown which neurons mediate the CPP. Identifying specific types of neurons that help activate CPP will provide a new target for therapies to restore breathing following high level cervical spinal cord injuries

Although early studies suggested that descending pathways from the rVRG to the phrenic motor neurons were exclusively monosynaptic (Ellenberger, Feldman, & Goshgarian, 1990; Moreno, Yu, &

Goshgarian, 1992; Dobbins & Feldman, 1994), later studies suggested that the CPP could also include polysynaptic pathways comprised of propriospinal neurons (Lane et al. 2008). Specifically, Lane et al., used a transynaptic tracing agent (pseudorabies virus) to identify spinal neurons presynaptic to phrenic motor neurons that also receive input from anterogradely labeled rVRG neurons in both uninjured and spinal cord injured rats. These results suggested that propriospinal neurons could serve as an anatomical substrate to restore respiratory function following injury. Functional studies have also provided evidence that spinal interneurons may mediate respiratory plasticity following spinal cord injury.

Specifically, Satkunendrarajah et al. 2018 showed that cervical excitatory (vGluT2+) neurons sustain breathing despite phrenic motor neuron loss and are required for adequate ventilation in a non- traumatic spinal cord injury mouse model of cervical myopathy. Moreover, increasing the excitability of these interneurons restored bursting activity to the ipsilateral diaphragm within hours of a C2Hx.

Our lab has previously shown that glutamatergic, ipsilaterally projecting Chx10+ V2a neurons are connected to accessory respiratory muscle motor neurons in the cervical spinal cord and are sufficient to increase their activity and enhance ventilation (Romer et al. 2017). In addition, these V2a neurons may also be important for recovery of breathing following spinal cord injury. Zholudeva et al. 2017 showed that V2a neurons show increased connectivity to phrenic motor neurons two weeks following a C2Hx

146 spinal cord injury, and transplantation of iPSC derived V2a neurons is correlated with increased diaphragm function one month following a C2Hx (Zholudeva et al., 2018). Therefore, this study investigates whether

V2a neurons mediates the CPP to promote recovery of breathing following spinal cord injury.

5.2 Methods

All animal procedures were performed in accordance with and approved by the Institutional Animal Care and Use Committees of Cincinnati Children’s Hospital Medical Center and University of Kentucky.

Animal models to alter excitability of V2a neurons

Cre/+ PNP-tTA/+ TetO-CHRM3/+ V2a-(Gq)DREADD mice (Chx10 ; ROSA ; Tg ) express the Gq excitatory DREADD receptor in brainstem and spinal cord V2a neurons, as well as the eye26. In these mice, the endogenous

Chx10 regulatory region drives expression of Cre recombinase, which removes a loxP flanked stop sequence and allows expression of the tetracycline controlled transactivator (tTA) in Chx10+ V2a neurons.

Binding of the tetracycline controlled transactivator (tTA) to the Tet operator (TetO) drives (Gq)DREADD expression in V2a neurons. Doxycycline, which blocks tTA activity, was not administered to any animals.

Treatment of V2a-(Gq)DREADD mice with clozapine-N-oxide (CNO) activates Gq signaling in V2a neurons, increasing their excitability, as previously described26. Non-DREADD expressing controls include both

(Chx10Cre/+; ROSAPNP-tTA/+; Tg+/+) and (Chx10+/+; ROSAPNP-tTA/+; TgTetO-CHRM3/+) mice.

Cre/+ PNP-CHRM4/+ V2a-(Gi)DREADD mice (Chx10 ; ROSA ) express the (Gi) inhibitory DREADD receptor in brainstem and spinal cord V2a neurons, as well as the eye. Treatment with CNO acutely decreases the excitability of V2a neurons, as previously described31. CNO was delivered by topical application to the exposed intraperitoneal cavity in all experiments. Our prior studies demonstrated that higher doses of

31 CNO are required to silence neurons in V2a-(Gi)DREADD mice (10.0 mg/kg*bw CNO) than are required

26 to increase the excitability of V2a neurons in V2a-(Gq)DREADD mice (1.0 mg/kg*bw CNO) , consistent with other reports32.

147

Although rare, we occasionally observe widespread expression of DREADDs (or other reporters) in Chx10Cre/+ animals, likely due to sporadic expression of Cre recombinase in the germline or early embryo. We used PCR to detect abnormal recombination in tail DNA from V2a-(Gi)DREADD animals as previously described31 and eliminated these animals from our experiments. In order to eliminate V2a-

(Gq)DREADD mice with abnormal recombination, we performed immunohistochemistry on tail tissue using an antibody that recognizes the HA-tag on the (Gq)DREADD (rabbit anti-HA at 1:1000, Cell Signaling

#3724).

C2Hx spinal cord injury surgeries

Lesions to one half of the spinal cord at C2 (C2Hx) were performed in mice under isoflurane anesthesia. Blunt dissection of the paravertebral muscles using cotton tip applicators exposed the back of the skull and the cervical vertebrae. Microscissors were used to perform a laminectomy. Once the spinal cord was exposed, a 30G needle was used to hemisect the left side of the spinal cord just caudal to the C2 dorsal root. The paravertebral muscles were sutured back together following the C2Hx. Dermal adhesive was used to close the skin incision. Immediately following the surgery, mice were injected subcutaneously with 1.0mL carprofen and placed in an incubator set at 29°C overnight. Each cage was supplied with nutritional gel and a water bottle. Mice also received subcutaneous injections of 1.0mL saline twice daily for the first two post-operative care days following surgery.

Terminal bilateral diaphragm EMG recordings

Terminal bilateral diaphragm EMG recordings were performed in separate cohorts of animals at the following time points: prior to, 4 hours, 1 day, or 15 days after the C2Hx surgery. Animals were anesthetized under 1% isoflurane/1% oxygen and placed supine on a heating pad. A 4cm lateral incision through the skin and abdominal muscle just below the xyphoid process exposed the intraperitoneal cavity and diaphragm. Bi-polar electrodes connected to an amplifier (BMA-400 AC/DC Bioamplifier) were

148 inserted into both hemidiaphragms (left and right) to record diaphragmatic activity via Spike2 Data

Analysis software (Cambridge Electronic Design Limited, Cambridge, England). Electrodes were grounded with an additional lead inserted into the abdominal muscle.

A 10-minute baseline of bilateral diaphragm activity was recorded. CNO was then topically applied to the exposed intraperitoneal cavity and the effect on diaphragm EMG was recorded for 60 minutes.

Three 15s nasal occlusions, each separated by a 10s break, were performed to induce the CPP at the end of the experiment. Diaphragm activity was allowed to return to the pre-nasal occlusion activity level before stopping the recording, removing the electrodes, and harvesting the brain and spinal cord. All mice were maintained at 1% isoflurane/1% oxygen for the entire recording.

Assessing the extent of the C2Hx

The harvested brains and spinal cords were placed in 4%PFA in 10x PB for 24h. The tissue was then transferred into 1x PBS and washed on a rocker O/N before being placed in cryoprotectant 30% sucrose O/N. Cervical segments 1-6 were embedded in OCT compound and stored at-80C until sectioned. Cervical tissue was sectioned at 20uM thickness using a cryostat. Every third section was stained with cresyl violet to measure spared spinal cord tissue33. Briefly, sectioned spinal cord tissue was dehydrated, placed in xylenes and rehydrated before staining in cresyl violet for 6 minutes. Tissue was then dehydrated again, placed in xylenes and coverslipped with permount. This cervical tissue was then imaged with the Zeiss Axio Scan.Z1 at C2 to assess the completeness and consistency of the C2Hx lesions. The spinal cord section with the greatest amount of damage was assessed for the extent of injury for each animal as previously described33. Briefly, ImageJ was used to trace the outline and measure the area of the spared gray matter and spared white matter. The total area of the spinal cord was estimated by multiplying the area of the gray matter or white matter in the uninjured half of the

149 spinal cord by 2. The percent of injury out of the total area of the spinal cord was separately calculated for the gray and white matter using the following equation:

% injury = ((2*HemicordArea)-(TotalSparedTissueArea))/(2*HemicordArea) * 100

Diaphragm EMG analysis in anesthetized mice

The diaphragm EMG signal was amplified (gain 2000x) and bandpass filtered (30-3000Hz) with a sampling frequency of 6.25kHz. EMG signals were further processed to remove DC noise and subsequently rectified and integrated over a 50ms window. Electrocardiogram (ECG) artifact was digitally filtered out using the “ECGDelete02” Spike2 script (Cambride Electronic Design) in traces where the ECG amplitude exceeded the diaphragm EMG amplitude. ECG was distinguished from EMG activity by its regular occurrence (~9-11Hz) and its characteristic shape with definitive QRST peaks34. The rectified and integrated signal was used to calculate the raw EMG peak amplitude for at least 30s before and after CNO treatment. These values were normalized to the raw EMG peak amplitude recorded during maximum ventilatory effort (nasal occlusion) in order to reduce intra-animal variability35.

Respiratory cycle and inspiratory time were also analyzed for each breath using the rectified and integrated signal from the same time period analyzed for peak amplitude. The regularity of breathing was assessed by calculating the coefficient of variation of bursting frequency (CVf) (Crone et al. 2012).

CVf for each animal reflects the average value over each 30s analysis period.

Cycle Triggered Averaging

We performed cycle triggered averaging (CTA) on rectified and integrated diaphragm EMG traces to correlate respiratory cycle phase activity (inspiration vs. expiration) from the ipsilateral diaphragm with that seen in the intact contralateral diaphragm following CNO administration. Spike2 was used to generate a memory channel that marked the onset of expiration (channel m1) in the contralateral diaphragm based on the falling threshold amplitude detected from the processed contralateral diaphragm channel for each

150 individual file. We next generated waveform averages for the contralateral and ipsilateral processed diaphragm channel for at least a 20s period. Each waveform was generated using the previously generated contralateral memory channel (m1) as the reference channel. The offset value was specified as 0.5s while the width was set to the maximum respiratory period in the file in order to encompass the entire respiratory cycle. Each waveform generated reflects the average and standard deviation for all breaths analyzed within the 20s analysis period. The final waveform graphs shown here highlight the inspiratory and expiratory periods. The inspiratory period was defined as the average duration of the inspiratory time during the analyzed period that was then offset from the onset of expiration. This method was also used to classify baseline ipsilateral diaphragm activity as spontaneous recovery or no spontaneous recovery two weeks following a C2Hx.

Surgical Implantation of Telemetry Devices to Record Diaphragm and ARM EMG in Conscious Mice

Telemetry implants were performed in P65 (and older) adult mice (see Chapter II, Section 2.2) prior to injury. Briefly, double channel telemetry devices (F20-EET) transmitters from Data sciences International were implanted to chronically record diaphragm and scalene ARM EMG from respiratory muscles of V2a-

(Gq)DREADD mice and non-DREADD expressing controls prior to and following a C2Hx spinal cord injury.

EMG Acquisition in Conscious Animals.

EMG acquisition and digital video recording was performed as described previously using DSI Ponemah

Physiology Platform Acquisition software v.5.20 (see Chapter II, section 2.2). Both (Gq)DREADD and non-

DREADD expressing control mice received 1.0 mg/kg*bw CNO.

Statistical Analysis

Diaphragm EMG peak amplitude and inspiratory time for the ipsilateral and contralateral diaphragm before and after CNO treatment were analyzed with separate paired t-tests. Data that failed the normality test were analyzed with the Wilcoxin signed rank test (indicated in figure legend when applicable). CVf

151 between the ipsilateral and contralateral diaphragm EMG under one condition only were measured using student t-tests (i.e. CVf between the ipsilateral and contralateral diaphragm post CNO). Data that failed the normality test were analyzed with the non-parametric equivalent Mann-Whitney U-test instead

(indicated in figure legend when applicable). A one-way ANOVA was used to compare the average amount of time it took for CNO to produce a response from the ipsilateral diaphragm in V2a-(Gq)DREADD mice among the different time points: 4 hours, 1 day and 15 days post C2Hx. A two-way ANOVA was used to compare the percentage of time spent in each cycle triggered average pattern among the different time points following the C2Hx spinal cord injury. Chi-square tests of independence were used to analyze how the presence of the (Gq)DREADD receptor in V2a neurons are related to two mutually exclusive outcomes

(i.e. tonic activity vs. no tonic activity). When the expected outcome from a Chi-square test was less than

5, a Fisher’s Exact test was used instead to accurately analyze a 2x2 contingency table with small sample sizes. Data are reported as average ± standard error of the mean. Significance level is set at p=0.05 with

(*)p<0.05.

5.3 Results

Increasing the excitability of V2a neurons restores bursting activity to a previously paralyzed diaphragm.

In order to test the impact of V2a neuron activity on diaphragm function after spinal cord injury,

Cre/+; tTA/+; CHRM3/+ we generated adult (P73-145) V2a-(Gq)DREADD mice (Chx10 ROSA TetO ) in which the excitability of V2a neurons in the brainstem and spinal cord is increased following treatment with CNO, as previously described (Romer et al. 2017). We lesioned the left side of the spinal cord at C2 to paralyze the diaphragm ipsilateral to the injury (Figure 1A). The location and completeness of each cervical lesion was assessed after the termination of the experiment by histological staining (cresyl violet) of cervical spinal cord sections (Figure 1B). The extent of injury to white and gray matter is expressed in percent of the total cord injured. There is no difference in the extent of injury between V2a-(Gq)DREADD mice at 4 hours following injury, V2a-(Gq)DREADD mice at 1 day following injury, or non-DREADD expressing

152 controls 1 day following injury (Gray matter: 36.5 ± 2.0% vs. 30.3 ± 4.4% vs. 27.1 ± 4.8%, respectively; one- way ANOVA, p=0.869; White matter: 29.9 ± 1.0% vs. 26.3 ± 5.0% vs. 29.2 ± 4.9%, respectively; one-way

ANOVA, p=0.937).

In freely breathing animals under isoflurane anesthesia, we exposed the diaphragm and performed electromyography (EMG) recordings from the ipsilateral and contralateral diaphragm (relative to the side of the C2Hx) one day following injury to assess diaphragm function (representative trace shown in Figure 1C, -CNO). We verified that the lesion was sufficient to produce complete paralysis of the diaphragm ipsilateral (but not contralateral) to the lesion during a 10-minute baseline recording prior to

CNO treatment. Animals without a functionally complete C2Hx were excluded from analysis (1/10 animals). We then topically applied CNO (1.0 mg/kg*bw) to increase the excitability of V2a neurons and assessed diaphragm activity for 1 hour. Rhythmic bursting activity was observed in the previously paralyzed ipsilateral diaphragm in 100% of animals following CNO treatment (n=9) (representative trace shown in Figure 1C, +CNO). This restored bursting activity continued until the animal was sacrificed and tissue was harvested. To demonstrate that the observed effects were due to the effects of CNO on

(Gq)DREADD expressing V2a neurons instead of off target effects of CNO, we performed the same experiments on non-DREADD expressing mice. The ipsilateral diaphragm remained paralyzed in 100% of non-DREADD expressing control animals treated with CNO (1.0 mg/kg*bw) (4/4 Chx10+/+; ROSAtTA/+;

TetOCHRM3/+ and 3/3 Chx10Cre/+; ROSAtTA/+; TetO+/+ mice) (representative trace shown in Figure 1F). Two control animals were excluded from analyses due to incomplete functional paralysis prior to CNO treatment.

In order to quantitate the degree of recovery in V2a-(Gq)DREADD mice and controls, the CPP was induced by nasal occlusion (3 x 15s, with 15s in between) in all mice at the end of the recording session

(representative trace shown in Figure 1G). As expected, nasal occlusion elicited bursting activity from the previously paralyzed diaphragm in 100% of Chx10+/+;ROSAtTA/+;TetOCHRM3/+ (n=4) and

153

Chx10Cre/+;ROSAtTA/+;TetO+/+ (n=3) non-DREADD expressing control mice 1 day after injury. Nasal occlusion also increased the recovered diaphragm EMG peak amplitude in 100% of V2a-(Gq)DREADD mice 1 day after injury (n=9). This procedure allowed us to normalize the activity of each hemidiaphragm to the maximal inspiratory drive produced during nasal occlusion and control for potential differences in EMG amplitude due to variations in electrode placement in different animals. We calculated the average EMG peak amplitude for each hemidiaphragm prior to and following CNO treatment and normalized each value to the peak amplitude for the same hemidiaphragm during nasal occlusion (Mantilla et al. 2011). Because the EMG activity during maximal drive is different in the ipsilateral versus contralateral diaphragm due to the injury, our analysis does not directly compare the ipsilateral and contralateral diaphragm EMG peak amplitude.

EMG recordings and nasal occlusion were also performed on uninjured control V2a-(Gq)DREADD mice to establish the EMG peak amplitude prior to injury. The EMG peak amplitude in both hemidiaphragms of uninjured mice reaches 63% of the maximal amplitude observed during nasal occlusion (Jensen et al., 2019). In C2Hx animals, the diaphragm ipsilateral to injury becomes paralyzed

(0%), whereas the contralateral diaphragm shows near maximal amplitude (97.5%). CNO treatment restores activity to the ipsilateral diaphragm to 59.4% of maximal amplitude (p<0.001), with a corresponding decrease in contralateral diaphragm activity to 59.9% maximal amplitude (similar to pre- injury levels) (p=0.032) (Figure 1D)(representative trace over a longer time span shown in Figure 1M).

However, the inspiratory time of the ipsilateral diaphragm post CNO is only 70.5% of the contralateral diaphragm post CNO (ipsi: 12.2 ± 0.7ms vs. contra: 17.3 ± 0.7ms, student t-test, p<0.002), indicating that the ipsilateral diaphragm does not fully recover to pre-injury levels of activity (Figure 1E). Non-DREADD expressing controls (4/4 Chx10+/+;ROSAtTA/+;TetOCHRM3 mice and 3/3 Chx10Cre/+;ROSAtTA/+;TetO+/+) do not show recovery of diaphragm activity following CNO treatment. Moreover, there is no change in diaphragm

EMG peak amplitude (p=0.852) or burst duration (p=0.108) in the contralateral hemidiaphragm following

154 treatment with CNO (n=4 Chx10+/+;ROSAtTA/+;TetOCHRM3) (Figure 1H). Because CNO does not differentially affect the ipsilateral diaphragm in Chx10+/+;ROSAtTA/+;TetOCHRM3 mice compared to

Chx10Cre/+;ROSAtTA/+;TetO+/+ mice, these two genotypes are pooled together for analysis. A Fisher’s Exact test shows that CNO treatment in V2a-(Gq)DREADD mice is significantly related to recovery of bursting activity from the ipsilateral hemidiaphragm (p=0.001), demonstrating that increasing the excitability of

V2a neurons restores function to the previously paralyzed hemidiaphragm.

We next tested whether we could achieve similar recovery within hours of a C2Hx to evaluate the potential of therapies targeting V2a neurons to restore breathing shortly after a traumatic injury. We used the same experimental paradigm described above to measure diaphragm EMG and increase the excitability of V2a neurons with CNO 4 hours after injury. All of the included animals (3/3) showed strong recovery of ipsilateral hemidiaphragm bursting activity following CNO treatment with an average normalized peak amplitude of 73±14% maximal activity (representative trace shown in Figure 1J). Since it is known that CPP pathways are not active in 100% of animals this soon after injury (Minor et al. 2006), we excluded animals from analysis that did not exhibit rhythmic bursting activity following nasal occlusion

(4/9 animals), as well as animals exhibiting functionally incomplete paralysis of the hemidiaphragm ipsilateral to the C2Hx prior to CNO treatment (2/9 animals). Unlike animals tested one day following injury, the contralateral diaphragm EMG peak amplitude did not decrease as the ipsilateral diaphragm recovered function (p=0.281) (Figure 1K). Similar to the results seen in V2a-(Gq)DREADD mice tested one day following injury, the burst duration of the ipsilateral diaphragm is lower than the burst duration of the contralateral diaphragm (ipsi: 15.9±0.8ms vs. contra: 23.4±2.0ms, p=0.0565) (Figure 1L), suggesting that recovery is not to pre-injury levels. However, these data clearly demonstrate that increasing the excitability of V2a neurons can restore rhythmic bursting activity to a previously paralyzed diaphragm as soon as 4 hours following a C2Hx spinal cord injury.

155

Figure 1: Increasing the excitability of V2a neurons restores bursting activity to a previously paralyzed diaphragm.

(A)Timeline of the experimental paradigm. (B) Representative image of three different C2Hx injured spinal cords stained with cresyl violet. The percentage of the total spinal cord gray matter is denoted in the bottom left hand corner. (C) Representative trace showing the contralateral (top) and ipsilateral (bottom) diaphragm before (left) and after (right) treatment with 1.0 mg/kg*bw CNO in a V2a-(Gq)DREADD animal one day post C2Hx. The diaphragm EMG peak amplitude (D) and inspiratory time (E) were measured and quantified in V2a-(Gq)DREADD mice before and after CNO treatment one day post C2Hx. (F) Representative trace showing the contralateral (top) and ipsilateral (bottom) diaphragm before (left) and after (right) treatment with 1.0 mg/kg*bw CNO in a Chx10Cre/+;ROSAtTA/+;TetOCHRM3/+ non-DREADD expressing control animal. (G) Representative trace from a Chx10Cre/+;ROSAtTA/+;TetOCHRM3/+ non-DREADD expressing control animal showing that nasal occlusion restores bursting activity to the previously paralyzed diaphragm. The diaphragm EMG peak amplitude (H) and inspiratory time (I) were measured and quantified in all non-DREADD expressing controls one day post C2Hx. (J) Representative trace showing the contralateral (top) and ipsilateral (bottom) diaphragm before (left) and after (right) treatment with 1.0 mg/kg*bw CNO in a V2a-(Gq)DREADD animal four hours after a C2Hx. The diaphragm EMG peak amplitude (K) and inspiratory time (L) were measured and quantified in V2a-(Gq)DREADD mice 4 hours after a C2Hx. (M) Example trace from a V2a-(Gq)DREADD animal 1 day following injury showing how a gradual increase in ipsilateral diaphragm EMG amplitude coincides with a gradual decrease of the contralateral diaphragm EMG amplitude. White bars = contralateral diaphragm. Black bars = ipsilateral diaphragm. Red X = paralyzed ipsilateral diaphragm where no bursting activity is detected. White arrowheads in (F) indicate ECG artifact. Black arrowheads in (F) indicate diaphragm EMG bursts. The ipsilateral and contralateral diaphragm data were analyzed separately with a parametric paired t-test (D, E- contralateral, H-I and K-L) or non-parametric Wilcoxin Signed-Rank test (E-ipsilateral), (*)p<0.05. Increasing the excitability of V2a neurons restores diaphragm function two weeks following a C2Hx.

Zholudeva et al. 2017 showed that V2a neuron connectivity to phrenic motor neurons is increased two weeks following a C2Hx. Therefore, we assessed the impact of altering V2a neuron function following

156 this period of respiratory circuit plasticity. V2a neuron excitability was increased two weeks following injury and we compared the effects on diaphragm function to our findings at 1 day and 4 hours after injury. Diaphragm EMG recordings confirmed that the ipsilateral diaphragm was still paralyzed two weeks following injury in 7/12 animals whereas 5/12 V2a-(Gq)DREADD mice showed spontaneous rhythmic bursting diaphragm activity before CNO treatment. We first assessed the effects of increasing the excitability of V2a neurons in the 7 animals that did not show spontaneous recovery. CNO restored rhythmic bursting activity to the previously paralyzed ipsilateral diaphragm in all seven animals

(representative trace shown in Figure 2A). Although CNO restored bursting activity to the ipsilateral diaphragm, the contralateral diaphragm EMG peak amplitude did not change (p=0.895) (Figure 2B). CNO did not restore diaphragm activity to the ipsilateral diaphragm or alter contralateral diaphragm EMG peak amplitude or inspiratory time in non-DREADD expressing controls, which include 2 Chx10+/+; ROSAtTA/+;

TetOCHRM3/+ mice and 2 Chx10Cre/+; ROSAtTA/+; TetO+/+ mice (diaphragm EMG peak amplitude: paired t-test, p=0.895; inspiratory time: paired t-test, p=0.181) (Figure 2D-E). There is no difference in the ipsilateral diaphragm maximal amplitude reached following treatment with CNO among V2a-(Gq)DREADD animals 4 hours, 1 day or 15 days following injury (one way ANOVA, p=0.339). Thus, comparable recovery of diaphragm function can be elicited within 2 weeks following injury in animals that did not display spontaneous recovery of function.

We next examined the effects of increasing the excitability of V2a neurons in the 5 V2a-

(Gq)DREADD mice that showed spontaneous recovery of rhythmic bursting activity in the ipsilateral diaphragm two weeks following injury (Figure 2F). All mice were monitored to confirm that chest movement ceased on the ipsilateral side immediately following the C2Hx, so it is unlikely that spontaneous activity is due to an incomplete hemisection. Moreover, there is no difference in the extent of injury between animals that did not show spontaneous recovery and animals that did spontaneous recovery 2 weeks following injury in V2a-(Gq)DREADD mice (no spontaneous recovery: 24.1 ± 5.6% gray matter and

157

22.6 ± 5.8% white matter vs spontaneous recovery: 29.9 ± 3.4% gray matter and 26.6 ± 4.1% white matter; p=0.458 and p=0.337) and non-DREADD expressing controls (no spontaneous recovery: 26 ± 8% gray matter and 14 ± 9% white matter vs spontaneous recovery: 33±5% gray matter and 28±5% white matter; p=0.492 and p=0.217). In contrast to the animals that did not show spontaneous recovery, increasing the excitability of V2a neurons had no effect on the ipsilateral or contralateral diaphragm EMG peak amplitude (ipsi: p=0.229; contra: p=0.182) (Figure 2G) or inspiratory time (ipsi: p=0.725; contra: p=0.520)

(Figure 2H) in these animals. Similarly, no effects of CNO were observed in the 5/5 non-DREADD expressing controls (4/4 Chx10+/+; ROSAtTA/+; TetOCHRM3/+ mice and 1/1 Chx10Cre/+;ROSAtTA/+;TetO+/+) that showed spontaneous recovery (Figure 2I-J). Fisher’s Exact test shows that genetic background (i.e.

DREADD expression) does not influence the proportion of animals that exhibit spontaneous recovery two weeks following injury (p=0.670).

Despite the lack of a significant impact on diaphragm EMG peak amplitude, we did observe an increase in muscle activity during the expiratory period in 1/5 V2a-(Gq)DREADD animals treated with CNO, a pattern hereafter referred to as “tonic activity” (Figure 2F). Following CNO treatment, we also observed both tonic activity and rhythmic bursting activity in 2/7 V2a-(Gq)DREADD mice that did not show spontaneous recovery of function two weeks post C2Hx, 6/9 V2a-(Gq)DREADD mice treated with CNO one day post C2Hx, and 3/3 V2a-(Gq)DREADD mice treated with CNO 4 hours post C2Hx. Tonic activity was also previously observed in 5/5 uninjured V2a-(Gq)DREADD mice following administration of CNO (Jensen et al. 2019). In contrast, tonic activity in the diaphragm was rarely observed in non-DREADD expressing mice following CNO treatment two weeks after injury (0/4 Chx10Cre/+; ROSAtTA/+; TetO+/+ and 1/6 Chx10+/+;

ROSAtTA/+; TetOCHRM3/+ mice) or one day after injury (0/4 Chx10+/+; ROSAtTA/+; TetOCHRM3/+ and 0/3 Chx10Cre/+;

tTA/+ +/+ ROSA ; TetO ). A Chi-square test of independence shows that (Gq)DREADD expression is significantly related to the presence of tonic activity following CNO (χ2(1) = 11.760, p<0.001). Thus, increasing the excitability of V2a neurons has a minimal impact on inspiratory diaphragm activity in mice that have

158 already recovered diaphragm function, but tonic activity during the expiratory phase may be elicited by increasing V2a neuron activity in healthy mice or following injury.

Figure 2: Increasing the excitability of V2a neurons restores diaphragm function two weeks following a C2Hx.

(A)Representative trace showing the contralateral (top) and ipsilateral (bottom) diaphragm before (left) and after (right) treatment with 1.0 mg/kg*bw CNO in a V2a-(Gq)DREADD animal that does not show spontaneous recovery two weeks after a C2Hx. The diaphragm EMG peak amplitude and inspiratory time were measured and quantified in V2a-(Gq)DREADD mice (B and C, respectively) and non-DREADD expressing controls (D and E, respectively) that did not show spontaneous recovery two weeks following a C2Hx. (F)Representative trace showing the contralateral (top) and ipsilateral (bottom) diaphragm before (left) and after (right) treatment with 1.0 mg/kg*bw CNO in a V2a-(Gq)DREADD animal that showed spontaneous recovery two weeks after a C2Hx. The diaphragm EMG peak amplitude and inspiratory time were measured and quantified in V2a-(Gq)DREADD mice (G and H, respectively) and non-DREADD expressing controls (I and J, respectively) that showed spontaneous recovery two weeks following a C2Hx. White bars = contralateral diaphragm. Black bars = ipsilateral diaphragm. Red X = paralyzed ipsilateral diaphragm where no bursting activity is detected. The ipsilateral and contralateral diaphragm data were analyzed separately with a parametric paired t-test (C and D) or non-paramteric Wilcoxin Signed-Rank test (B and E), (*)p<0.05.

159

Cycle triggered averaging reveals that the predominant pattern of diaphragm activity after activating V2a neurons is rhythmic bursting during inspiration.

We observed a waxing and waning of tonic activity throughout the hour long post-CNO recording period (Figure 3A) in all animals that showed tonic activity. In order to better characterize and quantify the different patterns of diaphragm activity observed following CNO treatment, we performed cycle triggered averaging (CTA) to correlate ipsilateral diaphragm activity with the rhythmic inspiratory burst activity observed in the intact contralateral diaphragm. Five different CTA patterns were observed in the ipsilateral diaphragm using this analysis: (1) paralysis characterized by a complete lack of EMG activity

(Figure 3B) (2) rhythmic bursting characterized by ipsilateral diaphragm activity only during inspiration and synchronous with contralateral diaphragm bursting activity (Figure 3C) (3) rhythmic bursting during inspiration mixed with expiratory tonic activity (Figure 3D) (4) tonic activity only during expiration and no activity during inspiration (Figure 3E) and (5) high amplitude tonic activity during both the expiratory and inspiratory phases of respiration (Figure 3F).

We measured the average amount of time each V2a-(Gq)DREADD mouse spent in each CTA pattern after CNO administration 4 hours, 1 day and 15 days following the C2Hx spinal cord injury. Animals who did not show bursting activity from the previously paralyzed ipsilateral diaphragm following nasal occlusion were excluded from analysis (4 V2a-(Gq)DREADD mice 4 hours post C2Hx). Only mice that did not show spontaneous recovery two weeks following the C2Hx were analyzed so that animals from all three time points exhibited a paralyzed diaphragm during the baseline recording. CTA pattern frequency analysis began as soon as the first non-paralyzed CTA pattern emerged after CNO administration.

Following this point, the beginning, end and duration of each CTA pattern was recorded. There is no difference among groups (4 hours, 1 day or 15 days) in the amount of time it took for the first non- paralyzed CTA pattern to emerge (4 hours: 14.2 ± 5.4 min. vs 1 day: 18.7 ± 4.5 min. vs 15 days: 8.7 ± 2.0 min., one-way ANOVA, p=0.266). However, there was a statistically significant interaction between the

160 time following a C2Hx (4 hours, 1 day or 15 days) and the percent of time each animal spent in the different

CTA categories (two-way ANOVA, p=0.047). Most of the time was spent in the rhythmic only bursting pattern for all three time points after injury (Figure 3G). However, the amount of time of time spent in the rhythmic only bursting pattern was significantly greater in the 15 day and 1 day post C2Hx group compared to the 4 hours post C2Hx group (two-way ANOVA, Holm-Sidak post-hoc test, p<0.05 in both comparisons).

In addition to the percent of time spent in each CTA pattern, we also measured the average duration of each CTA pattern before transitioning to a different pattern at each time point following the

C2Hx injury. The duration of the rhythmic only CTA pattern increases as the time following injury increases from 4 hours (6.5 ± 4.0 mins, n=3) to one day (12.9 ± 10.9 mins, n=9) to 15 days (25.9 ± 15.9 mins, n=6) post C2Hx (p<0.05) while the average duration of all other CTA patterns decreases, with a significant interaction between the time following the C2Hx and the duration spent in each CTA segment (two-way

ANOVA, p=0.006). One day and 15 days following injury, the rhythmic only CTA pattern has the greatest duration over any other CTA pattern during that same time point, lasting an average of at least 12 minutes before transitioning to a different pattern (p<0.05) (Figure 3H). Thus, rhythmic inspiration is the predominant pattern of ipsilateral diaphragm activity elicited by increasing the excitability of V2a neurons following a C2Hx injury.

Silencing V2a neurons impairs the ability to restore diaphragm function via nasal occlusion two weeks following injury In order to test the contribution of V2a neurons to recovery of respiratory function following injury, we performed a C2Hx on 15 V2a-(Gi)DREADD mice. This mouse line expresses the inhibitory

(Gi)DREADD and decreases the excitability (i.e. “silences”) V2a neurons in the spinal cord and brainstem following treatment with 10.0 mg/kg*bw CNO (Jensen et al. 2019). Two weeks following injury, we measured ipsilateral and contralateral diaphragm EMG to identify mice that did not show spontaneous

161

Figure 3: Cycle triggered averaging reveals that the predominant pattern of diaphragm activity after activating V2a neurons is rhythmic bursting during inspiration.

(A) Increasing the excitability of V2a neurons in V2a-(Gq)DREADD can cause tonic activity that waxes and wanes with rhythmic activity. (B-F) Cycle triggered averaging (CTA) analysis was performed to detect and characterize five different CTA patterns present after increasing the excitability of V2a neurons with CNO. Gray line = average and orange lines = standard deviation for all breaths analyzed during the 20s period. Inspiration is highlighted in gray while expiration is highlighted in yellow in the CTA graphs. Raw diaphragm EMG traces that correspond to each diaphragm CTA graph are shown below the CTA graphs in each example from the same V2a-(Gq)DREADD animal one day following injury. (B) The contralateral diaphragm bursting activity remains intact while the ipsilateral diaphragm is paralyzed. (C) Only rhythmic activity is detected in the ipsilateral diaphragm that is synchronous with bursting activity from the contralateral diaphragm. (D) Rhythmic bursting activity is interspersed with tonic activity that is present during the expiratory phase of the respiratory cycle in the ipsilateral diaphragm. The main ipsilateral bursting activity is still synchronous with the contralateral diaphragm. (E) Tonic motor unity activity from the ipsilateral diaphragm is observed only during the expiratory phase of the respiratory cycle. EMG activity is absent during inspiration. (F) High amplitude motor unit activity is present throughout the respiratory cycle. (G) Graph showing the percent of total time that is spent in each CTA pattern following CNO administration in V2a- (Gq)DREADD mice at each time point following the C2Hx injury. (H) Graph showing the average duration that is spent in each CTA pattern before transitioning to a different CTA pattern following CNO administration in V2a-(Gq)DREADD mice at each time point following the C2Hx injury. All values shown represent the average ± standard deviation. White bars = paralyzed CTA pattern. Black bars = rhythmic only CTA pattern. Orange bars = rhythmic+tonic CTA pattern. Green bars = expiratory only tonic CTA pattern. Blue bars = expiratory and inspiratory tonic CTA pattern. Statistics: (G) and (H) are analyzed with a two-way ANOVA. *p<0.05 among the rhythmic only CTA pattern and all other CTA patterns at that time point post C2Hx. # p<0.05 between the parameters specified by the black bars. recovery of ipsilateral diaphragm activity (5/15 animals) and performed nasal occlusion. Nasal occlusion has previously been used to elicit the crossed phrenic phenomenon (activation of rhythmic bursting from the previously paralyzed ipsilateral diaphragm) in rodents following a C2Hx (Hernandez-Torres et al.

162

2017). We then silenced V2a neurons by administration of 10.0 mg/kg*bw CNO and repeated the nasal occlusion (timeline shown in Figure 4A). The representative traces in Figure 4B displays diaphragm EMG activity during the maximal ventilatory effort elicited by nasal occlusion before (top) and after (bottom) administration of 10.0 mg/kg*bw CNO. Prior to silencing V2a neurons, nasal occlusion elicits robust rhythmic bursting activity in the ipsilateral diaphragm. However, the ipsilateral diaphragm EMG peak amplitude following nasal occlusion is dramatically reduced by 59± 5% (one-way t-test, p<0.001) after

CNO treatment. Notably, this reduction is significantly greater than the percent change observed in the intact contralateral diaphragm, which is only decreased by 14 ± 10% (n=5, student t-test, p=0.007) (Figure

4F). These results indicate that V2a neurons are critical components of pathways elicited by asphyxia to restore rhythmic diaphragm function below a C2Hx injury.

V2a neurons may be one of many mediators of spontaneous recovery

Spontaneous tonic activity during the baseline recording can be observed in the ipsilateral diaphragm EMG of some V2a-(Gi)DREADD animals (3/15) two weeks following a C2Hx injury prior to CNO treatment (Figure 5A, left). We silenced V2a neurons in V2a-(Gi)DREADD mice two weeks after a C2Hx to test whether V2a neurons contribute to this pattern of activity. This tonic activity was abolished in 3/3

V2a-(Gi)DREADD mice after silencing V2a neurons with CNO (Figure 5A, right). We can easily distinguish

EMG activity from ECG artifact (inset of Figure 5A, bottom) and show that ECG artifact alone remained following CNO treatment. These data indicate that V2a neurons contribute to tonic diaphragm activity during expiration that can arise during recovery from injury.

Spontaneous recovery of rhythmic bursting activity in the ipsilateral diaphragm synchronous with the contralateral diaphragm was observed in 7/15 V2a-(Gi)DREADD animals two weeks following a C2Hx.

Silencing V2a neurons drastically reduced the ipsilateral diaphragm EMG peak amplitude by 48.9% in 1/7

163

Figure 4: Silencing V2a neurons impairs the ability to restore diaphragm function via nasal occlusion two weeks following injury.

(A) Timeline of the experimental paradigm. (B) Some V2a-(Gi)DREADD animals did not show spontaneous recovery two weeks following injury. Nasal occlusion (gray box) was performed before (top trace) and after (bottom trace) treatment with 10.0 mg/kg*bw CNO. The percent change in diaphragm EMG peak amplitude is quantified in (C) White bar = contralateral diaphragm and black bar = ipsilateral diaphragm. Statistics: Data were analyzed with a t-test, *p<0.05.

164

V2a-(Gi)DREADD mice (representative trace in Figure 5B). However, in 6/7 animals, there was not a significant change in the diaphragm EMG peak amplitude before and after silencing V2a neurons (Ipsi: n=7, p=0.282; Contra: n=7, p=0.954) (representative trace in Figure 5C and quantified in Figure 5D). It is unlikely that variation in the extent of injury contributes to the three distinct baseline diaphragm activities observed in V2a-(Gi)DREADD mice because there is no difference in the percent of injury observed in the white matter (no spontaneous recovery: 38.3 ± 5.4% vs spontaneous tonic recovery: 33.4 ± 1.7% vs spontaneous rhythmic bursting recovery: 36.4 ± 6.1%, one-way ANOVA, p=0.897) or gray matter (no spontaneous recovery: 32.3 ± 6.5% vs spontaneous tonic recovery: 31.1 ± 1.1% vs spontaneous rhythmic bursting recovery: 30.0 ± 4.5%, one-way ANOVA, p=0.910). These results indicate that V2a neurons can contribute to spontaneous recovery in some animals, but that other animals rely on additional and/or different neurons.

Increasing the excitability of V2a neurons does not adversely affect respiratory rhythm generation.

Our previous studies demonstrated that V2a neurons are critical for regulating the frequency and regularity of breathing in neonatal animals, but not in healthy adult animals (Crone et al. 2012; Jensen et al. 2019). However, it is not known if increasing the excitability of V2a neurons impacts the regularity of respiratory rhythm following injury. To measure the regularity of respiratory rhythm, we calculated the coefficient of variation of bursting frequency (CVf) from the contralateral diaphragm before and after CNO treatment in V2a-(Gq)DREADD mice prior to injury, 4 hours post C2Hx, 1 day post C2Hx, and 2 weeks post

C2Hx (Crone et al. 2012; Jensen et al. 2019). We show that the regularity of the bursting activity from the contralateral diaphragm does not change after exciting V2a neurons in uninjured mice (n=5, p=0.329)

(Figure 6A), 4 hours following injury (n=3, p=0.857) (Figure 6B), one day (n=5, paired t-test, p=0.737)

(Figure 6C) or two weeks (n=7, Mann-Whitney U-test, p=0.484) (Figure 6D) following injury. Furthermore, there is no difference between the recovered ipsilateral and contralateral diaphragm CVf after 1.0 mg/kg*bw CNO 4 hours post C2Hx (n=3, student t-test, p=0.615), one day post C2Hx (n=5, p=0.373), or

165

Figure 5. V2a neurons may be one of many mediators of spontaneous recovery.

(A) Representative trace from a V2a-(Gi)DREADD animal that shows spontaneous tonic activity during the baseline recording two weeks following a C2Hx. Silencing V2a neurons with 10.0 mg/kg*bw CNO abolishes baseline motor unit activity. Blue inset demonstrates the ability to distinguish ECG (indicated by PQRST peaks) from EMG activity (blue arrow heads). (B) Representative trace from a V2a-(Gi)DREADD animal that exhibited spontaneous recovery of rhythmic bursting activity two weeks following injury. One (of seven) V2a-(Gi)DREADD mouse showed a decrease in EMG peak amplitude following treatment with 10.0 mg/kg*bw CNO. (C) Most V2a-(Gi)DREADD animals (6/7) that showed spontaneous recovery of rhythmic bursting activity two weeks following injury did not show a decrease in EMG peak amplitude from the ipsilateral or contralateral diaphragm following treatment with 10.0 mg/kg*bw CNO. Diaphragm EMG peak amplitude from all V2a-(Gi)DREADD animals that showed spontaneous recovery of bursting activity is quantified in (D). White bar = contralateral diaphragm and black bar = ipsilateral diaphragm. Statistics: (F) was analyzed with a student t-test and (G) was analyzed with a paired t-test, *p<0.05.

166 two weeks post C2Hx (n=7, p=0.805) following injury. Thus, increasing the excitability of V2a neurons does not adversely affect respiratory rhythm generation.

Figure 6: Increasing the excitability of V2a neurons does not adversely affect respiratory rhythm generation.

The coefficient of variation of breathing frequency (CVf) was analyzed in the ipsilateral and contralateral diaphragm of all V2a-(Gq)DREADD before and after CNO treatment (A) prior to injury (uninjured) (B) 4 hours after a C2Hx (C) 1 day after a C2Hx and (D) two weeks after a C2Hx. Paired t- tests were used to compare the contralateral diaphragm CVf pre to post CNO. Parametric student t-tests (A-C) and non-parametric Mann-Whitney U-tests (D) were used to compare the contralateral and ipsilateral diaphragm following CNO treatment. Repeatedly increasing V2a excitability following injury may restore diaphragm function back to pre-injury levels faster than spontaneous recovery alone

We have demonstrated that acutely activating V2a neurons is sufficient to restore diaphragm function in anesthetized mice following a C2Hx injury. However, these acute recordings are performed under anesthesia, which is a known inhibitor of diaphragm recovery following spinal cord injury

(Bezdudnaya et al. 2018). As a result, it is important to repeatedly perform diaphragm EMG recordings in conscious mice to monitor the recovery of diaphragm function following a C2Hx.

V2a-(Gq)DREADD mice and non-DREADD expressing control mice were implanted with F20-EET transmitters to chronically record diaphragm EMG throughout recovery. All diaphragm EMG peak amplitude values were normalized to pre-injury (-CNO) baseline values. There is no difference observed in the diaphragm EMG peak amplitude after increasing V2a excitability in V2a-(Gq)DREADD mice (n=3)

(Figure 7F-G, black circles) or following CNO treatment in a non-DREADD expressing control (n=1) (Figure

7F-G, blue boxes) prior to the C2Hx spinal cord injury, similar to what we observe in healthy anesthetized mice. A C2Hx was performed on these mice at Day 0, paralyzing the ipsilateral diaphragm and drastically

167 decreasing diaphragm EMG peak amplitude. Increasing V2a excitability in V2a-(Gq)DREADD mice increases diaphragm EMG peak amplitude by 108 ± 25% one day following injury, 40 ± 30% one week following injury, and only 15% two weeks following injury (Figure 7G, black circles). In contrast, CNO does not alter diaphragm EMG peak amplitude in the non-DREADD expressing control mouse one day, one week, or two weeks following injury (Figure 7G, blue circles). Interestingly, V2a-(Gq)DREADD animals reach pre-injury diaphragm EMG peak amplitude values during the baseline recording as soon as one week following injury, whereas the non-DREADD expressing control mouse has only reached half of that amplitude even two weeks following injury. Although preliminary, these data suggest that activating V2a neurons may restore diaphragm function back to pre-injury levels faster than spontaneous recovery in non-DREADD expressing controls can. More animals in the experimental and control groups presented here as well as

V2a-(Gq)DREADD mice that do not receive CNO need to be tested to finish this study.

Figure 7. Repeatedly increasing V2a excitability may train respiratory circuits to quickly restore diaphragm function. V2a-(Gq)DREADD mice (n=3) and a non-DREADD expressing control mouse (n=1) were implanted with F20-EET transmitters to chronically record diaphragm EMG activity. A representative traces of the ipsilateral diaphragm is shown (A) immediately preceding injury, (B) immediately following a C2Hx, and before and after treatment with 1.0 mg/kg*bw CNO 1 day (C), 7 days (D), and 14 days (E) after injury. (F) All diaphragm EMG peak amplitudes are normalized to pre-injury levels. Average normalized diaphragm EMG peak amplitudes are shown for V2a-(Gq)DREADD mice (black circles) and the non- DREADD expressing control (blue squares) after saline (vehicle treatment) (open shapes) and after CNO treatment (filled shapes). (G) The percent change pre to post CNO is shown for V2a-(Gq)DREADD mice (black circles) and the non-DREADD expressing control (blue circles) prior to injury, 1 day, 7 days, and 14 days following the C2Hx. Data are reported as mean ± SEM.

168

Spinal cord injury and activation of V2a neurons increases scalene ARM activity

Previous studies have shown that extradiaphragmatic muscles become active following spinal cord injury to enhance ventilation to compensate for impaired diaphragm function (Dougherty et al. 2012). Since V2a neurons are synaptically connected to ARM motor neurons as well as phrenic motor neurons, the effect of the C2Hx injury and increasing V2a excitability on inspiratory ARM activity was repeatedly assessed in the same mice using chronic EMG recordings. Preliminary data show that a C2Hx injury alone increases ipsilateral scalene ARM activity in V2a-(Gq)DREADD mice (n=3) (representative trace in Figure 8A, quantified in Figure 8B, white bars) and a non-DREADD expressing control mouse (representative traces not shown, quantified in Figure 8C, white bars). Ipsilateral trapezius ARM activation was not observed at any time point following a C2Hx, likely because the accessory nerve carrying axons from the trapezius motor neuron pool to the neuromuscular junction of the muscle itself was severed during the C2Hx.

Therefore, analysis of ARM activation was confined to the scalene muscle only.

First, the ipsilateral scalene bout frequency is increased due to the C2Hx alone (representative traces in Figure 8A-E (left traces) and quantified in Figure 8F, white bars). Scalene bout frequency is low prior to injury (0.3 bouts/min.), but a C2Hx increases scalene bout frequency to 3.2 ± 4.3 bouts/min. 1 day post C2Hx, 4.7 ± 2.2 bouts/min. 7 days post C2Hx, and 7.4 ± 4.1 bouts/min 14 days post C2Hx in V2a-

(Gq)DREADD animals (n=3). Scalene bout frequency then declines back to 0.6 ± 0.2 bouts/min. 28 days post C2Hx. A similar pattern of ARM activation is observed in non-DREADD expressing controls (n=2

Chx10Cre/+;ROSAtTA/+;TetO+/+) (representative traces not shown but quantified in Figure 8G, white bars).

Scalene bout frequency was increased from 0.7 bouts/min. (prior to injury) to 9.6 bouts/min 1 day post

C2Hx. Scalene activation then declined to 0.7 bouts/min. 7 days post C2Hx, 1.1 bouts/min. 14 days post

C2Hx, and 0.2 bouts/min. 28 days post C2Hx.

169

Next, the effect of increasing V2a excitability on ARM activity was assessed at each time point following a C2Hx (representative traces in Figure 8A-E (right) and quantified in Figure 8G, black bars). As shown previously in Romer et al. 2017, increasing V2a excitability with 1.0 mg/kg*bw CNO increases scalene activity in health mice prior to injury. A similar increase is observed 1 day post C2Hx, as shown by a 1,700% increase in scalene bout frequency. However, at 7 days post C2Hx, activating V2a neurons has a negligible effect on ARM recruitment (-1% decrease). Scalene bout frequency is then decreased by 86% on 14 days post C2Hx and 14% 28 days post C2Hx. In contrast, minimal change is observed in scalene activity following CNO treatment in non-DREADD expressing controls at all time points (Figure 8G). Thus, increasing V2a excitability increases ARM activity prior to and immediately following injury, but actually decreases ARM recruitment as the time following spinal cord injury increases.

Figure 8. The frequency of ipsilateral scalene activity changes following spinal cord injury and V2a neuron activation.

V2a-(Gq)DREADD mice (n=3) were implanted with F20-EET transmitters to chronically record scalene ARM EMG activity. Representative traces of scalene ARM activity following vehicle treatment (-CNO) (left traces) and 1.0 mg/kg*bw CNO (+CNO) (right traces) prior to injury (A), 1 day post C2Hx (B), 7 days post C2Hx (C), 14 day post C2Hx (D), and 28 days post C2Hx (E). (F) Scalene bout frequency is quantified before (white bars) and after CNO treatment (black bars) at during recovery after a C2Hx in V2a-(Gq)DREADD animals (n=3). (G) Scalene bout frequency is quantified before (white bars) and after CNO treatment (black bars) at during recovery after a C2Hx in a non-DREADD expressing control animals (n=2). Data are reported as mean ± SEM.

170

Ventilation is enhanced during scalene bouts

V2a neurons project to motor neurons throughout the spinal cord, including motor neurons that control respiratory muscle activity and limb function. Because spinal cord injury can cause muscle spasms

(Elbasiouny 2010), it was important to determine whether the observed scalene ARM activity (particularly after increasing V2a excitability) was “productive” muscle activity that enhances ventilation instead of

“unproductive” muscle activity that is characteristic of muscle spasms. Figure 9A shows that scalene activity enhances ventilation in a V2a-(Gq)DREADD mouse (+CNO treatment) one day following injury.

Preliminary data shows there is an increase in PIF (p=0.062) (Figure 9B), VT (p=0.151) (Figure 9C), f

(p=0.002) (Figure 9D), and MV (p=0.144) (Figure 9E) after increasing V2a excitability (n=3), suggesting that increasing V2a excitability increases productive scalene ARM activity. However, these data require n=6 to reach a power of 0.800. Importantly, neither forelimb nor hindlimb muscle spasms were observed after increasing V2a excitability at any time point prior to or following a C2Hx.

Figure 9. Ventilation is enhanced during scalene bouts.

Ventilation was measured before and after CNO treatment in V2a-(Gq)DREADD mice (n=3). (A) Representative trace showing whole breath plethysmography (WBP) and scalene EMG in a V2a-(Gq)DREADD animal following CNO treatment 1 day following a C2Hx spinal cord injury. Gray boxes outline scalene bouts and associated ventilation. Peak inspiratory flow (PIF) (B), tidal volume (VT) (C), breathing frequency (f) (D), and minute volume (MV) (E) was measured before and after treatment with 1.0 mg/kg*bw CNO. Gray dashed lines indicate individual animals. Black solid lines show the average ± SEM. Data are analyzed with a paired t-test.

171

Figure 10. Multiple subtypes of V2a neurons may contribute to recovery of diaphragm function following spinal cord injury. (A) Diagram highlighting different subsets of V2a neurons (by anatomical location) that may mediate diaphragm recovery observed after increasing the excitability of all cervical spinal V2a neurons. (#1) Brainstem V2a neurons projecting to the PreBotzinger Complex or rVRG. (#2) Contralateral reticulospinal V2a neuron projecting from the brainstem to the spinal cord. An ipsilateral reticulospinal V2a neuron is not shown because these projections would be disrupted following a C2Hx. (#3) Contralateral spinal V2a neuron that influences respiratory circuits on the ipsilateral side via a commissural interneuron that crosses the spinal cord. (#4) Ipsilateral spinal V2a neuron that receives descending input from the brainstem and projects directly or indirectly to phrenic motor neurons. (#5) Ipsilateral spinal V2a neuron that projects directly or indirectly to phrenic motor neurons and influences diaphragm function independent of descending brainstem drive. (B) #3, #4, or #5 V2a neurons may mediate rhythmic bursting of the diaphragm by relaying or amplifying rhythmic drive from the brainstem. Even if V2a neurons fire tonically, this increased excitatory input may directly or indirectly increase phrenic motor neuron excitability so they are able to fire rhythmically in response to descending brainstem drive.

5.4 Discussion

Our findings are the first to describe a developmental class of neurons that is critical for recovery of respiratory muscle function following spinal cord injury. We demonstrate that acutely increasing the excitability of the V2a class of ipsilaterally projecting excitatory neurons in the spinal cord and brainstem is able to restore respiratory activity to the diaphragm below a C2Hx injury. Repeatedly increasing V2a excitability may even strengthen respiratory circuits to promote a quicker recovery of respiratory function. Additionally, silencing V2a neurons impairs the induction of the crossed phrenic phenomenon

(as induced by nasal occlusion) and influences spontaneous recovery of diaphragm function weeks after injury. Our data supports a crucial role for V2a neurons in the circuits that contribute to recovery of respiratory function and indicate that strategies to alter the activity of V2a neurons may improve breathing following injury.

The crossed phrenic phenomenon has been an established model to study respiratory plasticity for over a century (Porter 1895; Goshgarian 2003; Ghali 2017). By silencing V2a neuron activity, we demonstrate that V2a neurons are necessary for full activation of the crossed phrenic phenomenon. The residual activity observed after decreasing V2a excitability with DREADDs may be because V2a neurons are not completely silenced in our experiments, or alternatively, additional neurons can partially

172 substitute for V2a function when they are silenced. The crossed phrenic phenomenon was proposed to be mediated primarily by projections from the rVRG that cross below the site of injury and directly contact phrenic motor neurons (Ellenberger and Feldman 1988; Ellenberger et al. 1990). Pre-phrenic rVRG neurons were later shown to belong to the V0 developmental class of neurons (commissural interneurons that cross the midline of the spinal cord) (Wu et al. 2017). However, more recent studies have implicated a role for propriospinal neurons in the crossed phrenic phenomenon (Lane et al. 2008; Alilain et al. 2008;

Ghali 2017; Satkunendrarajah et al. 2018). Our results are the first to demonstrate that ipsilaterally projecting excitatory neurons outside of the rVRG are critical for the crossed phrenic phenomenon.

A limitation of our studies is that we cannot distinguish whether recovery of diaphragm function is dependent on V2a neurons below the site of injury versus above the site of injury since both spinal and brainstem V2a neurons were targeted with DREADDs. It has been proposed that increased drive to the rVRG is responsible for activation of latent pathways mediating the crossed phrenic phenomenon

(Goshgarian et al. 1991; Moreno et al. 1992). However, it is unlikely that V2a neurons act by uniformly increasing respiratory drive at the level of the brainstem because we do not see an increase in contralateral diaphragm activity when V2a excitability is increased. In fact, a decrease in contralateral diaphragm activity is observed one day following injury, likely the result of a compensatory decrease in respiratory drive resulting from activation of the previously paralyzed ipsilateral diaphragm. Our data showing a regular respiratory rhythm in the contralateral diaphragm also argues against the possibility that increasing V2a excitability indirectly activates crossed phrenic pathways by causing asphyxia. Thus,

V2a neurons likely act downstream or in concert with rVRG pathways to enhance the crossed phrenic phenomenon without significantly affecting respiratory rhythm or drive. As a therapeutic approach, it could be advantageous to target pathways that enhance breathing without directly impacting respiratory rhythm generation. In order to test this, the excitatory (Gq)DREADD can be targeted to either cervical

173 spinal or brainstem V2a neurons using a Cre-dependent AAV injected into a Chx10Cre/+ mouse (see Chapter

III Methods).

One potential mechanism is that V2a neurons below the site of injury relay or amplify the rhythmic respiratory drive from the rVRG to phrenic motor neurons, either directly (Figure 10A, V2a neuron #4) or via commissural neurons (Figure 10A, V2a neuron #3). Alternatively, crossing axons from the contralateral rVRG may provide rhythmic respiratory drive directly to phrenic motor neurons ipsilateral to injury, but this drive may be insufficient to produce action potentials without additional tonic drive from excitatory V2a neurons (Figure 10A, V2a neuron #5) (hypothetical mechanism shown in Figure

10B). Our observation that silencing V2a neurons consistently blocked spontaneous recovery of tonic activity (see below) supports the latter hypothesis. However, these results do not rule out a role for V2a neurons in relaying rhythmic drive to phrenic motor neurons, as it is possible that respiratory V2a neurons can provide both tonic and rhythmic drive. For example, in the lumbar cord, some V2a neurons are rhythmically active during fictive locomotion whereas other V2a neurons fire tonically (Zhong et al. 2010;

Dougherty and Keihn 2016).

In order to distinguish between these possibilities, several experiments should be performed.

First, electrophysiology can be recorded from cervical spinal V2a neurons following a C2Hx to determine if they exhibit rhythmic, tonic, or a mixture of both intrinsic firing properties. Second, future studies should test whether spinal V2a neurons require rhythmic brainstem drive to restore bursting activity to the previously paralyzed diaphragm. Anesthetized V2a-(Gq)DREADD mice can be mechanically ventilated and undergo a complete C1 transection to completely sever descending bulbospinal input into cervical spinal

V2a neurons. If increasing V2a excitability restores rhythmic bursting activity from the diaphragm, this suggests that V2a neurons are not dependent on input from the brainstem rVRG. However, rhythmic activation may also result from rhythmic activation of respiratory muscle sensory afferents from the movement of the chest wall during mechanical ventilation. Therefore, if restored bursting activity still

174 persists after removing mechanical ventilation, this suggests that spinal V2a neurons are not dependent on descending brainstem input or sensory afferents to rhythmically activate the diaphragm. Finally, unilateral viral injections targeting the (Gq)DREADD receptor to either the ipsilateral or the contralateral side will determine whether V2a neurons project through commissural pathways to restore diaphragm activity.

Even though the mechanism of rhythmic bursting activity is currently unknown, increasing the excitability of V2a neurons still elicits predominantly rhythmic inspiratory activity in the diaphragm below the site of injury. However, that is not the only pattern of activity that we observe. Even within the same animal, rhythmic activity alternates with tonic activity or there is a mixture of tonic and rhythmic activity.

Although we observe spontaneous changes in the pattern of diaphragm activity at 4 hours, 1 day, and 2 weeks after injury, the reasons for these changes in pattern are not currently clear. Movement or changes in posture are not possible since the animals are anesthetized and immobile. The animals are freely breathing, so it is possible that fluctuations in blood oxygen or carbon dioxide trigger changes in breathing pattern. A previous study showed that tonic activity and/or a mixture of rhythmic and tonic activity could be observed in the ipsilateral diaphragm of C2 hemisected rats following treatment with serotonin or treatments that altered serotonergic innervation of the spinal cord (Warren et al. 2018). Thus, it is possible that serotonin elicits tonic activity in phrenic motor neurons by increasing the excitability of V2a neurons.

Alternatively, altering V2a activity may alter the release of serotonin by raphe neurons. A third possibility is that V2a neurons may alter the responsiveness of phrenic motor neurons to serotonin. Future experiments should investigate potential interactions between V2a neurons and serotonergic neurons in the control of respiratory pattern. It is important to understand how the different diaphragm activity patterns are generated because efficient breathing likely requires rhythmic only inspiratory activity.

As in the acute setting, increasing the excitability of V2a neurons can restore rhythmic bursting activity to a previously paralyzed diaphragm two weeks after a C2Hx injury. A previous study used

175 pseudorabies virus tracing to show that more V2a neurons are synaptically connected to phrenic motor neurons two weeks after a C2Hx than in uninjured animals, demonstrating that new V2a neurons are recruited into diaphragm circuits following injury (Zhouludeva et al. 2017). Given this observation, it was surprising to find that there is an increase in the proportion of rhythmic versus tonic activity induced by

CNO treatment at two weeks compared to one day after injury. If non-respiratory neurons were recruited into respiratory circuits, more tonic rather than rhythmic activity would be expected. This finding suggests that newly recruited V2a neurons either 1) received rhythmic respiratory drive prior to injury (e.g. they were involved in control of accessory respiratory muscles), 2) acquired rhythmic respiratory drive in the two weeks following injury, or 3) provide tonic drive to respiratory muscles (and rhythmicity is provided by other circuits). Importantly, although the ipsilateral diaphragm is paralyzed prior to CNO treatment in these experiments, we cannot rule out the possibility that activity dependent plasticity plays a role in promoting rhythmic versus tonic activity patterns. This is because diaphragm EMG recordings are performed under anesthesia (known to inhibit recovery of diaphragm function (Bezdudnaya et al. 2018), and it is possible that there is activity in ipsilateral phrenic motor neurons during the two weeks following injury while the animal is not under anesthesia. Future studies will need to investigate the potential of

V2a neurons to restore diaphragm function in cases of chronic injury since our studies did not assess the impact of increasing V2a neuron excitability beyond 2 weeks of age or evaluate the effects of chronically altering V2a neuron excitability.

Spontaneous recovery of diaphragm function has been observed as early as two weeks following a C2Hx injury in rodent models (Nwantwi et al. 1999; Golder et al. 2003; Fuller et al. 2006; Mantilla et al.

2014; Ghali 2017; Bezdudnaya et al. 2018). Here, we show that silencing V2a neurons can impair spontaneous recovery of diaphragm function two weeks following injury. Three animals showed spontaneous tonic activity (i.e. during both inspiration and expiration) in the diaphragm ipsilateral to injury that was blocked after silencing V2a neurons, demonstrating that V2a play a critical role in this stage

176 or type of recovery. Recordings of individual phrenic motor neuron discharge patterns following a C2 hemisection in rats demonstrated the emergence of tonically firing motor neurons after injury (Lee et al.

2013). It was proposed that this tonic activity could result from changes in intrinsic motor neuron properties and/or formation of new synaptic inputs to motor neurons after injury (Lee et al. 2013).

Anatomical evidence supports the appearance of novel connections between V2a neurons and phrenic motor neurons following injury (Zhouludeva et al. 2017). Our functional data support the hypothesis that spontaneous tonic activity before CNO treatment stems from the appearance of new connections between V2a neurons and phrenic motor neurons, since tonic activity is only observed in the baseline recording two weeks following injury (and not at 4 hours or one day) and silencing V2a neurons abolishes the tonic activity. However, tonic diaphragm activity does not require the formation of de novo circuits, evidenced by our observation that increasing the excitability of V2a neurons can generate tonic activity

(as well as rhythmic) as early as 4 hours after injury. This short time interval indicates that V2a neurons are already incorporated into the respiratory circuitry prior to injury that can restore diaphragm function.

Moreover, silencing V2a neuron activity does not disrupt respiratory rhythm generation, as evidenced by the regular, rhythmic inspiratory activity of the contralateral diaphragm even after CNO treatment, despite occasional tonic and rhythmic activity in some animals. This result is consistent with our previous study demonstrating that silencing V2a neurons in healthy adult mice at rest does not impair diaphragm function or the regularity of the respiratory rhythm (Jensen et al. 2019). Thus, V2a neurons likely contribute to (but are not required for) control of diaphragm activity in healthy animals, but likely become recruited into respiratory circuitry as a compensatory mechanism following injury.

We also found that silencing V2a neurons dramatically reduced diaphragm activity in a mouse that showed spontaneous recovery of rhythmic inspiratory activity. This result is consisted with a recent study demonstrating the importance of glutamatergic propriospinal neurons for maintaining breathing in a mouse model of cervical myelopathy, a non-traumatic spinal cord injury, as well as the C2Hx model

177

(Satkunendrarajah et al. 2018). Although their data suggested that commissural excitatory neurons might be responsible for recovery, our data demonstrate that ipsilaterally projecting V2a neurons can also promote recovery of diaphragm function. In addition, we found that silencing V2a neurons had no detectable effect in 6 other mice that showed spontaneous recovery of rhythmic inspiratory activity. The reason for this difference is not clear, although we did observe that the animal in which inspiratory activity was reduced upon silencing V2a neurons was the one that had the smallest EMG peak amplitude and duration of inspiratory activity prior to CNO treatment. The most likely explanation is that there are multiple mechanisms of spontaneous recovery and some animals rely largely on V2a neurons whereas other animals can utilize additional mechanisms (e.g. commissural excitatory propriospinal neurons). It is also possible that recovery proceeds through multiple stages- with V2a neurons being critical for early stages but dispensable at later stages. Future studies are required to test this further.

These experiments implicate V2a neurons as important mediators of recovery of diaphragm function at distinct, single time points following spinal cord injury. In other words, V2a neurons serve as neuromodulators to enhance diaphragm function. However, recovery of respiratory muscle activity is a dynamic process that should be continuously monitored to determine whether V2a neurons contribute to respiratory neuroplasticity. Moreover, these diaphragm EMG recordings were performed under anesthesia. Anesthesia has been shown to inhibit recovery of diaphragm function following spinal cord injury (Bezdudnaya et al. 2018) and alter pulmonary gas exchange and respiratory muscle mechanics

(Rehder 1979). Therefore, it is necessary to investigate the role of V2a neurons in respiratory muscle recovery in conscious mice following a C2Hx injury. Two important findings were discovered. First, increasing V2a excitability had the greatest impact on diaphragm EMG peak amplitude one day following injury (compared to 7 and 14 days post C2Hx). Second, repeatedly increasing V2a activity may be able to train the crossed phrenic phenomenon to permanently restore diaphragm function faster than non-

DREADD expressing controls show spontaneous recovery. These results are discussed below, with the

178 caveat that this study is underpowered and requires a greater sample size to draw more concrete conclusions (n=3 V2a-(Gq)DREADD mice and n=1 non-DREADD expressing control).

Preliminary evidence suggests that repeatedly increasing V2a excitability in V2a-(Gq)DREADD mice restores diaphragm EMG peak back to pre-injury levels faster than non-DREADD expressing controls show the same level of recovery. It is well established that continually activating locomotor circuits via spinal stimulation and training can restore limb function to spinal cord injury rodent models (Ichiyama et al.

2008; Hayashibe et al. 2016) and patients (Frood 2011). Therefore, it is possible that even the low frequency in which V2a neurons were excited following injury was sufficient to induce respiratory plasticity and permanently restore diaphragm function. The greatest effect of increasing V2a excitability was observed one day following injury, likely because the diaphragm was still paralyzed at this time point.

At later time points, some spontaneous recovery was observed and increasing V2a excitability had less of an effect on diaphragm EMG peak amplitude. These experiments show that the previously paralyzed diaphragm can reach a diaphragm EMG peak amplitude similar to pre-injury levels within one week following injury (faster than is observed in a non-DREADD expressing control mouse also treated with

CNO). Therefore, future experiments should chronically administer CNO immediately following injury to determine how quickly diaphragm function can reach pre-injury levels.

Finally, we have previously shown that altering V2a excitability can increase ARM activity in healthy mice and in SOD1(G93A) ALS model mice. Therefore, we investigated the effect of increased V2a excitability on ARM recruitment following spinal cord injury. There was an increase in ARM activation prior to CNO administration one day following a C2Hx. This is not surprising because ARMs are known to become recruited under conditions of increased oxygen demand or impaired diaphragm function. Further evidence supporting this is the enhanced ventilation associated with ARM recruitment. ARM recruitment then declined as more time following injury passed. This may result from the restored diaphragm function observed at these time points, rendering ARM activity less important for enhancing ventilation.

179

Increasing V2a excitability further increases scalene ARM activity one day following injury, a time when the greatest increase is observed in diaphragm EMG peak amplitude. Several possibilities may explain these results. First, it is possible that the respiratory system systemically increases descending brainstem drive to all respiratory motor neurons immediately following injury, including phrenic motor neurons and ARM motor neurons. Therefore, increasing the excitability of V2a neurons drastically increases activity of all respiratory muscles in an attempt to combat respiratory insufficiency.

Alternatively, ARMs are known to stabilize the chest well during eupnea and expand the chest wall during exercise in healthy animals to alter respiratory mechanics and facilitate diaphragm activity (Sieck and

Gransee 2012). Therefore, increasing V2a excitability may simply stabilize the chest wall enough to allow phrenic motor neurons to respond to descending brainstem drive to restore rhythmic bursting to the previously paralyzed diaphragm.

Surprisingly, increasing V2a excitability no longer increased ARM activity 7 days following injury.

It is possible that respiratory plasticity re-allocates resources to the main inspiratory muscle (the diaphragm) instead of ARMs during this period of recovery. Zhouludeva et al. 2017 showed that V2a neurons show increased connectivity to phrenic motor two weeks following injury, but connectivity to

ARM motor neurons was not tested. Therefore, if V2a neuron connectivity is preferentially increased to phrenic motor neurons, increasing V2a excitability may have a greater effect on diaphragm function than

ARMs, which inherently reduces the need for ARM activation. In order to test this, PRV should be injected into the scalene muscle in control mice and two weeks following a C2Hx in injured mice to determine if 1)

V2a connectivity to ARM motor neurons is also increased and 2) if a greater proportion of V2a neurons are synaptically connected to phrenic motor neurons or ARM motor neurons. Future studies should also test whether V2a neurons can improve expiratory ARM activity (such as the abdominals or obliques) to determine if V2a neurons can be targeted to improve expulsive behaviors (such as coughing and sneezing) that help clear the airways of infection.

180

Our studies implicate V2a neurons as important mediators of recovery of diaphragm function acutely following injury as well during the regenerative processes that contribute to spontaneous recovery of function weeks after injury. Importantly, altering V2a function does not impair breathing rhythm or produce significant motor deficits (Romer et al. 2017; Jensen et al. 2019), suggesting that these neurons may be safe to target therapeutically. Moreover, repeatedly activating V2a neurons may strengthen respiratory circuits to quickly restore diaphragm activity following spinal cord injury. Since V2a neurons are important for locomotion (Crone et al. 2008; Crone et al. 2009; Dougherty et al. 2013) and fine motor movements (e.g. reaching and grasping) (Azim et al. 2014; Ueno et al. 2018), V2a neurons may also play an important role in recovery of these motor behaviors as well.

181

Chapter VI:

Conclusions and Future Directions

182

These data reveal several important roles for V2a neurons in controlling respiratory muscle activity and ventilation in healthy mice and following ALS and spinal cord injury. First, ablating V2a neurons causes slow and irregular breathing in neonatal mice (Crone et al. 2012). Chapter II demonstrated that silencing V2a neurons in neonatal mice recapitulates this phenotype. However, silencing V2a neurons in adult mice does not alter the regularity of respiration and actually increases breathing frequency, suggesting that the role of V2a neurons in controlling breathing changes throughout development.

Second, decreasing brainstem and spinal cord V2a excitability in adult mice activates ARMs without impairing diaphragm activity, suggesting that ARMs are not activated as a compensatory mechanism for impaired diaphragm function. Moreover, it is unlikely that altered chemosensation is responsible for the activation of ARMs since heart rate and arterial oxygen saturation do not appear to change after silencing

V2a neurons. These data suggest that two different subtypes of V2a neurons exist in respiratory circuits: one subtype (Type I) that participates in an excitatory pathway to activate ARMs when they are needed and another subtype (Type II) that participates in an inhibitory pathway to prevent ARM activation when it is not needed to preserve energy.

Chapter III describes experiments that demonstrate the sufficiency of cervical spinal V2a neurons to pattern ARM activity, revealing that the two proposed subtypes of V2a neurons are both found in the cervical spinal cord. Moreover, activating V2a neurons increases diaphragm EMG peak amplitude during

ARM bouts whereas silencing V2a neurons does not alter diaphragm function, suggesting that V2a neurons are sufficient but not required for diaphragm function in healthy mice. These data suggest that

Type I V2a neurons may project to ARM motor neurons as well as phrenic motor neurons, whereas Type

II V2a neurons may only inhibit ARM activation at rest. Moreover, increasing V2a excitability on one side of the spinal cord bilaterally activates the left and right scalene muscles, suggesting that cervical spinal

V2a neurons may project to commissural interneurons. Taken together, these results provide additional evidence that spinal interneurons are important modulators of respiratory muscle activity and ventilation.

183

In addition to modulating respiratory muscle activity and ventilation in healthy mice, increasing and decreasing V2a excitability increases ARM activity and alters ventilation in SOD1(G93A) ALS model mice throughout disease progression (Chapter IV). Increasing or decreasing V2a excitability activates

ARMs at all stages of ALS disease progression; however, increasing and decreasing V2a excitability differentially affects ventilation throughout disease progression. Increasing V2a excitability drastically increases ventilation during ARM bouts at early stages of ALS disease progression, but has little effect at late stages. In contrast, silencing V2a neurons enhances ventilation the most during late stages of ALS.

While still preliminary, these data suggest that two different subtypes of V2a neurons may have different roles in controlling respiratory muscle activity and ventilation at different stages of ALS, which may result from selective connectivity to slow, intermediate, and slow motor units. This is important information for determining which subset of V2a neuron – Type I vs. Type II – should be targeted to improve ventilation at late stages of ALS.

Finally, Chapter V showed that V2a neurons are critical for inducing crossed phrenic pathways to mediate recovery of diaphragm function following a C2Hx spinal cord injury. Silencing V2a neurons even abolishes spontaneous tonic activity observed in the baseline recording two weeks following injury.

Moreover, repeatedly increasing V2a excitability following a C2Hx may train respiratory circuits to restore diaphragm function to the previously paralyzed diaphragm faster than spontaneous recovery alone.

Finally, the C2Hx injury alone increases ARM activity (likely as a compensatory mechanism to enhance ventilation when the diaphragm was impaired) and increasing V2a excitability further increased ARM activity one day following injury. These data provide additional evidence that interneurons – specifically

V2a neurons – are critical for promoting respiratory plasticity following spinal cord injury.

Taken together, these data suggest that V2a neurons are already incorporated into respiratory circuitry in healthy mice to regulate respiratory muscle (diaphragm and ARMs) activity and ventilation.

However, our studies in ALS model mice and C2 hemisected mice show that V2a neurons promote

184 recovery of respiratory muscle activity following disease and injury. Previous research has used anatomical tracing studies and functional studies (both indirect and direct) to show that interneurons are important mediators of respiratory muscle plasticity. However, we show that a single class of interneurons

– ipsilaterally projecting, glutamatergic V2a neurons – may be a promising neural target to improve breathing. In addition, these data implicate other potential roles of V2a neurons in controlling ventilation during different behaviors and conditions of high oxygen demand, which are discussed below.

Our studies focused on the role of V2a neurons in controlling the diaphragm and inspiratory accessory respiratory muscles. Even though expiration is a passive process that results from the relaxation of the diaphragm, active expiration is critical for coughing and sneezing, behaviors that keep the airways clear from obstruction and infection. Therefore, future studies should investigate the role of V2a neurons in controlling the abdominal, oblique, and internal intercostal expiratory muscle activity. Experiments should record expiratory muscle EMG and test 1) whether silencing V2a neurons impairs expiratory muscle activity and/or cough and 2) whether activating V2a neurons increases expiratory muscle activity and/or facilitates cough. This is especially important for clinical implications following neuromuscular disease or spinal cord injury since an impaired cough increases the risk for infection and respiratory failure.

V2a neurons are located throughout the neuraxis. Specifically, lumbar spinal V2a neurons modulate hindlimb motor neuron output (Crone et al. 2008; Crone et al. 2009; Dougherty and Kiehn 2010;

Dougherty et al. 2013) and should be investigated as targets for recovery of locomotor function following lower level (e.g. thoracic) spinal cord injury. Specifically, a contusion or a lower thoracic hemisection model of spinal cord injury can be performed in Chx10Cre/+ mice dually injected with a Cre-dependent AAV-

FLEX-(Gq)DREADD-mCherry virus and an AAV-FLEX-DTR-GFP virus (Azim et al. 2014) to target the excitatory DREADD receptor and the diphtheria toxin receptor (DTR) to V2a neurons in the lumbar spinal cord only. Three hypotheses should be tested. First, can increasing V2a excitability following spinal cord injury alone partially or fully restore locomotor activity (rhythmic or arrhythmic)? Second, does increased

185

V2a excitability need to be combined with locomotor training to induce recovery? Third, if recovery can be induced by either of the first two experiments, lumbar spinal V2a neurons should then be ablated by injection of diphtheria toxin to determine if removing V2a excitatory input abolishes recovery. These experiments can be performed in vivo (similar to Crone et al. 2009) as well as using electrophysiology in a brainstem/spinal cord ex vivo preparation in which the caudal medulla is stimulated to induce fictive locomotion.

The role of V2a neurons in controlling respiratory muscle activity, ventilation, and cardiovascular function during exercise should be tested in future studies. All of the experiments showing that V2a neurons are important for tightly regulating accessory respiratory muscle activity and ventilation in healthy mice were performed at rest (when the mouse was not moving). However, it is well known that one important role of ARMs is to enhance ventilation during exercise (Sieck and Gransee, 2012; Aliverti

2016). It has already been shown that V2a neurons are critical for ensuring the left-right alternation of locomotion in mice running at high speeds (Crone et al. 2009), but the role of V2a neurons in controlling

ARM activation and enhancing ventilation under these same conditions still needs to be tested. V2a neurons may be a prime candidate for matching ventilation to locomotion. For example, Pivetta et al.

2014 showed that V2a neurons in the lumbar spinal cord have ascending projections to brainstem regions.

Additional data presented in this thesis found that cervical spinal V2a neurons specifically project to respiratory centers in the brainstem. Moreover, activating all brainstem and spinal cord V2a neurons is sufficient to enhance ventilation even in the absence of ARM bouts (Romer et al. 2017). This thesis further showed that activating all brainstem and spinal cord V2a neurons also increased heart and blood pressure, which are physiological changes associated with the onset of exercise (Forster et al. 2012). Therefore, future studies should investigate how altering V2a excitability alters respiratory muscle activity, ventilation, cardiovascular output, chemosensation, and endurance in healthy mice during exercise.

186

The role of V2a neurons in controlling diaphragm and accessory respiratory muscle activity also needs to be explored further. Under healthy, normal conditions, ARMs are typically inactive because the diaphragm and external intercostals serve as primary pump muscles to drive inspiration. However, disease and injury significantly alter brainstem drive, interneuron connectivity, and morphology/signaling of motor neurons themselves to compensate for impaired ventilation. It has already been demonstrated that V2a neurons show increased connectivity to phrenic motor neurons two weeks following a C2Hx spinal cord injury (Zhouludeva et al. 2017). Functional data presented in this thesis suggest that V2a neurons are critical for restoring diaphragm function following a C2Hx. However, it is unclear how disease and injury affect connectivity of V2a neurons to ARM motor neurons. Do V2a neurons preferentially increase connectivity to phrenic motor neurons because the diaphragm is the main inspiratory muscle

(with the greatest mechanical advantage to restore breathing)? Or, do V2a neurons also show increased connectivity to ARM motor neurons to activate multiple respiratory muscles in an attempt to combat respiratory insufficiency? Preliminary data showed that increasing V2a excitability immediately following injury increased ARM activity, whereas increasing V2a excitability at later time points decreased ARM activation. These data suggest that the respiratory circuits controlling ARMs change during recovery, but the mechanism is currently unclear and should be investigated further.

This consideration is especially important when determining the best strategy to improve ventilation in ALS and spinal cord injury. For example, decreasing V2a excitability can increase ARM activity at late stages of ALS - but what if silencing V2a neurons actually impairs diaphragm function? Silencing

V2a neurons does not impair diaphragm function in healthy mice, but it does prevent the full induction of the CPP in C2 hemisected mice. Thus, silencing V2a neurons may activate ARMs and enhance ventilation during ARM activation at late stages of ALS, but this is not clinically relevant if it further impairs primary pump muscle activity. Therefore, additional studies should test how silencing and activating V2a neurons

187 in ALS model mice affect diaphragm function before a conclusion can be made as to what is a better method for improving ventilation: increasing or decreasing V2a excitability.

This question will be easier to answer if distinct subtypes of V2a neurons, marked by unique transcriptional profiles, morphology, or projection patterns, are identified in the cervical cord controlling respiratory circuits. This is important for two reasons. First, cervical spinal V2a neurons in the mouse control respiratory muscle motor neurons and forelimb motor neurons. Although we have not observed any motor deficits after increasing or decreasing V2a excitability in our experiments, identifying subsets of V2a neurons that only control respiratory circuits would identify a specific population of V2a neurons that can be targeted to improve ventilation without impacting movement. However, it is very possible that V2a neurons controlling respiration and locomotion are not distinct. Second, our data suggests that

V2a neuron Type I may control diaphragm and ARM activity whereas V2a neuron Type II may only control

ARM activity. If each subtype can be uniquely characterized, it may be possible to therapeutically target one V2a subtype over the other. For example, it would be beneficial to only silence Type II V2a neurons at late stages of ALS to activate ARMs and enhance ventilation without (potentially) impairing diaphragm function. On the other hand, it would be beneficial to only activate Type I V2a neurons following spinal cord injury to increase diaphragm and ARM activity. However, before drugs can be developed to increase or decrease V2a excitability, RNA sequencing is required to identify G-protein coupled receptor targets expressed in each subtype.

In order to more conclusively confirm the existence of Type 1 vs. Type II V2a neurons in the cervical spinal cord, several functional, tracing, and electrophysiology studies can be employed after identifying unique transcription factors that mark different subsets of V2a neurons. For example, Type I

V2a neurons may be Shox2+/Chx10+ whereas Type II V2a neurons may be Sp8+/Chx10+. It is possible to generate mice in which Cre recombinase is only expressed in Shox2+/Chx10+ or Sp8+/Chx10+ neurons. The existence of distinct V2a subtypes can then be functionally tested. For example, a Cre-dependent AAV

188 virus can be used to selectively deliver the (Gq)DREADD or (Gi)DREADD virus to cervical spinal V2a neurons.

If Type I V2a neurons participate in an excitatory spinal pathway to activate ARMs during conditions of high oxygen demand, we expect that increasing Type I V2a excitability would activate ARMs but silencing

Type I V2a neurons would not alter ARM activation. Vice versa, if Type II V2a neurons participate in an inhibitory spinal pathway to actively prevent ARM activation at rest, we propose that activating V2a neurons would not alter ARM activation whereas silencing Type II V2a neurons would increase ARM activity. At the end of these experiments, these mice can be injected with viral tracers into the scalene or trapezius muscle. Both monosynaptic (e.g. glycoprotein-deleted rabies virus) and polysynaptic viral tracers (pseudorabies virus) should be used in different cohorts of animals. For example, we propose that the mCherry fluorescent tag marking the Cre expressing Type I V2a neuron will co-express with the monosynaptic viral tracer whereas the Type II V2a neuron will not. In contrast, we hypothesize that the polysynaptic tracer will co-express with the Type II V2a neuron. The harvested tissue from this animal can then be stained for inhibitory interneuron markers (such as GAD67, GABA, etc.). If there are inhibitory interneurons in close proximity to Type II V2a neurons and both express the polysynaptic tracer, it suggests that Type II V2a neurons may participate in an inhibitory spinal pathway. Finally, electrophysiology in an isolated spinal cord preparation can be used to determine if increasing or decreasing V2a excitability increases ARM motor output. For example, Cre expressing cervical spinal Type

I or Type II V2a neurons can be transduced with the same Cre-dependent virus to target (Gq)DREADD expression to Type I vs. Type II V2a neurons. The mCherry reporter allows for the identification of each type of V2a neuron and extracellular single unit recordings can determine the firing rate of these neurons before and after CNO administration to verify that CNO depolarizes these neurons. Simultaneous spinal accessory nerve recordings should be performed to measure the output from trapezius motor neurons.

Based on our model, we propose that increasing V2a Type I excitability will result in increased spinal accessory nerve output whereas increasing V2a Type II excitability will have no effect.

189

In summary, we have revealed a key role for ipsilaterally projecting, glutamatergic V2a interneurons to control respiratory muscle activity in healthy mice, SOD1(G93A) ALS model mice, and spinal cord injured mice. Although future studies are required to elucidate how different subsets of V2a neurons control this activity, we conclude that V2a neurons are critical for promoting recovery of respiratory function following neuromuscular disease and spinal cord injury.

190

List of Abbreviations

AAV – Adeno Associated Virus

AAV-eGFP mice – Cervical spinal V2a neurons express enhanced Green Flourescent Protein (GFP) and cervical spinal V2a excitability is not altered following treatment with CNO. These mice serve as AAV injected non-DREADD expressing controls.

AAV-(Gi)DREADD mice – Cervical spinal V2a neurons only express the inhibitory DREADD receptor and cervical spinal V2a excitability is decreased following treatment with CNO.

AAV-(Gq)DREADD mice – Cervical spinal V2a neurons only express the excitatory DREADD receptor and cervical spinal V2a excitability is increased following treatment with CNO.

ALS – Amyotrophic lateral sclerosis

ARM – Accessory respiratory muscle

CNO – Clozapine-N-Oxide (synthetic ligand for DREADDs)

CPG – Central pattern generator

CPP - Crossed Phrenic Phenomenon/Pathway

CTA – Cycle Triggered Averaging

CVf – Coefficient of breathing frequency (used as a measurement of respiratory rhythm variability)

C2Hx – High level spinal cord injury where the spinal cord is hemisected at spinal segment C2

DREADDs – Designer Receptor Exclusively Activated by Designer Drugs (receptor for the synthetic ligand CNO) f – Breathing frequency or respiratory rate

IS – Irregularity score used to analyze breath-to-breath consistency from whole breath plethysmography

MV – Minute volume pFRG – Parafacial respiratory group

PIF – Peak inspiratory flow

RMSpeak – The Root Mean Square of the diaphragm signal was calculated from the raw EMG trace. The peak RMS value reflects the greatest value calculated for one diaphragm burst.

RTN – Retrotrapezoid nucleus in the brainstem that houses chemosensing Phox2b+ neurons

Ti – Inspiratory time

VT – Tidal volume

VT/Ti – Measure of respiratory drive

VRC – Ventral respiratory column that is composed of the PreBÖtzinger Complex, the BÖtzinger Complex, ventral respiratory group and dorsal respiratory group.

191 rVRG – Rostral ventral respiratory group cVRG – Caudal ventral respiratory group

V2a-(Gi)DREADD mice – Mouse line where all brainstem and spinal cord V2a neurons express the inhibitory DREADD receptor and V2a excitability is decreased following treatment with CNO.

V2a-(Gq)DREADD mice - Mouse line where all brainstem and spinal cord V2a neurons express the excitatory DREADD receptor and V2a excitability is increased following treatment with CNO.

WBP – whole breath plethysmography

192

References

Abbott, Stephen B.G., Ruth L. Stornetta, Melissa B. Coates, and Patrice G. Guyenet. 2011. “Phox2b- Expressing Neurons of the Parafacial Region Regulate Breathing Rate, Inspiration, and Expiration in Conscious Rats.” Journal of Neuroscience 31(45): 16410–22. Alaynick, W.A., T.M. Jessell, and S.L. Pfaff. 2011. “SnapShot: Spinal Cord Development.” Cell 146(1): 178– 178. Alilain, Warren J. et al. 2008. “Light-Induced Rescue of Breathing after Spinal Cord Injury.” Journal of Neuroscience 28(46): 11862–70. Aliverti, Andrea. 2016. “The Respiratory Muscles during Exercise.” Breathe 12(2): 165–68. Ampatzis, Konstantinos, Jianren Song, Jessica Ausborn, and Abdeljabbar ElManira. 2014. “Separate Microcircuit Modules of Distinct V2a Interneurons and Motoneurons Control the Speed of Locomotion.” Neuron 83(4): 934–43. http://dx.doi.org/10.1016/j.neuron.2014.07.018. . 2013. “Pattern of Innervation and Recruitment of Different Classes of Motoneurons in Adult Zebrafish.” Journal of Neuroscience 33(26): 10875–86. “Andersen, P et Al 2003 Sixteen Novel Mutations in SOD1 Gene in ALS a Decade of Discoveries, Defects, and Disputes.” Anderson, T.M. et al. 2016. “A Novel Excitatory Network for the Control of Breathing.” Nature Letter 536(4). Anderson, Tatiana M., and Jan-Marino Ramirez. 2017. “Respiratory Rhythm Generation: Triple Oscillator Hypothesis.” F1000Research 6(0): 139. Arnulf, Isabelle et al. 2000. “Sleep Disorders and Diaphragmatic Function in Patients with Amyotrophic Lateral Sclerosis.” American Journal of Respiratory and Critical Care Medicine 161(3 I): 849–56. Asmussen, E., and M. Nielsen. 1964. “Experiments on Nervous Factors Controlling Respiration and Circulation during Exercise Employing Blocking of the Blood Flow.” Acta Physiol Scand 60: 103–11. Azim, Eiman, Juan Jiang, Bror Alstermark, and Thomas M. Jessell. 2014. “Skilled Reaching Relies on a V2a Propriospinal Internal Copy Circuit.” Nature 508(7496): 357–63. Bach, Karen B., and Gordon S. Mitchell. 1996. “Hypoxia-Induced Long-Term Facilitation of Respiratory Activity Is Serotonin Dependent.” Respiration Physiology 104(2–3): 251–60. Baertsch, Nathan A., Liza J. Severs, Tatiana M. Anderson, and Jan Marino Ramirez. 2019. “A Spatially Dynamic Network Underlies the Generation of Inspiratory Behaviors.” Proceedings of the National Academy of Sciences of the United States of America 116(15): 7493–7502. Baker-Herman, Tracy L., and Gordon S. Mitchell. 2002. “Phrenic Long-Term Facilitation Requires Spinal Serotonin Receptor Activation and Protein Synthesis.” Journal of Neuroscience 22(14): 6239–46. Ballanyi, Klaus, Hiroshi Onimaru, and Ikuo Homma. 1999. “Respiratory Network Function in the Isolated Brainstem-Spinal Cord of Newborn Rats.” Progress in Neurobiology 59(6): 583–634. Barabino, Silvia M.L., Fabio Spada, Franco Cotelli, and Edoardo Boncinelli. 1997. “Inactivation of the Zebrafish Homologue of Chx10 by Antisense Oligonucleotides Causes Eye Malformations Similar to the Ocular Retardation Phenotype.” Mechanisms of Development 63(2): 133–43. Bareyre, FM et al. 2004. “The Injured Spinal Cord Spontaneously Forms a New Intraspinal Circuit in Adult Rats.” Nature Neuroscience 7(3): 269–77. Barnéoud, Pascal et al. 1997. “Quantitative Motor Assessment in FALS Mice: A Longitudinal Study.” NeuroReport 8(13): 2861–65. Basting, T.M. et al. 2015. “Hypoxia Silences Retrotrapezoid Nucleus Respiratory Chemoreceptors via Alkalosis.” Journal of Neuroscience 35: 527–43. Bellingham, Mark C., and Janusz Lipski. 1990. “Respiratory Interneurons in the C5 Segment of the Spinal Cord of the Cat.” Brain Research 533(1): 141–46. Bennett, J.R. et al. 2004. “Respiratory Muscle Activity during RME Sleep in Patients with Diaphragm

193

Paralysis.” Neurology 62: 134–237. Berlly, Michael, and Kazuko Shem. 2007. “Respiratory Management during the First Five Days after Spinal Cord Injury.” Journal of Spinal Cord Medicine 30(4): 309–18. Berlowitz, David J., Brooke Wadsworth, and Jack Ross. 2016. “Respiratory Problems and Management in People with Spinal Cord Injury.” Breathe 12(4): 328–40. Bezdudnaya, T, KM Hormigo, V Marchenko, and M Lane. 2018. “Spontaneous Respiratory Plasticity Following Unilateral High Cervical Spinal Cord Injury in Behaving Rats.” Experimental Neurology 305. Bochorishvili, Genrieta, Ruth L. Stornetta, Melissa B. Coates, and Patrice G. Guyenet. 2012. “Pre- Bötzinger Complex Receives Glutamatergic Innervation from Galaninergic and Other Retrotrapezoid Nucleus Neurons.” Journal of Comparative Neurology 520(5): 1047–61. Boillée, Séverine, Christine Vande Velde, and Don W W. Cleveland. 2006. “ALS: A Disease of Motor Neurons and Their Nonneuronal Neighbors.” Neuron 52(1): 39–59. Van Den Bosch, L., P. Van Damme, E. Bogaert, and W. Robberecht. 2006. “The Role of Excitotoxicity in the Pathogenesis of Amyotrophic Lateral Sclerosis.” Biochimica et Biophysica Acta - Molecular Basis of Disease 1762(11–12): 1068–82. Bourke, S.C., P.J. Shaw, and G.J. Gibson. 2001. “Respiratory Function vs Sleep-Disoredered Breathing as Predictors of QOL in ALS.” Neurology 57(11). Bouvier, Julien et al. 2015. “Descending Command Neurons in the Brainstem That Halt Locomotion.” Cell 163(5): 1191–1203. http://dx.doi.org/10.1016/j.cell.2015.10.074. Bretzner, Frédéric, and Robert M. Brownstone. 2013. “Lhx3-Chx10 Reticulospinal Neurons in Locomotor Circuits.” Journal of Neuroscience 33(37): 14681–92. Bruijn, Lucie I., Timothy M. Miller, and Don W. Cleveland. 2004. “Unraveling the Mechanisms Involved in Motor Neuron Degeneration in Als.” Annual Review of Neuroscience 27(1): 723–49. Butler, J.E., A.L. Hudson, and S.C. Gandevia. 2014. “The Neural Control of Human Inspiratory Muscles.” Progress in brain research 209. Butler, Jane E. 2007. “Drive to the Human Respiratory Muscles.” Respiratory Physiology and Neurobiology 159(2): 115–26. Butts, Jessica C. et al. 2017. “Differentiation of V2a Interneurons from Human Pluripotent Stem Cells.” Proceedings of the National Academy of Sciences of the United States of America 114(19): 4969– 74. Canning, Brendan J. et al. 2014. “Anatomy and Neurophysiology of Cough: CHEST Guideline and Expert Panel Report.” Chest 146(6): 1633–48. Capelli, Paolo, Chiara Pivetta, Maria Soledad Esposito, and Silvia Arber. 2017. “Locomotor Speed Control Circuits in the Caudal Brainstem.” Nature 551(7680): 373–77. http://dx.doi.org/10.1038/nature24064. Cappello, Valentina et al. 2012. “Analysis of Neuromuscular Junctions and Effects of Anabolic Steroid Administration in the SOD1G93A Mouse Model of ALS.” Molecular and Cellular Neuroscience 51(1– 2): 12–21. http://dx.doi.org/10.1016/j.mcn.2012.07.003. De Carvalho, Mamede, Susana Pinto, and Michael Swash. 2010. “Association of Paraspinal and Diaphragm Denervation in ALS.” Amyotrophic Lateral Sclerosis 11(1–2): 63–66. Castro-Moure, Federico, and Harry G. Goshgarian. 1996. “Reversible Cervical Hemispinalization of the Rat Spinal Cord by a Cooling Device.” Experimental Neurology 141(1): 102–12. 1997. “Morphological Plasticity Induced in the Phrenic Nucleus Following Cervical Cold Block of Descending Respiratory Drive.” Experimental Neurology 147(2): 299–310. De Castro, F. 1926. “Sur La Structure et l’innervation de La Glande Intercarotidienne (Glomus Caroticum) de l’homme et Des Mammiferes et Sur Un Nouveau Systeme de l’innervation Autonome Du Nerf Glossopharyngien.” Trav Lab Rech Biol 24: 365–432.

194

Chen, J. et al. 2002. “Chronic Hypoxia Upregulates Connexin43 Expression in Rat Carotid Body and Petrosal Ganglion.” Journal of Applied Physiology 92(4): 1480–86. Chen, Jia et al. 2007. “Effect of the Endothelin Receptor Antagonist Bosentan on Chronic Hypoxia- Induced Morphological and Physiological Changes in Rat Carotid Body.” American Journal of Physiology - Lung Cellular and Molecular Physiology 292(5): 1257–62. Cleland, Corey L, and Peter A Getting. 1993. “Cervical Spinal Cord ( C4-C6 ) of the Fluorocarbon-Perfused Guinea Pig.” : 307–11. Cleveland, Don W., and Jeffrey D. Rothstein. 2001. “From Charcot to Lou Gehrig: Deciphering Selective Motor Neuron Death in Als.” Nature Reviews Neuroscience 2(11): 806–19. Clovis, Yoanne M. et al. 2016. “Chx10 Consolidates V2a Interneuron Identity through Two Distinct Gene Repression Modes.” Cell Reports 16(6): 1642–52. http://dx.doi.org/10.1016/j.celrep.2016.06.100. Corda, M., C. Von Euler, and G. Lennerstrand. 1965. “Proprioceptive Innervation of the Diaphragm.” The Journal of Physiology 178(1): 161–77. Cotton, BA et al. 2005. “Respiratory Complications and Mortality Risk Associated Wtih Thoracic Spine Injury.” The Journal of Trauma 59. Courtine, Gregoire et al. 2008. “Recovery of Supraspinal Control of Stepping via Indirect Propriospinal Relay Connections after Spinal Cord Injury.” Nature Medicine 14(1): 69–74. Cregg, J.M. et al. 2017. “A Latent Propriospinal Network Can Restore Diaphragm Function after High Cervical Spinal Cord Injury.” Cell Reports 21: 654–65. Crone, Steven A. et al. 2008. “Genetic Ablation of V2a Ipsilateral Interneurons Disrupts Left-Right Locomotor Coordination in Mammalian Spinal Cord.” Neuron 60(1): 70–83. . 2012. “Irregular Breathing in Mice Following Genetic Ablation of V2a Neurons.” Journal of Neuroscience 32(23): 7895–7906. Crone, Steven A., Guisheng Zhong, Ronald Harris-Warrick, and Kamal Sharma. 2009. “In Mice Lacking V2a Interneurons, Gait Depends on Speed of Locomotion.” Journal of Neuroscience 29(21): 7098– 7109. Cruz, Martin Paspe. 2018. “Edaravone (Radicava): A Novel Neuroprotective Agent for the Treatment of Amyotrophic Lateral Sclerosis.” P and T 43(1): 25–28. Cui, Y. et al. 2016. “Defining PreBötzinger Complex Rhythm and Pattern Generating Neural Microcircuits in Vivo.” Neuron 91(3): 602–14. Dahlstrom, A., and K. Fuxe. 1964. “Demonstration of Monoamines in the Cell Bodies of Brain Stem Neurons.” Acta Physiol Scand 232: 1–55. Dalal, K, and AF DiMarco. 2014. “Diaphragm Pacing in Spinal Cord Injury.” Physical Medicine and Rehabilitation 25(3): 619–29. Datta, A, S Farmer, and J Stephens. 1991. “Central Nervous Pathways Underlying Synchroniziation of Human Motor Unit Firing Studied during Voluntary Contractions.” J Physiology 432. Decima, E., C. von Euler, and U. Thoden. 1969. “Intercostal-to-Phrenic Reflexes in the Spinal Cat.” Physiology. Denavit-Saubie, M. et al. 1994. “Maturation of Brainstem Neurons Involved in Respiratory Rhythmogenesis: Biochemical, Bioelectrical and Morphological Properties.” Biol Neonate 65: 171– 75. Devinney, Michael J, Adrianne G Huxtable, Nicole L Nichols, and Gordon S Mitchell. 2013. “Hypoxia Induced Phrenic Long Term Faciliation: Emergent Properties.” : 3–5. DeVivo, Michael J., Karin J. Black, and Samuel L. Stover. 1993. “Causes of Death during the First 12 Years after Spinal Cord Injury.” Archives of Physical Medicine and Rehabilitation 74(3): 248–54. Dharmadasa, Thanuja, and Matthew C. Kiernan. 2018. “Riluzole, Disease Stage and Survival in ALS.” The Lancet Neurology 17(5): 385–86. http://dx.doi.org/10.1016/S1474-4422(18)30091-7. Dick, T. E. et al. 2018. “Facts and Challenges in Respiratory Neurobiology.” Respiratory Physiology and

195

Neurobiology 258: 104–7. http://dx.doi.org/10.1016/j.resp.2015.01.014. DiMarco, Anthony F. 2009. “Phrenic Nerve Stimulation in Patients with Spinal Cord Injury.” Respiratory Physiology and Neurobiology 169(2): 200–209. Dobbins, E.G., and J.L. Feldman. 1994. “Brainstem Network Controlling Descending Drive to Phrenic Motoneurons in Rat.” J Comp Neurol 347: 64–86. Doi, Atsushi, and Jan Marino Ramirez. 2008. “Neuromodulation and the Orchestration of the Respiratory Rhythm.” Respiratory Physiology and Neurobiology 164(1–2): 96–104. Dougherty, B. J. et al. 2012. “Recovery of Inspiratory Intercostal Muscle Activity Following High Cervical Hemisection.” Respiratory Physiology and Neurobiology 183(3): 186–92. http://dx.doi.org/10.1016/j.resp.2012.06.006. Dougherty, Kimberly J. et al. 2013. “Locomotor Rhythm Generation Linked to the Output of Spinal Shox2 Excitatory Interneurons.” Neuron 80(4): 920–33. http://dx.doi.org/10.1016/j.neuron.2013.08.015. Dougherty, Kimberly J., and Ole Kiehn. 2010. “Firing and Cellular Properties of V2a Interneurons in the Rodent Spinal Cord.” Journal of Neuroscience 30(1): 24–37. Dubois, Christophe J. et al. 2018. “Role of the K+-CL- Cotransporter KCC2a Isoform in Mammalian Respiration at Birth.” eNeuro 5(5): 1–14. Dubreuil, Véronique et al. 2009. “Defective Respiratory Rhythmogenesis and Loss of Central Chemosensitivity in Phox2b Mutants Targeting Retrotrapezoid Nucleus Neurons.” Journal of Neuroscience 29(47): 14836–46. Duffin, J., and S. Iscoe. 1996. “The Possible Role of C5 Segment Inspiratory Interneurons Investigated by Cross-Correlation with Phrenic Motoneurons in Decerebrate Cats.” Experimental Brain Research 112(1): 35–40. Dutschmann, M., and T.E. Dick. 2012. “Pontine Mechanisms of Respiratory Control.” Comparative Physiology 2: 2443–69. Dutschmann, M., M. Morschel, M. Kron, and H. Herbert. 2004. “Development of Adaptive Behavior of the Respiratory Network: Implications for the Pontine Kolliker-Fuse Nucleus.” Respiratory physiology & neurobiology 143: 155–65. Dutschmann, M et al. 2014. “The Physiological Significance of Postinspiration in Respiratory Control.” Elsevier Neuroscience 212: 113–30. Eklöf-Ljunggren, Emma et al. 2012. “Origin of Excitation Underlying Locomotion in the Spinal Circuit of Zebrafish.” Proceedings of the National Academy of Sciences of the United States of America 109(14): 5511–16. Elbasiouny, Sherif M. 2010. “Management of Spasticity After Spinal Cord Injury.” Neurorehabil Neural Repair 24(1): 23–33. Eldridge, F.L., and P. Gill-Kumar. 1980. “Central Respiratory Effects of Carbon Dioxide, and Carotid Sinus Nerve and Muscle Afferents.” J Physiology 300: 75–87. Eldridge, F.L., D.E. Millhorn, and T.G. Waldrop. 1981. “Exercise Hyperpnea Nad Locomotion: Parallel Activation from the Hypothalamus.” Science 211: 844–46. Ellenberger, H.H., J.L. Feldman, and H.G. Goshgarian. 1990. “Ventral Respiratory Group Projections to Phrenic Motoneurons.” J Comp Neurol 302: 707–14. Ellenberger, HH, and JL Feldman. 1988. “Monosynaptic Transmission of Respiratory Drive to Phrenic Motoneurons from Brainstem Bulbospinal Neurons in Rats.” Journal of Neuroscience Research. Erickson, JT, and DE Millhorn. 1994. “Hypoxia and Electrical Stimulation of the Carotid Sinus Nerve Induce Fos-like Immunoreactivity within Catecholaminergic and Serotonergic Neurons of the Rat Brainstem.” The Journal of Comparative Neurology 384(2). Ericson, J. et al. 1997. “Pax6 Controls Progenitor Cell Identity and Neuronal Fate in Response to Graded Shh Signaling.” Cell 90(1): 169–80. Estenne, M, and A De Troyer. 1985. “Relationship between Respiratory Muscle Electromyogram and Rib

196

Cage Motion in Tetraplegia.” Am. Rev. Respir. Dys. 132: 53–59. Fallat, Robert J et al. 1979. “Spirometry Amyotrophic.” Archives of Neurology 36(2): 74–80. Feldman, Jack L., and Christopher A. Del Negro. 2006. “Looking for Inspiration: New Perspectives on Respiratory Rhythm.” Nature Reviews Neuroscience 7(3): 232–42. Feldman, Jack L., Christopher A. Del Negro, and Paul A. Gray. 2013. “Understanding the Rhythm of Breathing: So Near, Yet So Far.” Annual Review of Physiology 75(1): 423–52. Fenrich, Keith K., and P. Ken Rose. 2009. “Spinal Interneuron Axons Spontaneously Regenerate after Spinal Cord Injury in the Adult Feline.” Journal of Neuroscience 29(39): 12145–58. Fishburn, M.J., R.J. Marino, and J.F. Ditunno, Jr. 1990. “Atalectasis and Pneumonia in Acute Spinal Cord Injury.” Physical Medicine and Rehabilitation 71(3). Fogarty, M.J., C.B. Mantilla, and G.C. Sieck. 2018. “Breathing: Motor Control of Diaphragm Muscle.” Physiology 33(2): 113–26. Forster, Hubert V., Philippe Haouzi, and Jerome A. Dempsey. 2012. “Control of Breathing during Exercise.” Comprehensive Physiology 2(1): 743–77. Fournier, M., and G. C. Sieck. 1988. “Mechanical Properties of Muscle Units in the Cat Diaphragm.” Journal of Neurophysiology 59(3): 1055–66. Frood, Russell Thomas. 2011. “The Use of Treadmill Training to Recover Locomotor Ability in Patients with Spinal Cord Injury.” Bioscience Horizons 4(1): 108–17. Fujiwara, T, Y Hara, and N Chino. 1999. “Expiratory Function in Complete Tetraplegics: Study of Spirometry, Maximal Expiratory Pressure, and Muscle Activity of Pectoralis Major and Latissimus Dorsi Muscles.” American journal of physical medicine and rehabilitation 78(5). Fuller, D. D. et al. 2000. “Long Term Facilitation of Phrenic Motor Output.” Respiration Physiology 121(2– 3): 135–46. Fuller, David D., Francis J. Golder, E. B. Olson, and Gordon S. Mitchell. 2006. “Recovery of Phrenic Activity and Ventilation after Cervical Spinal Hemisection in Rats.” Journal of Applied Physiology 100(3): 800–806. Fuller, David D., Stephen M. Johnson, E. Burdette Olson, and Gordon S. Mitchell. 2003. “Synaptic Pathways to Phrenic Motoneurons Are Enhanced by Chronic Intermittent Hypoxia after Cervical Spinal Cord Injury.” Journal of Neuroscience 23(7): 2993–3000. Fuller, David D., Kun Ze Lee, and Nicole J. Tester. 2013. “The Impact of Spinal Cord Injury on Breathing during Sleep.” Respiratory Physiology and Neurobiology 188(3): 344–54. http://dx.doi.org/10.1016/j.resp.2013.06.009. Fuller, DD et al. 2005. “Cervical Spinal Cord Injury Upregulates Ventral Spinal 5-HT2A Receptors.” Journal of Neurotrauma 22(2). Le Gal, Jean Patrick, Laurent Juvin, Laura Cardoit, and Didier Morin. 2016. “Bimodal Respiratory- Locomotor Neurons in the Neonatal Rat Spinal Cord.” Journal of Neuroscience 36(3): 926–37. Gandevia, S. C., and J. C. Rothwell. 1987. “Activation of the Human Diaphragm from the Motor Cortex.” The Journal of Physiology 384(1): 109–18. Ghali, MG. 2017. “The Crossed Phrenic Phenomenon.” Neural regeneration research 12(6): 845–64. Ghali, Michael George Zaki. 2018. “Phrenic Motoneurons: Output Elements of a Highly Organized Intraspinal Network.” Journal of Neurophysiology 119(3): 1057–70. Giraudin, Aurore et al. 2012. “Spinal and Pontine Relay Pathways Mediating Respiratory Rhythm Entrainment by Limb Proprioceptive Inputs in the Neonatal Rat.” Journal of Neuroscience 32(34): 11841–53. Giraudin, Aurore, Marie Jeanne Cabirol-Pol, John Simmers, and Didier Morin. 2008. “Intercostal and Abdominal Respiratory Motoneurons in the Neonatal Rat Spinal Cord: Spatiotemporal Organization and Responses to Limb Afferent Stimulation.” Journal of Neurophysiology 99(5): 2626–40. Golder, Francis J. et al. 2003. “Respiratory Motor Recovery after Unilateral Spinal Cord Injury:

197

Eliminating Crossed Phrenic Activity Decreases Tidal Volume and Increases Contralateral Respiratory Motor Output.” Journal of Neuroscience 23(6): 2494–2501. ———. 2008. “Spinal Adenosine A2a Receptor Activation Elicits Long-Lasting Phrenic Motor Facilitation.” Journal of Neuroscience 28(9): 2033–42. Golder, Francis J., and Gordon S. Mitchell. 2005. “Spinal Synaptic Enhancement with Acute Intermittent Hypoxia Improves Respiratory Function after Chronic Cervical Spinal Cord Injury.” Journal of Neuroscience 25(11): 2925–32. Gordon, T., J. Hegedus, and T. Siu Lin. 2004. “Adaptive and Maladaptive Motor Axonal Sprouting in Aging and Motoneuorn Disease.” Neurological research 26(2). Gosgnach, Simon et al. 2017. “Delineating the Diversity of Spinal Interneurons in Locomotor Circuits.” Journal of Neuroscience 37(45): 10835–41. Goshgarian, HG. 2003. “Invited Review: The Crossed Phrenic Phenomenon: A Model for Plasticity in the Respiratory Pathways Folloiwng Spinal Cord Injury.” J Appl Physiol 94: 795–810. Goshgarian, HG, HH Ellenberger, and JL Feldman. 1991. “Decussation of Bulbospinal Respiratory Axons at the Level of the Phrenic Nuclei in Adult Rats: A Possible Substrate for the Crossed Phrenic Phenomenon.” Experimental Neurology 111(1): 135–39. Goshgarian, HG, and L Guth. 1977. “Demonstration of Functionally Ineffective Synapses in the Guinea Pig Spinal Cord.” Experimental Neurology 621: 613–21. Goshgarian, HG, XJ Yu, and JA Rafols. 1989. “Neuron and Glial Changes in the Rat Phrenic Nucleus Occurring within Hours after Spinal Cord Injury.” Comparative Neurology 284(4). Goulding, M. 2009. “Circuits Controlling Vertebrate Locomotion: Moving in a New Direction.” Nat Rev Neurosci 10(7): 507–18. Gray, P.A., J.C. Rekling, C.M. Bocchiaro, and J.L. Feldman. 2015. “Modulation of Respiratory Frequency by Peptidergic Input to Rhythmogenic Neurons in the PreBotzinger Complex.” Science 286: 1566– 68. Gray, Paul A. et al. 2010. “Developmental Origin of Prebötzinger Complex Respiratory Neurons.” Journal of Neuroscience 30(44): 14883–95. . 2013. “Transcription Factors Define the Neuroanatomical Organization of the Medullary Reticular Formation.” Frontiers in Neuroanatomy 7(APR): 1–21. Gregory, Susan A. 2007. “Evaluation and Management of Respiratory Muscle Dysfunction in ALS.” NeuroRehabilitation 22(6): 435–43. Grillner, Sten. 2006. “Biological Pattern Generation: The Cellular and Computational Logic of Networks in Motion.” Neuron 52(5): 751–66. Grillner, Sten, and Thomas M. Jessell. 2009. “Measured Motion: Searching for Simplicity in Spinal Locomotor Networks.” Current Opinion in Neurobiology 19(6): 572–86. Grillner, Sten, and Abdeljabbar El Manira. 2015. “The Intrinsic Operation of the Networks That Make Us Locomote.” Current Opinion in Neurobiology 31: 244–49. http://dx.doi.org/10.1016/j.conb.2015.01.003. Gurney, ME et al. 1994. “Motor Neuron Degeneration in Mice That Express a Human Cu, Zn Superoxide Dismutase Mutation.” Science 264: 1772–75. Guttmann, Ludwig, and JR Silver. 1965. “Electromyographic Studies on Reflex Activity of the Intercostal and Abdominal Muscles in Cervical Cord Lesions.” Guyenet, P.G. et al. 2013. “C1 Neurons: The Body’s EMTs.” American Journal of Physiology - Regulatory Integrative and Comparative Physiology 305: 187–204. —. 2014. “Regulation of Breathing and Autonomic Outflows by Chemoreceptors.” Comprehensive Physiology 4(4). Guyenet, P.G., and D.K. Mulkey. 2010. “Retrotrapezoid Nucleus and Parafacial Respiratory Group.” Respiratory physiology & neurobiology 173(3): 244–55.

198

Guyenet, Patrice G., and Douglas A. Bayliss. 2015. “Neural Control of Breathing and CO2 Homeostasis.” Neuron 87(5): 946–61. http://dx.doi.org/10.1016/j.neuron.2015.08.001. Haas, F et al. 1985. “Temporal Pulmonary Function Changes in Cervical Cord Injury.” Archives of Physical Medicine and Rehabilitation 66(3). Hachmann, Jan T. et al. 2017. “Electrical Neuromodulation of the Respiratory System After Spinal Cord Injury.” Mayo Clinic Proceedings 92(9): 1401–14. http://dx.doi.org/10.1016/j.mayocp.2017.04.011. Hachmann, JT et al. 2017. “Electrical Neuromodulation of the Respiratory System after Spinal Cord Injury.” Hadley, Scott D., Paul D. Walker, and Harry G. Goshgarian. 1999. “Effects of the Serotonin Synthesis Inhibitor P-CPA on the Expression of the Crossed Phrenic Phenomenon 4 h Following C2 Spinal Cord Hemisection.” Experimental Neurology 160(2): 479–88. Hägglund, Martin et al. 2013. “Optogenetic Dissection Reveals Multiple Rhythmogenic Modules Underlying Locomotion.” Proceedings of the National Academy of Sciences of the United States of America 110(28): 11589–94. Hardiman, Orla. 2011. “Management of Respiratory Symptoms in ALS.” Journal of Neurology 258(3): 359–65. Hatzipetros, Theo et al. 2015. “A Quick Phenotypic Neurological Scoring System for Evaluating Disease Progression in the SOD1-G93A Mouse Model of ALS.” Journal of Visualized Experiments 2015(104): 1–6. Hayashi, Marito et al. 2018. “Graded Arrays of Spinal and Supraspinal V2a Interneuron Subtypes Underlie Forelimb and Hindlimb Motor Control.” Neuron 97(4): 869-884.e5. https://doi.org/10.1016/j.neuron.2018.01.023. Hayashibe, M. et al. 2016. “Locomotor Improvement of Spinal Cord-Injured Rats through Treadmill Training by Forced Plantar Placement of Hind Paws.” Spinal Cord 54(7): 521–29. Hayes, Heather B. et al. 2014. “Daily Intermittent Hypoxia Enhances Walking after Chronic Spinal Cord Injury A Randomized Trial.” Neurology 82(2): 104–13. Hegedus, J., C. T. Putman, and T. Gordon. 2007. “Time Course of Preferential Motor Unit Loss in the SOD1G93A Mouse Model of Amyotrophic Lateral Sclerosis.” Neurobiology of Disease 28(2): 154– 64. Hegedus, J., C. T. Putman, N. Tyreman, and Tessa Gordon. 2008. “Preferential Motor Unit Loss in the SOD1 G93A Transgenic Mouse Model of Amyotrophic Lateral Sclerosis.” Journal of Physiology 586(14): 3337–51. Hernandez-Torres, V et al. 2017. “BDNF Effects on Functional Recovery across Motor Behaviors after Cervical Spinal Cord Injury.” Journal of Neurophysiology 117(2): 537–44. Hilaire, G., M. Khatib, and R. Monteau. 1983. “Spontaneous Respiratory Activity of Phrenic and Intercostal Renshaw Cells.” Neuroscience Letters 43(1): 97–101. Hilaire, Gérard et al. 2004. “Modulation of the Respiratory Rhythm Generator by the Pontine Noradrenergic A5 and A6 Groups in Rodents.” Respiratory Physiology and Neurobiology 143(2–3): 187–97. Hilaire, Gérard, and Bernard Duron. 1999. “Maturation of the Mammalian Respiratory System.” Physiological Reviews 79(2): 325–60. Hogue, I.B. et al. 2018. “Characterization of the Neuroinvasive Profile of a Pseudorabies Virus Recombinant Express the MTurquoise2 Reporter in Single and Multiple Injection Experiments.” Journal of Neuroscience Methods 308: 228–39. Hoh, Daniel J., Lynne M. Mercier, Shaunn P. Hussey, and Michael A. Lane. 2013. “Respiration Following Spinal Cord Injury: Evidence for Human Neuroplasticity.” Respiratory Physiology and Neurobiology 189(2): 450–64. Hökfelt, T., K. Fuxe, M. Goldstein, and O. Johansson. 1974. “Immunohistochemical Evidence for the

199

Existence of Adrenaline Neurons in the Rat Brain.” Brain Research 66(2): 235–51. Howland, David S. et al. 2002. “Focal Loss of the Glutamate Transporter EAAT2 in a Transgenic Rat Model of SOD1 Mutant-Mediated Amyotrophic Lateral Sclerosis (ALS).” Proceedings of the National Academy of Sciences of the United States of America 99(3): 1604–9. Hubner, C.A. et al. 2001. “Disruption of KCC2 Reveals an Essential Role of KCl Cotransport Already in Early Synaptic Inhibition.” Neuron 30: 515–24. Huckstepp, Robert T.R., Kathryn P. Cardoza, Lauren E. Henderson, and Jack L. Feldman. 2015. “Role of Parafacial Nuclei in Control of Breathing in Adult Rats.” Journal of Neuroscience 35(3): 1052–67. Ichiyama, Ronaldo M. et al. 2008. “Step Training Reinforces Specific Spinal Locomotor Circuitry in Adult Spinal Rats.” Journal of Neuroscience 28(29): 7370–75. Iizuka, M., H. Onimaru, and M. Izumizaki. 2016. “Distribution of Respiration-Related Neurons in the Thoracic Spinal of Neonatal Rats: Optical Imaging Study.” Elsevier Neuroscience 315: 217–27. Iizuka, Makito, Keiko Ikeda, Hiroshi Onimaru, and Masahiko Izumizaki. 2018. “Expressions of VGLUT1/2 in the Inspiratory Interneurons and GAD65/67 in the Inspiratory Renshaw Cells in the Neonatal Rat Upper Thoracic Spinal Cord.” IBRO Reports 5(May): 24–32. https://doi.org/10.1016/j.ibror.2018.08.001. Jensen, V.N., S.H. Romer, S.M. Turner, and S.A. Crone. 2017. “Repeated Measurement of Respiratory Muscle Activity and Ventilation in Mouse Models of Neuromuscular Disease.” Jove 122. Jensen, Victoria N. et al. 2019. “V2a Neurons Constrain Extradiaphragmatic Respiratory Muscle Activity at Rest.” eNeuro 6(4): 1–18. Johnson, Rebecca A., and Gordon S. Mitchell. 2013. “Common Mechanisms of Compensatory Respiratory Plasticity in Spinal Neurological Disorders.” Respiratory Physiology and Neurobiology 189(2): 419–28. http://dx.doi.org/10.1016/j.resp.2013.05.025. Johnson, S.M.., N. Koshiya, and J.C. Smitch. 2001. “Isolation of the Kernel for Respiratory Rhythm Generation in a Novel Preparation: The Pre-Botzinger Complex ‘Island.’” Journal of Neurophysiology 85(4). Jones, Sarah E. et al. 2012. “The Nucleus Retroambiguus as Possible Site for Inspiratory Rhythm Generation Caudal to Obex.” Respiratory Physiology and Neurobiology 180(2–3): 305–10. http://dx.doi.org/10.1016/j.resp.2011.12.007. Kam, Kaiwen et al. 2013. “Distinct Inspiratory Rhythm and Pattern Generating Mechanisms in the PreBötzinger Complex.” Journal of Neuroscience 33(22): 9235–45. Karunaratne, Asanka, Murray Hargrave, Alisa Poh, and Toshiya Yamada. 2002. “GATA Proteins Identify a Novel Ventral Interneuron Subclass in the Developing Chick Spinal Cord.” Developmental Biology 249(1): 30–43. Kiehn, Ole. 2016. “Decoding the Organization of Spinal Circuits That Control Locomotion.” Nature Reviews Neuroscience 17(4): 224–38. http://dx.doi.org/10.1038/nrn.2016.9. Kimura, Yukiko et al. 2013. “Hindbrain V2a Neurons in the Excitation of Spinal Locomotor Circuits during Zebrafish Swimming.” Current Biology 23(10): 843–49. http://dx.doi.org/10.1016/j.cub.2013.03.066. Kimura, Yukiko, Yasushi Okamura, and Shin Ichi Higashijima. 2006. “Alx, a Zebrafish Homolog of Chx10, Marks Ipsilateral Descending Excitatory Interneurons That Participate in the Regulation of Spinal Locomotor Circuits.” Journal of Neuroscience 26(21): 5684–97. Kinkead, Richard et al. 1998. “Cervical Dorsal Rhizotomy Enhances Serotonergic Innervation of Phrenic Motoneurons and Serotonin-Dependent Long-Term Facilitation of Respiratory Motor Output in Rats.” Journal of Neuroscience 18(20): 8436–43. Kirkwood, B Y P A, T A Sears, and R H Westgaard. 1981. “Recurrent Inhibition of Intercostal Motoneurones in the Cat.” J Physiology: 111–30. Krogh, A., and J. Lindhard. 1913. “The Regulation of Respiration and Circulation during the Initial Stages

200

of Muscular Work.” J Physiology 47: 112–36. Kuhnlein, P. et al. 2008. “Diagnosis and Treatment of Bulbar Symptoms in Amyotrophic Lateral Sclerosis.” Nature clinical practice 4: 366–74. Kumar, Prem, Nanduri R. Prabhakar, and Kumar Prem. 2012. “Peripheral Chemoreceptors Function And.” Comprehensive Physiology 2(1): 141–219. http://www.ncbi.nlm.nih.gov/pubmed/23728973. Lane, Michael A. et al. 2008. “Cervical Prephrenic Interneurons in the Normal and Lesioned Spinal Cord of the Adult Rat.” Journal of Comparative Neurology 511(5): 692–709. Larsson, B. et al. 2001. “Fibre Type Proportion and Fibre Size in Trapezius Muscle Biopsies from Cleaners with and without Myalgia and Its Correlation with Ragged Red Fibres, Cytochrome-c-Oxidase- Negative Fibres, Biomechanical Output, Perception of Fatigue, and Surface Electromyog.” European Journal of Applied Physiology 84(6): 492–502. Lazarenko, R.M. et al. 2009. “Acid Sensitivity and Ultrastructure of the Retrotrapezoid Nucleus in Phox2b-EGFP Transgenic Mice.” Comparative Neurology 517: 69–86. Lechtzin, Noah. 2006. “Respiratory Effects of Amyotrophic Lateral Sclerosis: Problems and Solutions.” Respiratory Care 51(8): 871–81. Lee, Jae K., and Binhai Zheng. 2008. “Axon Regeneration after Spinal Cord Injury: Insight from Genetically Modified Mouse Models.” Restorative Neurology and Neuroscience 26(2-3 AXONAL REGENERATO): 175–82. Lee, K et al. 2013. “Phrenic Motoneuron Discharge Patterns Following Chronic Cervical Spinal Cord Injury.” Experimental Neurology 249. ———. 2015. “Hypoxia Triggers Short Term Potentiation of Phrenic Motoneuron Discharge after Chronic Cervical Spinal Cord Injury.” Experimental Neurology 263: 314–24. Lee, Kun Ze, and David D. Fuller. 2011. “Neural Control of Phrenic Motoneuron Discharge.” Respiratory Physiology and Neurobiology 179(1): 71–79. http://dx.doi.org/10.1016/j.resp.2011.02.014. LEWIS, L. J., and J. M. BROOKHART. 1951. “Significance of the Crossed Phrenic Phenomenon.” The American journal of physiology 166(2): 241–54. Li, Peng, and Kevin Yackle. 2017. “Sighing.” Current Biology 27(3): R88–89. Li, Shengguo, Kamana Misra, Michael P. Matise, and Mengqing Xiang. 2005. “Foxn4 Acts Synergistically with Mash1 to Specify Subtype Identity of V2 Interneurons in the Spinal Cord.” Proceedings of the National Academy of Sciences of the United States of America 102(30): 10688–93. Lipski, J., R. E.W. Fyffe, and J. Jodkowski. 1985. “Recurrent Inhibition of Cat Phrenic Motoneurons.” Journal of Neuroscience 5(6): 1545–55. Liu, R. et al. 1999. “Increased Hydroxyl Radical Production and Apoptosis in PC12 Neuorn Cells Expressing the Gain of Function Mutant G93A SOD1 Gene.” Radiation Research 151: 133–41. Llado, J. et al. 2006. “Degeneration of Respiratory Motor Neurons in the SOD1 G93A Transgenic Rat Model of ALS.” Neurobiology of Disease 21(1): 110–18. Lois, J.H., C.D. Rice, and B.J. Yates. 2008. “Neural Circuits Controlling Diaphragm Function in the Cat Revealed by Transneuronal Tracing.” Journal of Applied Physiology 106(1): 138–52. Loveridge, B., R. Sanii, and H. I. Dubo. 1992. “Breathing Pattern Adjustments during the First Year Following Cervical Spinal Cord Injury.” Paraplegia 30(7): 479–88. Loveridge, BM, and HI Dubo. 1990. “Breathing Pattern in Chronic Quadriplegia.” Archives of Physical Medicine and Rehabilitation 71(7). Lu, D.C., T. Niu, and W.A. Alaynick. 2015. “Molecular and Cellular Development of Spinal Cord Locomotor Circuitry.” Frontiers in molecular neuroscience 8(25). Ludolph, A. C., and U. Knirsch. 1999. “Problems and Pitfalls in the Diagnosis of ALS.” Journal of the Neurological Sciences 165(SUPPL. 1): 14–20. Lyall, R. A. 2001. “Respiratory Muscle Strength and Ventilatory Failure in Amyotrophic Lateral Sclerosis.” Brain 124(10): 2000–2013.

201

Macefield, B Y Gary, and Simon C Gandeviat. 1991. “5 . During Quiet Breathing , in Which Subjects Were Relaxed and Distracted From.” : 545–58. Mantilla, Carlos B. et al. 2011. “Chronic Assessment of Diaphragm Muscle EMG Activity across Motor Behaviors.” Respiratory Physiology and Neurobiology 177(2): 176–82. http://dx.doi.org/10.1016/j.resp.2011.03.011. Mantilla, CB et al. 2014. “TrkB Kinase Activity Is Critical for Recovery of Respiratory Function after Cervical Spinal Cord Injury.” Experimental Neurology 0: 190–95. Marchenko, Vitaliy, Michael G.Z. Ghali, and Robert F. Rogers. 2015. “The Role of Spinal GABAergic Circuits in the Control of Phrenic Nerve Motor Output.” American Journal of Physiology - Regulatory Integrative and Comparative Physiology 308(11): R916–26. Marina, Nephtali et al. 2010. “Essential Role of Phox2b-Expressing Ventrolateral Brainstem Neurons in the Chemosensory Control of Inspiration and Expiration.” Journal of Neuroscience 30(37): 12466– 73. Marti-Fabregas, J. et al. 1995. “Respiratory Function Deterioration Is Not Time-Linked with Upper-Limb Onset in Amyotrophic Lateral Sclerosis.” Acta neurologica Scandinavica 92: 261–64. Mayeux R. 2003. “Epidemiology of Neurodegeneration.” Annual review of neuroscience 26: 81–104. McBain, Rachel A. et al. 2016. “Human Intersegmental Reflexes from Intercostal Afferents to Scalene Muscles.” Experimental Physiology 101(10): 1301–8. McCrea, David A., and Ilya A. Rybak. 2008. “Organization of Mammalian Locomotor Rhythm and Pattern Generation.” Brain Research Reviews 57(1): 134–46. McLean, DL et al. 2008. “Continuous Shifts in the Active Set of Spinal Interneurons during Changes in Locomotor Speed.” Nature Neuroscience 11(12): 1419–29. McMichan, John C. 1980. “Pulmonary Dysfunction Following Traumatic Quadriplegia.” Jama 243(6): 528. Meller, R, JM Babity, and DG Grahame-Smith. 2002. “5-HT2A Receptor Activation Leads to Increased BDNF MRNA Expression in C6 Glioma Cells.” Neuromolecular Med 1(3): 197–205. Mendell, Lorne M. 2005. “The Size Principle: A Rule Describing the Recruitment of Motoneurons.” Journal of Neurophysiology 93(6): 3024–26. Menelaou, Evdokia, and David L McLean. 2019. “Hierarchical Control of Locomotion by Distinct Types of Spinal V2a Interneurons in Zebrafish.” Nature communications 10(1): 4197. http://www.ncbi.nlm.nih.gov/pubmed/31519892%0Ahttp://www.pubmedcentral.nih.gov/articlere nder.fcgi?artid=PMC6744451. Merico, Antonio et al. 2018. “Effects of Combined Endurance and Resistance Training in Amyotrophic Lateral Sclerosis: A Pilot, Randomized, Controlled Study.” European Journal of Translational Myology 28(1): 132–40. Merrill, E. G., and J. Lipski. 1987. “Inputs to Intercostal Motoneurons from Ventrolateral Medullary Respiratory Neurons in the Cat.” Journal of Neurophysiology 57(6): 1837–53. Meznaric, Marija, and Erika Cvetko. 2016. “Size and Proportions of Slow-Twitch and Fast-Twitch Muscle Fibers in Human Costal Diaphragm.” BioMed Research International 2016: 10–15. Millhorn, DE, FL Eldridge, and TG Wadrop. 1980. “Prolonged Stimulation of Respiration by a New Central Neural Mechanism.” Respiration Physiology (February): 87–103. Minor, KH, LK Akison, HG Goshgarian, and NW Seeds. 2006. “Spinal Cord Injury-Induced Plasticity in the Mouse - the Crossed Phrenic Phenomenon.” Experimental Neurology 200(2): 486–95. Miscio, Giacinta et al. 2006. “The Cortico-Diaphragmatic Pathway Involvement in Amyotrophic Lateral Sclerosis: Neurophysiological, Respiratory and Clinical Considerations.” Journal of the Neurological Sciences 251(1–2): 10–16. Moreno, Dale E., Xiao Jun Yu, and Harry G. Goshgarian. 1992. “Identification of the Axon Pathways Which Mediate Functional Recovery of a Paralyzed Hemidiaphragm Following Spinal Cord Hemisection in the Adult Rat.” Experimental Neurology 116(3): 219–28.

202

Morgan, Barbara J. et al. 2014. “Quantifying Hypoxia-Induced Chemoreceptor Sensitivity in the Awake Rodent.” Journal of Applied Physiology 117(7): 816–24. Morin, D., and D. Viala. 2002. “Coordinations of Locomotor and Respiratory Rhythms in Vitro Are Critically Dependent on Hindlimb Sensory Inputs.” J Neurosci 22: 4756–65. Mulkey, D.K. et al. 2004. “Respiratory Control by Ventral Surface Chemoreceptor Neurons in Rats.” Nat Neuroscience 7: 1360–69. Nagai, M. et al. 2001. “Rats Expressing Human Cytosolic Copper-Zinc Superoxide Dismutase Transgenes with Amyotrophic Lateral Sclerosis: Associated Mutations Develop Motor Neuron Disease.” Journal of Neuroscience 21(23): 9246–54. Nair, Jayakrishnan, Kristi A. Streeter, et al. 2017. “Anatomy and Physiology of Phrenic Afferent Neurons.” Journal of Neurophysiology 118(6): 2975–90. Nair, Jayakrishnan, Tatiana Bezdudnaya, et al. 2017. “Histological Identification of Phrenic Afferent Projections to the Spinal Cord.” Respiratory Physiology and Neurobiology 236: 57–68. http://dx.doi.org/10.1016/j.resp.2016.11.006. Narula, Monica et al. 2014. “Afferent Neural Pathways Mediating Cough in Animals and Humans.” Journal of thoracic disease 6(Suppl 7): S712-9. http://www.ncbi.nlm.nih.gov/pubmed/25383205. Nashold, LJ et al. 2006. “Phrenic, but Not Hypoglossal, Motor Output Is Diminished in a Rat Model of Amyotrophic Lateral Sclerosis (ALS).” FASEB 20. Navarrete-Opazo, A., B. J. Dougherty, and G. S. Mitchell. 2017. “Enhanced Recovery of Breathing Capacity from Combined Adenosine 2A Receptor Inhibition and Daily Acute Intermittent Hypoxia after Chronic Cervical Spinal Injury.” Experimental Neurology 287: 93–101. http://dx.doi.org/10.1016/j.expneurol.2016.03.026. Navarrete-Opazo, Angela, Julio Alcayaga, Denisse Testa, and Ana Luisa Quinteros. 2016. “Intermittent Hypoxia Does Not Elicit Memory Impairment in Spinal Cord Injury Patients.” Archives of Clinical Neuropsychology 31(4): 332–42. Del Negro, Christopher A., Gregory D. Funk, and Jack L. Feldman. 2018. “Breathing Matters.” Nature Reviews Neuroscience 19(6): 351–67. http://dx.doi.org/10.1038/s41583-018-0003-6. Del Negro, Christopher A., Naohiro Koshiya, Robert J. Butera, and Jeffrey C. Smith. 2002. “Persistent Sodium Current, Membrane Properties and Bursting Behavior of Pre-Bötzinger Complex Inspiratory Neurons in Vitro.” Journal of Neurophysiology 88(5): 2242–50. Nichols, N.L. et al. 2013. “Ventilatory Control in ALS.” Respiratory physiology & neurobiology 189: 429– 37. Nichols, Nicole L. et al. 2013. “Intermittent Hypoxia and Stem Cell Implants Preserve Breathing Capacity in a Rodent Model of Amyotrophic Lateral Sclerosis.” American Journal of Respiratory and Critical Care Medicine 187(5): 535–42. Nichols, Nicole L., Irawan Satriotomo, Daniel J. Harrigan, and Gordon S. Mitchell. 2015. “Acute Intermittent Hypoxia Induced Phrenic Long-Term Facilitation despite Increased SOD1 Expression in a Rat Model of ALS.” Experimental Neurology 273: 138–50. http://dx.doi.org/10.1016/j.expneurol.2015.08.011. Niedermeyer, Shannon, Michael Murn, and Philip J. Choi. 2019. “Respiratory Failure in Amyotrophic Lateral Sclerosis.” Chest 155(2): 401–8. https://doi.org/10.1016/j.chest.2018.06.035. Nwantwi, KD et al. 1999. “Spontaneous Functional Recovery in a Paralyzed Hemidiaphragm Following Upper Cervical Spinal Cord Injury in Adult Rats.” NeuroRehabilitation and Neural Repair 13(4). Oka, Atsushi, Makito Iizuka, Hiroshi Onimaru, and Masahiko Izumizaki. 2019. “Inhibitory Thoracic Interneurons Are Not Essential to Generate the Rostro-Caudal Gradient of the Thoracic Inspiratory Motor Activity in Neonatal Rat.” Neuroscience 397: 1–11. https://doi.org/10.1016/j.neuroscience.2018.11.037. Okabe, Akihito et al. 2015. “KCC2-Mediated Regulation of Respiration-Related Rhythmic Activity during

203

Postnatal Development in Mouse Medulla Oblongata.” Brain Research 1601: 31–39. http://dx.doi.org/10.1016/j.brainres.2015.01.007. Oku, Yoshitaka, Haruko Masumiya, and Yasumasa Okada. 2007. “Postnatal Developmental Changes in Activation Profiles of the Respiratory Neuronal Network in the Rat Ventral Medulla.” Journal of Physiology 585(1): 175–86. Oliver, David. 1996. “The Quality of Care and Symptom Control - the Effects on the Terminal Phase of ALS/MND.” Journal of the Neurological Sciences 139(SUPPL.): 134–36. Pagliardini, Silvia et al. 2011. “Active Expiration Induced by Excitation of Ventral Medulla in Adult Anesthetized Rats.” Journal of Neuroscience 31(8): 2895–2905. de Paleville, Daniela G.L.Terson, William B. McKay, Rodney J. Folz, and Alexander V. Ovechkin. 2011a. “Respiratory Motor Control Disrupted by Spinal Cord Injury: Mechanisms, Evaluation, and Restoration.” Translational Stroke Research 2(4): 463–73. 2011b. “Respiratory Motor Control Disrupted by Spinal Cord Injury: Mechanisms, Evaluation, and Restoration.” Translational Stroke Research 2(4): 463–73. Park, Jung Hyun et al. 2010. “How Respiratory Muscle Strength Correlates with Cough Capacity in Patients with Respiratory Muscle Weakness.” Yonsei Medical Journal 51(3): 392–97. Di Pasquale, Eric, Fabien Tell, Roger Monteau, and Gérard Hilaire. 1996. “Perinatal Developmental Changes in Respiratory Activity of Medullary and Spinal Neurons: An in Vitro Study on Fetal and Newborn Rats.” Developmental Brain Research 91(1): 121–30. Persegol, L., R. Palisses, and D. Viala. 1987. “Different Mechanisms Involved in Supraspinal and Spinal Reflex Regulation of Phrenic Activity through Chest Movements.” Neuroscience 23(2): 631–40. Peter, James B. et al. 1972. “Metabolic Profiles of Three Fiber Types of Skeletal Muscle in Guinea Pigs and Rabbits.” Biochemistry 11(14): 2627–33. Petrof, Basil J., and Sabah N. Hussain. 2016. “Ventilator-Induced Diaphragmatic Dysfunction: What Have We Learned?” Current Opinion in Critical Care 22(1): 67–72. Petrov, Theodor, Christian Kreipke, Warren Alilain, and Kwaku D. Nantwi. 2007. “Differential Expression of Adenosine A1 and A2A Receptors after Upper Cervical (C2) Spinal Cord Hemisection in Adult Rats.” Journal of Spinal Cord Medicine 30(4): 331–37. Pinto, S., A. Pinto, and M. De Carvalho. 2007. “Do Bulbar-Onset Amyotrophic Lateral Sclerosis Patients Have an Earlier Respiratory Involvement than Spinal-Onset Amyotrophic Lateral Sclerosis Patients?” Europa mediocophysica 43: 505–9. Pinto, Susana, and Mamede De Carvalho. 2008. “Motor Responses of the Sternocleidomastoid Muscle in Patients with Amyotrophic Lateral Sclerosis.” Muscle and Nerve 38(4): 1312–17. Pivetta, Chiara, Maria Soledad Esposito, Markus Sigrist, and Silvia Arber. 2014. “Motor-Circuit Communication Matrix from Spinal Cord to Brainstem Neurons Revealed by Developmental Origin.” Cell 156(3): 537–48. http://dx.doi.org/10.1016/j.cell.2013.12.014. Polla, B., G. D’Antona, R. Bottinelli, and C. Reggiani. 2004. “Respiratory Muscle Fibres: Specialisation and Plasticity.” Thorax 59(9): 808–17. Polverino, Mario et al. 2012. “Anatomy and Neuro-Pathophysiology of the Cough Reflex Arc.” Multidisciplinary respiratory medicine 7(1): 5. http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3415124&tool=pmcentrez&rendertyp e=abstract. Porter, W. T. 1895. “The Path of the Respiratory Impulse from the Bulb to the Phrenic Nuclei.” The Journal of Physiology 17(6): 455–85. Posluszny, Joseph A. et al. 2014. “Multicenter Review of Diaphragm Pacing in Spinal Cord Injury: Successful Not Only in Weaning from Ventilators but Also in Bridging to Independent Respiration.” Journal of Trauma and Acute Care Surgery 76(2): 303–10. Prakash, Y. S. et al. 2000. “Phrenic Motoneuron Morphology during Rapid Diaphragm Muscle Growth.”

204

Journal of Applied Physiology 89(2): 563–72. Prendergast, Andrew, and Claire Wyart. 2016. “Locomotion: Electrical Coupling of Motor and Premotor Neurons.” Current Biology 26(6): R235–37. http://dx.doi.org/10.1016/j.cub.2016.02.021. Ramirez, Jan Marino, and Tatiana M. Anderson. 2017. “Respiratory Rhythm Generation: Triple Oscillator Hypothesis.” F1000Research 6(0): 1–10. Ramirez, Jan Marino, Tatiana Dashevskiy, Ibis Agosto Marlin, and Nathan Baertsch. 2016. “Microcircuits in Respiratory Rhythm Generation: Commonalities with Other Rhythm Generating Networks and Evolutionary Perspectives.” Current Opinion in Neurobiology 41: 53–61. http://dx.doi.org/10.1016/j.conb.2016.08.003. Ray, R.S. et al. 2011. “NIH Public Access.” Science 333: 83–88. Rehder, K. 1979. “Anesthesia and the Respiratory System.” Can Anaesth Soc J 26(6): 451–62. Remmers, By John E. 1973. “From the Department of Physiology , Dartmouth Medical School ,.” : 45–62. Renton, Alan E, Adriano Chiò, Bryan J Traynor, and Nat Neurosci. 2014. “State of Play in Amyotrophic Lateral Sclerosis Genetics HHS Public Access Author Manuscript.” Nat Neurosci 17(1): 17–23. https://europepmc.org/backend/ptpmcrender.fcgi?accid=PMC4544832&blobtype=pdf. Richter, Diethelm W., and Jeffrey C. Smith. 2014. “Respiratory Rhythm Generation in Vivo.” Physiology 29(1): 58–71. Rizzuto, Emanuele, Simona Pisu, Antonio Musarò, and Zaccaria Del Prete. 2015. “Measuring Neuromuscular Junction Functionality in the SOD1G93A Animal Model of Amyotrophic Lateral Sclerosis.” Annals of Biomedical Engineering 43(9): 2196–2206. Road, J.D. 1990. “Phrenic Afferents and Ventilatory Control.” Lung 168(3): 137–49. Romer, Shannon H. et al. 2017. “Accessory Respiratory Muscles Enhance Ventilation in ALS Model Mice and Are Activated by Excitatory V2a Neurons.” Experimental Neurology 287: 192–204. http://dx.doi.org/10.1016/j.expneurol.2016.05.033. Rosen, DR et al. 1993. “Mutations in Cu/Zn Superoxide Dismutase Gene Are Associated with Familial Amyotrophic Lateral Sclerosis.” Nature 362: 59–62. Rosenblueth, A, and T Ortiz. 1936. “THE CROSSED RESPIRATORY TO THE.” 574. Rybak, Ilya A., Kimberly J. Dougherty, and Natalia A. Shevtsova. 2015. “Organization of the Mammalian Locomotor CPG: Review of Computational Model and Circuit Architectures Based on Genetically Identified Spinal Interneurons.” eNeuro 2(5). Sandhu, MS et al. 2015. “Midcervical Neuronal Discharge Patterns during and Following Hypoxia.” J Neurophysiology 113(7): 2091–2101. Sant’Ambrogio, G., O. P. Mathew, J. T. Fisher, and F. B. Sant’Ambrogio. 1983. “Laryngeal Receptors Responding to Transmural Pressure, Airflow and Local Muscle Activity.” Respiration Physiology 54(3): 317–30. Satkunendrarajah, K. et al. 2018. “Cervical Excitatory Neurons Sustain Breathing after Spinal Cord Injury.” Nature 562: 419–22. Satriotomo, I., E.A. Dale, J.M. Dahlberg, and G.S. Mitchell. 2012. “Repetitive Acute Intermittent Hypoxia Increases Expression of Proteins Associated with Plasticity in the Phrenic Motor Nucleus.” Elsevier 237(1): 103–15. Saywell, S. A. et al. 2011. “Electrophysiological and Morphological Characterization of Propriospinal Interneurons in the Thoracic Spinal Cord.” Journal of Neurophysiology 105(2): 806–26. Schilero, Gregory J. et al. 2009. “Pulmonary Function and Spinal Cord Injury.” Respiratory Physiology and Neurobiology 166(3): 129–41. Schütz, Burkhard. 2005. “Imbalanced Excitatory to Inhibitory Synaptic Input Precedes Motor Neuron Degeneration in an Animal Model of Amyotrophic Lateral Sclerosis.” Neurobiology of Disease 20(1): 131–40. Sheikhbahaei, Shahriar, Alexander V. Gourine, and Jeffrey C. Smith. 2017. “Respiratory Rhythm

205

Irregularity after Carotid Body Denervation in Rats.” Respiratory Physiology and Neurobiology 246(May): 92–97. Shevtsova, Natalia A., Vitaliy Marchenko, and Tatiana Bezdudnaya. 2019. “Modulation of Respiratory System by Limb Muscle Afferents in Intact and Injured Spinal Cord.” Frontiers in Neuroscience 13(March): 1–11. Shimizu, Toshio et al. 2010. “Electrophysiological Assessment of Corticorespiratory Pathway Function in Amyotrophic Lateral Sclerosis.” Amyotrophic Lateral Sclerosis 11(1–2): 57–62. Sieck, G.C., L.F. Ferreira, M.B. Reid, and C.B. Mantilla. 2013. “Mechanical Properties of Respiratory Muscles.” Comparative Physiology 3(4): 1553–67. Sieck, G.C., and H.M. Gransee. Respiratory Muscles: Structure, Function and Regulation. Smith, Andrew C., and Maria Knikou. 2016. “A Review on Locomotor Training after Spinal Cord Injury: Reorganization of Spinal Neuronal Circuits and Recovery of Motor Function.” Neural Plasticity 2016. Smith, J.C., A.P.L. Abdala, H. Koizumi, and J.F.R. Paton. 2007. “Spatial and Functional Architecture of the Mammalian Brain Stem Respiratory Network: A Hierarchy of Three Oscillatory Mechanisms.” J Neurophysiology 98(6): 3370–87. Smith, Jeffrey C. et al. 1991. “Pre-Bötzinger Complex: A Brainstem Region That May Generate Respiratory Rhythm in Mammals.” Science 254(5032): 726–29. ———. 2013. “Brainstem Respiratory Networks: Building Blocks and Microcircuits.” Trends in Neurosciences 36(3): 152–62. http://dx.doi.org/10.1016/j.tins.2012.11.004. Smith, PEM et al. 1987. “Practical Problems in the Respiratory Care of Patients with Muscular Dystrophy.” Smittkamp, SE et al. 2010. “Measures of Bulbar and Spinal Motor Function, Muscle Innervation, and Mitochondrial Function in ALS Rats.” Behavioural Brain Research 211: 48–57. Smittkamp, Susan E. et al. 2010. “Measures of Bulbar and Spinal Motor Function, Muscle Innervation, and Mitochondrial Function in ALS Rats.” Behavioural Brain Research 211(1): 48–57. http://dx.doi.org/10.1016/j.bbr.2010.03.007. Song, G., H. Wang, H. Xu, and C.S. Poon. 2012. “Kolliker Fuse Neurons Send Collateral Projections to Multiple Hypoxia Activated and Non Activated Structures in Rat Brainstem and Spinal Cord.” Brain Structure 217: 835–58. Song, Jianren, Elin Dahlberg, and Abdeljabbar El Manira. 2018. “V2a Interneuron Diversity Tailors Spinal Circuit Organization to Control the Vigor of Locomotor Movements.” Nature Communications 9(1): 1–14. Stewart, Heather et al. 2001. “Electromyography of Respiratory Muscles in Amyotrophic Lateral Sclerosis.” Journal of the Neurological Sciences 191(1–2): 67–73. Streeter, KA et al. 2019. “Mid-Cervical Interneuron Networks Following High Cervical Spinal Cord Injury.” Respiratory physiology & neurobiology 271. Streeter, Kristi A. et al. 2017. “Intermittent Hypoxia Enhances Functional Connectivity of Midcervical Spinal Interneurons.” Journal of Neuroscience 37(35): 8349–62. Strickland, LM et al. 2019. “Chronic Intermittent Hypoxia and Hypercapnia Induces Respiratory Insufficiency in an Amyotrophic Lateral Sclerosis Mouse Model.” Physiology. Takakura, Ana Carolina Thomaz et al. 2006. “Peripheral Chemoreceptor Inputs to Retrotrapezoid Nucleus (RTN) CO2-Sensitive Neurons in Rats.” Journal of Physiology 572(2): 503–23. Talakad, N. et al. 2009. “Assessment of Pulmonary Function in Amyotrophic Lateral Sclerosis.” 51(2). Tamplin, Jeanette et al. 2011. “Assessment of Breathing Patterns and Respiratory Muscle Recruitment during Singing and Speech in Quadriplegia.” Archives of Physical Medicine and Rehabilitation 92(2): 250–56. http://dx.doi.org/10.1016/j.apmr.2010.10.032. Tankersley, Clarke G., Christine Haenggeli, and Jeffery D. Rothstein. 2007. “Respiratory Impairment in a

206

Mouse Model of Amyotrophic Lateral Sclerosis.” Journal of Applied Physiology 102(3): 926–32. Thaler, Joshua et al. 1999. “Active Suppression of Interneuron Programs within Developing Motor Neurons Revealed by Analysis of Homeodomain Factor HB9 the Concentration-Dependent Activity of Sonic Hedge- Hog (Shh) (Ericson et Al., 1997). Progenitor Cells within the Ventricular Zone I.” Neuron 23(4): 675–87. http://www.ncbi.nlm.nih.gov/pubmed/10482235. Timmers, Henri J.L.M., Wouter Wieling, John M. Karemaker, and Jacques W.M. Lenders. 2003. “Denervation of Carotid Baro- and Chemoreceptors in Humans.” Journal of Physiology 553(1): 3– 11. De Troyer, A., and A.M. Boriek. 2011. “Mechanics of the Respiratory Muscles.” Comparative Physiology 1(3). De Troyer, A., E. Brunko, D. Leduc, and Y. Jammes. 1999. “Reflex Inhibition of Canine Inspiratory Intercostals by Diaphragmatic Tension Receptors.” Physiology 514: 255–63. De Troyer, A., and A. Heilporn. 1980. “Respiratory Mechanics in Quadriplegia. The Respiratory Function of the Intercostal Muscles.” Am. Rev. Respir. Dis. 122: 591–600. De Troyer, A., P.A. Kirkwood, and T.A. Wilson. 2005. “Respiratory Action of the Intercostal Muscles.” Physiol Rev 85: 717–56. De Troyer, A.D. 1998. “The Canine Phrenic-to-Intercostal Reflex.” Physiology 3: 919–27. Turner, Martin R., and Matthew C. Kiernan. 2012. “Does Interneuronal Dysfunction Contribute to Neurodegeneration in Amyotrophic Lateral Sclerosis?” Amyotrophic Lateral Sclerosis 13(3): 245– 50. Ueno, M. et al. 2018. “Corticospinal Circuits from the Sensory and Motor Cortices Differentially Regulate Skilled Movements through Distinct Spinal Interneurons.” Cell Reports 23(5): 1286–1300. Urban, Daniel J., and Bryan L. Roth. 2015. “DREADDs (Designer Receptors Exclusively Activated by Designer Drugs): Chemogenetic Tools with Therapeutic Utility.” Annual Review of Pharmacology and Toxicology 55(1): 399–417. Urban, Mark W. et al. 2019. “Long-Distance Axon Regeneration Promotes Recovery of Diaphragmatic Respiratory Function after Spinal Cord Injury.” Eneuro 6(5): ENEURO.0096-19.2019. Urmey, W. et al. 1985. “Upper and Lower Rib Cage Deformation during Breathing in Quadriplegics.” Journal of Applied Physiology 60(2): 618–22. Valdez, Gregorio et al. 2012. “Shared Resistance to Aging and Als in Neuromuscular Junctions of Specific Muscles.” PLoS ONE 7(4). Vann, Nikolas C., Francis D. Pham, Kaitlyn E. Dorst, and Christopher A. Del Negro. 2018. “Dbx1 Pre- Bötzinger Complex Interneurons Comprise the Core Inspiratory Oscillator for Breathing in Unanesthetized Adult Mice.” eNeuro 5(3). Vazquez, RG et al. 2013. “Respiratory Management in the Patient with Spinal Cord Injury.” BioMed Research International. Viala, D., L. Persegol, and R. Palisses. 1987. “Relationship between Phrenic and Hindlimb Extensor Activities during Fictive Locomotion.” Neuroscience Letters 74(1): 49–52. Viemari, Jean Charles, Alfredo J. Garcia, Atsushi Doi, and Jan Marino Ramirez. 2011. “Activation of Alpha- 2 Noradrenergic Receptors Is Critical for the Generation of Fictive Eupnea and Fictive Gasping Inspiratory Activities in Mammals in Vitro.” European Journal of Neuroscience 33(12): 2228–37. Vinit, S, MR Lovett-Barr, and GS Mitchell. 2009. “Intermittent Hypoxia Induces Functional Recovery Following Cervical Spinal Injury.” Respiratory physiology & neurobiology 169(2): 210–17. Waldrop, T.G., R.M. Bauer, and G.A. Iwamotor. 1988. “Microinjection of GABA Antagonists into the Posterior Hypothalamus Elicits Locomotor Activity and Cardiorespiratory Activation.” Brain Research 444: 84–94. Waldrop, T.G., F.L. Eldridge, G.A. Iwamoto, and J.H. Mitchell. 19996. “Central Neural Control of Respiration and Circulation during Exercise.” Handbook of Physiology 12: 333–80.

207

Ward, Michael E., Guilio Vanelli, Mandana Hashefi, and Sabah N.A. Hussain. 1992. “Ventilatory Effects of the Interaction between Phrenic and Limb Muscle Afferents.” Respiration Physiology 88(1–2): 63– 76. Warren, Philippa M. et al. 2018. “Rapid and Robust Restoration of Breathing Long after Spinal Cord Injury.” Nature Communications 9(1). http://dx.doi.org/10.1038/s41467-018-06937-0. Wen, Ming-Han, Ming-Jane Wu, Stéphane Vinit, and Kun-Ze Lee. 2019. “Modulation of Serotonin and Adenosine 2A Receptors on Intermittent Hypoxia-Induced Respiratory Recovery Following Mid- Cervical Contusion in the Rat.” Journal of Neurotrauma 3004: 2991–3004. Wong-Riley, Margaret T T, and Qiuli Liu. 2008. “Neurochemical and Physiological Correlates of a Critical Period of Respiratory Development in the Rat.” Respiratory Physiology and Neurobiology 164(1–2): 28–37. Wong, Philip C. et al. 1995. “An Adverse Property of a Familial ALS-Linked SOD1 Mutation Causes Motor Neuron Disease Characterized by Vacuolar Degeneration of Mitochondria.” Neuron 14(6): 1105– 16. Wu, Jinjin et al. 2017. “A V0 Core Neuronal Circuit for Inspiration.” Nature Communications 8(1). http://dx.doi.org/10.1038/s41467-017-00589-2. Xiangzhe, L et al. 2019. “Blocking of BDNF-TrkB Signaling Inhibits the Promotion Effect of Neurological Function Recovery after Treadmill Training in Rats with Spinal Cord Injury.” Spinal Cord 57(1). Yates, BJ, JA Smail, SD Stocker, and JP Card. 1999. “Transneuronal Tracing of Neural Pathways Controlling Activity of Diaphragm Motoneurons in the Ferret.” Neuroscience 90(4): 1501–13. Yazawa, Itaru. 2014. “Reciprocal Functional Interactions between the Brainstem and the Lower Spinal Cord.” Frontiers in Neuroscience 8(8 MAY): 1–12. Yu, J., and M. Younes. 1999. “Powerful Respiratory Stimulation by Thin Muscle Afferents.” Respiration Physiology 117(1): 1–12. Zhang, W et al. 2016. “Hyperactive Somatostatin Interneurons Contribute to Excitatoxicity in Neurodegenerative Disorders.” Nature Neuroscience 19(4): 557–59. Zholudeva, L.V. et al. 2018. “Transplantation of Neural Progenitors and V2a Interneurons after Spinal Cord Injury.” Journal of neurotrauma. Zholudeva, Lyandysha V., Jordyn S. Karliner, Kimberly J. Dougherty, and Michael A. Lane. 2017. “Anatomical Recruitment of Spinal V2a Interneurons into Phrenic Motor Circuitry after High Cervical Spinal Cord Injury.” Journal of Neurotrauma 34(21): 3058–65. Zhong, Guisheng et al. 2010. “Electrophysiological Characterization of V2a Interneurons and Their Locomotor-Related Activity in the Neonatal Mouse Spinal Cord.” Journal of Neuroscience 30(1): 170–82. Zhong, Guisheng, Natalia A. Shevtsova, Ilya A. Rybak, and Ronald M. Harris-Warrick. 2012. “Neuronal Activity in the Isolated Mouse Spinal Cord during Spontaneous Deletions in Fictive Locomotion: Insights into Locomotor Central Pattern Generator Organization.” Journal of Physiology 590(19): 4735–59. Zimmer, M. Beth, Joshua S. Grant, Angelo E. Ayar, and Harry G. Goshgarian. 2015. “Ipsilateral Inspiratory Intercostal Muscle Activity after C2 Spinal Cord Hemisection in Rats.” Journal of Spinal Cord Medicine 38(2): 224–30. Zimmer, M. Beth, Kwaku Nantwi, and Harry G. Goshgarian. 2007. “Effect of Spinal Cord Injury on the Respiratory System: Basic Research and Current Clinical Treatment Options.” Journal of Spinal Cord Medicine 30(4): 319–30. Ziskind-Conhaim, L, and S Hochman. 2017. “Diversity of Molecularly Defined Spinal Interneurons Engaged in Mammalian Locomotor Pattern Generation.” Journal of Neurophysiology 118(6): 2956– 74.

208