The neural drive to breathe results from the integration of brainstem respiratory central pattern generators (rCPG) and the respiratory-related cortical networks (27–29). The rCPG is a hierarchical functional region; respiratory motions may result from integration among numerous brainstem nuclei (30–33). The main brainstem-associated respiratory nuclei, listed in descending order, include the periaqueductal gray (PAG) (34, 35), pontine respiratory group (PRG) (29, 36), retrotrapezoid nucleus (RTN)/parafacial respiratory group (pFRG) (37), pre-Bötzinger complex (pre-Bötc), Bötzinger complex (Bötc), ventral lateral medulla oblongata respiratory neuron group, and the nucleus tractus solitaries (NTS) (38). These nuclei harbor diverse neuron types and neurotransmitters, enabling them to perform distinct respiratory functions, as outlined in Table 1.

The respiratory rhythm originates from the ventral part of the medulla, specifically the pre-Bötc and Bötc. The pre-Bötc assumes the role of providing rhythmic excitation drive for respiration, while the Bötc engenders alternating inspiratory and expiratory patterns during normal breathing (44). The respiratory drives are conveyed to the ventral (VRG) and dorsal respiratory neuron groups (DRG) of the medulla oblongata (39). The rostral VRG is the largest group of inspiratory bulbospinal neurons and receives inspiratory drive from pre-Bötc neurons while being inhibited by expiratory Bötc neurons (39, 45). The rVRG transmits respiratory drive to phrenic motorneurons (PhMNs) via the bulbospinal respiratory neural pathway (46). The caudal VRG receives converging inputs, including those from the RTN/pFRG and BötC (39). The pons is proposed to interact with the medullary respiratory system (47). Respiratory adjustment primarily occurs in the midbrain, pons, and related cortical regions. The PAG mainly regulated the patterns of respiratory and contributed to arousal and control of sleep (34). Respiratory adjustments occur mainly in the PRG and the facial nucleus. The PRG, consisting of the Kölliker-Fuse nucleus (KF) and the parabrachial nuclei (PB), project upward to the amygdala and hypothalamus (48) and downward to the VRG through medullary raphe neurons (49). The RTN/pFRG serves as a conditioned oscillator of expiratory activity and serves as a significant generator of inspiratory rhythms (50, 51). Additionally, the lateral reticular nucleus also affects respiratory control and arousal (52).

Advanced respiratory centers connected to breathing include the cerebral cortex, thalamus, hypothalamus, hippocampus, extended amygdala, and limbic system (53–55). For details, please refer to this literature (29). Initial indications of a direct connection between the cortex and motor neurons were observed when the primary motor cortex was stimulated, inducing activation of the phrenic nerve in cats (56). Subsequently, the relationship between spontaneous breathing and the primary motor cortex was confirmed using transcranial electrical or magnetic cortical stimulation (57–60) and positron emission tomography (61). Furthermore, evidence suggests the existence of a rapid pathway from the supplementary motor area (SMA) to the diaphragm, with the SMA exerting a more pronounced excitatory effect on the diaphragm (62). The premotor area was thought to be involved in respiratory control during articulation (63). While our understanding of the brainstem respiratory center has advanced (39), the control mechanisms of the higher respiratory centers and their performance under varying conditions remain unclear.

The strongest evidence that potential respiratory neural pathways are activated in the weeks or months following high CSCI is the crossed phrenic phenomenon (CPP). CPP was first discovered by John Porter in 1895 through the C2 hemisection model (21). Since then, the neuroplasticity of diaphragm neuromotor control after CSCI has been studied. Cervical medullary hemisection interrupts the inspiratory drive of the ipsilateral PhMNs, paralyzes the ipsilateral diaphragm, and then achieves functional recovery of the ipsilateral diaphragm by severing the contralateral phrenic nerve to induce CPP (21, 95). The occurrence of CPP serves as a respiratory-related stressor leading to modulation of respiratory function, and the bulbospinal pathway contralateral to the injury is activated to cross over at the level of the phrenic nucleus within the spinal cord to innervate the ipsilateral PhMNs (4). Similarly, in rats with unilateral CSCI, the “crossed-spinal” pathways are also observed in the ipsilateral intercostal activity (77). Further studies have identified CPP as a state-dependent phenomenon that is induced by certain respiratory stressors, such as contralateral phrenic nerve dissection, hypercapnia, hypoxia, and asphyxia (96, 97). One study has identified CPP within a few days of injury (98), while other studies identify CPP within 2 weeks of C2 hemisection (99, 100). Beyond the C2 hemisection model, researchers have developed alternative models, including a cold block of the high cervical segment, C1 hemisection with the decerebrate brain, C2 hemisection with injury of the ventral lateral aspect, C4 hemisection, and contusion of the middle and high cervical segments, to replicate the plasticity of respiratory function observed in human SCI (Figure 3) (96). Recently, Vinit et al. demonstrated the presence of CPP by applying Transcranial Magnetic Stimulation (TMS) 1 h after the C2 injury, which resulted in evoked potentials in the diaphragm on the injured side (93).

The regenerative capacity of axons is influenced by the distance between the injury site and the cell body. In the case of CSCI, the corticospinal tract has a lower potential for axonal regeneration compared to the bulbospinal tract (45). Axon resection in the bulbospinal pathway, resulting from C2 lateral spinal cord injury in rats, leads to degenerative changes. However, it can also trigger regenerative processes such as lateral sprouting or medial axonal remobilization (101). The recruitment of Spinal Interneurons (SpINs) at the C1 level was found to be increased when interacting with the bulbospinal pathway (101). Additionally, other substances have been discovered to promote axonal regeneration related to respiration. Phosphatase and tensin homologs promote substantial long-distance regeneration of rVRG axons after hemisection (102–104), but have less ability to restore diaphragm function. Chondroitinase ABC (ChABC) inhibits the up-regulation of chondroitin sulfate proteoglycans around phrenic motor neurons after spinal cord injury (24). When combined with autologous nerve grafts, ChABC restores functionality in paralyzed diaphragms (24). Consequently, the activation of SpINs and the use of axonal growth-promoting drugs are potential therapeutic strategies. However, the extent of axonal regeneration necessary for functional recovery remains largely unknown in most cases (105).

Further research is required to investigate supraspinal neurons and SpINs following CSCI, including studies on temporal alterations, impacts on diaphragm movement, and performance in various injury models.

Following CSCI, patients may experience shallow and rapid breathing to maintain minute ventilation, but tidal volume tends to decrease under these circumstances (4, 100). Paralysis of the diaphragm due to SCI can result in defective breathing patterns such as paradoxical movements (106). During inspiration, the paralyzed diaphragm moves upward, while the abdomen moves inward as the accessory inspiratory muscles lift and the rib cage simultaneously expands, lowering intrathoracic pressure. This condition is more pronounced in the supine position.

Damage to phrenic neurons after SCI can induce the expression of proteases and atrophy-related genes, leading to immediate atrophy of all types of diaphragm fibers (5). Moreover, early use of a ventilator can accelerate the atrophy of the diaphragm (107), and diaphragm contractility can decrease by up to 40% within 8 weeks after injury (108). Following SCI, compensatory hypertrophy of the diaphragm occurs due to a reduction in the diaphragmatic contraction rate (109). However, individuals of Chinese tend to exhibit less diaphragmatic hypertrophy than individuals of European or American descent (110). Accessory muscles such as the deltoid and trapezius play a crucial role in respiratory function following CSCI (14). In addition, the abdominal muscles are paralyzed and other muscles such as the latissimus dorsi and pectoralis major are activated during coughing (111). During respiratory muscle strength tests, the rhomboid muscle is more involved during the maximum inspiratory pressure test and the pectoralis major and latissimus dorsi are more involved during the maximum expiratory pressure test, compared to non-injured individuals (112).

Pulmonary function often significantly improves within the initial 6-month period after injury and less so thereafter (113). Multiple factors, including the level, degree, and timing of injury, age, body position, and obesity, can influence changes in pulmonary function after CSCI (114, 115). Early improvements in respiratory function after SCI may be due to early edema subsidence, compensatory respiratory strategies, and altered respiratory muscle biomechanics. In contrast, for patients with chronic CSCI, age, persistent wheezing, and obesity are important factors affecting lung function. However, the degree and level of injury may not be one of the factors affecting pulmonary function in chronic spinal cord injury (116). A long-term follow-up found that age (>30 years) and BMI (>30 kg/m²) as important factors affecting lung function after CSCI (117). A retrospective study of 339 patients with CSCI revealed that smoking, persistent wheezing, obesity, and MIP, as well as SCI levels and integrity, were significant determinants of lung function (118).

Although there are different criteria for evaluating sleep breathing disorders, the incidence of sleep apnea in patients with SCI was higher than in the healthy population, and the incidence of patients with tetraplegia was higher than in the paraplegic population (119, 120). The exact mechanism underlying sleep apnea after CSCI remains unclear, although some researchers have suggested associations with upper airway muscle inactivation, autonomic dysregulation, reduced lung volume, and modifications in brainstem plasticity or chemosensitivity (11, 121).

Structural brain volume data can be obtained from T1-weighted images in MRI, and mathematical algorithms can be used to extract relevant brain features and perform statistical analysis of brain volume, morphology, and surface area (126). The reorganization of gray matter is caused by cell atrophy or apoptosis after axon transection, while the changes in white matter are caused by axon demyelination and deformation (127).

Following acute injury, the anatomy and organization of the spinal cord undergo significant changes. Several studies have shown that there is a reduction in the area and width of the cervical medulla, changes in the gray matter of the primary cortex and limbic system (50, 128), and a reduction in the dorsal pyramidal tract of the medulla oblongata and the white matter of the cerebellar peduncle (129) in the acute phase. These changes may result in clinical symptoms such as impairment of motor and sensory function.

In the subacute to chronic phase, atrophy and microstructural changes in the spinal cord proceed further. Progressive atrophy and microstructural changes occur in the thalamus, anterior cingulate gyrus, insula, pons, and secondary sensory cortex (130).

In the chronic phase after spinal cord injury, atrophy of the corticospinal tract and medial thalamic tract occurs, with reduced gray matter in the brainstem around the midbrain aqueduct, dorsal pons, and dorsal medulla oblongata myelin (131). A significant reduction in dorsal anterior spinal cord volume at the spinal medullary junction, is consistent with clinical histologic evidence (132). This may reflect the Wallerian degeneration of the associated axons after spinal cord injury. Extensive atrophy and microstructural changes are also observed in the cerebral tracts and sensory-motor cortical areas of the spinal cord at 2, 6, and 12 months after acute spinal cord injury (133). Furthermore, cortical gray and white matter volumes were reduced at the level of the medullary aqueducts compared to healthy controls at 12 months post-injury.

These findings suggest that structural changes in the spinal cord following acute spinal cord injury may be associated with respiratory dysfunction, particularly alterations in brainstem microstructure. However, further research is required to gain deeper insights into the mechanisms underlying these changes and their role in respiratory dysfunction.

Diffusion tensor imaging (DTI) is an imaging technique capable of quantifying atrophy, demyelination, and iron deposition within the spinal cord and cerebral cortex (134). It is widely used to assess spinal cord, brainstem, and brain alterations after CSCI (135–137). The diffusion of water molecules in neural tissues is mainly limited by cell membranes and myelin sheaths. When neurodegenerative lesions occur, the tissue produces more free water, leading to re-diffusion perpendicular to the white matter (124). Parameters commonly used in DTI include fractional anisotropy (FA), mean diffusivity (MD), and apparent diffusion coefficient (ADC). The definitions of these parameters can be found in more detail in the literature (124).

In the acute phase, a significant decrease in FA and an increase in MD at the level of injury can be observed after SCI, but changes in ADC are controversial (124, 138). Similarly, a decrease in FA values at the lesion level and in the upper cervical spinal cord can also be observed in the chronic phase (136, 139–142). A long-term follow-up study revealed worrisome results: gray matter below the C2/3 level decreased by 0.7% per month and white matter by 0.34% per month (143). At the level of the cerebral peduncle, individuals with cervical ASIA A/B SCI showed greater degrees of axonal damage and edema/tissue loss, but no statistical differences were found in whole-brain white matter compared to healthy individuals (144). Of concern is that spinal cord injuries also cause changes in the white matter of the brain. A DTI study found that chronic SCI has a wide range of neurodegenerative effects in the brain that are not limited to motor pathways (145). Meanwhile, patients with CSCI in the chronic phase continue to experience a slow decline in degenerative FA in the midbrain, pontine, and superior white matter of the medulla, while MD keeps increasing, probably because of cumulative cell membrane loss caused by delayed damage to glial cells or axons (146).

In addition, DTI can also be used to assess treatment efficacy. Gu et al. used DTI and immunohistochemistry to assess the efficacy of axonal regeneration, and DTI reflects the center of injury as well as the immediate condition of neural damage and the process of axonal regeneration (147). Zhang et al. used DTI to explore the efficacy of vocal therapy on respiratory function in patients with CSCI and found that neural networks related to respiration in the medulla and cortex became more active (148). To some extent, this reflects the treatment-facilitated recovery process in the brainstem and brain-related respiratory nerves. The combined application of DTI and fMRI enables central pathway monitoring and treatment evaluation (149).

In summary, scientific studies have provided valuable insights into the altered state of the spinal cord and brain following acute and chronic spinal cord injury. However, the understanding of these changes remains somewhat controversial, and further research is needed to explore the mechanisms and clinical implications in depth.

The aforementioned techniques are primarily utilized to evaluate structural changes. It is equally crucial to assess the functional reorganization of the brain and spinal cord. While fMRI can detect alterations in blood oxygen level-dependent (BOLD) signals, the pattern of these changes is expected to be utilized in assessing the ability to reorganize function (125, 150).

Acute SCI patients have decreased functional connections between the bilateral primary sensory cortex and motor cortex, while functional connections increase between primary sensorimotor cortex, premotor cortex, supplementary motor area, thalamus, and cerebellum (151). Patients with advanced SCI may have changes in cognitive-related areas, such as increased functional connections between the dorsal anterior cingulate cortex and the motor cortex (152). Various studies have found that the classic somatosensory pathway degenerates (153) and reorganizes (154, 155) after SCI, meaning that the spinal thalamic tract or dorsal tract transmits sensory signals to the thalamus, which is transformed and projected to the primary sensory cortex. Others have found that sensory input in patients with incomplete CSCI may be mediated by alternative pathways, which may consist of the ipsilateral cerebellum, pons, and contralateral posterior central gyrus (127). In patients with incomplete spinal cord injury, the visual cortex supplements sensory and motor inputs (156). Many studies have found that the cerebellum is important for activation after SCI (157). The functional connections between the cerebellum and primary motor cortex, primary sensory cortex, and primary auditory cortex can be found in patients with complete injury (151, 152, 158). In patients with chronic CSCI, whole-brain network connectivity is reduced, and increased connectivity is observed in the subnetworks of the sensory-motor cortex and cerebellum (159). Although the cerebellum is not a region of respiration in the traditional sense, we speculate that the cerebellum may play an important role in respiratory sensory and motor function after SCI.

Respiration-related fMRI investigations primarily focus on the influence of subcortical structures. Studies have shown that respiratory control is connected to brainstem and subcortical activity, with many indicating the maintenance of connectivity between cortical and brainstem regions (53, 54, 160). Pattinson et al. identified regions of the brainstem and thalamus in healthy subjects that respond to CO₂ stimulation by fMRI and diffusion fiber tracing. They explored a link between the thalamus and higher cortical tracts, suggesting that the thalamus can play a role in respiratory control, but the relationship between structure and function was not explored in this study (161). The brainstem could be studied by BOLD fMRI to observe characteristics of the respiratory center at rest and its corresponding mechanisms (162). There are fewer fMRI studies on respiration in patients with SCI, likely due to the challenging nature of conducting respiratory center fMRI owing to the unique anatomy and small size of the brainstem. Relevant atlases are lacking, and physiological noise from cerebrospinal fluid and arteries obscures measurements near the brainstem (150, 163, 164). Nonetheless, researchers have explored several methods for brainstem fMRI data processing (55, 164, 165).