Spinal cord injury (SCI) interrupts axonal connections between the brain and the spinal cord, and is characterized by a spectrum of sensorimotor and autonomic impairments. While spontaneous recovery is limited, recent studies have shown that functional improvements can be greatly augmented via rehabilitative approaches based on targeted epidural electrical stimulation that recapitulates the natural pattern of spinal activation. Nevertheless, these strategies depend on the presence of spared axonal connections, and are consequently constrained in the extent of elicitable recovery, and inapplicable in the case of very severe or anatomically complete SCI. It is therefore agreed that future therapies for SCI will require strategies to repair the injured spinal cord by stimulating severed axons to regenerate across the tissue lesions. Despite this need, axons from adult mammalian central nervous system (CNS) neurons are characterized by intrinsic incapacity of spontaneous regeneration. Research in the last several decades has uncovered multiple mechanisms underlying CNS regenerative failure, and a recent approach developed by our group has identified the requirements to induce experimental axon regrowth across anatomically complete SCI in rodents. Yet, while robust regeneration could be elicited with this and other strategies, restoring meaningful function after such injuries has been elusive. In the work presented in this thesis, we sought to build on our previous approach by identifying, and then providing, requirements that are missing for recovery. Concretely, we sought to determine whether restoring neurological function requires regeneration of specific subpopulations of neurons directed to their natural target region. To address these questions, we shifted our attention to a specific model of severe but incomplete SCI, following which natural reorganization of spinal circuits is associated with spontaneous recovery of walking. We performed projection-specific and comparative single-nucleus RNA sequencing to uncover the transcriptional phenotype and connectome of neuronal subpopulations involved in natural spinal cord repair, and identified a molecularly defined population of excitatory projection neurons in the thoracic spinal cord that extend axons to the lumbar spinal cord where walking execution centers reside. We optimized our previous strategy to provide sustained chemoattraction to the identified neuronal subpopulation. We showed that regrowing axons from these neurons across anatomically complete SCI and guiding them to their appropriate target region in the lumbar spinal cord restores walking in mice, whereas regeneration of axons simply across the lesion has no effect. Selective loss-of-function experiments further revealed that recovery is largely dependent on regeneration of the characterized neuronal subtype. These results demonstrate that mechanism-based repair strategies that recapitulate the natural topology of molecularly defined neuronal subpopulations can restore neurological functions following anatomically complete SCI.