Regenerative medicine for neurological diseases—will regenerative neurosurgery deliver?BMJ 2021; 373 doi: https://doi.org/10.1136/bmj.n955 (Published 23 June 2021) Cite this as: BMJ 2021;373:n955
- Terry C Burns, associate professor of neuroscience and neurosurgery1,
- Alfredo Quinones-Hinojosa, professor of neurosurgery2
- 1Department of Neurologic Surgery, Mayo Clinic, Rochester, MN, USA
- 2Neurosurgery Head and Neck Cancer Center, Mayo Clinic, Jacksonville, FL, USA
- Correspondence to Terry C Burns,
Regenerative medicine aspires to transform the future practice of medicine by providing curative, rather than palliative, treatments. Healing the central nervous system (CNS) remains among regenerative medicine’s most highly prized but formidable challenges. “Regenerative neurosurgery” provides access to the CNS or its surrounding structures to preserve or restore neurological function. Pioneering efforts over the past three decades have introduced cells, neurotrophins, and genes with putative regenerative capacity into the CNS to combat neurodegenerative, ischemic, and traumatic diseases. In this review we critically evaluate the rationale, paradigms, and translational progress of regenerative neurosurgery, harnessing access to the CNS to protect, rejuvenate, or replace cell types otherwise irreversibly compromised by neurological disease. We discuss the evidence surrounding fetal, somatic, and pluripotent stem cell derived implants to replace endogenous neuronal and glial cell types and provide trophic support. Neurotrophin based strategies via infusions and gene therapy highlight the motivation to preserve neuronal circuits, the complex fidelity of which cannot be readily recreated. We specifically highlight ongoing translational efforts in Parkinson’s disease, amyotrophic lateral sclerosis, stroke, and spinal cord injury, using these to illustrate the principles, challenges, and opportunities of regenerative neurosurgery. Risks of associated procedures and novel neurosurgical trials are discussed, together with the ethical challenges they pose. After decades of efforts to develop and refine necessary tools and methodologies, regenerative neurosurgery is well positioned to advance treatments for refractory neurological diseases. Strategic multidisciplinary efforts will be critical to harness complementary technologies and maximize mechanistic feedback, accelerating iterative progress toward cures for neurological diseases.
Stroke, Alzheimer’s disease, and traumatic injuries of the central nervous system (CNS) are among the most feared human maladies. Their capacity to detach individuals from their most basic functions of recall, communication, and movement, while erasing cognitive functions and distorting personality, all strike at the heart of human dignity.
The mammalian CNS is notoriously refractory to repair. Its intricate network has more than 100 billion neurons interconnected by over a quadrillion finely tuned synapses, which reside in interdependent synergy with astrocytes, oligodendrocytes, microglia, endothelial cells, and pericytes, creating a formidable task for regenerative science. To date, many of the underlying mechanisms that catalyze and propagate human CNS disease remain poorly understood.
Ramón y Cajal, widely considered the father of modern neuroscience, made a sobering observation in his 1928 text Degeneration and Regeneration of the Nervous System: “In adult centers, the nerve paths are something fixed, ended, immutable. Everything may die, nothing may be regenerated. It is for the science of the future to change, if possible, this harsh decree.”1
Decades passed before the first evidence emerged that new neurons could be added to the postnatal mammalian brain.2 Multiple progenitor cell populations are now known to reside in the adult human brain.34 Injury to the mammalian brain can trigger a tightly orchestrated, dynamic response that involves many cell types, including microglia, astrocytes, and endogenous neural progenitors.56 Genetic tools to inhibit innate responses may exacerbate the resultant damage.7 Experimental methods to promote neuro-regeneration have variously sought to reawaken mechanisms of CNS development, block adult inhibitory signals, add new cells through cell transplantation, or promote adaptive plasticity within remaining circuitry. Over the past 25 years, advances have sequentially unveiled embryonic stem cells, adult human neural stem cells (NSCs), and induced pluripotent stem cells (iPSCs), providing a versatile toolkit for repair of the diseased CNS.1
Emerging translational successes in other medical fields have added merit to bold projections that regenerative technologies will usher in a new era of clinical medicine, deploying cures in place of mere “treatments.” While neuroscience has yet to unveil a regenerative “standard of care” for any disease, a growing abundance of enabling science, technology, and industry investment are transitioning conversations from “if” to “when” and “for which diseases.” This review offers a synopsis of progress to date and confronts challenges ahead.
Definitions and scope
Broadly defined, regenerative neurosurgery provides access to the CNS or its supporting structures to preserve or restore neurological function. Indeed, established neurosurgical procedures may divert cerebrospinal fluid (CSF) for hydrocephalus, remove a tumor, repair an aneurysm, stabilize a spine, or modulate basal ganglia circuitry for Parkinson’s disease.
Within this broader definition, this review considers procedures penetrating CNS tissue with the specific goal to restore, replace, or regenerate CNS cell types, including those aptly described by Cajal as “fixed, immutable” almost a century ago. While cerebrovascular, intrathecal, intranasal, intra-ocular, and systemic routes are each under active evaluation for delivery of cell therapies, our focus on intraparenchymal procedures considers the neurosurgical epicenter of broader efforts. Similarly, complementary and emerging capabilities within the field of neuromodulation,8 including direct electrical stimulation,9 closed loop responsive stimulation,10 and optogenetic based paradigms11 to promote synaptic plasticity are synergistic, but beyond the scope of this review.
Why neurosurgery? The fragile brain and spinal cord are shielded physically by the skull and spinal column anatomy and chemically by the blood-brain barrier (BBB).12 As such, the surgical corridor provides physical access to deliver cells, sample pathology, and dynamically interact with otherwise inaccessible biology. While trophic cell types can migrate into diseased tissue (fig 1), higher local dose is achieved with direct implantation and cell replacement remains unachievable without parenchymal delivery. Such need for neurosurgical procedures remains a barrier to translation given associated risk, discomfort, and expense. Many alternatives are being pursued, including systemic, cerebrovascular intra-arterial, transnasal, intraorbital, and intrathecal delivery of biologics, as well as CNS homing and BBB disruptive technologies. Neurosurgical stereotactic targeting remains the most definitive and accurate avenue for targeted therapeutic delivery, and can be deployed to deliver candidate cellular therapies, neurotrophic factors, gene therapies, or combinations thereof, such as delivery of genetically engineered cellular therapies. The following sections explore each of these potentially potent resources for CNS regeneration.
Sources and selection criteria
We searched online databases including Pubmed, Google Scholar, and Clinicaltrials.gov from their earliest available records until 4 November 2019. Keywords included “cell therapy”, “stem cells”, “regenerative”, “spinal cord”, “brain”, as well as disease entities: “stroke”, “Parkinson’s”, “ALS”, “spinal cord injury”, “neurodegeneration”, and therapeutic avenues: “fetal”, “mesenchymal pluripotent”, “oligodendrocyte”, “progenitor”, “gene therapy,” and delivery: “infusion”, “implant”, “transplant”, “stereotactic”, and “convection enhanced delivery”. Given the diversity of indications and delivery methods and our topic focus on regenerative neurosurgery, inclusion criteria included use of a regenerative product and delivery via penetration of CNS tissue in human subjects. Studies not published in English were excluded. Additional studies cited by these publications were considered when pertinent. Preclinical studies were considered based upon rigor of methodology and reporting,210 evidence of use as the basis for clinical trials, and pertinence to illustrate key translational principles or provide important context. Two relevant studies published during peer review were added during revision.117158
The use of cellular therapies to restore neurological function remains a widely coveted and eagerly anticipated promise of neuroregenerative surgery. Neurosurgery offers an avenue for directed cell implantation into the sites of disease within the brain and spinal cord. Given the “immutable” nature of the human brain, scientists and clinicians have routinely turned to the developmental source of the cells lost by disease and injury (box 1), as epitomized by fetal dopamine cell implants pioneered in the 1980s and 90s.13 The rise of stem cell technologies over the past 25 years has energized clinical hopes of regenerative cell therapies. In this article we focus on therapeutic strategies for the CNS requiring neurosurgical entry into the substance of the CNS (fig 1).
Fetal neurons: ethical and practical limitations
Neurons from pluripotent stem cells
human embryonic stem cells (ESCs): ethical concerns and risk of immunosuppression
induced pluripotent stem cells (iPSCs): from adult human somatic cells
somatic nuclear cell transfer (SCNT): closer resemblance to ESCs
Neural stem cells
Restricted in vivo to subventricular zone (SVZ) and subgranular zone (SGZ)
Can migrate to pathology
Glial progenitor cells
Long term engraftment potential
Remarkable migratory capacity
Potential capacity to outcompete diseased cell types
Mesenchymal stem cells
Accessible from adipose tissue
Innate tropism and trophic effects
Potential efficacy with systemic or intrathecal delivery
Neuronal cell replacement
Certain neurodegenerative diseases are characterized in part by symptomatic loss of specific neuronal subpopulations, including dopaminergic neurons from the substantia nigra in Parkinson’s disease,14 medium spiny GABAergic neurons from the striatum in Huntington’s disease,15 and motor neurons from the spinal cord and motor cortex in amyotrophic lateral sclerosis (ALS).16 The full manifestation of each disease is more complex than select neuronal loss; however, targeted cell replacement has been explored to ameliorate major functional deficits associated with disease progression. In addition to neuronal or neuroendocrine17 cell “replacement,” supplementing neuronal circuitry with additional GABAergic interneurons could attenuate hyperactive circuits in preclinical models of epilepsy.18
Diseases characterized by more widespread loss of multiple cell types, including stroke,19 spinal cord injury,20 and Alzheimer’s disease21 have not escaped consideration for neuronal replacement or supplementation. However, the formidable challenges of recreating complex neuronal circuitry encourage efforts to preserve at-risk neurons and their synaptic connections.
Signaling molecules in the adult brain inhibit widespread sprouting of new neuronal projections from existing neurons. However, if immature neurons can sufficiently resist inhibition,22 remnant developmental cues can guide projections through structurally preserved tissue to appropriate adult targets.23 Nevertheless, long distance projections are typically challenging to establish in the adult brain, even using juvenile neurons.24 This reality has made neuronal cell therapies most helpful when replacement neurons restore neurotransmitter balance, rather than recreate complex circuitry, as exemplified by fetal dopaminergic neurons for Parkinson’s disease. Implanted neurons generate functional synapses, express appropriate autoregulatory receptors, and help maintain appropriate synaptically active levels of dopamine.
To date, neuronal cell therapies have typically harnessed cells developmentally destined to generate the desired neuronal subtypes. These cells may be procured from the developing brain of fetal donors or generated from stem cells in vitro by mimicking developmental signals or genetic induction:
• Fetal neurons—Tissue from carefully microdissected human fetal brain enabled pioneering neuronal replacement studies. Multiple fetal donors are typically needed to provide sufficient cells for a single patient, therefore practical as well as ethical challenges preclude fetal cells from providing a long term solution. Robust protocols have emerged to generate large numbers of functional post-mitotic regionally specified neurons from pluripotent stem cells.25
• Neurons from pluripotent stem cells—Pluripotent stem cells can be expanded extensively and differentiated into essentially any cell type. As such, pluripotent stem cells provide an almost unlimited supply of cells for scientific and regenerative purposes. Human embryonic stem cells (ESCs) are derived from the inner cell mass of the developing blastocyst. Proponents argue that generating ESCs is an attractive alternative to discarding unwanted surplus embryos from IVF clinics. Moreover, methods are now available to isolate ESCs without destroying the originating embryo.26 Nevertheless, pre-existing ethical concerns regarding the origins of ESCs, as well as the desire for autologous therapies to avert immunosuppression, have each motivated the search for alternate sources of pluripotent cells.
Epigenetic reprogramming technology has enabled the generation of iPSCs from adult human somatic cells.2728 Before the discovery of iPSCs, alternative candidate sources of autologous pluripotent stem cells were parthenogenesis or somatic cell nuclear transfer (SCNT). Parthenogenesis involves activation of an oocyte without fertilization to generate diploid progeny homozygous for all nuclear genes.29 SCNT, first exemplified by Dolly the sheep, is a technically challenging procedure wherein the nucleus of an oocyte is replaced with the nucleus of a somatic (non-germ) cell. The donor nuclear DNA is epigenetically reprogrammed by the oocyte, yielding a pluripotent cell composed of donor nuclear DNA and mitochondrial DNA of the oocyte host. Advantages of SCNT include potential to correct mitochondrial defects while maintaining autologous nuclear DNA. SCNT also erases previous epigenetic memory more completely than iPSC reprograming, yielding an autologous pluripotent stem cell that most closely mirrors that of ESCs.30
Direct reprograming can generate neurons from other cell types without passing through a pluripotent stem cell state. Resultant neurons maintain the epigenetic age of the originating cell and show promise for in vitro studies of neurodegenerative processes and individualized drug screening.31 Deployed within the diseased rodent brain, reprogramming has generated new neurons from glia in situ.32
Neural stem cell (NSC) therapies
The initial isolation of NSCs from rodent and human brains revolutionized thinking regarding the reparative potential of the CNS.33 It has now been shown that new neurons are born throughout life in the human hippocampal dentate gyrus,3435 contributing cellular and synaptic substrate for learning and memory. A second population of NSCs exists in the subventricular zone (SVZ) of rodents where they generate thousands of new neurons in the olfactory bulb and contribute functionally to olfactory discrimination (fig 2). Relatively limited numbers of NSCs exist in the human subventricular wall; studies to determine if such cells give rise to new neurons in the human olfactory bulb have yielded conflicting results.3637 However, in contrast with rodent counterparts, evidence from carbon dating of neuronal nuclei suggests human SVZ derived neurons may populate the adjacent striatum.38 NSCs have understandably been among the most widely considered candidates for translational cell therapies across several neurological diseases.3 In addition to the homeostatic generation of new neurons in the select locations, NSCs from the SVZ can be mobilized to sites of injury.35 A functional role for these cells is supported by studies showing that absence of endogenous NSCs worsens outcomes in animal brain injury models,7 while increased endogenous NSC mobilization improves outcomes.3940 Limited reports have described cells expressing markers of immature neurons associated with vasculature in the peri-infarct region of human patients.41 However, rigorous studies have failed to find any notable evidence of newly born neurons at short or late time points after human stroke.42 Perhaps in light of their forebrain specification, NSCs are also challenging to coax into midbrain, hindbrain, and spinal cord fates without genetic modifications, even in vitro.43 Nevertheless, the myriad trophic functions attributed to NSCs provide several candidate mechanisms through which implanted or mobilized NSCs and their progeny could improve outcomes in response to injury.
Glial cell transplantation
In contrast with NSCs that are largely restricted to the SVZ and subgranular zone (SGZ), glial progenitor cells (GPCs) persist throughout the white matter, where they can give rise to new astrocytes and oligodendrocytes (collectively termed “glia”) throughout life. Contrary to the Latin origin for glia, which literally means “glue,” implying a structural role, glia play indispensable roles in neuronal performance and synaptic regulation. Oligodendrocytes insulate and metabolically support neurons. When myelin is lost, for example, in multiple sclerosis, neuronal function declines. If myelination is not restored either via generation of new GPC derived oligodendrocytes or remodeling of remaining oligodendrocytes, the denuded neurons ultimately die.44 Early clinical trials have already shown the safety and feasibility of their use for childhood leukodystrophies4546 and spinal cord injury.4748 Astrocytes regulate neuronal activity through gap junctions, modulate neurotransmitter levels, limited direct synaptic communication,49 and robust regulation of synaptic plasticity.50 Multiple neurodegenerative diseases—including ALS515253, Huntington’s disease,5455 Parkinson’s disease,56 and Alzheimer’s disease5758—are increasingly attributed to astrocytic dysfunction.
Accordingly, glial cells are emerging as promising candidates for CNS cell replacement. Unlike neurons, whose many projections and synapses can be exceptionally difficult to replace with high fidelity, glia can be removed and replaced almost at will without evident ill effects. In this regard, GPCs echo the regenerative potential of cell replacement for the hematopoietic system.59 Like hematopoietic stem cells, GPCs can respond dynamically to their micro-environment, filling vacant niches and differentiating into appropriate progeny in response to endogenous cues.60 By harnessing these capabilities, GPCs may help restore endogenous reparative capacity61 or rescue otherwise fatal disease.62 Indications for glial cell replacement could extend to virtually any disease that involves white matter or affects astrocyte function. Examples may span replacement of defective astrocytes in Huntington’s disease,63 rejuvenation of oligodendrocytes and GPCs damaged by cancer therapies,64 and repair of white matter lost owing to leukodystropies,454665 autoimmune demyelination,66 chronic hypoxia,67 or aging68 (fig 3).
Another advantage of GPCs is their migratory capacity. During development, GPCs derive from the medial ganglionic eminence and migrate widely throughout the cerebral hemispheres. Stem cell derived GPCs embody this phenotype and can migrate all the way from the brain to the caudal end of the spinal cord in preclinical models. Given their versatility, migratory capacity, long term engraftment potential, and potential capacity to outcompete diseased cell types, GPC therapies hold promise.
Mesenchymal stem cell (MSC) therapies
Certain observed supportive or “trophic” activities of NSCs, as discussed above, may also be innate to stem cell populations derived from outside the CNS, including bone marrow mononuclear cells (BMMCs) and co-isolated expandable progenitors termed mesenchymal stem cells or marrow stromal cells. MSCs have generally shown greater therapeutic value,69 can be genetically modified, and can be prepared from individual patients from multiple sites including adipose tissue. Like NSCs, MSCs exhibit both innate “tropism” (ie, attraction toward pathology) and “trophic” (ie, supportive therapeutic) properties. Commonly cited therapeutic mechanisms include immunomodulation, metabolic support, and release of therapeutic cytokines—including those that promote vasculature and impede fibrosis.70 When autologous MSCs were used for patients with ALS, cells from patients with best response had elevated levels of vascular endothelial growth factor (VEGF), angiogenin, and transforming growth factor beta (TGF-β), with cells from such patients also yielding greater effects in SOD1 mice.71 As a result, it may be possible either to screen and identify optimal therapeutic MSCs or to engineer MSCs with the desired expression levels. Additionally, the secretome of MSCs could be dynamically regulated in response to local pathology. However, multiple studies have shown MSCs to provide therapeutic benefit, even if most cells never reach the CNS (distributing instead to other sites, including the lungs),72 perhaps through secreted or induced factors.
Extracellular vesicles have gained prominence as likely candidates for several trophic effects attributed to certain stem cells, including MSCs,73 with efficacy even after lyophilized storage.74 Additionally, MSC extracellular vesicles may help recruit endogenous stem cells after localized delivery.75 Multiple clinical studies have shown the safety and feasibility of MSC based therapies for ALS, traumatic brain injury, and stroke, though efficacy has yet to be established in randomized clinical trials. Increasing scrutiny surrounds the rigor of data generated using MSCs. Indeed, a recent study of MSCs for cardiac injury revealed that dead MSCs and a chemical lesion induced innate immune mediated therapeutic efficacy similar to live MSCs.76 Given many uncertainties surrounding mechanisms of trophic cell therapies, rigorous controls are essential for critical evaluation and progress. Similarly, reports claiming in vivo trans-differentiation of bone marrow derived cells into neurons were subsequently found to be explainable by cell fusion or transfer of thymidine analog labels from cells that died after implantation to dividing host cells.77 Understanding mechanisms of apparent MSC trophic functions of MSCs may reveal novel opportunities to augment therapy.
Delivering neurotrophic factors to preserve neuronal circuitry
Neurotrophins include a family of signaling molecules required for the development and survival of certain neuronal populations. Glial cell derived neurotrophic factor (GDNF) is one such neurotrophin available in limited quantities during the development of substantia nigra dopaminergic neurons, which continues to be expressed in adults at lower levels.78 Intrastriatal administration of recombinant human GDNF in monkeys treated with MPTP doubled striatal dopamine levels and improved bradykinesia, rigidity, and postural instability.79
Many open label clinical studies of continuous intraputamenal GDNF infusion showed safety, improvement in off-medication sub-scores for Unified Parkinson’s Disease Rating Scale (UPDRS), decreased on-medication dyskinesias, and increased 18F-DOPA uptake in the putamen.808182 An RCT of 41 patients with moderate Parkinson’s disease compared 10 infusions of GDNF into the bilateral putamen via a subcutaneous cranial port with a placebo infusion over 40 weeks. It found a 25-100% increase in 18F-DOPA uptake compared with vehicle placebo infused patients (P<0.0001). However, although the primary functional outcome measure of OFF state UPDRS motor score improved in both groups, no significant difference was seen between them (-17.3+/-17.6% in treatment v -11.8+/-15.8% in placebo; P=0.41).83 In an open label continuation study, both the original placebo (n=20) and drug treated patients received GDNF (n=21), and further cumulative improvements from baseline were seen in both groups (-26.7+/-20.7% GDNF/GDNF v 27.6 +/-23.6% placebo/GDNF; P=0.96), prompting investigators to suggest an 80 week placebo controlled design may be needed to evaluate the functional impact of GDNF infusion. Indeed, preplanned comparison of 80 week GDNF/GDNF outcomes with 40 week placebo outcomes (before GDNF infusion) revealed greater absolute reduction in off-state motor scores (-9.6 +/-6.7 v -3.8+/-4.2 points P=0.0108) and activity of daily living scores (-6.9 +/-5.5 vs -1.0 +/-3.7, P=0.0003). No safety concerns were identified.84
Despite suggestion of benefit with GDNF, more is not necessarily better. A single high concentrations infusion of GDNF induced local neuronal apoptosis in rats.85 Continuous daily infusion of high dose GDNF in monkeys yielded multifocal Purkinje cell loss in some animals, with decreased food intake and weight loss.86 By contrast, once a month infusion was well tolerated without cerebellar or other toxicities.87 Neurturin (NTN)88 and cerebral dopamine neurotrophic factor (CDNF)89 are also implicated in dopaminergic neuron protection. An ongoing clinical trial in Sweden and Finland (NCT03775538) is testing the safety of two concentrations of CDNF versus placebo with monthly infusions for six months.
Gene therapy is an attractive therapeutic option to treat neurological disorders affecting the CNS, prompting an expanding array of preclinical and clinical studies. Gene therapy can involve stable or inducible therapeutic gene expression and may be targeted to specific cell types.90 Viral vectors can convert cells into therapeutic factories. Adeno-associated viruses (AAV) can transduce both dividing and non-dividing cells and have an excellent safety track record in recent clinical trials for multiple neurological conditions, including Parkinson’s disease.91
Early phase studies used AAV2 to deliver aromatic acid decarboxylase, the rate limiting enzyme for conversion of L-dopa to dopamine. An open label safety feasibility study showed improved PET tracer uptake (average 30% and 75%, in low and high dose groups; P=0.009; P=0.004, respectively),92 and decreased total (-31%) and motor (-36%) off-medication UPDRS at 6 months. A separate concurrent phase I trial of six patients yielded similar results with persistence of the improved PET tracer uptake for up to 96 weeks.93 Multiple phase II trials are ongoing. By contrast, early clinical evaluation of AAV2 mediated delivery of NTN failed to increase 18F-DOPA uptake94 despite subsequent postmortem evidence of ongoing expression of NTN. One long term follow-up of 53 patients enrolled across four phase I or II studies with up to five years’ follow-up suggested general stability of disease; with stability or modest improvements compared with baseline.95 Collectively, these data show the feasibility of AAV for CNS gene delivery. Clinical trials are ongoing to evaluate AAV6 and AAV9 for treatment of other neurological disorders, including lysosomal storage disorders and spinocerebellar ataxia.96
Combination cell and gene therapy: ex vivo gene therapy
Ex vivo gene therapies implant cells genetically engineered to secrete therapeutic compounds or augment the functional capabilities of implanted cells. Enhanced functions may affect differentiation, resilience, expression of trackable or conditional transgenes, immune resistance, or inducible suicide constructs to augment safety.97 As a flexible platform, ex vivo gene therapy allows rigorous quality control of genetically engineered cells prior to patient treatment, minimizing opportunities for uncontrolled viral integration into host cells. Modifying well established cell therapies can accelerate translation of novel gene therapy constructs. The longevity and distribution of therapy can be tailored via choice of cellular vehicle. For example, the injury tropic, yet transient nature of MSCs can deliver decreasing levels of a therapy throughout a lesioned area after implantation—kinetics that may be more cumbersome to achieve with direct infusion or viral delivery.
Multiple progenitor cell types tend to migrate toward sites of injury or malignancy,98 facilitating delivery of compounds that are unstable, systemically toxic, or unable to penetrate the CNS.99 One such pilot clinical study used cytosine deaminase-expressing NSCs to convert the CNS penetrant prodrug 5-fluorocytosine to 5-fluorouracil.100
The ethical bar of safety for translation of neuroregenerative therapies
A dynamic tension can exist between attempts to optimize therapies in preclinical models and the necessity for human clinical feedback to identify required improvements. While clinical efficacy has been notoriously challenging to predict in preclinical models, demonstrating an acceptable level of preclinical safety must be balanced with the need to harness the human model to iteratively optimize novel therapies. Increasing potential risk, implanted cells are expected to persist for the duration of the patient’s life so cannot be readily “discontinued” like a poorly tolerated medical therapy. Indeed, significant dyskinesias were noted after the NIH funded fetal dopaminergic cell trials for Parkinson’s disease.101
Procedural risks and surgical sham controls
Risks of neurosurgical procedures can include wound infection, seizure, and hemorrhage.92102 Cumulative risks of around 1-2% per procedure create ethical challenges for sham controlled study designs seeking to mitigate against the potentially powerful placebo effect of surgery.103 While neurological risks of CNS penetration can be averted by sham procedures restricted to skin and bony work, CNS penetration itself can have impacts important to distinguish from the therapeutic product. CNS injury by even an empty needle may promote release of brain derived neurotrophic factor (BDNF) and GDNF by injury activated monocytes and neuronal sprouting.104
Low versus high dose cell injections could help infer efficacy of the viable cellular product. A dose-response study of human teratocarcinoma derived neurons in a rat stroke model yielded dose dependent behavioral benefits, and suggested a higher percentage of surviving cells with higher numbers of implanted cells,105 perhaps via paracrine effects. Conversely, larger grafts could compromise cell survival and distribution by limiting nutrient supply.106 Dead cell controls have been used for preclinical, but not clinical studies.76 If dead cells could provide a real biological benefit by activating plasticity or immune responses in human patients, such knowledge could empower iterative efforts to further augment newly hypothesized mechanisms of action.76 One study, mentioned above, utilized a vehicle control for 40 weeks before offering the active GDNF infusion therapy to all patients. However, that significant benefit ultimately required 80 weeks in the first cohort, by which time the control group had already received GDNF, illustrating the challenge of designing clinical trials for novel therapies.84
Risk of tumorigenesis
Stem cells are defined in part by their capacity for self-renewal. Appropriately regulated self-renewal is critical to maintain an available pool of endogenous stem cells, but unchecked self-renewal is tumorigenic (fig 4). The most reliable way to avoid tumor formation is to implant a pure population of post-mitotic cells. This is not possible when replacing astrocytes or GPCs that maintain proliferative capacity throughout life. Expanded cells are at risk of accumulating mutations and cytogenetic abnormalities. Guidelines for assuring genetic integrity of stock cell lines have been published.107 Excluding pluripotent cells from differentiated cultures is essential to avoid teratoma formation.108 Cells isolated during development or neural lineage committed cells from pluripotent stem cells can generate larger grafts than desired.109 This could suggest autocrine signaling from densely engrafted cells may over-ride signals from the host microenvironment (fig 4B). Karyotypically normal somatic stem cells from adult stem cell niches have generally been regarded as safe. However, even these can demonstrate potential to “de-differentiate” into a germ cell-like phenotype that generates embryonic yolk sac tumors after focal,110111 but not systemic112 delivery. As such, cellular products subject to prolonged in vivo culture or expressing high levels of defined pluripotency markers (such as Oct4) should be rigorously scrutinized in immune deficient animals, using supratherapeutic cell numbers with prolonged survival assays prior to clinical trials.
Unregulated stem cell clinics around the world pose an ongoing threat to unsuspecting patients and the field alike. Lapses in quality control can replace hopes in stem cell therapies with dismay when tumors113 or hypertrophic inflammatory tumor-like lesions are identified.114
Immunosuppression carries well recognized risks, including infections, organ toxicity, and elevated risk of malignancy, adding complexity to clinical trials—particularly those with sham controls. Though long considered “immune privileged,” CNS antigen trafficking through the recently identified lymphatic system facilitates adaptive immune responses.115 Debris from grafted cells and temporary BBB disruption from the needle tract likely facilitate T cell mediated immune rejection. As such, at least a few months of immunosuppression are almost uniformly provided for trials transplanting fetal neurons into the CNS. Some protocols maintain lifelong immunosuppression, though viable dopaminergic neurons have been demonstrated up to 14 years after withdrawal of immunosuppression when cells were implanted as a cell suspension116 that may be less immunogenic than tissue chunks.
Autologous cell transplants genetically matched to the recipient do not require immunosuppression and remain an increasingly tangible goal of cell therapies.117 Autologous MSCs frequently deployed to elicit trophic benefits are relatively easy to generate and require no immunosuppression.85 Interestingly, even non-autologous MSC trials frequently also forgo immunosuppression, given active trophic effect of even transiently present cells.118
Traumatic or acute ischemic injuries may benefit from off-the-shelf allogenic or “universal donor” cells. Yamanaka’s group in Japan has generated banks of “universal donor,” wherein a limited number of iPSC lines can provide an immunologically compatible graft to a significant percentage of ethnically homogeneous populations.119
Alternatively, “immune edited” cells may be engineered to lack immune activating HLAs120 or express immune inhibitory antigens such as CD47.121 Innate safeguards in the event of neoplastic transformation may still be preserved by persistent expression of HLA-C.122 The efficacy and risks of immune edited cells remain to be evaluated in clinical trials.
Potential applications of regenerative neurosurgery
In this section, we provide examples of translational trajectories highlighting regenerative efforts for Parkinson’s disease, ALS, stroke, and spinal cord injury (fig 1).
Parkinson’s disease is clinically characterized in part by tremor, rigidity, and bradykinesia. These result from dysfunction and loss of dopaminergic nigrostriatal neurons, leading to decreased synaptic release of striatal dopamine.123 The mainstay of therapy is L-dopa, which is taken up by remaining dopaminergic neurons and converted to dopamine facilitating dopaminergic neurotransmission. As the numbers of functional dopaminergic synapses decline, L-dopa becomes less effective. With higher doses of L-dopa required and fewer synapses to store and release dopamine, patients experience increasing fluctuations between “ON” states of sometimes excessive dopamine release with associated dyskinesias, and “OFF” states of bradykinesia.
After many years of animal studies, clinical trials over the past three decades have provided proof of principle that grafted dopaminergic neurons can effectively reinnervate the striatum, restoring dopaminergic neurotransmission in the striatum and longlasting improved motor function.124125 Although other non-motor features of Parkinson’s persist,126 some patients in early open label trials no longer required L-dopa with stably improved symptoms 15+ years after implantation.125 A meta-analysis of 11 such trials showed improved off-UPDRS (effect size, d=1.7; P<0.05). Improved response was correlated with younger patients, bilateral graft placement, and persistent viability of grafts on 18F-dopa PET.124
To determine if fetal dopamine cell therapy could meet level 1 evidence standards, two separate double blind placebo controlled studies were performed, both of which failed to meet primary endpoints.127128 These failures, together with frequent symptomatic dyskinesias in 15%127 and 56%128 of patients, prompted a detailed re-evaluation of dopamine cell therapy strategies. Details of cell preparation, graft purity and yield, immunosuppression, patient selection, and primary outcome measures were each scrutinized129 and preclinical studies attempted to model and understand graft induced dyskinesias.130
After roughly a decade’s hiatus, a new wave of cell therapy trials began with Transeuro—a randomized trial evaluating fetal cell transplantation employing “best practices gleaned from expert consensus.” These included use of cell suspension, rather than tissue chunks, which can include more immunogenic cells like endothelium and may compromise homogeneous cell distribution131; enrolling patients exhibiting good response to L-dopa without significant baseline L-dopa induced dyskinesis; optimizing yield of nigrostriatal (A9) dopaminergic neurons132 while avoiding potentially dyskinesia inducing serotonergic neurons133; and minimizing in vitro culture time to minimize loss of fragile A9 DA neurons. A10 neurons from the ventral tegmental area may be more resilient, but lack D2 autoreceptors important for autoregulation upon striatal implantation.134
During this time, progress was also made in development of protocols for use of pluripotent stem cell derived dopaminergic neurons.25135 Several clinical trials are ongoing, and others are about to start. Parthenogenic stem cells were first implanted by an Australian team in 2016.136 Based on promising primate data,137 the first embryonic stem cell derived therapy trial for Parkinson’s disease launched in China in 2017.138 In October 2018 the first human transplant of iPSC derived dopaminergic neurons was performed.139 The first autologous iPSC implantation was reported in 2020.117 Multiple additional trials of pluripotent stem cell based transplants are anticipated.140 Progress and coordination has been facilitated by “G-force,” an international group of translational scientists and clinicians coordinating their parallel efforts to minimize future translational mis-steps.140
Amyotrophic lateral sclerosis ALS is a fatal neurodegenerative disease characterized by progressive loss of upper and lower motor neurons. As more cells become dysfunctional and die, patients become unable to move, swallow, or speak. In such patients, ability to move even a single finger could have a tremendously positive impact on communication, assisted mobility, and quality of life. To date, the safety and feasibility of implanting NSCs into the spinal cord have been demonstrated in clinical trials, but efficacy of any neuroregenerative strategy has yet to be established.141
In a first trial of intraspinal implantation of autologous bone marrow mononuclear cells into the T4/5 thoracic spinal cord, implantation appeared safe. Histology at autopsy showed greater than fourfold surviving motor neurons in treated segments, compared with distant untreated segments. Consistent with data from preclinical studies wherein implanted cells surrounded motor neurons and produced GDNF,142 most motor neurons at the treated level showed a “nest” of adjacent CD90+ cells consistent with implanted cells. These neurons did not contain ubiquitinated inclusions, in contrast with adjacent motor neurons that showed typical intracellular pathology of ALS.143 Consistent findings were observed in 3/3 patients analyzed at autopsy (of 11 implanted),143 congruent with prior preclinical findings.
Recent efforts have focused on implantation of NSCs into the spinal cord. Modest functional gains have been observed in several uncontrolled open label studies. A phase I clinical trial administered 100 000 cells in five to 10 intraspinal locations for each of 12 patients. One patient experienced progressively improved ALS functional rating scale scores from 30 at time of implantation to 46 at 9 months. No safety concerns were encountered.144 Another phase I trial showed modest improvement of lower extremity function or ambulation in three of the six study subjects after implantation of allogenic human NSCs into the upper lumbar spinal cord. Subsequent work progressed to five injections within the unilateral cervical cord, including in some patients who previously received both lumbar and cervical injections for a maximum total of 1.5 million cells. No changes in motor neuron number or appearance were mentioned in any reports to date, though autopsy analysis showed no evidence of rejection or postoperative neuroinflammation.145 Cumulative post-hoc analysis of 21 patients enrolled in phase I or II studies of fetal NSC implantations for ALS into the lumbar and/or cervical cord suggested improvement in the revised ALS functional rating scale at 24 months compared with matched cohorts from two historical datasets 30.1 ± 8.6 (cell implants, n=21) versus 24.0 ± 10.2 (ProACT data set, n=1108; P=0.048); and 30.7 ± 8.8 (cell implants, n=20) versus 19.2 ± 9.5 (ceftriaxone historical control, n=177; P=0.0023).146 No survival benefit was observed and the retrospective analysis limits conclusions regarding functional outcomes.
Building upon preclinical data showing improved motor neuron survival in a rat model of ALS147 and aging,148 results are awaited from a recently completed clinical trial using intra-cord delivery of GDNF-overexpressing NSCs for patients with ALS (NCT02943850). Of note, neurotrophic factors including GNDF are taken up from synaptic nerve terminals via retrograde axonal transport.149 This explains the improved survival of primate midbrain dopaminergic neurons with GDNF delivery into the striatum.150 Indeed, intraspinal GDNF-NSCs do not maintain functional muscle projections of rat ALS spinal motor neurons.151 However, intramuscular GDNF-MSCs promoted spinal motor neuron survival, and motor endplate function152 and was further improved with VEGF-MSC coadministration.153 To date, it remains unclear if spinal GDNF delivery could promote survival of corticospinal projection motor neurons. However, anterograde transport of GDNF was shown to be feasible upon delivery of GDNF expressing adeno associated virus to the red nucleus, which has extensive projections to spinal grey matter.154 Preclinical data show improved behavior and survival of SOD1G93A rats with implantation of GDNF expressing NSCs into the cerebral cortex,97 an outcome not achieved with intraspinal delivery of NSCs or GDNF-NSCs. Given the multiple sites of disease and potential need for therapies acting at muscle, spinal cord, and brain, the road from demonstrating safety of individual approaches to a clinically meaningful therapy may require both perseverance and a willingness to consider early phase combination therapies for ALS.
Stroke remains the leading cause of disability, for which improved therapies are urgently sought. For the persistent majority who do not receive or benefit from acute revascularization procedures, there remains an unmet need to combat secondary injury through neuroprotection or regain function through neuroregeneration.
Most survivors experience some variable degree of neurological and functional recovery. Inability to predict the degree of spontaneous recovery hampers functional baseline comparisons and increases trial enrollment needed for adequate statistical power. Such pragmatic considerations have contributed in part to all phase II studies of stereotactic cell implantation for stroke to date being restricted to patients with chronic motor deficits. In the first such study, patients at least one year from stroke with stable motor deficits over two months were randomized to rehabilitation only control, versus implantation of 5 or 10 million teratocarcinoma derived neurons at 25 sites divided among five trajectories. With only four controls and seven patients per treatment group, this observer blinded study trial did not meet the primary efficacy measure of change in European Stroke Scale (ESS) motor score. However, improvement was seen in multiple secondary measures. Action Research Arm Test gross hand movement scores were improved in comparison with control (P=0.017) and baseline (P=0.001) values; Everyday Memory scores improved relative to controls (P=0.012) and baseline (mean change 13 (95% confidence interval (CI) 4.9 to 21.2), P=0.004).102
More recent studies have evaluated stereotactic MSC or NSC implantation into the basal ganglia of patients with chronic basal ganglia stroke. In a non-randomized phase I/IIa trial, patients with chronic basal ganglia stroke were implanted with Notch-1 overexpressing allogenic MSCs (SB623, SanBio). Results showed significant improvement from baseline on multiple clinical outcome scores, including ESS: mean increase 6.88 (95% CI, 3.5 to 10.3; P<0.001), that plateaued at 12 months with no subsequent deterioration by two years.118155 A significant positive correlation was noted between the size of fluid attenuated inversion recovery (FLAIR) signal on a 1 week postoperative magnetic resonance image (MRI) and improvement at 12 months (ESS, P<0.001), the mechanistic basis of which remains uncertain. Results of a subsequent randomized study have yet to be published, but reportedly did not meet primary outcome measures. As such, this product is not being pursued further for stroke. However, work with SB623 is proceeding for traumatic brain injury, for which positive results of a phase II trial were reported in abstract form.156 The Pisces phase I trial implanted 2-20 million immortalized allogenic NSCs (ReNeuron, Inc) into the putamen of 11 men with stable deficits 6-60 months after ischemic stroke, yielding a median improvement of two points (range 0-5) on the NIH Stroke Scale score at two years.157 An open label single arm multicenter trial (Pisces-2) implanting 20 million cells 2-13 months after stroke failed to meet primary outcome measures relating to arm movement at 3 months. Twenty five per cent of patients improved by >6 points (of 57 total) on the total Action Research Arm Test score at 12 months, all of whom maintained at least some upper limb movement at baseline.158
Parallel work has seen completion of phase II clinical trials for intravenous159160 and intra-arterial161 delivery of autologous159161 or allogenic160 bone marrow mononuclear cells159 and somatic stem cells.160161 These have demonstrated safety with promising indications of efficacy, including improved outcomes from previously stable baselines.
Meta-analysis of preclinical studies has shown significant benefit of stem cell therapy for stroke; parallel evaluation of human studies reveals safety with less clear evidence of benefit.162163164 Inhibition of apoptosis was determined as the best predictor of functional outcome across preclinical data.69 Consistent with this finding, analyses of MSC therapy in both preclinical and clinical studies have suggested MSC efficacy is time dependent, with improved outcomes observed with early delivery.165166 Intra-arterial delivery at the time of endovascular thrombectomy may provide a pragmatic window of opportunity. Conversely, the rapid improvement of select patients after intraparenchymal cell therapy for chronic stroke155 motivates efforts to modulate innate immunity and neuronal plasticity.
Spinal cord injury
Spinal cord injury combines some of the greatest challenges facing the field of neuroregeneration. Lesions are heterogeneous and include transections, contusions, hemorrhage, and compression, with complex pathophysiology spanning ischemia, inflammation and demyelination, and glial scarring. Injuries occur without warning in patients without risk factors. Damaged circuitry includes spinal corticospinal tracts—the longest CNS projections in the CNS.
In preclinical studies, stem cell derived remyelination for spinal cord injury has been reported to improve function.167 Trophic benefits of NSC implants have also been described that promote neurite outgrowth,168 with long term persistence of graft derived astrocytes that functionally contribute to the blood-cord barrier and uptake of? glutamate.169 A smaller number of reports have suggested ability of implanted stem cell derived cells to create long distance connections and improve locomotor performance.170
Much preclinical work for SCI has been performed in rodents,171 which, like cats,172 have a robust spinal cord central pattern generator (CPG) that augments recovery from spinal cord injury. Though more difficult to directly study in humans, the presence of such a CPG has been shown,173 and increasingly harnessed via epidural stimulation, to facilitate recovery of function in patients following spinal cord injury.174 Neuroanatomic and neurophysiological differences between humans and animal models limit ability of promising preclinical findings to predict clinical efficacy.175 For example, humans and other primates may have greater capacity than rodents for recovery after a hemicord injury, because of formation of extensive corticospinal detour circuits below the level of injury.176 Encouragingly, human spinal cord progenitor cell derived progeny may generate appropriate and functionally relevant synaptic connections in the injured primate spinal cord.20
Extracellular matrix (ECM) substrates may promote neuronal survival and function in cavitary or scared CNS lesions.177 Normal adult synapses are supported by a specialized ECM mesh termed perineuronal nets178 that support synaptic stability and help maintain both memories and cognitive function by impeding excess plasticity.179 Enzymatic degradation of chondroitin sulfate—one of the components of both glial scars and perineural nets has been eyed as an avenue to promote regenerative plasticity yielding benefit in a primate model of C7 hemisection injury.180
To date, cell therapy trials for spinal cord injury remain in their relative infancy. Multiple early phase clinical trials have reported safety of intraspinal implantation of autologous mesenchymal stem cells,181 autologous Schwann cells,182 and allogenic NSCs.183 Failure to meet primary outcome measures of efficacy at 12 months prompted early termination of the Stem Cells Inc phase II Study of Allogenic NSCs trial.183 As reported with MSC implants into stroke, a safety and feasibility study of 29 patients noted occasional FLAIR signals around the injection site that resolved within 6-12 months,183 albeit without reported correlation with functional benefit. Geron undertook a highly publicized phase I trial of ESC derived oligodendrocyte progenitor cells for subacute thoracic spinal cord injury, with five patients implanted. The technology was acquired by Asterias, and another 25 patients with subacute cervical cord injury were implanted in the SCiStar phase I/IIa dose escalation trial. Interim safety results were reported without peer review by the company at one year, noting no unexpected serious adverse events, and that 96% of patients improved at least one level and 33% of patients improved two or more levels.184
How can the study of regenerative neurosurgery improve?
Evidence objectively supporting regenerative neurosurgery interventions remains extremely limited and is of poor quality. Only two double blind placebo controlled clinical trials have taken place,127128 both of which missed primary outcome targets. Demonstrating benefit in clinical trials for neurological diseases has proven challenging even in the best of circumstances. Despite ample promise,185 endovascular stroke interventions failed in multiple large randomized studies, partly because of pitfalls spanning patient selection, study methodology, and design and outcome measures utilized.186 Ultimately, level 1 evidence was established with now expanding indications. Without such rapid and unequivocal feedback from hundreds of centers wherein rapidly reopened vessels beget functionally salvaged patients, the predictably numerous pitfalls for regenerative CNS therapies will undoubtedly be more challenging to identify and overcome.187 Cell therapy trials are complex, expensive, and labor intensive. Imperfect preclinical models may create detours.188 Resilience of the multidisciplinary teams and their sponsors will be needed to persevere and apply hard earned insights in the face of predictably repeated failures.
Innovative surrogate outcome measures
Additional technologies are needed to provide the neuroregenerative equivalent of an angiogram that objectively documents success of thrombectomy for stroke. Perhaps the closest to date is 18F-DOPA PET, which helps quantify functional dopamine physiology in Parkinson’s disease. Although MRI based methods to visualize cell loss are improving,189 mechanistically pertinent surrogates are lacking for stroke, spinal cord injury, ALS, and most other neurological diseases; novel molecular imaging targets offer future promise.190 Innovations harnessing neurosurgical access to the brain could provide unparalleled opportunities to interact functionally with the degenerating and regenerating brain. Emerging micro-scale electrode arrays, including NeuroPixel,191 can record from hundreds of individual neurons simultaneously, detailing connectivity and circuit function. Microdialysis,192 microperfusion,193 solid state electrochemical probes,194 carbon nanotubes,195 and closed loop systems196 may quantify neurotransmitters cytokines, matrix modulating enzymes, and metabolic biomarkers197 in concert with neuronal activity, affording a rich milieu for transformative innovations.
Synergy with neuromodulation
Future development of cell based neuroregenerative technologies may be further augmented by emerging closed loop,196 neuromodulation,174 and neural prosthetic or brain-computer interface based strategies,198 augmented by artificial intelligence.199 Optimally deployed, these technologies have potential to enhance the differentiation,200 distribution,201 and functional integration of implanted cells,202 while also enhancing the functional utility of remaining circuitry.203
Efforts to identify and harness individual molecular therapeutic targets for neurological disease have largely failed. iPSCs themselves exist owing to Yamanaka’s ingenuity in emulating a “successful” pluripotency via multiple simultaneous transcription factors. If he had tried only one transcription factor at a time, iPSCs would not exist. Prudence would urge that the field follow Yamanaka’s precedent to emulate success—anticipating success trying one candidate therapy at a time may be unrealistic. Plasma from juvenile animals can augment neurogenesis, synaptic density, and cognitive performance of aged animals.204 Eliminating senescent cells from obese animals can improve neurogenesis and ameliorate anxiety.205 Exercise206 and caloric restriction207 may improve, just as stress208 and systemic inflammation209 hamper neurogenesis and cognition. Cell and other regenerative therapies should be similarly optimized through systematic integrative efforts.
The future of regenerative neurosurgery is bright but faces important challenges. This decade will be a time not only to optimize experimental paradigms, but also to invest as a society in our future to change Ramon y Cajal’s “harsh decree.” We have presented a synopsis of current cellular and rejuvenative therapies—some of which could have utility against multiple diseases. It will be critical that the operating room becomes a seamless extension of the laboratory in a way that is both ethically and fiscally responsible to accelerate translation. Patients are looking for cures and searching for hope—sometimes turning to unregulated and dangerous sources. As such, it is imperative to harness, responsibly and safely, neurosurgical opportunities to accelerate progress toward cures. Our ability to make use of the increasingly available technologies to benefit patients is limited only by our collective imagination and tenacity.
Questions for future research
Which patients have the greatest potential to benefit from specific candidate neuroregenerative therapies?
What biomarkers could provide early feedback regarding the mechanistic impact of therapies to empower iterative improvements?
How can we make safe and effective autologous cell therapies cost effective to avoid immune reactions without immunosuppression?
How can we rationally and rigorously develop multimodal neurorestorative therapies that leverage synergistic mechanisms of action?
Series explanation: State of the Art Reviews are commissioned on the basis of their relevance to academics and specialists in the US and internationally. For this reason they are written predominantly by US authors
Contributors: Both authors conceived the scope and content of the manuscript. TCB wrote the manuscript with intellectual contribution from AQH. Both authors critically reviewed the manuscript and approved it for publication. The authors acknowledge the editing assistance of Superior Medical Experts (who assisted in the early draft of the manuscript but had no input into the selection of literature or final wording, and have no commercial links to the subject matter), and Kirsten Burns. Illustration services were provided by Caitlin Mock. Diogo Garcia and Gaetano De Biase provided assistance with preparation of tables. The authors acknowledge many colleagues and collaborators for fruitful discussions, as well as patients whose courageous participation in clinical trials inspire and empower regenerative neurosurgery.
Competing interests: We have read and understood TheBMJ policy on declaration of interests and declare the following interests: TCB serves on the scientific advisory board of Neurametrix and as a consultant for Alector.
In 2020, after the current manuscript was submitted, AHQ became founder of Dome Therapeutics, a company dedicated to stem cell therapies for neurological disorders. The company currently has no products or technologies in development and has no commercial revenue.
Funding: TCB was supported by NIH K12 NRDCP, NINDS NS19770, the Minnesota Partnership for Biotechnology and Genomics, Mayo Clinic Center for Regenerative Medicine, Lucius & Terrie McKelvey, and Regenerative Medicine Minnesota. AQH was supported by the Mayo Clinic Professorship, the Mayo Clinic Clinician Investigator award, the Florida Department of Health Cancer Research Chair Fund, as well as the National Institutes of Health (R43CA221490, R01CA200399, R01CA195503, R01CA216855).
Provenance and peer review: commissioned; externally peer reviewed.