The ability to repair or promote regeneration within the adult human brain has been envisioned for decades. Until recently, such efforts mainly involved delivery of growth factors and cell transplants designed to rescue or replace a specific population of neurons, and the results have largely been disappointing. New approaches using stem-cell-derived cell products and direct cell reprogramming have opened up the possibility of reconstructing neural circuits and achieving better repair. In this Review we briefly summarize the history of neural repair and then discuss these new therapeutic approaches, especially with respect to chronic neurodegenerative disorders.
The mammalian central nervous system (CNS), unlike many other organs, has a limited capacity for self-repair, and the ability to overcome this limitation would transform how we could treat patients with a vast array of neurological disorders. The relative lack of neurogenesis, failure of axonal regeneration and non-permissive local environment in the adult CNS have been major—and literal—barriers to CNS repair. Nevertheless, in recent years it has become clear that much can be done in the mature CNS to foster innate regeneration.
Diseases or disorders of the CNS can be developmental, inherited or acquired and acute or chronic, and each presents a different set of challenges when developing repair strategies. Both acute damage and chronic neurodegenerative disorders are characterized by an inter-related macroglial and microglial/inflammatory reaction that involves changes in the extracellular matrix and/or alterations in the blood–brain barrier. Therefore, agents that improve this environment could be used to help to protect neurons from further damage and to promote circuit reformation and function. Such disease-modifying therapies—using, for example, trophic factors or cell transplants that release such factors—could in theory be used alone in the early stages of disease or combined with approaches designed to replace lost cells through cell transplantation, re-activation of endogenous neurogenesis or directed reprogramming in situ. In this Review, we discuss three main strategies for restoring function in the damaged or diseased CNS: (i) cell rescue using neurotrophic factors or cell-based approaches with putative-disease modifying effects; (ii) cell replacement therapies using transplants of exogenously derived and in vitro cultured cells; and (iii) cell replacement using endogenous neural precursor cells or strategies that involve the direct reprogramming of resident cells within the brain (Fig. 1).
Neurotrophic factors and disease modification
Enhancing the regenerative potential of the CNS in the face of a disease process can be done in two major ways. The first is to use approaches that restore and maintain the functional integrity of cells that are dysfunctional but not lost. This typically involves neurotrophic factors, some of which could be delivered using cell-based treatments. However, grafted cells could also be used to protect dysfunctional cells from ongoing insults such as metabolic stress. The second strategy is to minimize environmental barriers to repair, for example, by modifying the extracellular matrix and/or glial and immune reactions1,2.
Exercise-based therapies and cognitive training have been shown experimentally to promote endogenous neurogenesis and the local production of growth factors such as BDNF in the brain3. Whether these changes in experimental animals can translate into clinical benefits is largely unknown, but such interventions do have the potential to be used in combination with cell replacement or cell rescue therapies to optimize their reparative benefits.
The use of trophic factors has been explored for over 20 years, most notably in amyotrophic lateral sclerosis (ALS) and Parkinson’s disease4,5 (Table 1). These initial studies suffered from major problems with delivery—the factors either failed to reach their neuronal targets or were rapidly inactivated. This led to a new round of trials using intraparenchymal delivery, most notably involving delivery of GDNF into the striata of patients with moderately advanced Parkinson’s disease. In two open-label studies, from two independent centres, this produced positive clinical responses, with post mortem and imaging data supporting its mode of action6,7,8,9. However, a subsequent double-blind placebo-controlled trial in 34 patients with Parkinson’s disease found that patients treated with GDNF showed no significant benefit over placebo-treated patients after 6 months10.
The reasons for the discordant results from these trials have been discussed extensively11, including possible placebo effects, although these would not explain the reported changes on PET imaging and at post mortem in patients treated with GDNF infusions. Furthermore, in the double-blind study, the group treated with saline did not improve, which also argues against any placebo effects and suggests that the trial failed because the active treatment did not work. This could have been due to the dose of GDNF and/or the way in which it was delivered, both of which differed from those used in the open-label studies. Given this uncertainty, further trials have been undertaken, including a double-blind approach with a new convection-enhanced delivery system that delivers GDNF more effectively. However, so far positive outcomes have not been achieved (https://scienceofparkinsons.com/2016/07/07/the-gdnf-trial-bristol-initial-results/), which suggests that the use of trophic factors alone may not work to treat such chronic neurodegenerative disorders. This is reinforced by parallel studies involving the delivery of other GDNF-like factors12, such as neurturin using an adeno-associated virus (AAV) delivery system (AAV-NTN). Here, the initial open-label studies showed efficacy but two subsequent double-blind placebo control studies (which differed in terms of where, and how, the AAV-NTN was delivered) failed to show efficacy13,14,15.
So, does this mean that repairing the brain with trophic factors is not a viable way forward? This is likely to be the case when the disease has advanced beyond a certain stage; once a patient has lost a critical number of cells or fibres, it is unrealistic to expect any growth factor to work, as there is nothing left for it to regenerate. Given that most neurodegenerative disorders present clinically when there has already been extensive cell or fibre loss, the early use of these agents will be required, and a post hoc analysis of some of the data from GDNF and AAV-NTN trials supports this view14,15. Moreover, the disease state may also interfere with growth factor function. For example, GDNF may not be able to act through its normal receptor in the presence of α-synuclein pathology in Parkinson’s disease because the pathology causes alterations in NURR1 signalling, which is needed for GDNF to produce its therapeutic effect16. Thus, it may be necessary to enhance levels of NURR1 before using GDNF to treat Parkinson’s disease for it to work efficiently, necessitating the adoption of combination approaches in any future trials. Finally, the optimal dose of any growth factor and how to best deliver it, including the volume of distribution needed to maximise its benefits, are still important questions. To date, GDNF and AAV-NTN have been delivered in ways that are insufficient to cover the structure being treated, and many of the failures seen clinically may relate to this. Post mortem data from both the original GDNF and subsequent AAV-NTN studies provide some support for this idea9,15.
Overcoming environmental hurdles to regeneration
Promoting neuronal survival and outgrowth is only one aspect of the problem in repairing the CNS. The other is enhancing the environment to support that regenerative process. This essentially distils down to modifying the glial elements and/or the extracellular matrix (ECM), both of which have been reported to produce signals that inhibit repair in the adult CNS1,2.
In terms of the glial cells, it has long been known that both astrocytes and oligodendrocytes express cell surface molecules that inhibit regenerating axons, although the expression and extent of these varies as a function of disease state (for example, being greater in patients with acute traumatic injuries than in chronic neurodegenerative disorders). The premise that glial cells produce inhibitory factors that can be blocked has led to early clinical trials, most notably using antibodies against the myelin-derived molecule NOGO in patients with spinal cord injury (http://www.sci-research.uzh.ch/en/research/ClinicalTrials/Nogo.html). However, the concept that glia are ‘bad’ for CNS repair has been challenged, in that the glial scar or glial reaction to injury or disease involves a heterogeneous collection of cells, which have a variety of functions. So, under some conditions the glial response may be beneficial for repair, possibly through effects on the inflammatory response to CNS disease1. Conversely under other circumstances, the opposite may occur. For example, it has recently been reported that microglia can induce astrocytes to become neurotoxic17. This concept of ‘gliopathy’ is not new and has been explored in models of ALS, where transplanted glial cells have been used to buffer extracellular glutamate and thus protect motor neurons18,19. Whether such an approach could be used to treat ALS clinically is unknown, but this work highlights the potential use of non-neuronal cell transplants to protect neurons in disease states.
More recently there has been interest in the ECM of the brain, and especially peri-neuronal nets (PNNs). PNNs are unevenly distributed across the brain and spinal cord and appear to have many functions during development and in the adult brain, including a role in neural plasticity. This role has been shown to extend to neurodegenerative disorders, most notably in transgenic models of Alzheimer’s disease, where removal of PNNs, either using chrondoitinase-ABC or genetically, has been shown to improve cognitive function without influencing the underlying disease process, possibly by enhancing neurite outgrowth and synaptic plasticty20,21. These results suggest that the use of this, or similar, agents may have benefits when combined with cell therapies in which synaptic integration with host cells may be restricted by local PNNs and/or a disease-driven increase in expression of inhibitory ECM molecules.
Replacement therapies using exogenous cells
When a neurodegenerative disorder or injury reaches the point at which a critical number of neurons has been lost, the replacement of these neurons is essential for functional improvement. Lost cells could be replaced by transplantation into single or multiple brain regions to partially recreate complex circuits either locally or over long distances.
For this to work, a number of important questions need to be addressed. First, what is the optimal cell type(s) to be transplanted and at what developmental stage should they be collected for grafting? Second, what CNS location(s) are needed for optimal recovery and fibre outgrowth? Third, how many cells are needed for effective repair, and finally, should other interventions be done either at the same time as grafting or after transplantation to enhance survival, integration or function of the cells?
Cell replacement therapy using human fetal allografts
Parkinson’s disease is one of the least complex neurodegenerative disorders to target with cell-based therapies, because local degeneration of dopamine neurons in the midbrain is critical for the clinical expression of many features of this disorder. It has therefore been the subject of a relatively large number of clinical transplantation trials using developing midbrain dopamine cells derived from the human fetal ventral midbrain22. Although the outcome of these trials has been highly variable23, they have shown that at least a sub-group of patients with Parkinson’s disease that receive such transplants at different centres do well24,25,26,27. However, the results have also shown that, over time, a minority of the transplanted cells become affected by the disease process and show protein aggregation28,29, and that some patients can experience troublesome side effects such as graft-induced dyskinesias30,31,32 relating to graft composition and capacity to innervate the whole putamen22.
Clinical trials using transplants of primary human fetal striatal tissue have also been performed in patients with Huntington’s disease33,34,35, on the premise that the medium spiny neurons of the striatum are the neurons that are primarily lost in this disease36. In this case, the neurons are placed homotopically into the striatum to reconstruct basal ganglia circuits. To date the effects in patients have been modest at best33,34,35, with evidence for poor graft survival and integration37.
Other neurodegenerative and CNS disorders have more complex pathologies, involving several cell types and/or cells in many regions of the CNS, and it is hard to see how these could be treated using fetal tissue transplants. However, stem cells may be a more promising source of cells for transplantation in such disorders, because they can be differentiated into a wide range of neuronal phenotypes.
Cell replacement therapy using stem cells
A major challenge for developing any cell replacement therapy is the ability to reliably and robustly generate large numbers of relevant cells. Recent developments using human embryonic stem cells (hES cells) and induced pluripotent stem cells (iPS cells) have now made this possible38 and clinical trials using this approach are imminent, for example, in patients with Parkinson’s disease39,40.
Stem-cell-derived dopamine neurons have also been used to better understand graft function, innervation and integration in pre-clinical animal models of Parkinson’s disease. For example, comparative studies with human fetal ventral midbrain tissue have shown that both hESC and hiPSC-derived dopaminergic progenitors are very similar to their fetal counterparts41,42 and mediate functional recovery with equivalent potency over a similar time course43. New methodologies have also been used to study their mechanism of action, as well as their ability to appropriately innervate and integrate into host circuitry. These studies have shown that both fetal and stem-cell-derived dopaminergic neurons transplanted to the rat brain can innervate the correct target structures when placed either locally at the site of action (in the striatum) or back in the midbrain, which requires more extensive axonal outgrowth from the cells42,43. Furthermore, these cells not only innervate but also integrate into host neuronal circuitry44, and produce functional recovery38,45. Similarly, transplantation of stem-cell-derived medium spiny neurons have now been shown to survive and function in rodent models of striatal degeneration (used to model Huntington’s disease)46,47,48.
Another advantage of using stem-cell-derived transplants is the ability to remove or minimize unwanted neuronal subtypes; for example, serotonergic neurons that may mediate the graft-induced dyskinesias seen in patients that have received transplants of human fetal ventral midbrain tissue to treat Parkinson’s disease. Furthermore, it is possible to study the effect of grafting pure populations of neurons, or a combination of neuronal and or astrocytic cell types. Indeed, deficient astrocyte function may contribute to the loss of neurons in many neurodegenerative disorders, and the absence of such cells may affect neuronal survival in transplants, as has been shown in fetal striatal grafts in patients with Huntington’s disease49. As an alternative to possible co-grafting with glial cells, other agents or biomaterials engineered to express growth factors, immune-modulating components or similar substances to provide a supportive and protective microenvironment could be delivered together with the transplanted neurons; this could even include some form of scaffold for larger lesions such as stroke50.
Genetic modifications or correction of cells could also be used before transplantation to make the grafted cells resistant or refractory to the disease pathology. This may be of critical importance when the transplants are derived from the patients’ own cells42, or in diseases such as ALS, where motor neuron loss is driven in part by astrocytes. However, even unrelated healthy donor cells can be affected by proteinopathies in some neurodegenerative disorders, as has been seen in trials of fetal tissue transplantation for Parkinson’s disease and Huntington’s disease28,29,51. In these cases, making cells resistant to the disease by removing the normal protein onto which the pathological species templates could be helpful.
Additionally, as genetic engineering tools become more refined, stem cells could be modified to enhance aspects of their behaviour. Such techniques could include modifying chemotactic interactions between the grafted cells and their progeny to enhance migration52; making cells refractory to the inhibitory cues of the glial scar; increasing their innervation potential by enhanced expression of polysialic acid53 (all of which would need to done so that the enhanced outgrowth was targeted specifically to the area for repair); and/or decreasing their immunogenicity54,55. Based on cell transplants in animal models, the possibility of engineering stem cell progeny so that their function can be modified using optogenetic or chemogenetic interventions is now also being considered45,56,57.
Cell replacement therapy for circuit reconstruction
Owing to the complexity of the organization of the cerebral cortex, replacing neurons in this part of the brain is a more challenging task, and is thus still at the pre-clinical stage. New technologies for studying cell integration and function have shown that transplanted cells mature, integrate into local host circuitry, form functional synapses, mediate recovery and provide circuit repair in animal models of Alzheimer’s disease, stroke and cortical lesions58,59,60,61,62. Notably, it has recently been shown that the exact brain-wide input connectome of neurons in the mouse visual cortex can be replicated by transplanted neurons59. This is important, as aberrant connectivity could lead to deleterious consequences such as epileptic seizures. Thus, the adult mammalian brain is more plastic than previously thought in terms of accommodating new neurons and building new functional circuitry.
Replacement therapies using endogenous cells
Brain repair from endogenous sources has been considered for some time63, especially after the (re-)discovery of adult neurogenesis and proof of its existence in the human brain64,65, prompting attempts to recruit new neurons from endogenous neurogenic niches to achieve repair in the injured brain66,67 (Fig. 2). Simply recruiting new neurons from active neurogenic sites appears intuitively to be the most promising approach, and the easiest to achieve. However, in most cases, the subtypes of neurons needed for repair are different from those generated endogenously. Hence, attempts to recruit neuroblasts to form neurons at other sites have met with only limited success, with many of the neuroblasts streaming to the injury site but then failing to fully differentiate into appropriate neuronal types that can survive in the long term67,68 (Fig. 2). Adult neurogenesis is very species-specific69, though, so endogenous striatal neurogenesis that has been said to occur in the human (but not rodent)70,71 brain may prove useful and be more easily manipulated to generate the relevant neuronal subtypes lost in the striatum after stroke or in neurodegenerative disorders72,73.
A further approach to minimize neuronal loss or help repair and replacement is through the recruitment of stem-cell-derived astrocytes. These differ from the local parenchymal astrocytes and release an array of beneficial growth factors74. However, the extent to which this occurs in human patients and can be used for therapy is currently unknown.
Excitingly, attempts to turn stem-cell-derived astrocytes or other parenchymal glial cells into functional neurons by forced expression of neurogenic fate determinants have progressed rapidly. This idea was first tested in the nervous system in 2002, by turning glial cells into neurons in vitro75. The first attempts to achieve this in an injured brain in vivo in 2005 were exciting as proof-of-principle experiments, but disappointing in terms of quantitative outcome76,77,78. However, outcomes have improved recently79,80,81, and there is now encouraging data on fate conversion of local non-neuronal cells towards a neuronal phenotype of the sort affected by the disease process.
A single neurogenic transcription factor (acting also as potent fate determinant in development) can in many cases be sufficient to convert glial cells into fully functional neurons82. For example, neurogenin 2 or NEUROD1 can produce glutamatergic neurons from proliferating glial cells in the adult cerebral cortex, including in the highly inflamed brain environment found after traumatic brain injury79,83,84. The key to generating highly efficient neuronal conversion protocols is not necessarily to add more transcription factors, but rather to keep the induced neurons alive by protecting them from death and promoting their maturation. Many neurons induced by direct reprogramming from glial cells die because of the local generation of excessive reactive oxygen species (ROS), in a process known as ferroptosis79,84. By keeping the new neurons alive and reducing ROS levels, conversion efficiencies of over 90% have been achieved, with almost all glial cells, transduced with the neurogenic transcription factor, acquiring a neuronal identity within 1–3 weeks84. Importantly, these neurons also acquired a pyramidal neuronal morphology with the identity of deep layer neurons, at least at the transcriptional level. This approach has also shown promise in other brain areas, including in the minimally injured striatum (using SOX2 and growth factor treatment), the midbrain and the spinal cord85, and in the cerebral cortex using NEUROD1 and an amyloid plaque deposition model83. In these latter approaches, a mix of glutamatergic or GABAergic (γ-aminobutyric-acid-releasing) neurons has been obtained, which may be beneficial if several types of neurons need to be replaced or synaptic plasticity increased by providing new GABAergic interneurons, as shown through transplantation studies86,87. Synaptic plasticity and new circuit formation using these approaches could also be further promoted by co-administration of rehabilitation therapies (see above).
The promising outcome of direct reprogramming approaches raises two important issues: which glial cells to target and which vectors to use. Targeting proliferating glial cells has obvious advantages88, but they are heterogeneous1,76,89. Specific targeting of glial subtypes has been achieved using subtype-specific promoters that either drive reprogramming factors directly83,90,91 or drive cell-type-specific Cre expression in transgenic mice80,81,85. This approach has now been shown to work with AAVs that not only have been used in patients clinically but also can preferentially infect glial cells92,93,94. AAVs have been used to reprogram oligodendrocyte progenitor cells80,95 to a fate that is dependent on the transcription factor combination—for example, generating fast-spiking parvalbumin-expressing interneurons in the intact striatum95. Moreover, these neurons have been shown to receive input from local interneurons80. Finally, recent evidence suggests that striatal and midbrain astrocytes and NG2 glia can be converted to neurons with functional GABAergic- and dopaminergic-like phenotypes in a dopamine depletion model96. While these pioneering studies show that newly induced neurons can connect locally and result in behavioural changes, their long-distance inputs and outputs and contribution to the neuronal network still remain to be explored.
Direct neuronal reprogramming in the adult brain has come a long way in the past few years, but it remains to be seen how these exciting new approaches will translate to the clinic. While there are no major obstacles to turning adult human cells into functioning subtype-specific neurons, as has been shown, for example, with fibroblasts or pericytes in vitro97,98,99,100, neuronal reprogramming in vivo in the diseased brain has proven more challenging79. Thus, investigation of differences in the starting cells that affect the outcome of neuronal reprogramming, as well as influences in the local (disease) environment, is one of the next requirements to further this approach79. Moreover, tools need to be developed for non-invasive reprogramming techniques, either using systemic application of viral vectors (as is the case for some AAVs92,93,94) or using small molecules that have been shown to efficiently achieve direct neuronal reprogramming in vitro79,101.
Direct reprogramming also brings new standards to the field, as it is important to compare the induced cell type to the endogenous counterpart that is being replaced or instructed (for example, iPS cells to ES cells). This can be done using functional assessments and genome-wide expression analysis. Such techniques are urgently needed for the field of neuronal replacement, as it is so far unknown how authentic the transplanted or reprogrammed neurons really are, compared to their endogenous counterparts. Raising these standards in basic research and then translating them to the clinic will be a challenge for the future.
Conclusion and future perspectives
A broad approach will be needed to deliver effective repair strategies to the human brain that combine established approaches with new technologies, and ultimately may involve combination therapies. This progress will be possible only if we better understand disease states and the contribution of all cells and the ECM to the processes of damage and cell loss, as well as the innate regenerative ability of the adult brain and its capacity to be reprogrammed. By combining approaches to optimize the disease environment and synaptic plasticity while providing new neurons for replacement therapy, new circuits could be rebuilt with functional benefits across a range of disorders, and this could eventually develop into a more standard treatment. However, this will need to be an iterative approach with an ongoing dialogue between preclinical work and clinical trials. It will also be crucial to avoid the temptation to short-circuit this approach with premature claims and commercialization, which have the potential to derail this whole field.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
R.A.B. is funded by the NIHR Biomedical Research Centre in Cambridge, Cure PD, PDUK, European Research Council under the European Union’s Seventh Framework Programme: FP/2007-2013 NeuroStemcellRepair (no. 602278), Wellcome Trust MRC Stem Cell Institute and MRC UKRMP PSCP. He has received consultancy payments from FCDI and LCT. M.G. is funded by the German Research Foundation (CRC870, SPP1738, 1757, EXC1010 Synergy), The Ministry of Science and Research (MAIV), ERANET and the ERC (ChroNeuroRepair). M.P. receives funding from the New York Stem Cell Foundation, the European Research Council under the European Union’s Seventh Framework Programme: FP/2007-2013 NeuroStemcellRepair (no. 602278) and ERC Grant Agreement no. 30971, the Swedish Research Council and the Strategic Research Area Multipark at Lund University. M.P. is a New York Stem Cell Foundation Robertson Investigator. We thank D. Daft for her help in the preparation of this manuscript.
Nature thanks P. Arlotta, J. Takahashi and the other anonymous reviewer(s) for their contribution to the peer review of this work.