Introduction

The progressive degeneration of motor neurons that occurs in amyotrophic lateral sclerosis (ALS), the most common adult motor neuron disease (MND), affects almost all cellular components of the neuromuscular system, including cortical and spinal motor neurons, interneurons Crabé et al. (2020), glial cells Van Harten et al. (2021); Vahsen et al. (2021), as well as muscle Loeffer et al. (2016). Motor axons, contained within peripheral nerves, serve as the final common relay for transmission of motor signals that control all voluntary muscle contraction and movement. However, one of the earliest characteristic pathological features of ALS involves “die-back” of motor axon terminals Fischer et al. (2004) and breakdown of neuromuscular junctions (NMJs), the specialized synapses between motor axon terminals and muscle fibres. Effectively, this represents a single point of failure that permanently blocks motor signal transmission, irrespective of the condition of central motor circuits involved in coordination of movement signals. This results in an inexorable progression of muscle weakness, atrophy and, ultimately, complete paralysis, eventually leading to premature death. The median survival time in ALS, from initial onset of symptoms to death, typically as a result of respiratory complications, is only 20-48 months Chiò et al. (2009) and ALS has an estimated global mortality of 30,000 patients per year Mathis et al. (2019).

ALS is typically classified into either familial (fALS) or sporadic (sALS) forms of the disease, based on whether or not patients have an identified family history of the disease; between 5-10% of total ALS cases fall into the former category, fALS, with the remaining 90-95% consisting of sALS cases Mathis et al. (2019). To date, over 20 monogenic mutations that cause ALS have been identified, however these still only account for 45% of fALS cases and only 7% of sALS cases Mejzini et al. (2019). The downstream cellular and molecular pathomechanisms underlying neurodegeneration in ALS are extremely complex and include dysregulation of proteostasis, autophagy, RNA metabolism and axon transport, as well as excitotoxicity, oxidative stress and neuroinflammation Mejzini et al. (2019). Given the disparate causes and complex disease mechanisms, development of an effective therapy has proven extremely challenging and there are currently no effective treatments capable of arresting the progressive paralysis that occurs in ALS Brown and Al-Chalabi (2017). Even emergent gene therapy approaches, such as antisense oligonucleotides (ASOs) that have shown early promise in clinical trials Miller et al. (2013), are unlikely to benefit the majority of patients with sporadic ALS and will not replace degenerated motor neurons or restore motor function once it has been lost.

Similarly, efforts to develop cell therapy approaches for ALS have, so far, primarily been aimed at slowing the degeneration of motor neurons in the spinal cord by providing neurotrophic factor (NTF) support through intraspinal engraftment of foetally-derived neural stem cells (NSCs) Feldman et al. (2014); Goutman et al. (2018) or intrathecal delivery of autologous mesenchymal stem cells (MSCs) that are modified to overexpress NTFs Cudkowicz et al. (2022); Berry et al. (2019). Whilst these approaches were shown to significantly slow disease progression in transgenic animal models of ALS Xu et al. (2006); Yan et al. (2006), clinical trials in ALS patients have shown only a modest therapeutic effect in the case of intraspinal NSC grafts Glass et al. (2012) and no overall benefit of intrathecal MSC delivery Cudkowicz et al. (2022) Given the highly invasive surgical laminectomy required to engraft NSCs into the ventral horn of the spinal cord, only localized populations of motor neurons in the lumbar and/or cervical region were targeted, however, the surgical procedure was generally well tolerated and the approach was proven to be safe in clinical trials Feldman et al. (2014); Glass et al. (2016). Moreover, these clinical trials have provided evidence that specific muscle functions can be preserved for longer in some ALS patients Mazzini et al. (2019). Importantly, this pioneering approach provides a precedent for implementation of an allogeneic stem cell-based therapy and also shows that ALS patients can tolerate a 6 month period of immunosuppression Mazzini et al. (2015), which appears to be sufficient to confer long term survival of engrafted cells Tadesse et al. (2014). Nonetheless, it appears unlikely that this approach will be able to restore motor function once it has been lost, since the NSCs do not replace lost motor neurons, and any therapeutic effect has so far been shown to be transient Goutman et al. (2018); Mazzini et al. (2019). Therefore, there remains an urgent unmet need to develop novel therapies that can rescue muscle innervation and maintain muscle function in ALS patients.

We have previously demonstrated a novel proof-of-concept strategy to overcome muscle denervation and restore control of muscle contraction in a nerve injury model of muscle paralysis that could have major therapeutic value for restoring function of any targeted muscle or group of muscles in ALS patients Bryson et al. (2014). Briefly, optogenetically-modified replacement motor neurons, derived from murine embryonic stem cells (mESCs), were engrafted into distal branches of peripheral nerves supplying specific lower hindlimb flexor and extensor target muscles in wild-type mice that had undergone a nerve ligation injury. Our results showed that the engrafted motor neurons were able to project axons from the graft site to the target muscles where they formed de novo NMJs. Due to the ectopic location of the engrafted motor neurons, outside the central nervous system (CNS), they do not receive endogenous motor signals, and must therefore be exogenously activated. Expression of the blue-light sensitive channelrhodopsin-2 (ChR2) protein Nagel et al. (2003) in the engrafted motor neurons conferred the ability to selectively activate these engrafted neurons and thereby control the contractile function of the target muscle using acute optical stimulation Bryson et al. (2014). The aim of this neural replacement strategy is therefore to provide a biological interface capable of rendering any target muscle receptive to control signals transmitted by optical stimulation to engrafted motor neurons Bryson et al. (2016). Importantly, we have recently developed a prototype 64-channel stimulation and recording device capable of controlling multiple independent intraneural graft sites that could be used to elicit coordinated function of large numbers of muscles, in order to restore useful motor functions Liu et al. (2022).

This novel approach to restore control of paralyzed muscles in ALS patients, using a combination of cell replacement and optical stimulation, has several key advantages over existing cell replacement and electrical stimulation strategies, including: i) the ability to engraft motor neurons peripherally, in close proximity to targeted muscles, which greatly accelerates the rate of reinnervation and reduces the period of denervation, consequently ameliorating denervation-induced muscle atrophy; ii) avoidance of engrafting replacement cells into the neurotoxic environment that exists within the CNS of ALS patients and the necessity for reinnervating axons of CNS-engrafted motor neurons to overcome the inhibitory CNS:PNS barrier in order to exit the CNS and grow the often long distances to target muscles; iii) specificity of optical stimulation to the engrafted ChR2⁺ motor neurons avoids painful off-target activation of sensory afferents or aberrant activation of endogenous motor axons associated with electrical nerve stimulation (ENS); and critically, iv) the ability to recruit motor units in correct physiological order using optical nerve stimulation (ONS) Llewellyn et al. (2010), avoids the problem of rapid muscle fatigue associated with ENS and incorrect, non-physiological motor unit recruitment. Furthermore, ENS-mediated control of muscle function depends on the presence of surviving motor axons and, since these are progressively lost during the course of disease progression in ALS, the ability of ENS to induce muscle contraction is steadily eroded. More importantly, it has recently been shown that ENS, applied to the phrenic nerve to assist respiratory function in two separate clinical trials in ALS patients McDermott et al. (2016); Gonzalez-Bermejo et al. (2016) accelerated diaphragm muscle denervation, which likely contributed to a significant reduction in patient life-span Guimarães-Costa et al. (2019). Therefore, it is unlikely that ENS will ever be suitable for artificial control of critical muscle function in ALS patients. In contrast, optogenetic stimulation Henderson et al. (2009), in combination with neural replacement, represents a safe alternative approach to artificially restore innervation and function of paralyzed muscles in ALS.

In the current study, we sought to optimize critical elements of this novel therapeutic strategy and to determine whether ChR2⁺ motor neurons can be used to successfully restore innervation, induce muscle contraction and prevent atrophy of targeted muscles in a model of ALS using the highly aggressive SOD1^G93A mouse model of ALS.

Results

Host-vs-graft rejection causes loss of most intraneural ESC-MNs allografts

ChR2⁺ motor neurons were derived from our previously characterized mouse embryonic stem cell (mESC) line Bryson et al. (2014) and differentiated using a well-established protocol Wichterle et al. (2002). Since these donor ChR2⁺ motor neurons were generated from an mESC line originating from the 129S1/SvImJ mouse strain (Supplementary Table 1) and recipient mice SOD1^G93A mice were on a congenic C57BL/6J genetic background, they constitute an allogeneic source of donor cells. Importantly, allogeneic cells are likely to provide a more cost-effective, off-the-shelf cell therapy platform, compared to generating individual, patient-specific batches of cells suitable for autologous engraftment. Therefore, an important initial objective of this study was to identify an immunosuppression regimen capable of preventing host-vs-graft rejection of allogeneic ChR2⁺ motor neurons to enable their innervation of target muscles in recipient SOD1^G93A mice.

In the absence of any immunosuppression, the survival rate of ChR2⁺ motor neuron grafts at 35 d post engraftment, was extremely low (<5%) in SOD1^G93A mice. However, in the rare cases where the engrafted ChR2⁺ motor neurons did survive, we observed robust intramuscular axon growth and partial reinnervation of NMJs, even at very late-stage disease (Figure 1A-C and Video 1). Furthermore, acute optical stimulation of the engrafted ChR2⁺ motor neurons was able to induce tetanic contraction of the target muscle in these SOD1^G93A mice (Figure 1D), although these were weak in magnitude. Interestingly, assessment of shorter timepoints revealed that engrafted ChR2⁺ motor neurons were capable of surviving for up to 14 days (data not shown). It is therefore likely that the poor long-term survival was due to host-vs-graft rejection of the engrafted cells, rather than disease-related toxicity. These findings suggest that if the engrafted motor neurons can evade the host immune response, they can functionally reinnervate target muscles for a therapeutically relevant timescale in this highly aggressive model of ALS.

Figure 1

Tacrolimus (FK506) overcomes ESC-MN allograft rejection but inhibits muscle innervation

In an effort to enhance graft survival, we first tested the calcineurin inhibitor (CNI), tacrolimus (FK506), which is not only routinely used to facilitate solid organ allograft survival in humans but has also been reported to promote axon regeneration after nerve injury in rats Gold et al. (1995). Cohorts of wild-type and SOD1^G93A mice (Supplementary Table 2) were treated with FK506 (5mg, kg^-1, d^-1; dose selected based on evidence of axon regeneration studies) from the time of intraneural ChR2⁺ motor neuron engraftment; graft survival and muscle reinnervation were then assessed 30-35d post-engraftment. Although FK506 did enable robust ChR2⁺ motor neuron allograft survival, at the graft site, in all animals examined (Supplementary Table 2), we identified three major problems with this immunosuppressant agent. Firstly, although the engrafted motor neurons were able to survive and project axons along peripheral nerve branches and within the targeted lower hindlimb muscles (Figure 2A), FK506 completely prevented muscle fibre reinnervation, evidenced by lack of response to acute ONS (data not shown) and complete absence of ChR2⁺/YFP⁺ NMJs (Figure 2B) in both SOD1^G93A (n = 8) and nerve-ligated wild-type mice (n = 4). Secondly, exuberant growth of carry-over pluripotent stem cells led to intraneural tumour formation in most mice (Figure 2C and Supplementary Table 2), which resulted in overt hindlimb motor deficits (Figure 2D and Video 2). Thirdly, in partial agreement with previous reports in rats Aydin et al. (2004), FK506 prevented body mass increase and/or induced body mass decline in a subset (44.4%; n = 8/18) of SOD1^G93A mice (Figure 2E and Figure 2-figure supplement 1A); this effect was less evident in wild-type mice (Figure 2-figure supplement 1B). Onset of body mass decline in B6.SOD1^G93A mice is highly consistent and typically occurs at 115 days in Hayworth and Gonzalez-Lima (2009), indicating that, at this relatively high dose, FK506 may be preferentially toxic or may exacerbate disease phenotype in SOD1^G93A mice.

Figure 2

Since peripheral neuropathy is a known adverse event associated with calcineurin inhibitors Arnold et al. (2013), it is possible that FK506 treatment alone may adversely affect endogenous or engrafted motor axons. Indeed, examination of the cross-sectional area distribution of total (ie. sensory and motor) and motor neuron axons in branches of the sciatic nerve in FK506-treated SOD1^G93A and WT mice (Figure 3A) revealed a significant loss of axons that affected most axonal calibres in the tibial nerve in wild-type and SOD1^G93A mice (Figure 3B and C). A more pronounced loss of total and motor axons, spanning medium to large sized axonal calibres, was observed in the common peroneal nerve branch (Figure 3D and E), which indicates that FK506 can not only exacerbate ongoing motor axon loss in SOD1^G93A mice but can also induce motor axon loss even in wild-type mice. Importantly, these neuropathy-like effects appear to be specific to FK-506, since an alternative immunosuppressant, H57-597 mAb (discussed in detail below), did not significantly alter total or motor axon size distribution or total axon counts (Figure 3-figure supplement 1), compared to untreated SOD1^G93A mice.

Figure 3

Since FK506 is known to suppress myoblast proliferation and differentiation Hong et al. (2002) and can cause rare cases of myopathy in humans Breil and Chariot (1999), it is also possible that the FK506-dependent prevention of muscle reinnervation by engrafted ChR2⁺ motor neurons is due to a muscle specific effect. In any case, these findings clearly show that FK506 is unsuitable as an immunosuppressant to support ChR2⁺ motor neuron allograft survival and, indeed, suggest that long-term administration at this relatively high dose should be avoided in ALS patients. Importantly, however, the complete protection of engrafted ChR2⁺ motor neurons by FK506 did confirm that host-vs-graft rejection was responsible for the poor long-term graft survival observed in the absence of immunosuppression, rather than disease related neurotoxicity.

T-cell modulatory immunosuppression confers graft survival and target muscle innervation

In light of the deleterious effects of FK506, and given our aim of conferring compatibility of allogeneic ChR2⁺ motor neurons as a universal cell therapy for ALS, we sought to identify a more specific form of immunosuppression that avoids the negative effects of FK506 yet supports long term graft survival. Therefore, we investigated the T-cell receptor-beta (TC-beta) targeting monoclonal antibody, mAb H57-597, which has previously been shown to effectively promote long-term heart allograft survival in mice Miyahara et al. (2012) In addition, since our findings indicated that immunosuppression results in a greater risk of tumour formation from carry-over pluripotent stem cells, differentiated ChR2⁺ motor neurons were also pre-treated with mitomycin-C (MMC; 2 μg/ml for 2 hrs) prior to engraftment, to eliminate tumorigenic cells Magown et al. (2016), to further enhance the translational potential. MMC-treated ChR2⁺ motor neurons were unilaterally engrafted into the tibial nerve of symptomatic SOD1^G93A mice (aged 95.7 ±4.6 days) in conjunction with transient H57-597 mAb delivery (1mg, kg-1; i.p. injection at 0, 1-, 3-, 7- and 14-days post-engraftment). The extent of reinnervation and the ability to optically control the function of the triceps surae (TS) muscle group in the lower hindlimb of SOD1^G93A mice was then assessed at late-stage disease (133 ±6.9 d; n = 12) by determining the physiological response of the reinnervated muscles to acute ONS of the engrafted motor neurons followed by histological analysis of muscle and nerve. Histological analysis confirmed that in mAb H57-597 treated animals, engrafted ChR2⁺ motor neurons were present in all recipient SOD1^G93A mice (n > 84 to date). Importantly, there was significant axonal projection within intramuscular nerve branches and robust reinnervation of muscle fibres in the targeted TS muscle (Figure 4A and 4B; Video 3 and 4). As we previously reported in a wild-type mouse nerve ligation model Bryson et al. (2014), some de novo NMJs in SOD1^G93A mice exhibited signs of immaturity, including polyinnervation (Figure 4C), as well as collateral and terminal sprouting of motor axons (Figure 4D). Since the peripherally-engrafted reinnervating motor neurons are inactive during the post-engraftment period and progressive muscle atrophy is ongoing, only 10.5% of endplates are innervated by engrafted ChR2⁺ motor neurons (Figure 4E). Importantly, these reinnervated endplates are functional, since acute in vivo ONS of the engrafted ChR2⁺ motor neurons in the exposed sciatic nerve of late-stage disease SOD1^G93A mice at 133 ±7.2 days of age (37.7 ±5.1 days post-engraftment; n = 12), induced positive contractile responses in all animals, although the amplitude of the maximal contractile force was still weak (0.8 ± 0.2 g; n = 11).

Figure 4

Motor neuron subtype identity does not affect response to optical stimulation

In an effort to increase the amplitude of the contractile response of the target muscle to optical stimulation we next tested whether engraftment of motor neurons with a fast-firing subtype identity may be more suitable than engraftment of predominantly slow-firing motor neurons by using alternative differentiation protocols.32 The stronger, predominantly fast-twitch, gastrocnemius component of the TS muscle group is usually innervated by fast-firing/fast-fatigable (FF) motor neurons, which are known to have the capacity to innervate many more individual muscle fibres per motor unit than slow-firing motor neurons, which normally innervate a much smaller number of weaker, slow-twitch muscle fibres, predominantly within the soleus and plantaris regions of the triceps surae. MMC-treated ChR2⁺ motor neurons, differentiated to yield FF subtype identity motor neurons, were engrafted into the tibial nerve of (106 ±7.2 days) SOD1^G93A mice, in combination with transient H57-597 mAb treatment. Maximum isometric muscle contraction of the TS muscle in response to acute optical stimuli was then determined at the same age (133.9 ±7.2 days, n = as previous grafts of predominantly slow-firing motor neurons (133 ±6.9 days, n = 11). This physiological analysis revealed that the motor neuron subtype identity did not significantly affect amplitude of the muscle response to acute optical stimulation (Figure 4-figure supplement 1A) and that the maximum contractile force elicited by ONS remained extremely weak compared to supra-maximal ENS (Figure 4-figure supplement 1B). This result implies that, unlike during normal neuromuscular development, motor neuron subtype identity is not an important determinant of muscle fibre innervation in the mixed fibre type triceps surae muscle. This finding has significance for future clinical translation, since only a single subtype of motor neuron may be required to innervate a variety of different muscles. Only MMC-treated motor neurons with a slow-firing medial motor column identity were used in the subsequent experiments reported here. Since modification of motor neuron subtype identity did not enhance contractile response, our next challenge was to identify an effective method to enhance reinnervation and force generating capacity of the targeted muscle in response to optical stimulation.

Optical stimulation training significantly enhances target muscle force generation

Spinal motor circuit development Milner and Landmesser (1999); Li et al. (2005) and NMJ formation/maintenance Sanes and Lichtman (1999) are known to be activity-dependent processes, thus, without regular stimulation, although peripherally engrafted ChR2⁺ motor neurons may survive, they are unlikely to form mature NMJs and will therefore have little effect on declining muscle function and atrophy in SOD1^G93A mice. Moreover, paralysis and atrophy of affected muscles proceed unchecked in the SOD1^G93A mouse model of ALS. We therefore investigated whether regular activation of the engrafted ChR2⁺ motor neurons, in conjunction with H57-597 mAb treatment, could enhance NMJ maturation and maximize the ability of target muscles to generate contractile force in response to acute optical stimulation in late-stage SOD1^G93A mice. To do this, we adapted a wireless, fully implantable optical stimulation system Montgomery et al. (2015), in order to implement a daily optical stimulation training regimen for engrafted mice. First, we modified the printed circuit board (PCB) design to simplify assembly (Figure 5-figure supplement 1A, implemented a new encapsulation method to ensure long term survival of the devices after implantation (Figure 5-figure supplement 1B) and incorporated an RF signal switch and pulse controller to deliver precisely timed RF pulses to power a 470nm light emitting diode (LED) connected to the implantable device (Figure 5-figure supplement 1C).

The modified optical stimulation devices were then surgically implanted in SOD1^G93A mice concomitantly with intraneural engraftment of ChR2⁺ motor neurons, with the trailing LED positioned in close apposition to the graft site (Figure 5A). Commencing at 14 d post-engraftment, to allow growing ChR2⁺ motor axons sufficient time to reach the target muscle, the mice were transferred to a custom built chamber located above a resonance-frequency induction cavity for 1 hr per day, in order to undertake optical stimulation training (OST; Figure 5B and Video 5), using a bespoke pulse pattern that was empirically determined to elicit maximum contraction (Figure 5-figure supplements 2, 3 and 4), followed by a 2 s rest interval. Following daily OST in engrafted SOD1^G93A mice for 21 days, isometric muscle tension physiology was performed at late-stage disease to determine the maximal contractile force of the TS muscle elicited by acute ONS of the exposed sciatic nerve graft site. In confirmation of our hypothesis, there was a highly significant, 9.4-fold, increase in the maximal tetanic force (7.5 ±0.94 g versus 0.8 ±0.2 g; p = ≤0.000001) elicited by ONS in the engrafted OST group of SOD1^G93A mice at late-stage disease, compared to age-matched untrained SOD1^G93A mice (132.4 ±6.8 days versus 133 ±6.9 days; n = 7 and 11, respectively; Figure 5C-G), although OST did not alter contractile rate characteristics (Figure 5-figure supplement 5A-C). Moreover, quantitative comparison of the maximum force elicited by ONS compared to ENS of TS muscles in late-stage SOD1^G93A mice (Figure 5H), showed that in mice that underwent OST, acute ONS elicits up to 22.7% (± 4.5) of total residual muscle force produced by supra-maximal ENS, which activates both endogenous and engrafted motor neurons (Figure 5I), in contrast to only 1.46% (± 0.18) in untrained SOD1^G93A mice (Figure 5-figure supplement 6A and 6B). This represents a remarkable >13-fold improvement in force generation. In engrafted SOD1^G93A mice that did not undergo OST, the extremely weak twitch contractions in response to ONS precluded interrogation of individual motor unit force values and determination of motor units number estimates (MUNE) in most mice. In contrast, in SOD1^G93A mice that underwent OST, the significantly increased contractile response to ONS enabled clear separation of individual motor unit values (Figure 5J), enabling MUNE values to be determined (Figure 5-figure supplement 5D). Furthermore, as we previously reported in nerve-ligated WT mice,22 delivery of repetitive ONS pulses (250 ms bursts of 20 Hz illumination, every 1 s, for 180 s duration) to induce rhythmic, submaximal contraction of the TS muscle did not induce rapid muscle fatigue, whereas equivalent pulses of ENS stimulation of the contralateral TS muscle did result in rapid muscle fatigue (Figure 5K). This observation has significant implications for the ability to control repetitive motor functions in ALS patients.

Figure 5

Optical stimulation training prevents atrophy of reinnervated muscle fibres

Finally, having established that optical stimulation training significantly enhances the maximal force elicited by optical stimulation of engrafted ChR2⁺ motor neurons in late-stage SOD1^G93A mice, we examined whether long-term optical stimulation training could also prevent atrophy of reinnervated muscle fibres. Since NMJs comprise an extremely small volume of the entire muscle, high-resolution 3D imaging of the entire muscle to determine muscle fibre innervation status and fibre diameter information is not feasible. Therefore, we developed a novel technique, termed digital Cross-sectional area Analysis from Longitudinal Muscle Sections (dCALMS), in order to assess these properties. This involved 3D reconstruction and analysis of regions of interest, obtained from 30μm thick longitudinal TS muscle sections (Figure 6A) from ChR2⁺ motor neuron engrafted, late-stage SOD1^G93A mice that had undergone OST. Each region contained at least one NMJ innervated by a ChR2⁺ motor neuron, along with randomly captured neighbouring fibres (Figure 6B, Figure 6-figure supplement 1 and Video 6). The 3D reconstructions were then digitally re-sliced in the transverse orientation (Figure 6C, top panel), in order to obtain data on muscle fibre cross-sectional area (CSA) using a semi-automated process (Figure 6C, lower panel). The dCALMS analysis revealed that the average CSA of muscle fibres innervated by engrafted ChR2⁺ motor neurons were similar in size to fibres still innervated by residual endogenous motor neurons (922.3 versus 1018.3 μm²; p = ≤0.85; Figure 6D). Importantly, the CSA of muscle fibres innervated by ChR2⁺ motor neurons was significantly greater than fibres with completely denervated endplates (average CSA = 525.4 μm2; p = ≤0.00001) or fibres whose innervation status could not be determined (average CSA = 668.8 μm2; p ≤0.0003), since the endplate was outside the scanned region of interest.

Figure 6

Discussion

This study shows for the first time that replacement stem cell-derived motor neurons can robustly and reliably reinnervate target muscles in the highly aggressive SOD1^G93A mouse model of ALS, even when engrafted after onset of overt symptoms; moreover, the restored innervation can be maintained even until extremely late-stage disease. Furthermore, our findings suggest that engrafted ChR2⁺ motor neurons can not only provide an interface to safely and selectively control the function of targeted muscle but, also, that regular optical stimulation training (OST) can be used to: i) reinforce connectivity between engrafted motor neurons and muscle fibres; ii) significantly enhance the maximal force elicited by optical stimulation of the targeted muscle; and iii) prevent atrophy of muscle fibres that are reinnervated by engrafted motor neurons. The highly significant improvements in muscle innervation, atrophy prevention and maximum contractile force, as a result of the daily OST regimen, confirms that stimulation-induced activity is necessary to maximize connectivity between engrafted motor neurons and their target muscles.

The prevailing view in the ALS field is that spinal motor neuron pathology first manifests at the nerve terminal and that muscle fibres are likely to actively contribute to degeneration of endogenous motor neurons in ALS Scaricamazza et al. (2021). Our findings clearly demonstrate that affected muscles, even in the highly aggressive SOD1^G93A ALS model, remain receptive to reinnervation by healthy engrafted motor neurons, even until late-stage disease. Moreover, once target muscles have been reinnervated, this approach enables implementation of a safe muscle training/exercise regimen that can be used to preserve muscle integrity and prevent irreversible muscle wasting that otherwise occurs as a result of progressive neurodegeneration in ALS Mora (1989). Since skeletal muscles are not simply biomechanical actuators, but have complex functions in overall metabolic homeostasis, thermoregulation, venous return and maintenance of blood volume, the ability to prevent muscle atrophy using OST, is likely, by itself, to have major health benefits for ALS patients. In contrast, the use of electrical nerve stimulation (ENS) to control muscle function or, indeed, ENS-based exercise programs is likely to accelerate degeneration of remaining motor axon terminals Guimarães-Costa et al. (2019), and is therefore unlikely to be safe. Unlike ENS, the highly selective nature of ONS does not activate or interfere with endogenous motor neuron function and due to its ability to recruit motor units in physiological order, ONS has the added significant benefit of avoiding rapid muscle fatigue Llewellyn et al. (2010). The ability to combat muscle atrophy, using OST, could extend the ability of targeted to execute functionally useful movements, potentially indefinitely in ALS patients.

Since differentiation methods that yield either fast-firing or slow-firing motor neurons did not appear to affect the ability of engrafted motor neurons to innervate the mixed fibre type TS muscle group in SOD1^G93A mice, motor neuron subtype identity appears to be redundant in this case. Therefore, a single type of motor neuron, produced at scale, could potentially be used to target a large number of different muscles in each patient. This has advantages in terms of simplifying the regulatory approval process, since the donor motor neurons would be produced in exactly the same way, irrespective of the graft site or recipient.

Indeed, another key finding of this study that could streamline future translation of this therapeutic strategy, is the identification of a TCR-β targeting antibody, H57-597 mAb, as an effective mediator of allograft survival. Existing T-cell targeting monoclonal antibodies, such as OKT-3, have been clinically approved Page et al. (2013) and, importantly, this form of immunosuppression overcomes severe adverse effects that we observed with the commonly used CNI-based immunosuppressant, tacrolimus (FK506). Our data shows that H57-597 mAb treatment was well tolerated during transient administration from symptom-onset up until late-stage disease in SOD1^G93A mice, however, the aggressive disease progression in this model precludes investigation of longer-term effects and it remains to be seen how long-term TCR-β-based immunosuppression may be tolerated in ALS patients. A similar immunosuppression regimen involving a CD25 (IL2) targeting monoclonal antibody, Basiliximab, along with glucocorticoid and tacrolimus maintenance therapy, has been shown to be safe but poorly tolerated in ALS patients, as an independent investigative approach Fournier et al. (2018), as well as part of a separate clinical trial assessing intraspinal grafts of neural precursor cells in ALS patients Mazzini et al. (2019). Therefore, immunosuppression per se, is not an impediment to this strategy.

The use of allogeneic donor cells, with a safe and effective immunosuppression regimen, means that a future cell therapy could potentially be universally compatible with all ALS patients, which would significantly reduce costs and simplify the regulatory approval process, compared to individually tailored autologous cell grafts. Of course, further studies will be required to ensure that human-compatible, induced pluripotent stem cell (iPSC)-derived donor motor neurons are able to function in the same manner as allogeneic mouse ESC-derived motor neurons. The generation of HLA-matched super donor hiPSC lines may further mitigate the need for immunosuppression Turner et al. (2013), however, immunogenicity of the ChR2 protein could mean that some form of immunosuppression may be necessary for any optogenetic therapy in the peripheral nervous system, including viral delivery Maimon et al. (2018).

The anatomical separation of specific nerve branches in humans means that this approach could be used to target and independently control large numbers of muscles in each individual ALS patient. However, it will be first necessary to demonstrate safety and efficacy in a relatively low risk muscle to restore a simple motor function. For example, the common peroneal nerve is highly accessible from a neurosurgical perspective and existing ENS devices, developed to correct foot drop for other neurological disorders Hausmann et al. (2015), could be readily adapted to assist ambulatory function in early-stage ALS patients. In the longer term, existing multichannel ENS devices, which have been developed to control more complex ADLs in high-level spinal cord injury (SCI) patients Memberg et al. (2014), could also be adapted into a minimally invasive, transcutaneous optical stimulation device Maimon et al. (2017). This combinatorial therapeutic strategy, comprising allogeneic donor cells, an effective immunosuppression regimen and optical stimulation device, is highly compatible with the rapidly evolving field of brain computer interface (BCI) technology. BCI could be used to decode a paralyzed patient’s intention to perform a given movement in order to control the activity of the engrafted motor neurons, which provide the necessary interface to execute the intended movement via a wearable optical stimulation device. This novel approach would entirely bypass the severe damage that occurs throughout the entire CNS in ALS patients and enable autologous control of movement. Moreover, this strategy also has broad utility for a wide range of other neurogenic causes of paralysis, such as spinal cord injury and stroke.

Although the findings of this study clearly demonstrate that our combinatorial cell therapy approach is effective in a highly aggressive mouse model of ALS, further investigation is required in order to confirm that the strategy can be applied to alternate model of ALS, with a longer lifespan, in order to fully explore the long-term efficacy of the approach, particularly in terms of chronic allograft survival using a transient immunosuppression approach. It is possible that some form of maintenance therapy may also be required to confer long term graft survival. Of course, the biggest challenge will be to demonstrate that human optogenetically-modified motor neurons, derived from either induced pluripotent stem cells (iPSCs) or human ESCs, are capable of reinnervating target muscles in the same manner as we have demonstrated for mESC-derived motor neurons. It will also be necessary to scale this cell therapy strategy up, using larger animal models that more accurately recapitulate human-scale anatomy.

Despite these remaining challenges, the findings of this study provide strong support for this novel cell therapy, which, if successful, could finally begin to deliver major health benefits for ALS patients.

Methods and Materials

Detailed methods are provided in the Supplementary material.

mESC motor neuron differentiation

The Channelrhodopsin2-YFP expressing mESCs (Clone C9G) used in this study were generated as previously describedBryson et al. (2014). Motor neuron differentiation was performed according to a standard protocol developed by Wichterle et al Wichterle et al. (2002) for production of predominantly slow-firing medial motor column (MMC) identity motor neurons and an updated “caudalized-ventralized” (CV) protocol, also developed by the Wichterle group Peljto et al. (2010), that produces a higher proportion of motor neurons with fast-firing properties. Following differentiation on Day 5 (or Day 7 for the C-V protocol), embryoid bodies (EBs) containing differentiated motor neurons were dissociated and total cells were resuspended in PBS at a concentration of 50,000 cell/μl. Where indicated, mitomycin-C (1μg/μl) was added to the EBs for 2 hrs prior to dissociation. Nile Blue A (0.0002% final concentration) was added to the cell suspension and the cells were kept on ice until engraftment. Single nucleotide polymorphism (SNP) analysis of mESC lines (Clone C9G, HBG3 and a C57BL/6J mESC line for reference) was carried out by Charles River Laboratories. See the Supplementary materials for full details 1.

Mice

All procedures and experiments involving animals were carried out under License from the UK Home Office in accordance with the Animals (Scientific Procedures) Act 1986 (Amended Regulations 2012), following ethical approval from the UCL Queen Square Institute of Neurology Animal Welfare Ethical Review Body (AWERB), and in accordance with the ARRIVE guidelines. B6.CgTg(SOD1*G93A)1Gur/J mice (The Jackson Laboratory, stock number 004435) were bred specifically for this study by mating presymptomatic male transgene carriers with congenic C57BL/6J females (Charles River Laboratories)1.

Intraneural engraftment of mESC-derived ChR2+ motor neurons

Surgical engraftment of ChR2⁺ motor neurons was performed under aseptic conditions, as previously described Bryson et al. (2014). Briefly, 1μl of dissociated EB cell suspension (50,000 cells) was injected into the tibial nerve close to the trifurcation point of the sciatic nerve, using a 5μl Hamilton syringe equipped with a customized 33G needle. Where indicated, an implantable optical stimulation device was inserted through the same surgical incision and positioned subcutaneously under the skin on the back; the trailing LED was fixed with sutures to the overlying muscles at the graft site during wound closure. Immunosuppression, as indicated, was initiated at the time of surgical engraftment. See supplementary materials for full details1.

Implantable optical stimulation devices and power transmission system

Optical stimulation devices were largely produced as described by Montgomery et al. Montgomery et al. (2015), with minor, but essential, modifications to the PCB design and encapsulation method. Similarly, the power transmission system used to activate the implanted LED devices was also largely as described by Montgomery et al. Montgomery et al. (2015), however, a Solid State Switch (MiniCircuits; ZX80-DR230-S+), controlled by a USB-TTL Interface (Prizmatix), was used to generate specific optical stimulation patterns. Engrafted SOD1^G93A mice that underwent daily optical stimulation training were placed in the stimulation chamber for 1 hr/day from post-engraftment day 14 until termination of the experiment at late-stage disease (132.4 ±6.8 days). See the Supplementary materials for full details1.

Isometric muscle tension physiological analysis

At the experimental end-point, engrafted SOD1^G93A mice underwent isometric muscle tension physiology, in order to accurately determine the contractile properties of the triceps surae muscle in response to acute optical stimulation, as previously described Bryson et al. (2014), with the following modifications: a PowerLab 4/30 stimulation and recording system (AD Instruments) was used to deliver bespoke electrical stimulation signals, either as direct constant voltage pulses applied to the nerve via bipolar electrodes for electrical nerve stimulation (ENS), or used as a 5V TTL signal to activate a 470nm LED light-source (CoolLED; pe100), delivered to the exposed sciatic nerve via a liquid lightguide for optical nerve stimulation (ONS); stimulation program available on request. LabChart software (AD Instruments) was used for automated data analysis of contractile parameters. See the Supplementary materials for full details1.

Nerve and muscle histology and image analysis

See Supplementary materials for full details, including automated motor/sensory axon CSA analysis1, axon counts and innervation analysis method1. The digital CSA analysis of longitudinal muscle sections (dCALMS) method is also described in full in the Supplementary materials1.

Quantification and statistical analysis

Sample sizes

The number of mice (n) is provided in the figures and/or figure legends; also see Supplementary Tables 2 and 3.

Statistical analysis

All data are presented as mean ± SEM unless otherwise indicated. GraphPad Prism 9 (Prism) was used for statistical analyses. No out-liners or data points were eliminated. Differences between two groups were assessed using multiple two-tailed unpaired t tests. Differences between more than two groups were assessed by using one-way or two-way analysis of variance (ANOVA) with multiple comparison correction, or mixed model effects analysis, as stated in the figure legends. Significance was defined as *P ≤0.05, **P ≤0.01, ***P ≤0.001, or ****P ≤0.0001. See supplementary Methods for further details 1.

Data availability

The data that support the findings of this study are available from the corresponding authors, upon reasonable request.

Acknowledgements

The authors are grateful to Profs Elizabeth Fisher, Jennifer Morgan (UCL) and Victor Tybulewicz (The Crick Institute) for helpful discussions about immuno-compatibility of allogeneic cells in mice and adverse effects of CNIs on muscle, and Dr Henry Lancashire for assistance with modified PCB production. In addition, we are grateful to Prof Ada Poon and Dr Yuji Tanabe (Stanford University) for assistance with establishing their wireless optical stimulation system.

Acknowledgements

JBB was supported by a Motor Neurone Disease Association Lady Edith Wolfson Senior Non-clinical Fellowship (Bryson 965-799), as well as an NIHR BRC UCL Excellence Fellowship award (BRC371/ED/AT/101310), in addition to grant support from the Rosetrees Trust (ref: M643) and an Early Career Researcher Award from the Richard Stravitz Foundation. This study was supported in part by the UK Medical Research Council (MRC) research project grant (MR/R011648/1), awarded to JBB, AD and LG (principal investigator) and a Thiery Latran Foundation project grant (JBB and LG). LG is supported by Brain Research UK and holds The Graham Watts Senior Research Fellowship (LG).

Appendix 1

Accompanying Files

3-D reconstruction of innervated endplates from SOD1^G93A mice in the absence of immunosuppression.

FK506 facilitates graft survival but allows intraneural tumour formation that causes severe motor dysfunction.

3D reconstruction of an entire triceps surae muscle group from a late-stage SOD1^G93A mouse, after ChR2⁺ motor neuron engraftment showing extent of reinnervation.

3D reconstruction of individual endplates (red) reinnervated by engrafted ChR2⁺ motor neuron (green) in a 135d SOD1^G93A mouse (35d post-engraftment) in combination with transient H57-597 mAb treatment.

Daily optical stimulation training significantly enhances elicited muscle force in SOD1^G93A mice.

3D visualization of longitudinal muscle section from an engrafted SOD1^G93A mouse along with “dCALMS” muscle fiber analysis technique.

Figure 2—figure supplement 1.

Figure 3—figure supplement 1.

Figure 4—figure supplement 1.

Figure 5—figure supplement 1.

Figure 5—figure supplement 2.

Figure 5—figure supplement 3.

Figure 5—figure supplement 4.

Figure 5—figure supplement 5.

Figure 5—figure supplement 6.

Figure 6—figure supplement 1.