Optogenetic Control of Muscles: Potential Uses and Limitations

Mechanism of muscle contraction

Skeletal, smooth, and cardiac muscle contractions rely on calcium signaling, which is initiated by various mechanisms. In skeletal muscles, action potentials are triggered by acetylcholine release from the motor neuron, which binds to nicotinic acetylcholine receptors on the muscle cell surface. ⁴ These receptors are cation-selective ion channels, and their opening allows sodium influx, initiating spreading depolarization along the cell. This triggers the release of calcium from the sarcoplasmic reticulum. Calcium binds to troponin C, exposing the myosin-binding site; the resulting myosin attachment forms a cross-bridge, which shortens the sarcomere. After muscle cell contraction, calcium is taken up into the sarcoplasmic reticulum until the actin-binding site is no longer exposed and cross-bridge formation stops allowing sarcomere relaxation. Fast and slow skeletal muscle fibers differ in their metabolic activity, allowing some muscles to be fatigue resistant but slower to respond, while other muscles can contract rapidly and powerfully, but fatigue after short periods of time.

Although skeletal muscles can rapidly fatigue with repeated use, possibly due to alterations in calcium signaling, which may disrupt excitation–contraction coupling, this is not the case for cardiac muscle. ^4,5 Cardiomyocytes are fatigue resistant and cannot form tetanic contractions. The endogenous pacemaker activity of the sinoatrial node is modulated by neuronal input, but depolarization spreads in a coordinated pattern, meaning individual cardiomyocytes do not require neuronal input to initiate contraction. Smooth muscle contraction requires less energy than skeletal muscle, making it resistant to fatigue. However, the varied innervation patterns of smooth muscle (found in large arteries and airways, as well as in the gut and bladder) makes it a particular challenge for indirect optogenetic activation, due to the need to coordinate both spatial and temporal activation.

Direct versus indirect optical stimulation

Indirect optical stimulation of muscles, by transducing the motor neuron with an optogenetic construct, has been used as an alternative to electrical stimulation. In animal models, this has been shown to restore diaphragm electromyographic (EMG) activity after spinal cord injury ⁶ and provide fine control over movement. ⁷ Closed-loop optical stimulation of the peripheral nerves in the lower limb was able to control the ankle joint position, for example. ⁸ Moreover, optical stimulation was able to produce tetanic contraction, or alternating contraction and relaxation, with substantially less fatigue than electrical stimulation, due to the physiological motor unit recruitment. During physiological contractions, smaller and less fatigable groups of fibers are recruited first by the motor nerve, with larger fibers being recruited with increasing stimulation. Electrical stimulation produces reverse-order recruitment, where large and fast-fatiguing groups of muscle fibers are activated first, because of their lower input resistance. This leads to much earlier fatigue during muscle contraction, a problem which is not apparent during optical stimulation. ⁹

On the other hand, transducing a nerve to produce indirect optical stimulation of a muscle may not have the required specificity, as it is challenging to transduce and stimulate only the target motor neuron(s), potentially leading to off-target stimulation of afferent nerves, for example. Stimulation of a whole motor neuron can also make precise spatially varying control of the innervated muscle challenging. Recent evidence also suggests that retrograde transmission of a vector can lead to immune-mediated neuronal degeneration in transduced axons starting at 8 weeks postinjection, ¹⁰ with anti-opsin antibodies appearing in the serum at 6 weeks, concomitant with a loss of light response. This could lead to safety concerns, as well as long-term reduction in effectiveness of neuronal stimulation.

Lastly, indirect stimulation is unsuitable where the target motor neuron is damaged or vulnerable to further damage. This covers many neuromuscular disorders, including motor neuron disease, nerve avulsion, and spinal cord injury. Conditions such as denervation atrophy and certain recurring cardiac arrhythmias can be treated with electrical stimulation from implanted devices but replacing this with direct optogenetic stimulation is potentially a safer and less damaging way of restoring muscle function.

Mechanism of direct optogenetically induced muscle contraction

Direct optogenetic stimulation of a muscle can be achieved by transducing the muscle fibers themselves. To ensure that only the targeted cells are transduced, the capsid and promoter must be chosen carefully; this will be discussed in later sections. The most commonly used opsin, channelrhodopsin-2 (ChR2), is a nonselective cation channel. When stimulated by blue light, ChR2 opens and allows the movement of positive ions across the cell membrane, activating intracellular signaling cascades to produce contraction of the muscle cells. ¹¹ Bruegmann et al. showed that optical stimulation can cause tetanic contractions in skeletal muscles, which are 84% as strong as those produced by electrical stimulation. ¹⁰ Additionally, ChR2 expression enabled light-induced EMG responses in the tongue muscle of mice, recapitulating the endogenous contraction, which is lost during obstructive sleep apnea. ¹² However, strong or prolonged ChR2 stimulation causes refractoriness to depolarization in the heart ¹³ and pH changes inside the cell in blood vessels, ⁵ possibly due to inactivation of sodium channels.

Allowing the cell to repolarize between action potentials creates a more physiological response and normal muscle cell contraction in skeletal muscles. ^10,14 Therefore, the stimulation protocol must be optimized for effective muscle contraction. For example, Bruegmann et al. ¹⁵ produced varying contractile responses in explanted transduced or transgenic larynges and soleus muscles. Continuous illumination for 2 s produced an initial contraction, which rapidly tapered off. Pulsed illumination produced sustained contractions, with the contractile force increasing with greater pulse frequencies until a saturation point (around 40 Hz using 10 ms pulses). This model shows that optical activation can be modulated to cause varying degrees of muscle cell responses.

Optogenetic versus electrical muscle stimulation

Direct optogenetic muscle stimulation has been shown to produce broadly similar muscle contractions to electrical stimulation.

Bruegmann et al. ¹⁰ showed that direct optical stimulation can cause tetanic contractions in skeletal muscles, which are 84% as strong as those produced by electrical stimulation. On the other hand, Magown et al. ¹¹ found optical stimulation of the triceps surae or soleus muscles appeared weaker than electrical stimulation of the innervating nerve. Equivalent frequencies of light pulses or electrical pulses, from 5 to 50 Hz, generated lower amplitude EMG (in millivolts, measured with surface electrodes) with longer durations (in milliseconds). They speculated that optical stimulation might cause depolarization in fewer fibers, and that this is more asynchronous than electrical stimulation, because depolarization occurs first in fibers closest to the light source. However, this apparent weakness could simply be a result of poor transduction—in Bruegmann et al., ¹⁵ virally transduced animals had much weaker contraction of the larynx compared with transgenic animals using the same optical stimulation paradigm, likely because only about 10% of the fibers were transduced by the virus, compared with all fibers in transgenic animals.

In this case, optical stimulation is not necessarily weaker, but it may stimulate fewer muscle cells. Transduction efficiency may be improved by intramuscular (IM) rather than intravenous (IV) administration, or by optimizing the capsid and construct. Other studies report that direct optical stimulation can generate as much force as electrical stimulation in vivo, or even greater force ex vivo in gastric smooth muscle, ¹⁶ without the concomitant fatigue.

Optogenetic stimulation also potentially has some key advantages over electrical stimulation. The improved cellular specificity of optogenetics affords a more physiological stimulation ¹⁶ and may allow normal motor unit recruitment. This is in contrast to the reverse-order recruitment seen with electrical stimulation, which is applied to a nerve controlling the muscle of interest. Direct optogenetic stimulation using muscle-specific promoters to restrict transgene expression to muscle cells may avoid unwanted stimulation of afferent fibers, making it more comfortable than electrical stimulation, which can stimulate both afferent and efferent fibers in a mixed nerve. Moreover, the amount of current needed to directly stimulate a muscle electrically can damage tissue and cause substantial discomfort. This is not the case for optical stimulation, ¹³ provided the light stimulation protocol is optimized. ⁵ Lastly, if ChR2 is expressed in skeletal muscles close to the skin, contraction can be produced by transdermal illumination, ¹⁴ obviating the need for implanted electrodes and their associated risks and complications.

Transgenic animals

Transgenic models are commonly used in proof-of-concept optogenetic studies. Since these models are created by direct genetic manipulation of a fertilized egg, they cannot be directly translated to human or clinical use, and are not discussed further in this review.

Viral delivery: AAVs

Adeno-associated viruses (AAVs) are commonly used as gene therapy vectors. The transgene, under the control of a promoter, is packaged into a viral capsid for delivery. In optogenetics, the transgene is an opsin—usually ChR2—and the capsid serotype is chosen for its preferential tropism to the target cell type; in this study, we focus on muscle cells.

AAVs are currently the only viral vector approved to deliver gene therapies in humans due to their safety profile. Dosing is a compromise between efficacy and toxicity. Typical doses range from 8 × 10¹⁰ copies per kg of body weight ¹⁷ to 6 × 10¹⁴ copies per kg ¹⁸ in preclinical models, and human and nonhuman primate (NHP) trials have used up to 10¹⁴ copies per kg. ^{3,19 –21} Most toxicity testing in animal models is encouraging, since adverse events are rare; however, animal mortality has been reported at 2.14 × 10¹⁰ copies per kg when spinally injected in rats, and patient deaths have occurred using 3 × 10¹⁴ copies per kg. ²²

Targeting AAV delivery to muscles

Opsins

The first and most commonly used microbial opsin family are the channelrhodopsins. As noted above, channelrhodopsins are cation channels, and illumination results in an influx of positively charged ions initiating depolarization. ² Conversely, illumination of opsins within the bacteriorhodopsin and halorhodopsin families, outward proton pumps and inward chloride pumps, respectively, results in hyperpolarization. Numerous microbial opsins have been discovered and engineered to have specific characteristics, including varied kinetics, ³⁸ shifted absorption/action spectra, and increased light sensitivity and photocurrent amplitudes. ³⁹ Less frequently used, particularly in muscle applications, are opsins derived from eukaryotes. These are light-activated G protein-coupled receptors (GPCRs), traditionally responsible for animal vision and circadian rhythm modulation. One example is the hOPN5 human neuropsin, an endogenous opsin found in the mammalian brain and eye. hOPN5 is a GPCR, and when stimulated by UV light, it instigates a signaling cascade, which increases intracellular calcium.

When stimulated, transgenic mouse hearts increased the rate of spontaneous beating, while in the smooth muscle cells of the small intestine, light stimulation generated force increases of varying degrees depending on the stimulation protocol. ⁴⁰ In transgenic mouse intestines, this increased contractility, and in whole hearts, increased the rate of spontaneous beating.

Largely due to their simplicity and ease of engineering, research in muscle-based optogenetics more commonly uses microbial opsins. In this study, we focus on opsins used in muscle cells and the translation from preclinical to clinical applications.

The most common opsins in direct optogenetic control of muscle are ChR2 and ChR2(H134R), a ChR2 variant with a single point gain-of-function mutation at position H134R. ² This mutation enables greater channel conductivity and increased photocurrents. Both are blue light sensitive, that is, they are activated by ∼470 nm wavelengths, and have induced contractions in rodent skeletal muscle, cardiac muscle, ²³ and smooth muscle. ⁴¹ The use of opsins sensitive to red light, such as ChrimsonR, ReaChR, and ChRmine, ⁴² is rare in muscle optogenetics, but will likely increase, following trends observed in neural applications, and success in in silico studies. ²⁴ ChrimsonR is the only opsin used in humans, but not in muscles (partial restoration of visual function was achieved following illumination of ChrimsonR-expressing retinal ganglion cells). ⁴³ Opsins with different spectral sensitivity provide two primary benefits. First, spectrally distinct opsins, that is, blue light- and red light-sensitive opsins, can be simultaneously expressed in a single muscle or neighboring muscles, and independently activated.

In rodents, spectrally distinct depolarizing opsins ChR2(H134R) and CsChrimson (red light sensitive) have been used for indirect optogenetic control of agonists and antagonist muscles innervated by a single mixed peripheral nerve. ⁴⁴ Spectrally distinct depolarizing opsins and hyperpolarizing opsins expressed in a single muscle or functionally synchronized muscles, may enable fine control of both contraction and relaxation. Spectrally distinct depolarizing (blue light-sensitive ChR2) and hyperpolarizing opsins (yellow light-sensitive NpHR, a halorhodopsin), have been suggested for synchronized contraction and relaxation of the detrusor muscles and the urethral sphincter, respectively, for control of urination in the context of bladder dysfunction. ⁴¹

Different wavelengths of light penetrate tissues to different extents. With the use of spectrally distinct opsins, it may be possible to independently control different muscle regions. All visible light wavelengths exhibit reductions in intensity at increased tissue depths, however, the degree of attenuation is greater with shorter wavelengths (e.g., blue light, ∼430–500 nm) versus longer wavelengths (e.g., red light, ∼630–710 nm). As such, opsins activated by longer wavelengths and lower light intensities are less susceptible to reduced efficacy at greater tissue depths. They will likely be vital for clinical applications of optogenetics, where light must penetrate tissues several orders of magnitude larger compared with rodents. ²⁴

For many applications of optogenetics in muscle, fast opsin kinetics are ideal (see Table 1 for a summary of some opsin kinetics). Ultra-fast opsins such as ChETA ³⁸ and Chronos that quickly activate with illumination onset, experience minimal desensitization during illumination, and recover quickly allow rapid, repetitive, and maximal depolarization or hyperpolarization, and subsequent muscle contraction and relaxation, respectively. In neurons, the fast on/off kinetics of ChETA and Chronos provide more rapid repolarization, stronger and more sustained spikes, and greater spike synchronicity ⁴⁵ ; this allows higher stimulation rates compared with ChR2. In muscles, increased stimulation frequency would allow a faster and more sustained recruitment of fibers, leading to tetanic contractions.

Opsins with exceedingly slow kinetics, known as step function opsins (SFOs), may also aid optogenetics by reducing light intensity and energy requirements. Extremely slow off-kinetics results in ion channels remaining open for over 30 min after blue light stimulation, or until the opsin is deactivated with orange light.

SFO activation can induce stable and subthreshold depolarization that sensitizes cells to excitation and potentially allows disease-attenuated activity to be effective. These opsins can be coexpressed in muscle fibers with spectrally distinct “standard” activating opsins, for example, ChrimsonR. The elevated “resting threshold” of the targeted cells reduces light intensity and power requirements for the subsequent activation of the “standard” opsin. New SFOs are being developed with ultra-high light sensitivity and could therefore contribute to the efficacy of direct optogenetic activation of deep and/or large volumes of muscle, since greater light sensitivity means less light would be required to penetrate deep tissues.

Inhibitory opsins have also been developed, which can prevent or attenuate action potentials in electrically excitable cells. Although these opsins were developed primarily for neural mapping, they have applications for muscle-directed therapies, particularly in the heart. Proton pump rhodopsins have been used for this purpose, but channelrhodopsins are faster and require less light for a similar effect.

GtACR is a light-activated anion channel with a preference for chloride. The effect of GtACR on cardiomyocytes depends on the cell's membrane potential and the balance between intracellular and extracellular chloride. When GtACR is first activated, chloride ions flow into the cell, depolarizing it. Sustained or increasing illumination leads to sustained depolarization, preventing repolarization and further action potentials. ⁴⁶ In fruit fly hearts, the heart rate could be increased with lower intensity illumination, then stopped entirely with increased illumination. ⁴⁷ In cultured rat cardiomyocytes, activation of GtACR decreases action potential amplitude, which could be used to treat long-QT syndrome in the heart. ⁴⁸

Another chloride-conducting channelrhodopsin, iLMO4, immediately stopped contractility in engineered heart tissue grown from transduced guinea pig cardiomyocytes. ⁴⁹ Potential applications of this could relate to the control of some abnormal heart rhythms involving overexcitation, but chloride-conducting channels must be used with extreme care, since they have paradoxical effects depending on how much chloride the cell contains at any given time.

Light-activated potassium channels can also be used to hyperpolarize cells, leading to electrical “deactivation.” Since neurons and muscle cells maintain a high intracellular concentration of potassium, opening these channels leads to a rapid outflow of potassium, reducing the membrane voltage and preventing action potential formation. In cultured rabbit cardiomyocytes transduced with the light-activated potassium channel SthK, a single light pulse prevents electrically evoked action potentials for several minutes. ⁵⁰ Another opsin called WiChR has an even stronger preference for potassium over sodium, and in stem cell-derived cardiomyocytes, light activation inhibited the spontaneous contraction normally seen in these cells by completely blocking action potentials. ⁵¹ Potassium conductance is a more tractable mechanism for “switching off” unwanted electrical activity in muscles, although this technique requires further development.

Neutralizing antibodies and AAV-based gene therapy

Circulating antibodies develop within hours of exposure to AAVs. They can limit successful transduction when AAV-mediated gene therapy is delivered intravenously or intramuscularly. Since AAV is a naturally occurring virus, human populations have high rates of anti-AAV antibodies resulting from prior infection with high serotype crossreactivity. ²⁷ Clinical trials often exclude subjects with a total anti-AAV antibody titer of 1:50 or more, even though these antibodies may or may not be capable of neutralizing the vector. ⁶⁰ Administration of an AAV-based gene therapy to a previously seronegative patient will likely result in the development of antibodies, which may preclude them from receiving similar therapies in the future. ⁶¹ In the context of optogenetic control of muscles, the immunogenicity of the AAV capsid can interfere with successful transduction, and the opsin itself can be targeted by the immune system, resulting in anti-ChR2 antibody formation. ⁸

Although antibodies have long been assumed to be the main reason for therapeutic failure, the presence or development of antibodies does not necessarily correspond with a loss of expression or a lack of functional improvement. ¹⁰ This could be because existing antibodies are not capable of neutralizing the capsid or transgene, or are not present at a sufficient titer to clear the vector. Alternately, transduction failure can be mediated by other components of the immune system. After IM injection of an AAV containing a human FIX (HFIX) transgene to correct hemophilia B in mice, Nayak et al. ⁵⁹ showed that anti-hFIX antibodies did not affect serum hFIX levels but did reduce coagulation times, indicating that antibody interference with therapeutic efficacy can occur without affecting transgene expression. This presumably occurs through antibody-mediated removal of a secreted protein, as in the case of hFIX, but may also occur with membrane-bound proteins like channelrhodopsin.

In NHPs, femoral artery administration of an AAVrh74 vector carrying GALGT2 or μDystrophin produced initially strong gene and protein expression in the gastrocnemius, which then declined substantially between 12 and 24 weeks. ¹⁹ Interestingly, the use of prednisone in some groups partially rescued the RNA expression and protein levels of the transgene, without affecting DNA levels. This demonstrates that viral entry into host cells does not necessarily lead to therapeutic success, and that suppression of key immune pathways may enhance treatment efficacy. Prednisone suppresses leukocyte migration, with T cells being a likely cellular component to mediate this effect; this will be discussed in a later section.

In a clinical trial using AAV1 to deliver α-1 antitrypsin intramuscularly, all subjects developed anti-AAV1 antibodies over 3 months. ³ Higher doses produced greater increases in serum AAT, but there was no clear relationship between serum AAT, preexisting antibodies, and anti-capsid antibody development. Kuranda et al. ⁶¹ used ex vivo human blood to investigate immune responses to AAVs. Human blood exhibited an innate immune system response to AAV capsids, characterized by interferon-gamma (IFNγ) secretion from natural killer cells in seronegatives and tumor necrosis factor alpha secretion from capsid-specific memory CD8+ (cytotoxic) T cells in seropositives. These innate responses may be why AAV-based gene therapy can fail even in seronegative individuals. AAV2 capsids also triggered memory B cell differentiation in seropositive donors, indicating that both humoral and cellular immune responses occur after AAV administration.

Lastly, anti-IL-1β and anti-IL6 antibodies were administered to mice to prevent antibody development, on the basis of allowing readministration of an AAV vector. Three weeks after AAV8 administration, the treated animals had lower anti-AAV8 antibody levels—but the degree of transduction was unaffected. Clearly, therapeutic failure involves factors beyond antibody development.

Beyond antibodies: Cell-mediated responses to AAV-based gene therapy

While circulating antibodies develop rapidly, the T cell response to a novel antigen develops more slowly, and is more technically challenging to measure accurately. T cells clear virally infected cells from tissues by secreting perforin, granzyme, and IFNγ after recognizing elements of the capsid proteins or the viral genome. Posttransduction T cell activity is another obstacle for AAV-based gene therapy. Depending on their differentiation status, T cells can help or hinder gene therapies; cytotoxic T cells destroy and remove transduced cells, ⁶⁰ while regulatory T cells (Tregs), which normally limit autoimmune responses, may promote vector tolerance. ²⁰ This means any immune suppression treatments must be administered cautiously, targeting specific pathways to prevent therapeutic rejection and promote transgene persistence.

Cramer et al. ¹⁹ investigated T cell and antibody responses to an AAVrh74 vector carrying GALGT2 or μDystrophin given IM or IV in NHPs. After 24 weeks, IFNγ responses to the capsid or transgene were negative in almost all animals, although this extended time point may have failed to capture an earlier response. The use of prednisone in some animals enhanced transgene expression at the RNA and protein levels, possibly due to its inhibition of programmed cell death protein 1 (PD1). PD1 expression, often stimulated by viral infection, can lead to T cell exhaustion, a state characterized by poor effector function. The authors claim that preventing T cell exhaustion with prednisone administration was a key mechanism behind transgene persistence. However, Gernoux et al. ⁶² and Flotte et al. ³ argue that T cell exhaustion can improve capsid tolerance. In NHPs and humans, the various types of T cell display unique cell surface markers, allowing specific identification of Tregs, which are activated, exhausted, or both.

Based on this, previously seronegative NHPs receiving IM delivery of AAV1-rhAAT developed muscle infiltration of exhausted T cells after 21 days, and exhausted activated T cells after 60 days; at the 60-day time point, all animals had increased serum AAT. ⁶² This indicates that modulating T cell activation status could improve transgene persistence in the muscle.

In an early clinical trial, patients given AAV1-AAT IM exhibited an increase then plateau in serum AAT, concomitant with the development of both antibody and T cell responses. ³

A later trial further investigated T cell responses to IM-delivered AAV1-based gene therapy for AAT deficiency. ²⁰ In three patients (two previously seronegative and one seropositive), both serum AAT and anti-capsid IFNγ peaked, declined, then plateaued—despite the development of high-titer anti-capsid antibodies. Viral DNA was rapidly cleared from the blood and persisted in the muscle, while intact AAV1 capsids and T cells were present in muscle biopsies at 3 and 12 months postadministration. Despite the immune activity in these patients, protein expression of the transgene was maintained, although the patients' serum AAT was still well below a normal level. This study indicates that a gene therapy, which escapes initial contact with the humoral immune system—aided by IM administration rather than IV—may persist long term, although this has not been consistently found in other human trials or animal models.

On the other hand, Mendell et al. ³⁴ showed variable and transient IFNγ responses in human patients after IM injection of an AAV1 vector carrying α-sarcoglycan with a tMCK promoter. An early immune response, comprising both T cell and antibody responses, seemed to correspond with a loss of long-term transgene expression. However, these responses were highly variable, making it difficult to draw strong conclusions.

Along with targeting specific immune pathways, the timing of any immune therapy as an adjunct to AAV-based gene therapy must be carefully considered. Several studies have identified a pattern of expression loss around 5 weeks posttreatment. ^3,17 Based on this, Samelson-Jones et al. ⁶³ investigated T cell and antibody responses using AAV2-hFIX gene therapy in NHPs, given intraarterially. The immune suppressant drugs, rapamycin and mycophenolate mofetil (MMF), were given to all animals, then three received early antithymocyte globulin (ATG) at the time of vector administration, while the other three received late ATG at 5 weeks postadministration. In both groups, anti-AAV and antitransgene antibodies were suppressed until the withdrawal of immune suppressant drugs at 8 weeks. At this point, animals in the early ATG group developed antibodies against AAV and against hFIX, while only one animal in the late ATG group developed anti-hFIX antibodies. hFIX protein levels were also sustained longer in the late ATG group.

The authors state that early ATG contributed to antibody development, but do not address the possibility that antibody development may have occurred regardless of ATG and was only suppressed by MMF and rapamycin during the period of administration. However, they also state that immune suppression can actually exacerbate therapeutic rejection, something other groups have previously noted. ⁶⁴ Clearly, timing of immune suppression is crucial, and both antibodies and T cells must be studied together, since they communicate with each other and conspire to prevent transduction.

Immune suppression

Many studies, including those discussed in previous sections, use short-term immune suppression to enhance AAV-mediated gene therapy, with variable results. Although many of these studies do not specifically investigate muscle expression or delivery, their results highlight the potential usefulness of immune suppression in delivering AAV-based gene therapies, and the complexities involved, which make it challenging to predict the outcome of such therapies.

Steroids, such as prednisone and dexamethasone, inhibit T cell proliferation and cytokine production by preventing phosphorylation of signaling molecules used by the T cell receptors. Other drugs used to prevent transplant organ rejection have also been trialed for this purpose. However, these drugs may improve or worsen transgene persistence, ⁶⁴ either overall or specifically in muscle. For example, prednisone reduces cytotoxic T cell infiltration of muscle after IV AAV injection in NHPs, ¹⁹ but does not affect IM-delivered AAV-mediated transgene expression in mouse muscle. ²⁷ In some models, immune suppression delays therapeutic failure without preventing it entirely.

Meliani et al. ⁶⁵ showed promising results for IV rapamycin nanoparticles in completely preventing neutralizing antibody formation in mice. The strength of this approach is in suppressing antigen-specific responses, by coadministrating the immune suppressant with the vector. A single dose of rapamycin-carrying nanoparticles was given with an IV “inoculation” of an AAV8 vector carrying a reporter gene. Twenty-one days later, another dose of rapamycin was given with an AAV8-hFIX vector intravenously. Rapamycin slightly increased activated Treg cells in the liver, indicating promotion of vector tolerance, while reducing a subclass of B cells mediating long-term antibody production. The coadministration of rapamycin with the first vector allowed the second vector, carrying the hFIX transgene, to produce hFIX serum levels equivalent to those in mice receiving only one dose of the hFIX vector. These results were replicated in NHPs.

Additionally, Nayak et al. ⁵⁹ showed that a “tolerizing” treatment of rapamycin and IL-10 prevented the formation of inhibitory antibodies against a hFIX transgene (given intramuscularly in hemophilia B mice), if the exogenous hFIX protein was given concurrently.

The authors conclude that this treatment promoted a shift from cytotoxic T cells to Treg cells, which improved functional outcomes. Although these studies did not investigate muscle-specific responses, it demonstrates that the right immune suppression protocol could facilitate readministration of a vector, a goal which has eluded the field thus far.

Immune suppression to enhance AAV-based gene therapy is limited by the temporary nature of the drug effects; long-term immune suppression carries severe side effects and risks, which are unsuitable in many patient populations. Additionally, some immunosuppressive drugs interact with each other in undesirable ways. ⁵⁹ Ultimately, a combination of immune modulation and vector optimization may be needed for successful long-term AAV-based gene therapies.

Hardware

A major consideration for successful optogenetic activation of muscle is light delivery to the target tissue. Animal research started with “tethered” hardware, using external light sources connected to optical fibers surgically implanted in rodents. Naturalistic behavior was impossible and tissue damage was common, due to the stiffness mismatch between hardware and surrounding soft tissues. Early “wireless” hardware used bulky, externally mounted stages bonded to bony structures, so animal movement was restricted. Advances in material science, electrical engineering, and microfabrication techniques, optogenetics technology have enabled miniaturized, battery-free, minimally invasive hardware, using materials mechanically compliant with the surrounding tissue. This hardware can be wireless and fully implantable and in animals, allows for largely naturalistic behavior.

Despite these recent advances in technology for animal experiments, the technology required for human optogenetics remains in its infancy. Key issues include scaling of technology from rodents to humans and the low tissue penetration of visible light through larger human muscles. Optogenetics in humans is currently restricted to clinical trials for retinitis pigmentosa, where visual function is partially restored. ⁴³ Unlike all but the most superficial skeletal muscle cells, retinal ganglion cells are optically accessible, enabling stimulation of the retinal ganglion cells with light projected by supplementary goggles. For light delivery to most skeletal muscles, optogenetic control will need either implanted light sources or external sources delivering light at wavelengths that can penetrate into deep tissues, such as red light that is paired with red-shifted opsins.

Additionally, while indirect optogenetic stimulation of muscle may be able to adapt existing technology used in electrical stimulation of peripheral nerves, by substituting electrical for light stimuli, direct muscle optogenetics will likely require significantly greater technological innovation.

Clearly, the optical stimulator design will depend on the therapeutic application as well as the location of the muscle to be stimulated. Some superficial muscle may be able to be stimulated transcutaneously using wearable devices, but deep muscles will likely require surgically implanted light sources. Options include LEDs near or within the targeted muscle, or remote light sources connected to optical fibers situated in or near the target muscle. These need to be integrated with any required sensing and control hardware, plus a power source. For instance, implanted stretchable strain gauges encircling the bladder to identify pathological voiding behavior have been proposed to control the delivery of optogenetic therapy of incontinence, ⁶⁶ whereas external inertial measurement units monitoring ankle joint velocity and angle, could be used as part of a “closed-loop” system for controlling foot and ankle position to prevent foot drop. ⁸

Both transcutaneous and implanted stimulation approaches face the challenge of illuminating muscles whose volumes are several orders of magnitude greater than rodents, where most research has been done to date. Single LEDs or optical fibers will likely be insufficient for all but the smallest of muscles. LED arrays, or optical fibers with splitters, multiple mirrored exposed points, or tapered ends are potential solutions to maximize the field of illumination. However, these methods come at the cost of optical power, as each light emitting site will reflect only a portion of the power generated by a single primary light source. The use of multiple light sources, on the other hand, increases power requirements and subsequent risk of tissue damage.

Another approach that has been explored is the use of embedded nanomaterials that convert wavelengths such as focused ultrasound or near-infrared (NIR) light to visible wavelengths. These could be combined with implanted or transcutaneous devices for direct optogenetic activation of muscle. Lanthanide-doped upconversion nanoparticles (UCNPs) locally convert NIR light into the required wavelengths of light for opsin activation. A small UCNP device implanted in the spinal cord evoked hind limb EMG activity in ChR2-expressing rodents with transdermal pulsed NIR stimulation. ⁶⁷ Alternatively, zinc sulfide nanoparticles codoped with trace amounts of silver and cobalt can be injected into the blood stream, charged with 400 nm light when passing through superficial vessels, and subsequently triggered with ultrasound to emit blue light in deep tissues to stimulate ChR2. ⁶⁸

In summary, while there remain many challenges in developing technology for clinical applications of direct optogenetic muscle control, the outlook is positive and developments are progressing rapidly. Successful, clinically acceptable technology must address the specific needs to each therapeutic application, and this requires careful consideration of the anatomy (target muscle size, location, and type), stimulation requirements (light penetration, intensity, temporal profile, and duty cycle), and control elements (sensors, control systems). Altogether, these will define the power requirements and possible methods of powering the system (battery, inductive power, etc.). Such systems must also be safe and comfortable for patients to use and meet all the required regulatory standards for medical devices.

Any AAV-mediated gene therapy will involve the limitations discussed above. Some specific limitations exist for optogenetic therapies. Along with the challenges discussed above, the opsin itself, being membrane bound, presents an immunogenic risk in sites, which are not immune privileged, including muscle tissue. ¹⁰ Additionally, concerns exist regarding potential inactivation of the opsin channel due to repeated use. ⁶⁹ This is difficult to study in vivo since a “wearing off” effect is commonly ascribed to a loss of transduction, and because preclinical studies using repeated stimulation generally investigate early time points. Lastly, opsin selection must be informed by the kinetic profile. Fast on/off times can facilitate rapid repolarization, and red-light responsiveness can improve tissue penetration, but some opsins (e.g., ChRmine) have a lower calcium ion conductance than others (e.g., CatCh). Each opsin must be tested in vivo for the relevant experimental paradigm.

Controlling cardiac function

Certain types of cardiac arrhythmia can be treated with implantable electrical pacemakers or defibrillators. Pacemakers deliver small and targeted impulses to one or more chambers of the heart, usually without discomfort. Although these devices are well tolerated, an optogenetic approach could reduce the power requirements, enabling wireless devices with smaller batteries. ⁷⁰ Cardioverters and defibrillators deliver strong electrical shocks, either synchronized to the cardiac cycle (cardioversion) or delivered at any point in the cardiac cycle (defibrillation). These electrical discharges cause significant pain and myocardial damage. An optogenetic strategy could produce a return to sinus rhythm while avoiding incidental stimulation of efferent nerves and surrounding muscles. Recent preclinical work provides proof of concept for optogenetic pacing and/or defibrillation, normalizing the electrical activity of the heart without damaging the tissue. In primary cells from transgenic ChR2-expressing mice, optical stimulation increased the rate of spontaneous beating. ¹³

In a two-dimensional layer modeling a syncytium, illuminating any local area triggered spreading depolarization to neighboring cells, similar to the pacemaker function of the sinoatrial node. This work demonstrated the feasibility of an optogenetic pacemaker.

Transgenic models provide proof of concept, but optogenetic pacemakers in humans would require gene transfer. Vogt et al. demonstrated this approach in mice, using AAV delivery of ChR2 through the jugular vein. ²³ This transduced just over half the cardiomyocytes, with minimal off-target expression, at 4–10 weeks after transduction. Subsequent optical pacing by illumination of the ventricular epicardium was successful in almost three-quarters of animals. Nonresponsive animals were found to have insufficient transduction in postmortem imaging. Variability in transduction efficiency is a major limitation of AAV-based gene therapy but optogenetic control of the cardiac cycle could form part of clinical treatments in the future. Nussinovitch and Gepstein ⁷⁰ successfully paced rat hearts transduced with AAV9-carrying ChR2 under a CAG promoter using ventricular injection; the cardiomyocytes were transduced at the injection site after 2 weeks.

Ventricular illumination increased sinus rhythm in an open chest model, and biventricular pacing was successful in resynchronizing explanted Langendorff hearts during simulated left bundle branch block. Existing cardiac resynchronization therapy devices have a limited number of pacing sites; optogenetic pacing could be a superior system with the right hardware.

Optogenetic defibrillators could be safer than electrical defibrillators for patients who are prone to certain types of arrhythmia. An optical platform was developed in transgenic mouse hearts, successfully mapping the cardiac conduction pathway and achieving defibrillation or cardioversion with a high degree of spatial control and much lower energy requirements compared with electrical systems. ⁷¹ Bruegmann et al. created an optogenetic defibrillator to simultaneously depolarize a large amount of ventricular tissue, using both transgenic and AAV-transduced mice expressing ChR2 in cardiomyocytes. ²⁴ Successful defibrillation of experimentally induced ventricular arrhythmias was demonstrated for up to 1 year after gene transfer. Subsequent experiments had similar success with optogenetic cardioversion, ⁷² where depolarization was synchronized to the cardiac cycle to encourage a return to sinus rhythm. Additionally, AAV-transduced mice expressed ChR2 in 8–43% of cardiomyocytes 6–8 months after transduction.

Light was applied directly to the ventricles in ex vivo preparations, or through the pericardium in vivo. Cardioversion of atrial fibrillation was successful in about three-quarters of AAV-transduced animals and all transgenic animals, compared with spontaneous recovery rates of 43% and 9%, respectively. Notably, if fewer than 17% of cardiomyocytes had been transduced by the AAV vector, cardioversion was unsuccessful; this highlights the importance of strong and reliable expression.

A similar study delivered the red-shifted opsin ReaChR to the heart in rats using AAV9, under the control of a promoter encoding cardiac troponin T. ⁷³ This transduced 93% of cardiomyocytes at 4–6 weeks after IV delivery. Ventricular tachycardias were electrically induced in Langendorff perfused hearts, then light delivery successfully terminated 57–97% of these abnormal rhythms, depending on the subtype. The authors noted that a return to sinus rhythm after light delivery was preceded by prolonged action potential duration. This has been noted elsewhere and may be mechanistically relevant in the termination of some arrhythmias. ⁷⁴

Recently, a series of proof-of-concept experiments showed the feasibility of combining implantable cardiac devices with optogenetics. Ausra et al. ⁷⁵ developed a recording and light delivery device in transgenic mice. The device could accurately detect the heart rate in freely moving animals and provided optical coverage of the entire heart to a depth of up to 3 mm. Pacing was also achieved in vivo, and light emitted from the device increased the amplitudes of the QRS complex in these animals. Additionally, Nyns et al. ⁷⁶ developed an optical cardioversion system using implanted LEDs and cardiac cycle monitor in the rat heart. The right atrium of adult rats was transduced with AAV9-carrying ReaChR, with a promoter specific for atrial cardiomyocytes called natriuretic peptide A, using a gene painting method. Transduction was effective in 78% of the targeted cells 4 weeks later. After electrically induced atrial fibrillation, the automated detection system triggered the LED and terminated the abnormal rhythm in 96% of episodes.

Taken together, these studies show the potential for optogenetics to replace electrical stimulation of the heart in applications where electrical stimulation is currently used, including pacing and cardioversion. Optical stimulation could be safer and more specific than electrical stimulation, although an implantable device to deliver the light would still be required. Gene transfer methods are sufficiently flexible to enable targeting of single chambers or the entire heart, and can allow depolarization or hyperpolarization, depending on the clinical need. ⁷⁷ However, transduction failure is unpredictable even in animal models, and although rodent hearts appear to sustain long-term transduction—possibly because of the low rate of cell turnover in the heart—sustained transgene expression is harder to achieve in skeletal muscle.

Restoring motor function and control

Optical stimulation could theoretically be used in any muscle with insufficient or inappropriate contraction, using excitatory or inhibitory opsins, respectively. Direct optical stimulation of a muscle to prevent atrophy could replace electrical stimulation in cases where the nerve supplying a muscle is diseased or damaged. Magown et al. ¹¹ demonstrated this by expressing ChR2 in the skeletal muscles of transgenic mice, successfully stimulating contraction of denervated triceps surae muscles in vivo by transdermal light application, and attenuating denervation atrophy. Although direct transdermal light stimulation of muscles in larger animals, including humans, is more technically challenging due to the required depth of light penetration, optimization of the gene therapy could develop this into a clinical treatment.

Direct muscle stimulation with optogenetics could also be combined with implantable devices to produce or enhance muscle contraction in neuromuscular disorders. For example, electrical stimulation of paralyzed laryngeal muscles is being developed but this approach has side effects, including the unwanted concurrent stimulation of muscles used to open and close the vocal cords. Bruegmann et al. ¹⁵ trialed optical stimulation in both transgenic and AAV-transduced animals. In explanted larynges, opening or closing of the vocal cords was achieved by targeting light pulses to the posterior or superior cricoarytenoid muscles, respectively, potentially providing finer control that was possible with electrical stimulation. Recently, AAV-transduced mouse diaphragms expressing ChR2 exhibited light-triggered contractions, which restored respiration after transection of the phrenic nerves. ⁷⁸

Optogenetic targeting of smooth muscle poses additional technical challenges, since smooth muscle contraction involves opening L-type voltage-gated calcium channels, which require longer and larger depolarizations than skeletal muscle. Vogt et al. ¹⁶ applied optogenetics to a transgenic mouse model of gastroparesis, in which delayed gastric emptying follows damage to the vagus nerve. In primary gastric smooth muscle cells, patch clamp experiments confirmed that light triggered depolarization and contraction. In ex vivo strips of stomach muscle, light caused muscle contraction, which was stronger than that produced by electrical stimulation. In whole explanted stomachs, illumination of the entire organ increased intragastric pressure and food propulsion.

A limitation identified in this study involves the unwanted transduction of the vascular smooth muscle cells of gastric arterioles, potentially compromising blood supply to the stomach. Theoretically, stimulation parameters could be varied to achieve gastric emptying without affecting gastric arterioles, or a construct may one day be developed to transduce the phasic smooth muscle cells of the stomach without transducing the tonic smooth muscle cells of the vessels, but neither approach has been demonstrated to date.