Spatial and temporal variability in response to hybrid electro-optical stimulation

Abstract

Hybrid electro-optical neural stimulation is a novel paradigm combining the advantages of optical and electrical stimulation techniques while reducing their respective limitations. However, in order to fulfill its promise, this technique requires reduced variability and improved reproducibility. Here we used a comparative physiological approach to aid the further development of this technique by identifying the spatial and temporal factors characteristic of hybrid stimulation that may contribute to experimental variability and/or a lack of reproducibility. Using transient pulses of infrared light delivered simultaneously with a bipolar electrical stimulus in either the marine mollusk Aplysia californica buccal nerve or the rat sciatic nerve, we determined the existence of a finite region of excitability with size altered by the strength of the optical stimulus and recruitment dictated by the polarity of the electrical stimulus. Hybrid stimulation radiant exposures yielding 50% probability of firing (RE₅₀) were shown to be negatively correlated with the underlying changes in electrical stimulation threshold over time. In Aplysia, but not in the rat sciatic nerve, increasing optical radiant exposures (J cm⁻²) beyond the RE₅₀ ultimately resulted in inhibition of evoked potentials. Accounting for the sources of variability identified in this study increased the reproducibility of stimulation from 35% to 93% in Aplysia and 23% to 76% in the rat with reduced variability.

1. Introduction

Traditional electrical techniques have long served as the method of choice for neural activation and monitoring, offering clinical efficacy and safety with well-characterized and modifiable parameters. Despite widespread successful application of electrical stimulation in both the research and clinical arenas, electrical stimulation is fundamentally limited by the unwanted spread of current away from the stimulation site. Recent innovations in electrical stimulation design have achieved fascicular and sub-fascicular selectivity. The flat interface nerve electrode (FINE) gently reshapes the nerve to target electric fields, and can achieve fascicular selectivity in stimulation and recording with negligible changes to nerve morphology [1, 2]. However, the perineurium surrounding the individual fascicles has high impedance, resulting in a uniform voltage distribution and preventing sub-fascicular selectivity. Intra-fascicular electrode arrays such as the Utah slanted electrode array (USEA) and polymer longitudinal intrafascicular electrode (polyLIFE) penetrate the perineurium to allow direct electrical contact within individual fascicles [3, 4]. While these intra-fascicular arrays achieve remarkable selectivity, there are concerns regarding the long-term safety of these approaches.

Many researchers are now developing optical methods of neural activation to circumvent the limitations of traditional electrical techniques and/or to complement the benefits of this well-established approach. The ability to modulate and even evoke action potentials in neurons using light has been known for decades [5–8], but recently this method of neural stimulation has become much more promising with the discovery of higher resolution stimulation via caged compounds [9], optogenetics [10] and infrared neural stimulation (INS) [11]. Whereas stimulation of caged compounds and microbial or plant opsins inherently requires exogenous additives and/or engineered neurons, INS is a neural stimulation modality in which pulsed infrared light will generate a propagating action potential within an endogenous neural system. With INS, there is also the lack of current spread, stimulation artifact and necessity of contact with the neural tissue that limits electrical techniques [12]. However, there is a restricted range of radiant exposure (J cm⁻²), H, that will safely stimulate without the risk of thermally induced tissue damage [13].

Recently, hybrid neural stimulation was developed as a new stimulation modality combining traditional electrical techniques with novel infrared nerve stimulation methods [14]. The combination of the two techniques utilizes their respective advantages while avoiding their primary limitations. Specifically, hybrid stimulation combines the safety, established characteristics and demonstrated clinical utility of electrical stimulation with the spatial selectivity of INS. While hybrid stimulation does not provide the contact- and artifact-free aspects of INS, the high spatial selectivity of INS remains and will enhance clinical neural interfaces. Additionally, sub-threshold electrical currents should also reduce the problem of electrode corrosion over time. The essence of hybrid stimulation is to combine a sub-threshold electrical stimulus over a broad area, and then bring a spatially selective location to threshold by adding a sub-threshold pulse of infrared light. In doing so, both the electrical current and optical radiant exposures are reduced, effectively achieving spatial selectivity with reduced risk of tissue damage. Previously, hybrid stimulation was shown to reduce optical radiant exposures (J cm⁻²) by approximately a factor of 3 when compared to INS alone [14]. By offering reduced threshold radiant exposures, hybrid nerve stimulation is attractive for biomedical applications requiring spatial selectivity where laser power constraints and tissue damage are primary concerns. However, further development of this technology will require that the reliability and repeatability of hybrid stimulation be improved.

The experiments demonstrating feasibility of hybrid stimulation in the rat sciatic nerve showed large variations in the reduction of optical radiant exposures [14]. In these experiments, the electrical threshold was set at a chosen sub-threshold current and the additional optical radiant exposure required to achieve stimulation threshold was determined as a percent of the optical threshold radiant exposure when it was applied alone. The reduction in optical radiant exposures and their variability were both shown to increase as the applied electrical stimulus approached threshold. For an electrical stimulus at 95% of the threshold current, the additional optical energy required for stimulation ranged from 6% to 60% of the optical stimulation threshold. In addition, our unpublished data show a lack of reproducibility with hybrid stimulation from animal to animal. Initial attempts at hybrid stimulation were successful in 23% of rat sciatic nerves and 35% of Aplysia californica buccal nerves (unpublished data).

The objective of this study was to identify common factors that play a role in and may be controlled to enhance the reproducibility of hybrid electro-optical stimulation. Using this methodology, we will identify relevant sources of variability in an experimentally tractable and relatively simple neurobiological system. Then we will test these variability sources in a more clinically relevant model, where the complexity of the neural system may obscure their detection. Accordingly, the experimental procedures may differ slightly between the two model neural systems; however, the purpose of this study is to analyze and assess the overarching trends rather than the minor differences in stimulation protocols. To accomplish these goals, our choices of neural systems are the buccal nerve of the invertebrate marine mollusk Aplysia californica and the sciatic nerve of the vertebrate mammal Rattus norvegicus (rat). The Aplysia buccal ganglion provides a tractable, robust nervous system with large identified neurons and relatively few axons per nerve [15, 16]. These advantages facilitate the systematic empirical exploration of potential factors underlying the reproducibility of hybrid stimulation. The myelinated rat sciatic nerve is a more clinically relevant model for hybrid stimulation, but it is less robust than Aplysia nerves, and the fundamental interaction between the optical and electrical stimuli is confounded by the presence of myelin and a less stable nerve preparation. Therefore, we identify and characterize factors contributing to the reproducibility of hybrid stimulation in the Aplysia buccal nerve and then evaluate those factors in the rat sciatic nerve to determine whether similar trends are observed. In this study, we will investigate both spatial and temporal factors that may be controlled to reduce variability and enhance reproducibility.

There are two aspects of the spatial component that we address: (1) the relative locations of the optical and electrical stimuli and (2) the size of the excitable region as a function of the optical stimulus strength. The mechanism of INS was shown to involve a thermal gradient [17]. Thus, it is assumed that the thermal gradient and the electrical current path must overlap spatially. However, what is not known is where this overlap may occur, or how the two fields may affect each other. The activating function, which describes the transmembrane potentials leading to the electrical activation of a neuron, results in neurons closest to the cathode being activated first (with larger axons recruited before smaller axons) [18, 19]. Experimentally, stimulation threshold current is shown to increase with increasing distance from the cathode [20]. Given that the electrical stimulus preferentially targets neurons nearest the cathode, we hypothesize that hybrid stimulation will require the lowest optical pulse energies when the optical stimulus is located along the electrical current path and adjacent to the cathode. Like electrical stimulation, increasing INS radiant exposures results in an increase in magnitude of the evoked response, suggesting recruitment of additional axons [21]. Therefore, we expect that for a given sub-threshold electrical stimulus, an increase in the sub-threshold optical stimulus will yield an increase in the size of the excitable region for hybrid stimulation.

It has long been known that electrical stimulation thresholds vary over time [22]. In examining temporal factors, we seek to evaluate how brief fluctuations (minutes) and long-term trends (minutes to hours) in electrical stimulation thresholds affect optical pulse energies for hybrid stimulation. Correct measures of optical energies for hybrid stimulation require an accurate determination of the electrical ‘priming’ stimulus at the time of the measurement. If one incorrectly assumes that the electrical stimulation threshold is stationary over a fixed period of time, then hybrid stimulation performance will suffer. To address this issue, we measure threshold optical energies for hybrid stimulation while monitoring electrical thresholds over an extended period of time. We hypothesize that if the electrical threshold is known at any point in time, then the additional optical energy required for stimulation can be predicted for a given sub-threshold stimulus. Additionally, we believe that changes in threshold radiant exposures for the optical component of hybrid stimulation will be positively correlated with the changes in the underlying electrical stimulation threshold.

2. Materials and methods

2.1. Aplysia californica preparation and electrophysiology

Aplysia californica (n = 26) weighing 190–250 g (Marinus Scientific, Newport Beach, CA) were maintained in an aerated aquarium containing circulating artificial seawater (ASW) (Instant Ocean; Aquarium Systems, Mentor, OH) kept at 16–17 °C. The animals were fed dried seaweed every 1–3 days.

Aplysia were anesthetized with an injection of 333 mM MgCl₂ (50% of body weight) prior to dissection. Once anesthetized, animals were dissected and the buccal ganglia were removed and pinned in a recording dish and immersed in Aplysia saline (460 mM NaCl, 10 mM KCl, 22 mM MgCl₂, 33 mM MgSO₄, 10 mM CaCl₂, 10 mM glucose, 10 mM HEPES, pH 7.6). Once dissected and pinned, Aplysia nerves were left untreated so as not to reduce spontaneous activity. We chose not to discard data from trials where spontaneous activity occurred, as excitability varies with the level of activity. This is an inherent biological factor that we wanted to assess in our study. For each experiment, the nerve of interest (either buccal nerve 2 (BN2) or buccal nerve 3 (BN3)) was anchored in place by pinning the protective sheath around the nerve to the Sylgard base (Dow Corning, Midland, MI) of the recording dish. Once securely pinned, the nerve to be investigated was suctioned into a nerve-recording electrode to monitor the response to stimulation (figure 1(a)). Nerve suction recording electrodes were made by hand-pulling polyethylene tubing (1.27 mm outer diameter; PE90; Becton Dickinson) over a flame to the desired thickness. Recording electrodes were suction-filled with Aplysia saline prior to suctioning of the nerve. Nerve signals were amplified (×1000) and band-pass filtered (300–500 Hz) using an ac-coupled differential amplifier (model 1700; A-M Systems), digitized (Axon Digidata 1440A; Molecular Devices, Sunnyvale, CA) and recorded (Axograph X; Axograph Scientific).

Figure 1.

Experimental setups used for the (a) Aplysia californica buccal nerve (50×) and (b) rat sciatic nerve (20×) experiments in this study. RN = radular nerve; CBC = cerebrobuccal connective; BN3 = buccal nerve 3; BN2 = buccal nerve 2; BN1 = buccal nerve 1; EN esophageal nerve.

2.2. Rat preparation and electrophysiology

All rat experiments were performed following protocols approved by the Institutional Animal Care and Use Committee. Female Sprague-Dawley rats (n = 9) weighing 150–200 g (Charles River) were anesthetized with continuously inhaled isoflurane (induction: 3% isoflurane, 2.0 LPM oxygen; maintenance: 2–2.5% isoflurane, 1.5 LPM oxygen). A rectal probe and heating pad (catalog 40–90-8, FHC, Bowdoin, ME) were used to maintain the rat at a target body temperature of 35–37 °C throughout the experiment. The lateral sides of the animals’ back legs were shaved and the sciatic nerve exposed proximal to the knee via an incision in the overlying muscle. The muscular fascia over the nerve was removed while the nerve’s epineurial layer was left intact. Saline was added periodically to keep the nerve from dehydrating throughout the experiment. A custom Sylgard platform was anchored to a micromanipulator and placed below the sciatic nerve with minimal added tension to minimize motion of the nerve due to the animal’s respiration (figure 1(b)).

Evoked muscle action potentials were recorded using paired needle electrodes inserted in the areas of the biceps femoris and gastrocnemius muscles. EMG signals were amplified (×1000), band-pass filtered (300–1000 Hz), digitized and acquired using the same setup as for Aplysia.

2.3. Endpoint definition

Analysis of hybrid stimulation requires an appropriately defined endpoint. In Aplysia, we defined our endpoint as the visible detection of single and/or compound extracellular nerve spikes in response to stimulation (figure 2(a)). Similarly, the endpoint for rat experiments was visibly identified single and/or compound muscle action potentials in response to stimulation (figure 2(b)). For both species, we also required that the evoked potentials were frequency locked with the repeating stimulus (i.e. constant delay following a presented stimulus pulse) to distinguish evoked responses from spontaneous activity.

Figure 2.

Evaluation of output of system. To evaluate electrical, optical and hybrid stimulation, we looked for the presence of single and/or compound extracellular nerve potentials in the Aplysia californica buccal nerve and single and/or compound muscle potentials in the innervated muscles of the rat sciatic nerve. A representative recording from (a) the Aplysia californica buccal nerve and (b) the innervated muscle (biceps femoris) of the rat sciatic nerve.

2.4. Electrical and optical stimulation

Extracellular stimulating electrodes were made from thin-wall borosilicate capillary glass (catalogue 615000; A-M Systems, Everett, WA) pulled to resistances of about 0.2Ω (PP-830; Narishige). For each Aplysia experiment, two electrodes were capillary filled with Aplysia saline and placed on either side of the nerve in contact with the nerve sheath. This created a bipolar stimulus, with the pipettes oriented transverse to the longitudinal axis of the nerve. Pipettes were positioned such that their angle of approach to the nerve was as shallow as was allowed by the edge of the recording dish. For rat experiments, two glass pipettes were filled with normal saline and placed in contact with the nerve along the nerve’s longitudinal axis. The stimulating pipette arrangement for each species was chosen based on consistency of stimulation thresholds and ability to achieve reliable supra-threshold stimulation on each nerve tested. Monophasic currents were supplied by a bipolar stimulus isolator (A365R; WPI) and passed between the two pipettes in each preparation. Electrical stimulation was defined as the minimal current that would yield five consecutive evoked potentials in response to pulsed stimuli.

For optical stimulation, both a holmium:yttrium-aluminum-garnet (Ho:YAG) solid state laser (SEO Laser 1–2-3, Schwartz Electro-Optics, Orlando, FL) and a tunable pulsed diode laser were used (Capella; Lockheed-Martin-Aculight, Bothwell, WA). We chose two different lasers due to the established performance in peripheral nerves offered by the Ho:YAG and the ease of use and INS-specific design of the Capella. While the Capella was used in our previous demonstration of hybrid nerve stimulation, the Ho:YAG is the laser of choice for much of the INS literature pertaining to peripheral mammalian nerves [11–13, 17, 23, 24]. However, the Capella offers vastly improved ease of use and greatly reduced pulse-to-pulse variability when compared with the Ho:YAG. The Capella is also known to work exceptionally well for INS in a wide array of excitable tissues including the cochlea, somatosensory cortex, embryonic heart, cardiomyocytes and the vestibular system [25–29]. While the Ho:YAG provides pulses of infrared light (λ = 2.12 μm) having fixed pulse duration (τ_p = 0.25 ms), the Capella has slightly tunable wavelength (λ = 1.855–1.875 μm) and a variable pulse duration. The important parameter for INS is penetration depth in tissue (as pulse duration was shown to have negligible effects [17]); therefore, we set the Capella to have a wavelength of λ = 1.875 μm for all experiments to match the absorption (i.e. penetration depth) of the Ho:YAG laser [30].

For Aplysia experiments, laser output was coupled into either a flat-polished 100 or 200 μm diameter optical fiber (Ocean Optics, Dunedin, FL). For each experiment, the tip of the optical fiber was immersed in the Aplysia saline bath and brought into contact with the nerve sheath. The optical fiber was then slowly retracted with a micromanipulator and gently translated back and forth transverse to the nerve until the optical fiber was just out of contact with the nerve sheath. For radiant exposures presented in this study, the laser-irradiated area is assumed to be a circular spot on the incident surface of the nerve sheath having diameter equal to that of the optical fiber (i.e. 0.0314 mm² for a 200 μm fiber and 0.00785 mm² for a 100 μm fiber). For simplicity, as the optical fiber is just out of contact with the nerve sheath, this assumes no divergence of the beam from the tip of the optical fiber to incident surface of the nerve sheath. For rat experiments, laser output was coupled into a flat-polished 400 μm diameter optical fiber (Ocean Optics, Dunedin, FL). The fiber diameter for rat experiments was chosen to match the 400–600 μm optical fibers used in mammalian peripheral nerve studies, while smaller fibers were used in Aplysia studies to scale with the size of the Aplysia buccal nerves [14, 23, 24]. The optical fiber was positioned 500 μm from the incident surface of the nerve at an angle just off of vertical with a layer of saline just covering the surface of the nerve. The laser-spot size was measured using the knife-edge technique where two perpendicular measurements were taken along the axes of the presumed circularly shaped laser spot, yielding an irradiated area of 0.19 mm² [31]. Pyroelectric energy detectors were used to measure pulse energies from the tip of the optical fiber for the Ho:YAG laser (J25, Coherent-Molectron Inc., Santa Clara, CA) and Capella laser (PE50BB-SH-V2, Ophir Optronics Ltd).

For INS alone, optical stimulation threshold was defined as the minimum radiant exposure that would yield five consecutive evoked potentials in response to pulsed stimuli. In the Aplysia buccal nerve, using the Capella laser coupled to a 200 μm optical fiber that was retracted just out of contact with the nerve, threshold radiant exposures averaged 8.93 J cm⁻² with a 95% confidence interval of 8.72–9.14 J cm⁻² (25 measurements from 7 nerves). In the rat sciatic nerve, using the Ho:YAG laser coupled to a 400 μm optical fiber, threshold radiant exposures averaged 1.12 J cm⁻² with a 95% confidence interval of 0.92–1.32 J cm⁻² (12 measurements from 8 nerves).

Previous published studies found threshold radiant exposures in mammalian peripheral nerves ranging from 0.32 to 1.77 J cm⁻² [11–14, 17, 23, 24]. However, directly comparing these values with published data is difficult. Ongoing studies in our lab show stimulation thresholds in the rat sciatic nerve from 0.7 to 1.3 J cm⁻² (unpublished). In the cochlea, stimulation thresholds are on the order of mJ cm⁻² [32]. To make direct comparisons, it is imperative that certain factors be controlled; in particular, spot-size determination and measures of threshold must be the same. Radiant exposures are highly dependent on the spot-size. Differences in the way spot-sizes are calculated or measured between studies propagate into large differences in reported radiant exposures (due to the squared term in the denominator). In addition to variations in experimental preparations (i.e. neural model system, in vivo, ex vivo or in situ), thresholds may vary based on the definition of the endpoint for a given study, for example, whether the threshold is defined by the appearance of muscle or nerve action potentials, or by a visibly identified muscle twitch [12, 14, 32]. A noteworthy aspect of this study is that no visible damage or loss of function (as indicated by the response to electrical stimulation) was noted as a result of stimulation with the radiant exposures used. This is particularly relevant to Aplysia, where optical- and hybrid-evoked potentials remained steady over several hours of stimulation (not shown).

All nerve stimulation was coordinated through computer software (AxoGraph X; AxoGraph Scientific, Sydney, Australia) and applied at a repetition rate of 2 Hz. In both preparations, electrical pulses of 100 μs were used. Optical pulse durations were 250 μs for the Ho:YAG and 2–3 ms for the Capella lasers, respectively. This is due to the fixed pulse duration of the Ho:YAG and the minimum pulse duration of the Capella required to achieve optical energies for stimulation. Since the underlying mechanism of INS has been shown to be thermally mediated and dependent on a temperature gradient [17], as long as the pulse duration is significantly shorter than the thermal diffusion time (~100 ms), the laser pulse can be considered as an input delta function to the system. For hybrid stimulation, pulses were synchronized such that they ended concurrently. This allowed for the total charge and total thermal deposition to occur simultaneously. Nerve recordings were triggered and acquired for 10 ms prior to stimulation through 140 ms post stimulation.

2.5. Experimental methods for spatial factors

To investigate spatial factors contributing to the reproducibility of hybrid stimulation, sub-threshold pulses of electrical current (90% of electrical stimulation threshold) were applied simultaneously with optical pulses of a set magnitude. During hybrid stimulation, the optical fiber was translated across the nerve between the stimulating pipettes using a micromanipulator (see supplementary movie available at stacks.iop.org/JNE/9/036003/mmedia). A CMOS color USB camera and accompanying software (catalog 59–367; Edmund Optics, Barrington, NJ) were used to record the position of the optical fiber. A LED was triggered by computer software to flash synchronously with the laser pulse so that we could reconstruct the exact position of the optical fiber at the time of stimulation. The center of the tip of the optical fiber was plotted and correlated with the presence or absence of stimulation as indicated by an evoked potential on the nerve recording.

2.6. Experimental methods for temporal factors

Temporal factors were examined by investigating how fluctuations in electrical stimulation threshold over time affect the optical component of hybrid stimulation. Threshold currents were measured every 2–3 min for 1–3 h to monitor underlying changes in electrical stimulation with time and to assure that hybrid stimulation was not inducing alterations in threshold currents. One hour of each trial was an experimental period where radiant exposures eliciting hybrid stimulation were measured along with electrical stimulation threshold currents. Every 2–3 min during this experimental period, electrical stimulation threshold currents were first measured and then the stimulus current was reduced to 90% of electrical stimulation threshold. For the Aplysia experiments, five pulses of five different radiant exposures were then systematically applied with the sub-threshold current pulses. For the rat experiments, eight pulses of five different radiant exposures were applied. The order in which the radiant exposures were applied was determined by a random sequence generator so as to limit any conditioning effects or bias. Each hybrid stimulus pulse was recorded as either a 1 or 0 as determined by the presence (1) or absence (0) of a visibly identified nerve (Aplysia) or muscle (rat) action potential. This process was repeated every 2–3 min for the duration of the experimental period.

2.7. Data analysis

For spatial data, movie files were analyzed with custom software (Matlab r2010b; Mathworks, Natick, MA). Locations of successful stimulation were compared using non-parametric statistical tests. The two-sample Kolmogorov–Smirnov test compares two empirical distributions and responds to both the overall shape and location of the distributions. While this test will indicate if the distributions are statistically different, it will not tell whether it is due to the relative size or location of the distributions. To distinguish whether differences are due to changes in size or location of the region of excitability (ROE), we also performed the Mann–Whitney test, which is a non-parametric test that determines if the median of one data set is greater than another. The interquartile range was used as a measure of the size of the ROE.

Temporal data were aggregated using Matlab with statistical analysis performed in Microsoft Excel (part of Microsoft Office Professional Plus 2010) and Slide Write Plus Version 6 (Advanced Graphics Software, Inc., Encinitas, CA). For each radiant exposure, the number of ones was divided by the sum of ones and zeros to achieve a probability of firing. The cumulative distribution function (CDF) of the standard normal distribution,

$$F\left(x;\mu ,{\sigma }^2\right)=\frac{1}{2}\left[1+\mathrm{e}\mathrm{r}\mathrm{f}\left(\frac{x-\mu }{\sigma \sqrt{2}}\right)\right],x\in \mathfrak{R},$$ (1)

where $x$ is a random variable with mean $\mu$ and variance ${\sigma }^2$, was then fitted to the data to determine the radiant exposure yielding 50% probability of firing (RE₅₀). While the RE₅₀ is not practically useful for stimulation, we use this approach as a generally well-accepted model for making comparisons and identifying thresholds [29, 33–36]. We seek to establish a methodology and identify pertinent considerations for successful hybrid stimulation rather than prescribe optimal conditions for stimulation.

3. Results

3.1. Existence of a bounded excitable region

When translating the optical fiber back and forth across the nerve, it was determined that there exists a finite region between the cathode and anode where hybrid stimulation is possible (figure 3). This was observed in all of the nerves tested for both Aplysia (n = 42) and the rat (n = 13). However, in two rat sciatic nerves, some experimental trials yielded locations of successful hybrid stimulation extending outside of this finite region. During these trials, the electrical stimulation threshold was more variable. Occasionally the electrical component of hybrid stimulation approached electrical stimulation threshold, raising the overall excitability of the nerve. For both Aplysia and the rat, there were variations in the size and shape of evoked responses between animals, nerves and locations within a single nerve. This suggests that multiple different axons were recruited over the course of the experiments. In each species, there were ROEs consisting of only a single evoked unit and others that exhibited different units depending on the location of the optical fiber and the intensity of the optical stimulus. No apparent differences in ROE were observed when comparing the Capella and Ho:YAG within a single nerve (figures 4(a) and (b)) or across animals (figures 4(c) and (d)) for Aplysia or the rat. However, our yield with the Ho:YAG in the rat sciatic nerve was greater due to more reliable optical stimulation. With no obvious differences between the lasers other than overall yield, we placed greater emphasis on the Capella for the remaining Aplysia experiments (due to its ease of use and consistent pulse energies) and the Ho:YAG for the rat (due to the superior results it provided for myelinated nerve fibers).

Figure 3.

(a) A finite ROE exists between the cathode and anode where the combination of sub-threshold electrical and optical stimuli will achieve neural activation in an Aplysia nerve. Outside of this ROE, stimulation does not occur. (b) Evoked electrical response to hybrid stimulation recorded from the distal nerve. (c) Absence of evoked response outside of ROE. Hybrid stimulus parameters used: 675 μA (100 μs), 4.58 J cm⁻² (3 ms). Electrical stimulation threshold was 750 μA. In (b) and (c), the LED and electrical stimulation artifacts are indicated by the shaded region.

Figure 4.

A finite ROE exists between the cathode and anode where the combination of sub-threshold electrical and optical stimuli will achieve neural activation. ROEs for the Capella and Ho:YAG within the same Aplysia nerve are shown in (a) and (b), respectively. Typical ROEs observed in the rat sciatic nerve are shown for the Capella (c) and Ho:YAG (d).

3.2. Size of the ROE

After identifying the existence of a finite ROE, we investigated how the strength of the optical stimulus altered its size. With electrical current at 90% of electrical stimulation threshold, we compared ROE size for optical stimuli of 1.78 and 4.71 J cm⁻² using the Capella in Aplysia and 0.29–1.18 J cm⁻² with both the Ho:YAG and Capella lasers in the rat. These values were chosen to cover a range of optical radiant exposures that, in the absence of the electrical stimulus, are sub-threshold for stimulation in their respective neural systems. Locations of hybrid stimulation were binned and plotted as a probability histogram by dividing the number of stimuli evoking a response by the total number of attempts for each bin (figures 5(a), (b), (d) and (e)). After confirming that the ROE median was the same for each radiant exposure (using the Mann–Whitney test), the two-sample Kolmogorov–Smirnov test was applied to determine if the sizes of the distributions were significantly different.

Figure 5.

ROE size as a function of radiant exposure in the buccal nerve of Aplysia californica (a)–(c) and the rat sciatic nerve (d)–(f).

In Aplysia, a total of 28 trials were acquired from 3 nerves (3 different animals). In the rat, a total of 26 trials were acquired from 4 nerves (4 different animals). Equal radiant exposures from the same nerve and animal were combined into one data set. In Aplysia, a statistically significant increase (p < 0.05) in ROE size with increasing radiant exposure was observed for all nerve tested (figure 5(c)). For the rat, the results indicated a statistically significant increase in ROE size (p < 0.05) for one of the four animals tested (figure 5(f)) and an insignificant increase (p > 0.05) for the remaining nerves. However, combining the results from all four rat nerves shows a linear increase in ROE size across the radiant exposures tested. The lack of statistical significance in three of the four rat nerves tested is likely due to the limited range of radiant exposures tested in each nerve. However, the center of each ROE showed a greater probability of firing at the higher radiant exposure in all nerves (not shown).

3.3. Effects of stimulus polarity

It was hypothesized that the polarity of the electrical stimulus would shift the location of the ROE. To test this, the ROE was identified as before, and then the polarity was reversed (while keeping the electrodes in place) and the new ROE was found. In Aplysia, this experiment was repeated using both the Capella and Ho:YAG lasers with a constant optical stimulus (2.42–4.71 J cm⁻²) across a total of 8 nerves from 7 animals yielding 11 polarity pairs. The Mann–Whitney test was used to evaluate whether a shift in the ROE median occurred with a change in polarity. For all polarity pairs, a reversal in polarity showed a statistically significant shift (p < 0.05) in the ROE median such that the ROE was located adjacent to the cathode (figure 6). This demonstrates that, for a given electrode arrangement, two unique ROEs may be achieved by simply reversing the direction of the current path. In the rat sciatic nerve, effects of polarity were investigated using both the Ho:YAG and Capella lasers in a total of six nerves from four animals. A statistically significant shift in the ROE median was observed in three of the six nerves tested. Of the three nerves not showing a statistically significant shift in the ROE median, two exhibited successful hybrid stimulation with only one polarity. While statistically significant shifts in the ROE median were observed in half of the nerves tested, changes in location were not as dramatic as in the Aplysia.

Figure 6.

Changing the polarity of a sub-threshold electrical stimulus (90% of electrical stimulation threshold) in the Aplysia buccal nerve yields two distinct regions of excitability (ROEs) with both the (a) Capella (λ= 1.875 μm; τ_p = 3 ms; H = 4.97 J cm⁻²) and (b) Ho:YAG (λ= 2.120 μm; τ_p = 0.25 ms; H = 2.67 J cm⁻²) lasers. The location of the ROE is adjacent to the location of the cathode. The dark-colored circles represent locations of successful hybrid stimulation when the cathode is located on the left side of the nerve. The light-colored circles represent locations of successful hybrid stimulation when the polarity is reversed and the cathode is located on the right side of the nerve.

3.4. Effects of electrical stimulation threshold on hybrid stimulation

Electrical stimulation threshold currents as well as the RE₅₀ for hybrid stimulation were monitored in the same nerve to determine if fluctuations in the former affect the latter. The RE₅₀ for hybrid stimulation was determined by first generating probabilities of firing at a given radiant exposure for each time point (by dividing the number of stimulation attempts evoking a response by the number of total attempts) and then fitting those probabilities to a CDF (equation (1)). The RE₅₀ was defined as the radiant exposure providing a 50% probability of firing as indicated by the CDF fit.

For the Aplysia, 5 pulses of 5 radiant exposures (using the Capella laser) yielded 25 total data points every 2 min. These data were not sufficient for a reliable CDF fit at each time point, so a sliding window was applied to fit a CDF to 6 min windows of data. Figure 7(a) provides an example of the changes in thresholds for electrical stimulation and the optical component of hybrid stimulation over an hour. Each of the four Aplysia buccal nerves tested had a statistically significant (p < 0.05) negative correlation between thresholds for electrical stimulation and the optical component of hybrid stimulation. In the rat, 8 pulses of 5 radiant exposures (using the Ho:YAG laser) yielded 40 total data points every 3 min. A sliding window was applied to fit a CDF to 6 min windows of data. Of the two nerves tested, one exhibited a statistically significant (p < 0.05) negative correlation between thresholds for electrical stimulation (figure 8(a)) and the optical component of hybrid stimulation and the other showed an insignificant (p > 0.05) negative correlation.

Figure 7.

Electrical stimulation threshold and RE₅₀ for hybrid stimulation as a function of time in an Aplysia californica buccal nerve. (a) Results from one nerve showing a negative correlation (r² = −0.47, p < 0.05) between thresholds for electrical stimulation and the RE₅₀ for hybrid stimulation measured every 2 min. (b) Probability of firing as a function of radiant exposure using data accumulated from all animals. The slope of the CDF fit at 50% probability indicates the amount of variability in hybrid stimulation radiant exposures yielding stimulation over time. Effects of adjusting the electrical priming current every 2 min versus every 20 min are also shown. More frequent adjustments to the priming current increase the slope of the CDF fit, thus reducing variability in threshold radiant exposure for the optical component of hybrid stimulation. Note that the y-intercept for the 20 min adjustment plot is greater than 0, suggesting that there is a small probability of firing even with 0 J cm⁻² of optical stimulus. This is due to rare occasions where the electrical stimulation threshold fell below the previously set sub-threshold stimulus before the next adjustment was made.

Figure 8.

Electrical stimulation threshold and RE₅₀ for hybrid stimulation as a function of time in the rat sciatic nerve. (a) Results from one nerve showing a negative correlation (r² = −0.66, p < 0.05) between threshold for electrical stimulation and the RE₅₀ for hybrid stimulation measured every 2 min. (b) Probability of firing as a function of radiant exposure in each animal using all data acquired over 1 h. The slope of the CDF fit at 50% probability indicates the amount of variability in threshold measurements over time. There is more variability between animals in the rat than in Aplysia (figure 7(b)).

To evaluate the consistency over time of the RE₅₀ for hybrid stimulation, all of the data acquired from a given nerve were compiled and each radiant exposure was converted to a probability of firing. The probability of firing as a function of radiant exposure was then fit to a CDF. In Aplysia, a total of four nerves from four animals (n = 610 data points at each radiant exposure) yielded a 50% probability of firing at 1.34 J cm⁻² with a 95% confidence interval between 1.13 and 1.55 J cm⁻² (figure 7(b)). Here, the confidence interval is indicative of variability in hybrid stimulation RE₅₀ over the hour of measurements, where a narrow confidence interval (and increased slope of the CDF fit) indicates less variability. A subsequent set of experiments was performed in Aplysia to determine if increasing the interval between adjustments to the sub-threshold electrical stimulus yielded an increase in the confidence interval (i.e. an increase in variability). For these experiments, the electrical stimulation threshold was measured every 2 min, but the sub-threshold electrical stimulus used for hybrid stimulation was only set to 90% of electrical stimulation threshold at the 0, 20 and 40 min time points. A total of five nerves from three animals (n = 610–900 data points per radiant exposure) yielded a 50% probability of firing of 1.86 J cm⁻² with a 95% confidence interval between 1.40 and 2.33 J cm⁻². When comparing the 2 and 20 min adjustment intervals, the 95% confidence interval for the 20 min adjustment is roughly twice that of the 2 min adjustment. This is also shown in figure 7(b) as a shallower slope in the probability of firing as a function of radiant exposure for the 20 min adjustment. A noteworthy aspect of figure 7(b) is that the y-intercept for the 20 min adjustment plot is greater than 0, suggesting that there is a small probability of firing even with 0 J cm⁻² of optical stimulus. This is due to rare occasions where the electrical stimulation threshold fell below the previously set sub-threshold stimulus before the next adjustment was made.

Figure 8(b) shows the results of aggregating data from each rat for the purpose of assessing threshold radiant exposure consistency. Rather than compiling the data from both animals, each animal is plotted separately. The results indicate that threshold variability is more prominent in the rat than in Aplysia. Animal 1 has RE₅₀ of 0.13 J cm⁻² with a 95% confidence interval of 0.10–0.16 J cm⁻², whereas animal 2 has RE₅₀ of 0.25 J cm⁻² and a 95% confidence interval of 0.17–0.33 J cm⁻².

3.5. Hybrid inhibition

In the course of evaluating temporal factors affecting the RE₅₀ for hybrid stimulation in Aplysia, it was discovered that at higher radiant exposures, the probability of firing began to decrease rather than asymptotically approach 100% as expected. To further investigate this phenomenon, the electrical stimulus was set to 90% of electrical stimulation threshold every 2 min and five pulses of five radiant exposures were applied in the manner described above; however, for this experiment the radiant exposures were higher than those used for identifying the RE₅₀. The results from four nerves from two animals (n = 600 data points per radiant exposure) are shown in figure 9. Interestingly, if we define stimulation as >50% probability of firing, then with an electrical priming stimulus of 90% of electrical stimulation threshold, stimulation will occur for radiant exposures from 1.34 to 4.79 J cm⁻² rather than >1.34 J cm⁻² as was initially expected. This raised the question as to whether higher radiant exposures actually inhibit neuronal firing, or whether another mechanism is activated at these radiant exposures. We applied an electrical stimulus at 110% of electrical stimulation threshold and then added the optical stimulus (three nerves from three animals). In each trial, the electrically evoked unit was inhibited by the optical stimulus (figure 10). Radiant exposures for inhibition of the electrically evoked unit averaged 7.13 ± 0.51 J cm⁻² over 12 trials. It is important to note that all of these radiant exposures are below optical stimulation threshold radiant exposures and that this process is completely reversible. If radiant exposures are reduced, then the evoked response returns. Hybrid inhibition was investigated in the rat but was not observed.

Figure 9.

There is a limited window of radiant exposures for successful hybrid stimulation in Aplysia. A >50% probability of firing with an electrical stimulus at 90% of electrical stimulation threshold requires radiant exposures from 1.34 to 4.79 J cm⁻². Evoked responses to a range of radiant exposures were acquired every 2 min for 1 h. These data were aggregated to achieve a probability of firing for each radiant exposure. The increasing and decreasing phases of the plot were then each fitted to a CDF.

Figure 10.

Optical stimulation of sufficient radiant exposure will inhibit electrically evoked action potentials. In both (a) and (b), a supra-threshold stimulus (110% of threshold) is applied (100 μs, 567 μA). In (a), the optical stimulus (3 ms) is 5.73 J cm⁻², whereas in (b), the optical stimulus is 6.49 J cm⁻². Note how the electrically evoked action potential is present in (a) but not in (b). The electrical stimulation artifact is indicated by the shaded region.

4. Discussion

Reducing the optical energy required to stimulate excitable tissues may facilitate clinical translation of infrared neural interfaces due to the reduced likelihood of thermal tissue damage, and by making the design criteria for laser sources less restrictive. The purpose of this study was to assess potential factors that might contribute to variability in hybrid electrooptical stimulation, as well as to create a methodology for reliable and reproducible hybrid stimulation. We approached this task by comparing trends seen in two different neurobiological systems—the tractable and well-characterized Aplysia californica buccal ganglion and the myelinated and more clinically relevant rat sciatic nerve. Given the variability and lack of reproducibility we previously experienced, this approach allowed for identification of factors in the more experimentally tractable system that could subsequently be applied to the more clinically relevant preparation. Some concern may arise as to the translation of hybrid stimulation between an unmyelinated, invertebrate nerve and amyelinated, mammalian nerve. However, this study shows that the information gathered from experiments in Aplysia directly led to improved understanding and performance of hybrid stimulation in the rat sciatic nerve. Although some aspects of the experimental protocol differ between the two preparations (i.e. orientation of stimulating pipettes, source of optical stimulation, endpoint definition), overarching trends were clearly evident across both species. Prior to both adopting the methods used in this study and controlling for the spatial and temporal factors we have assessed, our efficacy for hybrid stimulation in the Aplysia buccal nerve and the rat sciatic nerve was 35% and 23%, respectively (unpublished data). In this paper, we define efficacy as a nerve demonstrating a hybrid stimulation event where a sub-threshold electrical stimulus and sub-threshold optical stimulus are combined to achieve an evoked response. We attempt to determine whether or not sub-threshold electrical and optical stimuli were combined to achieve supra-threshold stimulation. At the conclusion of this study, we now have an efficacy of 93% (42/45 nerves) in the Aplysia buccal nerve and 76% (13/17 nerves) in the rat sciatic nerve.

Relative mechanical stability between the target neural tissue, optical fiber and electrodes was imperative to achieving reliable and reproducible hybrid stimulation. This allowed for consistent location of the stimuli throughout a given experiment by minimizing nerve movement due to optical fiber movement, fluid flow (Aplysia) or animal respiration (rat). Stabilization challenges are likely to be alleviated as hybrid stimulation progresses to multi-modality nerve cuff stimulators where microfabricated cuffs will be able to adapt to changes in nerve shape and movement.

The orientation of the stimulating glass pipettes is also an important part of the physical setup that must be taken into account. In the rat, electrical stimulation was more reliable with the pipettes oriented along the longitudinal axis of the nerve than in a transverse configuration. For electrical stimulation of myelinated nerves, it is necessary to induce longitudinal axonal currents, which may explain the reason that pipettes oriented longitudinally to the nerve were most effective. Recent models of intrafascicular stimulation support these observations. As a function of position relative to nodes of Ranvier, bipolar stimulation with a longitudinal configuration was shown to have less variability in threshold currents as compared to a transverse configuration [37]. While Aplysia nerves are unmyelinated, and thus do not possessnodes of Ranvier, they do exhibit clustering of voltage-gated sodium channels that may aid in the conduction of action potentials along the nerves [38]. However, it was found in Aplysia nerves that electrical stimulation was more reliable with the pipettes oriented transverse to the nerve. Due to the thick outer sheath protecting the nerve, placing the glass pipettes along the longitudinal axis of the nerve may result in electrical current dissipating into the bath rather than penetrating to the axons. When placing the pipettes transverse with respect to the midline of the nerve, the current may take a more direct path through the axonal tissue.

The choice of laser is also a contributor to the reproducibility of hybrid stimulation. The two lasers used in this study differ in many respects, but are expected to perform equally from the point of view of thermal laser–tissue interaction. However, the Ho:YAG laser yielded greater reproducibility in the rat than did the Capella. To understand how this may have occurred, we must examine the two laser sources. The Capella used for this study is a diode laser, which is chopped to produce square pulses having tunable pulse duration at a center wavelength of 1.875 μm. The Ho:YAG laser is a pulsed solid-state laser at 2.12 μm, which produces a 250 μs pulse (full width at half maximum), exhibiting an initial rising phase followed by a decay, with spikes in output energy throughout the pulse duration. The mechanism by which pulsed infrared light produces neural activation is known to be thermally mediated, and directly associated with the absorption of infrared light by water in tissue [17]. Which attribute of the laser contributes most significantly to the thermal gradient is the most relevant issue. A comparison of the absorption coefficient as a function of wavelength for pure water reveals that 1.875 μm and 2.12 μm have similar absorption coefficients (μ_a= 26.9 cm⁻¹ and μ_a= 24.01 cm⁻¹, respectively) [30]. Although tissue is predominantly water, these values may differ slightly in our preparation and are known to be temperature dependent. However, it is unlikely that the differing wavelengths of the lasers is the source of the Ho:YAG laser’s superior reproducibility in myelinated peripheral nerves. A second obvious difference is the pulse durations of the two lasers. However, there is conflicting evidence as to whether pulse duration plays a role in optical stimulation thresholds [17, 32]. A third possibility is that the broad spectral width of the Capella (15–20 nm, FWHM) causes much of the laser’s output to occur at wavelengths that are not optimal for optical stimulation of peripheral, myelinated nerves. In applications with more direct access to the target neural tissue, the effects of spectral width are minimized due to all of the light being absorbed at the site of neuronal activation. However, in peripheral nerves, where the optical energy must penetrate through connective tissue and myelin surrounding the axons, longer wavelengths emitted by the Capella may be absorbed before they ever reach the axons. Thus, stimulation thresholds would be higher and quickly approach damage thresholds. The differing temporal pulse structure has not been investigated, but may also contribute to the relative effectiveness of the lasers. Whereas the Capella is a chopped diode laser exhibiting a square pulse, the Ho:YAG laser has a temporal structure in which the optical energy varies and includes numerous energy spikes throughout the pulse duration [39]. This could result in higher peak power and peak irradiance for the Ho:YAG laser.

There are two broad categories of factors that affect the reproducibility of hybrid stimulation related to the interaction of the optical and electrical stimuli. In the first category are spatial factors, where the relative location of the two stimuli determines the efficacy of stimulation. Our initial working hypothesis was that for a given sub-threshold radiant exposure, hybrid stimulation would be possible for all locations between the cathode and anode of a bipolar stimulus. The results of this study have shown that hypothesis to be false. In figure 3, it is clear that there is a finite ROE for the combination of a constant sub-threshold radiant exposure delivered simultaneously with an electrical stimulus that is 90% of electrical stimulation threshold. While figure 3 is drawn from data in the Aplysia buccal nerve, figure 4 shows that the same results were seen in the rat sciatic nerve as well. Therefore, successful and reproducible hybrid stimulation calls for accurate placement of the optical fiber relative to the site of electrical stimulation.

This raises the question of where the ROE is located. This answer is clearer in Aplysia, where the ROE was consistently located adjacent to the cathode. Within a single nerve, the location of the ROE was effectively ‘steered’ by reversing the polarity of the electrical stimulus. In the rat sciatic nerve, half of the nerves showed a statistically significant shift in ROE location upon polarity reversal, though the effect was not as dramatic as in Aplysia. In the other trials, the ROE location either did not shift, or hybrid stimulation was ineffective when the polarity was reversed. However, in cases of successful hybrid stimulation, different evoked potentials were recruited for each stimulus polarity. This suggests that hybrid stimulation offers two forms of selectivity, as both the position of the optical stimulus and the polarity of the electrical stimulus dictate the units recruited. The results also imply that ROE location in the rat sciatic nerve is influenced more by whether or not optical stimulation is possible rather than by the direction of current flow. Anecdotal evidence reveals that there are ‘sweet spots’ on the sciatic nerve where optical stimulation is most effective; in particular, these spots are found just proximal to the branch point of the fascicles, but also at some additional locations along the nerve trunk. This could potentially be due to thinning of the epineurium, proximity of fascicles to the irradiated surface or to increased concentration of nodes of Ranvier in these locations.

The existence of a finite ROE with the potential for shifting location in response to polarity reversal must be taken into account for reproducible hybrid stimulation. Much of the previously observed variability is also likely to be due to the relationship between ROE size and applied radiant exposure. The results indicate an approximately linear increase in ROE size over the range of radiant exposures tested (figure 4(f)). Thus, the center of the ROE will have the lowest threshold radiant exposures when combined with a given sub-threshold electrical stimulus. If this is not accounted for (as was the case in [14]), then variability in measured thresholds is certainly expected. Furthermore, with the highest probability of firing at the center (figure 5), it is likely that an optical stimulus located along the periphery of the ROE will induce a reduced firing rate.

A second category of factors contributing to the reproducibility of hybrid stimulation is temporal factors. These factors include how the electrical stimulation threshold and the hybrid stimulation RE₅₀ change with time and relative to one another. We initially expected that the excitability of a nerve to the combination of electrical and hybrid optical stimuli would follow a similar temporal pattern. However, figures 7 and 8 illustrate a negative correlation between the electrical and hybrid optical stimuli in both Aplysia and rat. If the sub-threshold electrical stimulus is set and the underlying electrical stimulation threshold subsequently decreases (so that an electrical stimulus approaches the stimulation threshold), one would expect the threshold for the optical component of hybrid stimulation to be reduced as well. However, the results did not show this to be true. Thus, we may conclude that the underlying mechanisms of optical and electrical stimulation are dissimilar. If the mechanisms were similar, we would expect a positive correlation between thresholds for electrical stimulation and the optical component of hybrid stimulation. Instead, our data show that as the nerve becomes more excitable to electrical stimulation, its excitability in response to optical stimulation decreases. In the rat, an unexpected decay of electrical threshold currents over time was observed (figure 8). This decay may be a sign of increased excitability in response to surgery or trauma.

The underlying electrical stimulation threshold must be taken into account to reduce variability and enhance the reproducibility of hybrid stimulation. Whenever short-term fluctuations (minutes) in threshold radiant exposures are present, controlling for these fluctuations yields overall long-term (1 h) threshold radiant exposures that are consistent (figures 7(b) and 8(b)). If electrical stimulation threshold is not controlled over time (as was the case in [14]), then the variability of measured thresholds for the optical component of hybrid stimulation will increase. This is evident in figure 7(b). When the sub-threshold electrical stimulus was only set to the chosen magnitude every 20 min, the threshold for the optical component of hybrid stimulation increased and its 95% confidence interval (indicative of the variability) showed greater than a twofold increase. It should be noted that while the inter-rat variability represented in figure 8(b) is much greater than in Aplysia (figure 7(b)), the overall variability and reduction in INS threshold are much lower than what was previously reported. Taking the minimum bound of the 95% RE₅₀ confidence interval for animal 1 and the maximum bound for animal 2 yields an RE₅₀ for hybrid stimulation ranging from 12% to 29% of the radiant exposures required for optical stimulation alone, as opposed to the roughly 30–80% in the previous study.

In the course of investigating temporal factors affecting hybrid stimulation, it was discovered that elevated radiant exposures (although still below threshold radiant exposures for optical stimulation alone) resulted in a decline in the probability of firing (figure 9). Sub-threshold radiant exposures for optical stimulation alone were also shown to inhibit electrically evoked potentials (figure 10). These results indicate that the potential exists for full hybrid electro-optical control of neural tissue, making it possible to selectively excite or inhibit axons. Preliminary results indicate a spatially confined region of inhibition surrounded by excitation (either hybrid or electrically evoked), suggesting that this is not an artifact, but is a spatially discrete phenomenon, although it may be due to a different mechanism than the excitatory effect. Without an elucidated mechanism of INS, it is difficult to conclude how pulsed infrared light inhibits electrically evoked potentials. Recently, it was shown that intracellular calcium increases in response to optical stimulation of cardiomyocytes [26]. It is conceivable that for hybrid stimulation, supra-threshold radiant exposures may cause an increase in intracellular calcium that activates calcium-dependent potassium channels, thus hyperpolarizing the cell. Further studies will be required to test this hypothesis.

We previously showed the proof-of-concept potential for combined optical and electrical stimulation of neural tissue [14]. This study extends that work by outlining some potential sources of variability that may be controlled to provide reproducible hybrid stimulation. The results presented here also demonstrate the potential of combining optical and electrical stimulation techniques by providing further evidence for selectivity as well as the ability to inhibit neuronal firing. Finally, the study demonstrates the translational value of parallel studies in invertebrates and vertebrates. The key aspects of the methodology to capitalize on the potential of hybrid electro-optical stimulation are summarized as follows.

• The optical stimulus, electrical stimulus and target tissue should be mechanically stabilized and controlled relative to one another.

• The laser and target neural anatomy must be taken into account to determine the maximum possible expected reproducibility.

• For constant electrical priming current, the optical stimulus must be located within the ROE.

• For constant electrical priming current, the size of the ROE depends on the strength of the optical stimulus.

• Variability in the electrical stimulation threshold induces variability in the RE₅₀ for hybrid stimulation. This variability can be reduced by frequent adjustments to maintain a constant sub-threshold electrical stimulus relative to the electrical stimulation threshold.

• There is a range of radiant exposures for which hybrid stimulation has >50% probability of firing. Radiant exposures below or above this range have <50% probability of firing.

Having taken these points into account, we have improved our efficacy by threefold in both the Aplysia californica buccal nerve and the rat sciatic nerve. There are other potential sources of variability that could be controlled to bring our current efficacy up to 100%. In Aplysia, the three nerves that did not show hybrid stimulation were from animals with questionable health, but were included in the success rate calculations for completeness. In myelinated peripheral nerves, the efficacy of optical stimulation is crucial to the success of hybrid stimulation. Elucidating the mechanism of INS will provide a priori knowledge of where on the nerves to stimulate (e.g., near the nodes of Ranvier). Improving the efficacy of optical stimulation will in turn improve the efficacy and reduce variability of hybrid stimulation. Knowing the mechanism of INS will also provide a clearer understanding of the interaction between electrical and optical stimuli that drives hybrid stimulation. In this study, we have demonstrated that mechanical stabilization of the nerve, electrodes and optical fiber is of utmost importance. Even with the efforts we have taken to stabilize the system, there is potentially still movement-inducing variability. To address this issue, we envision a hybrid stimulation cuff that moves with the nerve and is thus able to hold the stimuli in place relative to the nerve. However, our results thus far have provided the ability to begin assessing the clinical utility of hybrid neural stimulation. We believe that the concepts and techniques presented in this study will facilitate the application of spatially selective neural interfaces where thermal tissue damage and/or laser design constraints are currently of concern.